- refactor core library

- refactor python and wasm bindings
- update python pipeline evaluate step to use the core library
- add multi-threading support to the core library, cli application and python library
- first version of unseeded data synthesis in rust (supported on core library, cli and python pipeline)
- general performance optimizations on the core library, python pipeline and web application

.eslintignore

@@ -13,5 +13,4 @@ node_modules/
 pkg/
 target/
 .eslintrc.js
-babel.config.js
-packages/webapp/src/workers/sds-wasm/worker.js
+babel.config.js

.pnp.cjs

@@ -13697,10 +13697,10 @@ function $$SETUP_STATE(hydrateRuntimeState, basePath) {
         }]
       ]],
       ["sds-wasm", [
-        ["file:../../target/wasm#../../target/wasm::hash=932617&locator=webapp%40workspace%3Apackages%2Fwebapp", {
-          "packageLocation": "./.yarn/cache/sds-wasm-file-aa35777e98-1550beb7ae.zip/node_modules/sds-wasm/",
+        ["file:../../target/wasm#../../target/wasm::hash=3f33a3&locator=webapp%40workspace%3Apackages%2Fwebapp", {
+          "packageLocation": "./.yarn/cache/sds-wasm-file-6dcf65ffd1-dd8b934db1.zip/node_modules/sds-wasm/",
           "packageDependencies": [
-            ["sds-wasm", "file:../../target/wasm#../../target/wasm::hash=932617&locator=webapp%40workspace%3Apackages%2Fwebapp"]
+            ["sds-wasm", "file:../../target/wasm#../../target/wasm::hash=3f33a3&locator=webapp%40workspace%3Apackages%2Fwebapp"]
           ],
           "linkType": "HARD",
         }]
@@ -15482,7 +15482,7 @@ function $$SETUP_STATE(hydrateRuntimeState, basePath) {
             ["react-is", "npm:17.0.2"],
             ["react-router-dom", "virtual:d293af44cc1e0d0fc09cc0c8c4a3d9e5fccdf4ddebae06b8fad52a312360d8122c830d53ecc46b13c13aaad8c6ae7dbd798566bd5cba581433425b2ff3f7540b#npm:6.0.2"],
             ["recoil", "virtual:d293af44cc1e0d0fc09cc0c8c4a3d9e5fccdf4ddebae06b8fad52a312360d8122c830d53ecc46b13c13aaad8c6ae7dbd798566bd5cba581433425b2ff3f7540b#npm:0.5.2"],
-            ["sds-wasm", "file:../../target/wasm#../../target/wasm::hash=932617&locator=webapp%40workspace%3Apackages%2Fwebapp"],
+            ["sds-wasm", "file:../../target/wasm#../../target/wasm::hash=3f33a3&locator=webapp%40workspace%3Apackages%2Fwebapp"],
             ["styled-components", "virtual:d293af44cc1e0d0fc09cc0c8c4a3d9e5fccdf4ddebae06b8fad52a312360d8122c830d53ecc46b13c13aaad8c6ae7dbd798566bd5cba581433425b2ff3f7540b#npm:5.3.3"],
             ["ts-node", "virtual:d293af44cc1e0d0fc09cc0c8c4a3d9e5fccdf4ddebae06b8fad52a312360d8122c830d53ecc46b13c13aaad8c6ae7dbd798566bd5cba581433425b2ff3f7540b#npm:10.4.0"],
             ["typescript", "patch:typescript@npm%3A4.4.3#~builtin<compat/typescript>::version=4.4.3&hash=32657b"],

.vsts-ci.yml

@@ -1,33 +0,0 @@
-name: SDS CI
-pool:
-  vmImage: ubuntu-latest
-
-trigger:
-  batch: true
-  branches:
-    include:
-      - main
-
-stages:
-  - stage: Compliance
-    dependsOn: []
-    jobs:
-      - job: ComplianceJob
-        pool:
-          vmImage: windows-latest
-        steps:
-          - task: CredScan@3
-            inputs:
-              outputFormat: sarif
-              debugMode: false
-
-          - task: ComponentGovernanceComponentDetection@0
-            inputs:
-              scanType: 'Register'
-              verbosity: 'Verbose'
-              alertWarningLevel: 'High'
-
-          - task: PublishSecurityAnalysisLogs@3
-            inputs:
-              ArtifactName: 'CodeAnalysisLogs'
-              ArtifactType: 'Container'

Cargo.lock

@@ -42,6 +42,12 @@ dependencies = [
  "winapi",
 ]
 
+[[package]]
+name = "autocfg"
+version = "1.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cdb031dd78e28731d87d56cc8ffef4a8f36ca26c38fe2de700543e627f8a464a"
+
 [[package]]
 name = "base-x"
 version = "0.2.8"
@@ -103,6 +109,50 @@ dependencies = [
  "wasm-bindgen",
 ]
 
+[[package]]
+name = "crossbeam-channel"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "06ed27e177f16d65f0f0c22a213e17c696ace5dd64b14258b52f9417ccb52db4"
+dependencies = [
+ "cfg-if",
+ "crossbeam-utils",
+]
+
+[[package]]
+name = "crossbeam-deque"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6455c0ca19f0d2fbf751b908d5c55c1f5cbc65e03c4225427254b46890bdde1e"
+dependencies = [
+ "cfg-if",
+ "crossbeam-epoch",
+ "crossbeam-utils",
+]
+
+[[package]]
+name = "crossbeam-epoch"
+version = "0.9.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4ec02e091aa634e2c3ada4a392989e7c3116673ef0ac5b72232439094d73b7fd"
+dependencies = [
+ "cfg-if",
+ "crossbeam-utils",
+ "lazy_static",
+ "memoffset",
+ "scopeguard",
+]
+
+[[package]]
+name = "crossbeam-utils"
+version = "0.8.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d82cfc11ce7f2c3faef78d8a684447b40d503d9681acebed6cb728d45940c4db"
+dependencies = [
+ "cfg-if",
+ "lazy_static",
+]
+
 [[package]]
 name = "csv"
 version = "1.1.6"
@@ -307,6 +357,25 @@ version = "2.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "308cc39be01b73d0d18f82a0e7b2a3df85245f84af96fdddc5d202d27e47b86a"
 
+[[package]]
+name = "memoffset"
+version = "0.6.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5aa361d4faea93603064a027415f07bd8e1d5c88c9fbf68bf56a285428fd79ce"
+dependencies = [
+ "autocfg",
+]
+
+[[package]]
+name = "num_cpus"
+version = "1.13.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "05499f3756671c15885fee9034446956fff3f243d6077b91e5767df161f766b3"
+dependencies = [
+ "hermit-abi",
+ "libc",
+]
+
 [[package]]
 name = "once_cell"
 version = "1.8.0"
@@ -499,6 +568,31 @@ dependencies = [
  "rand_core",
 ]
 
+[[package]]
+name = "rayon"
+version = "1.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c06aca804d41dbc8ba42dfd964f0d01334eceb64314b9ecf7c5fad5188a06d90"
+dependencies = [
+ "autocfg",
+ "crossbeam-deque",
+ "either",
+ "rayon-core",
+]
+
+[[package]]
+name = "rayon-core"
+version = "1.9.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d78120e2c850279833f1dd3582f730c4ab53ed95aeaaaa862a2a5c71b1656d8e"
+dependencies = [
+ "crossbeam-channel",
+ "crossbeam-deque",
+ "crossbeam-utils",
+ "lazy_static",
+ "num_cpus",
+]
+
 [[package]]
 name = "redox_syscall"
 version = "0.2.10"
@@ -582,6 +676,9 @@ dependencies = [
  "lru",
  "pyo3",
  "rand",
+ "rayon",
+ "serde",
+ "serde_json",
 ]
 
 [[package]]
@@ -589,6 +686,7 @@ name = "sds-pyo3"
 version = "1.0.0"
 dependencies = [
  "csv",
+ "env_logger",
  "log",
  "pyo3",
  "sds-core",

docker-compose.yml

@@ -1,8 +1,8 @@
-version: "3"
+version: '3'
 services:
-    webapp:
-        build:
-            context: .
-            dockerfile: webapp.dockerfile
-        ports:
-            - 3000:80
+  webapp:
+    build:
+      context: .
+      dockerfile: webapp.dockerfile
+    ports:
+      - 3000:80

package.json

@@ -7,7 +7,7 @@
 		"build:": "yarn workspaces foreach -ivt run build",
 		"start:": "yarn workspaces foreach -piv run start",
 		"lint:": "essex lint --fix --strict",
-		"build:lib-wasm": "cd packages/lib-wasm && wasm-pack build --release --target no-modules --out-dir ../../target/wasm",
+		"build:lib-wasm": "cd packages/lib-wasm && wasm-pack build --release --target web --out-dir ../../target/wasm",
 		"prettify": "essex prettify",
 		"rebuild-all": "cargo clean && run-s clean: && cargo build --release && run-s build:lib-wasm && yarn install && run-s build:"
 	},

packages/cli/Cargo.toml

@@ -7,7 +7,7 @@ repository = "https://github.com/microsoft/synthetic-data-showcase"
 edition = "2018"
 
 [dependencies]
-sds-core = { path = "../core" }
+sds-core = { path = "../core", features = ["rayon"] }
 log = { version = "0.4" }
 env_logger = { version = "0.9" }
 structopt = { version = "0.3" }

packages/cli/src/main.rs

@@ -1,8 +1,11 @@
 use log::{error, info, log_enabled, trace, Level::Debug};
 use sds_core::{
-    data_block::{block::DataBlockCreator, csv_block_creator::CsvDataBlockCreator},
-    processing::{aggregator::Aggregator, generator::Generator},
-    utils::reporting::LoggerProgressReporter,
+    data_block::{csv_block_creator::CsvDataBlockCreator, data_block_creator::DataBlockCreator},
+    processing::{
+        aggregator::Aggregator,
+        generator::{Generator, SynthesisMode},
+    },
+    utils::{reporting::LoggerProgressReporter, threading::set_number_of_threads},
 };
 use std::process;
 use structopt::StructOpt;
@@ -26,6 +29,14 @@ enum Command {
             default_value = "100000"
         )]
         cache_max_size: usize,
+        #[structopt(
+            long = "mode",
+            help = "synthesis mode",
+            possible_values = &["seeded", "unseeded"],
+            case_insensitive = true,
+            default_value = "seeded"
+        )]
+        mode: SynthesisMode,
     },
     Aggregate {
         #[structopt(long = "aggregates-path", help = "generated aggregates file path")]
@@ -104,6 +115,12 @@ struct Cli {
         help = "columns where zeros should not be ignored (can be set multiple times)"
     )]
     sensitive_zeros: Vec<String>,
+
+    #[structopt(
+        long = "n-threads",
+        help = "number of threads used to process the data in parallel (default is the number of cores)"
+    )]
+    n_threads: Option<usize>,
 }
 
 fn main() {
@@ -118,6 +135,10 @@ fn main() {
 
     trace!("execution parameters: {:#?}", cli);
 
+    if let Some(n_threads) = cli.n_threads {
+        set_number_of_threads(n_threads);
+    }
+
     match CsvDataBlockCreator::create(
         csv::ReaderBuilder::new()
             .delimiter(cli.sensitive_delimiter.chars().next().unwrap() as u8)
@@ -131,15 +152,20 @@ fn main() {
                 synthetic_path,
                 synthetic_delimiter,
                 cache_max_size,
+                mode,
             } => {
-                let mut generator = Generator::new(&data_block);
-                let generated_data =
-                    generator.generate(cli.resolution, cache_max_size, "", &mut progress_reporter);
+                let mut generator = Generator::new(data_block);
+                let generated_data = generator.generate(
+                    cli.resolution,
+                    cache_max_size,
+                    String::from(""),
+                    mode,
+                    &mut progress_reporter,
+                );
 
-                if let Err(err) = generator.write_records(
-                    &generated_data.synthetic_data,
+                if let Err(err) = generated_data.write_synthetic_data(
                     &synthetic_path,
-                    synthetic_delimiter.chars().next().unwrap() as u8,
+                    synthetic_delimiter.chars().next().unwrap(),
                 ) {
                     error!("error writing output file: {}", err);
                     process::exit(1);
@@ -153,28 +179,24 @@ fn main() {
                 sensitivity_threshold,
                 records_sensitivity_path,
             } => {
-                let mut aggregator = Aggregator::new(&data_block);
+                let mut aggregator = Aggregator::new(data_block);
                 let mut aggregated_data = aggregator.aggregate(
                     reporting_length,
                     sensitivity_threshold,
                     &mut progress_reporter,
                 );
-                let privacy_risk =
-                    aggregator.calc_privacy_risk(&aggregated_data.aggregates_count, cli.resolution);
+                let privacy_risk = aggregated_data.calc_privacy_risk(cli.resolution);
 
                 if !not_protect {
-                    Aggregator::protect_aggregates_count(
-                        &mut aggregated_data.aggregates_count,
-                        cli.resolution,
-                    );
+                    aggregated_data.protect_aggregates_count(cli.resolution);
                 }
 
                 info!("Calculated privacy risk is: {:#?}", privacy_risk);
 
-                if let Err(err) = aggregator.write_aggregates_count(
-                    &aggregated_data.aggregates_count,
+                if let Err(err) = aggregated_data.write_aggregates_count(
                     &aggregates_path,
                     aggregates_delimiter.chars().next().unwrap(),
+                    ";",
                     cli.resolution,
                     !not_protect,
                 ) {
@@ -183,9 +205,7 @@ fn main() {
                 }
 
                 if let Some(path) = records_sensitivity_path {
-                    if let Err(err) = aggregator
-                        .write_records_sensitivity(&aggregated_data.records_sensitivity, &path)
-                    {
+                    if let Err(err) = aggregated_data.write_records_sensitivity(&path, '\t') {
                         error!("error writing output file: {}", err);
                         process::exit(1);
                     }

packages/core/Cargo.toml

@@ -18,4 +18,7 @@ getrandom = { version = "0.2", features = ["js"] }
 log = { version = "0.4", features = ["std"] }
 csv = { version = "1.1" }
 instant = { version = "0.1", features = [ "stdweb", "wasm-bindgen" ] }
-pyo3 = { version = "0.15", features = ["extension-module"], optional = true }
+pyo3 = { version = "0.15", features = ["extension-module"], optional = true }
+rayon = { version = "1.5", optional = true }
+serde = { version = "1.0", features = [ "derive", "rc" ] }
+serde_json = { version = "1.0" }

packages/core/src/data_block/block.rs

@@ -1,12 +1,16 @@
 use super::{
-    record::DataBlockRecord,
     typedefs::{
-        AttributeRows, AttributeRowsMap, CsvRecord, CsvRecordSlice, DataBlockHeaders,
-        DataBlockRecords, DataBlockRecordsSlice,
+        AttributeRows, AttributeRowsByColumnMap, AttributeRowsMap, ColumnIndexByName,
+        DataBlockHeaders, DataBlockRecords,
     },
     value::DataBlockValue,
 };
-use std::collections::HashSet;
+use fnv::FnvHashMap;
+use itertools::Itertools;
+use serde::{Deserialize, Serialize};
+use std::sync::Arc;
+
+use crate::{processing::aggregator::typedefs::RecordsSet, utils::math::uround_down};
 
 #[cfg(feature = "pyo3")]
 use pyo3::prelude::*;
@@ -15,189 +19,162 @@ use pyo3::prelude::*;
 /// The goal of this is to allow data processing to handle with memory references
 /// to the data block instead of copying data around
 #[cfg_attr(feature = "pyo3", pyclass)]
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct DataBlock {
-    /// Vector of strings representhing the data headers
+    /// Vector of strings representing the data headers
     pub headers: DataBlockHeaders,
     /// Vector of data records, where each record represents a row (headers not included)
     pub records: DataBlockRecords,
 }
 
 impl DataBlock {
+    /// Returns a new DataBlock with default values
+    pub fn default() -> DataBlock {
+        DataBlock {
+            headers: DataBlockHeaders::default(),
+            records: DataBlockRecords::default(),
+        }
+    }
+
     /// Returns a new DataBlock
     /// # Arguments
-    /// * `headers` - Vector of string representhing the data headers
+    /// * `headers` - Vector of string representing the data headers
     /// * `records` - Vector of data records, where each record represents a row (headers not included)
     #[inline]
     pub fn new(headers: DataBlockHeaders, records: DataBlockRecords) -> DataBlock {
         DataBlock { headers, records }
     }
 
-    /// Calcules the rows where each value on the data records is present
-    /// # Arguments
-    /// * `records`: List of records to analyze
+    /// Returns a map of column name -> column index
     #[inline]
-    pub fn calc_attr_rows(records: &DataBlockRecordsSlice) -> AttributeRowsMap {
+    pub fn calc_column_index_by_name(&self) -> ColumnIndexByName {
+        self.headers
+            .iter()
+            .enumerate()
+            .map(|(column_index, column_name)| ((**column_name).clone(), column_index))
+            .collect()
+    }
+
+    /// Calculates the rows where each value on the data records is present
+    #[inline]
+    pub fn calc_attr_rows(&self) -> AttributeRowsMap {
         let mut attr_rows: AttributeRowsMap = AttributeRowsMap::default();
 
-        for (i, r) in records.iter().enumerate() {
+        for (i, r) in self.records.iter().enumerate() {
             for v in r.values.iter() {
                 attr_rows
-                    .entry(v)
+                    .entry(v.clone())
                     .or_insert_with(AttributeRows::new)
                     .push(i);
             }
         }
         attr_rows
     }
-}
-
-/// Trait that needs to be implement to create a data block.
-/// It already contains the logic to create the data block, so we only
-/// need to worry about mapping the headers and records from InputType
-pub trait DataBlockCreator {
-    /// Creator input type, it can be a File Reader, another data structure...
-    type InputType;
-    /// The error type that can be generated when parsing headers/records
-    type ErrorType;
 
+    // Calculates the rows where each value on the data records is present
+    /// grouped by column index.
     #[inline]
-    fn gen_use_columns_set(headers: &CsvRecordSlice, use_columns: &[String]) -> HashSet<usize> {
-        let use_columns_str_set: HashSet<String> = use_columns
-            .iter()
-            .map(|c| c.trim().to_lowercase())
-            .collect();
-        headers
-            .iter()
-            .enumerate()
-            .filter_map(|(i, h)| {
-                if use_columns_str_set.is_empty()
-                    || use_columns_str_set.contains(&h.trim().to_lowercase())
-                {
-                    Some(i)
-                } else {
-                    None
-                }
+    pub fn calc_attr_rows_with_no_empty_values(&self) -> AttributeRowsByColumnMap {
+        let mut attr_rows_by_column: AttributeRowsByColumnMap = AttributeRowsByColumnMap::default();
+
+        for (i, r) in self.records.iter().enumerate() {
+            for v in r.values.iter() {
+                attr_rows_by_column
+                    .entry(v.column_index)
+                    .or_insert_with(AttributeRowsMap::default)
+                    .entry(v.clone())
+                    .or_insert_with(AttributeRows::new)
+                    .push(i);
+            }
+        }
+        attr_rows_by_column
+    }
+
+    /// Calculates the rows where each value on the data records is present
+    /// grouped by column index. This will also include empty values mapped with
+    /// the `empty_value` string.
+    /// # Arguments
+    /// * `empty_value` - Empty values on the final synthetic data will be represented by this
+    #[inline]
+    pub fn calc_attr_rows_by_column(&self, empty_value: &Arc<String>) -> AttributeRowsByColumnMap {
+        let mut attr_rows_by_column: FnvHashMap<
+            usize,
+            FnvHashMap<Arc<DataBlockValue>, RecordsSet>,
+        > = FnvHashMap::default();
+        let empty_records: RecordsSet = (0..self.records.len()).collect();
+
+        // start with empty values for all columns
+        for column_index in 0..self.headers.len() {
+            attr_rows_by_column
+                .entry(column_index)
+                .or_insert_with(FnvHashMap::default)
+                .entry(Arc::new(DataBlockValue::new(
+                    column_index,
+                    empty_value.clone(),
+                )))
+                .or_insert_with(|| empty_records.clone());
+        }
+
+        // go through each record and map where the values occur on the columns
+        for (i, r) in self.records.iter().enumerate() {
+            for value in r.values.iter() {
+                let current_attr_rows = attr_rows_by_column
+                    .entry(value.column_index)
+                    .or_insert_with(FnvHashMap::default);
+
+                // insert it on the correspondent entry for the data block value
+                current_attr_rows
+                    .entry(value.clone())
+                    .or_insert_with(RecordsSet::default)
+                    .insert(i);
+                // it's now used being used, so we make sure to remove this from the column empty records
+                current_attr_rows
+                    .entry(Arc::new(DataBlockValue::new(
+                        value.column_index,
+                        empty_value.clone(),
+                    )))
+                    .or_insert_with(RecordsSet::default)
+                    .remove(&i);
+            }
+        }
+
+        // sort the records ids, so we can leverage the intersection alg later on
+        attr_rows_by_column
+            .drain()
+            .map(|(column_index, mut attr_rows)| {
+                (
+                    column_index,
+                    attr_rows
+                        .drain()
+                        .map(|(value, mut rows_set)| (value, rows_set.drain().sorted().collect()))
+                        .collect(),
+                )
             })
             .collect()
     }
 
     #[inline]
-    fn gen_sensitive_zeros_set(
-        filtered_headers: &CsvRecordSlice,
-        sensitive_zeros: &[String],
-    ) -> HashSet<usize> {
-        let sensitive_zeros_str_set: HashSet<String> = sensitive_zeros
-            .iter()
-            .map(|c| c.trim().to_lowercase())
-            .collect();
-        filtered_headers
-            .iter()
-            .enumerate()
-            .filter_map(|(i, h)| {
-                if sensitive_zeros_str_set.contains(&h.trim().to_lowercase()) {
-                    Some(i)
-                } else {
-                    None
-                }
-            })
-            .collect()
+    /// Returns the number of records on the data block
+    pub fn number_of_records(&self) -> usize {
+        self.records.len()
     }
 
     #[inline]
-    fn map_headers(
-        headers: &mut CsvRecord,
-        use_columns: &[String],
-        sensitive_zeros: &[String],
-    ) -> (CsvRecord, HashSet<usize>, HashSet<usize>) {
-        let use_columns_set = Self::gen_use_columns_set(headers, use_columns);
-        let filtered_headers: CsvRecord = headers
-            .iter()
-            .enumerate()
-            .filter_map(|(i, h)| {
-                if use_columns_set.contains(&i) {
-                    Some(h.clone())
-                } else {
-                    None
-                }
-            })
-            .collect();
-        let sensitive_zeros_set = Self::gen_sensitive_zeros_set(&filtered_headers, sensitive_zeros);
-
-        (filtered_headers, use_columns_set, sensitive_zeros_set)
+    /// Returns the number of records on the data block protected by `resolution`
+    pub fn protected_number_of_records(&self, resolution: usize) -> usize {
+        uround_down(self.number_of_records() as f64, resolution as f64)
     }
 
     #[inline]
-    fn map_records(
-        records: &[CsvRecord],
-        use_columns_set: &HashSet<usize>,
-        sensitive_zeros_set: &HashSet<usize>,
-        record_limit: usize,
-    ) -> DataBlockRecords {
-        let map_result = |record: &CsvRecord| {
-            let values: CsvRecord = record
-                .iter()
-                .enumerate()
-                .filter_map(|(i, h)| {
-                    if use_columns_set.contains(&i) {
-                        Some(h.trim().to_string())
-                    } else {
-                        None
-                    }
-                })
-                .collect();
-
-            DataBlockRecord::new(
-                values
-                    .iter()
-                    .enumerate()
-                    .filter_map(|(i, r)| {
-                        let record_val = r.trim();
-                        if !record_val.is_empty()
-                            && (sensitive_zeros_set.contains(&i) || record_val != "0")
-                        {
-                            Some(DataBlockValue::new(i, record_val.into()))
-                        } else {
-                            None
-                        }
-                    })
-                    .collect(),
-            )
-        };
-
-        if record_limit > 0 {
-            records.iter().take(record_limit).map(map_result).collect()
+    /// Normalizes the reporting length based on the number of selected headers.
+    /// Returns the normalized value
+    /// # Arguments
+    /// * `reporting_length` - Reporting length to be normalized (0 means use all columns)
+    pub fn normalize_reporting_length(&self, reporting_length: usize) -> usize {
+        if reporting_length == 0 {
+            self.headers.len()
         } else {
-            records.iter().map(map_result).collect()
+            usize::min(reporting_length, self.headers.len())
         }
     }
-
-    #[inline]
-    fn create(
-        input_res: Result<Self::InputType, Self::ErrorType>,
-        use_columns: &[String],
-        sensitive_zeros: &[String],
-        record_limit: usize,
-    ) -> Result<DataBlock, Self::ErrorType> {
-        let mut input = input_res?;
-        let (headers, use_columns_set, sensitive_zeros_set) = Self::map_headers(
-            &mut Self::get_headers(&mut input)?,
-            use_columns,
-            sensitive_zeros,
-        );
-        let records = Self::map_records(
-            &Self::get_records(&mut input)?,
-            &use_columns_set,
-            &sensitive_zeros_set,
-            record_limit,
-        );
-
-        Ok(DataBlock::new(headers, records))
-    }
-
-    /// Should be implemented to return the CsvRecords reprensenting the headers
-    fn get_headers(input: &mut Self::InputType) -> Result<CsvRecord, Self::ErrorType>;
-
-    /// Should be implemented to return the vector of CsvRecords reprensenting rows
-    fn get_records(input: &mut Self::InputType) -> Result<Vec<CsvRecord>, Self::ErrorType>;
 }

packages/core/src/data_block/csv_block_creator.rs

@@ -1,4 +1,4 @@
-use super::{block::DataBlockCreator, typedefs::CsvRecord};
+use super::{data_block_creator::DataBlockCreator, typedefs::CsvRecord};
 use csv::{Error, Reader, StringRecord};
 use std::fs::File;
 

packages/core/src/data_block/csv_io_error.rs

@@ -0,0 +1,36 @@
+use csv::Error;
+use std::fmt::{Display, Formatter, Result};
+
+#[cfg(feature = "pyo3")]
+use pyo3::exceptions::PyIOError;
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+/// Wrapper for a csv::Error, so the from
+/// trait can be implemented for PyErr
+pub struct CsvIOError {
+    error: Error,
+}
+
+impl CsvIOError {
+    /// Creates a new CsvIOError from a csv::Error
+    /// # Arguments
+    /// * `error` - Related csv::Error
+    pub fn new(error: Error) -> CsvIOError {
+        CsvIOError { error }
+    }
+}
+
+impl Display for CsvIOError {
+    fn fmt(&self, f: &mut Formatter<'_>) -> Result {
+        write!(f, "{}", self.error)
+    }
+}
+
+#[cfg(feature = "pyo3")]
+impl From<CsvIOError> for PyErr {
+    fn from(err: CsvIOError) -> PyErr {
+        PyIOError::new_err(err.error.to_string())
+    }
+}

packages/core/src/data_block/data_block_creator.rs

@@ -0,0 +1,166 @@
+use super::{
+    block::DataBlock,
+    record::DataBlockRecord,
+    typedefs::{CsvRecord, CsvRecordRef, CsvRecordRefSlice, CsvRecordSlice, DataBlockRecords},
+    value::DataBlockValue,
+};
+use std::{collections::HashSet, sync::Arc};
+
+/// Trait that needs to be implement to create a data block.
+/// It already contains the logic to create the data block, so we only
+/// need to worry about mapping the headers and records from InputType
+pub trait DataBlockCreator {
+    /// Creator input type, it can be a File Reader, another data structure...
+    type InputType;
+    /// The error type that can be generated when parsing headers/records
+    type ErrorType;
+
+    #[inline]
+    fn gen_use_columns_set(headers: &CsvRecordSlice, use_columns: &[String]) -> HashSet<usize> {
+        let use_columns_str_set: HashSet<String> = use_columns
+            .iter()
+            .map(|c| c.trim().to_lowercase())
+            .collect();
+        headers
+            .iter()
+            .enumerate()
+            .filter_map(|(i, h)| {
+                if use_columns_str_set.is_empty()
+                    || use_columns_str_set.contains(&h.trim().to_lowercase())
+                {
+                    Some(i)
+                } else {
+                    None
+                }
+            })
+            .collect()
+    }
+
+    #[inline]
+    fn gen_sensitive_zeros_set(
+        filtered_headers: &CsvRecordRefSlice,
+        sensitive_zeros: &[String],
+    ) -> HashSet<usize> {
+        let sensitive_zeros_str_set: HashSet<String> = sensitive_zeros
+            .iter()
+            .map(|c| c.trim().to_lowercase())
+            .collect();
+        filtered_headers
+            .iter()
+            .enumerate()
+            .filter_map(|(i, h)| {
+                if sensitive_zeros_str_set.contains(&h.trim().to_lowercase()) {
+                    Some(i)
+                } else {
+                    None
+                }
+            })
+            .collect()
+    }
+
+    #[inline]
+    fn normalize_value(value: &str) -> String {
+        value
+            .trim()
+            // replace reserved delimiters
+            .replace(";", "<semicolon>")
+            .replace(":", "<colon>")
+    }
+
+    #[inline]
+    fn map_headers(
+        headers: &mut CsvRecord,
+        use_columns: &[String],
+        sensitive_zeros: &[String],
+    ) -> (CsvRecordRef, HashSet<usize>, HashSet<usize>) {
+        let use_columns_set = Self::gen_use_columns_set(headers, use_columns);
+        let filtered_headers: CsvRecordRef = headers
+            .iter()
+            .enumerate()
+            .filter_map(|(i, h)| {
+                if use_columns_set.contains(&i) {
+                    Some(Arc::new(Self::normalize_value(h)))
+                } else {
+                    None
+                }
+            })
+            .collect();
+        let sensitive_zeros_set = Self::gen_sensitive_zeros_set(&filtered_headers, sensitive_zeros);
+
+        (filtered_headers, use_columns_set, sensitive_zeros_set)
+    }
+
+    #[inline]
+    fn map_records(
+        records: &[CsvRecord],
+        use_columns_set: &HashSet<usize>,
+        sensitive_zeros_set: &HashSet<usize>,
+        record_limit: usize,
+    ) -> DataBlockRecords {
+        let map_result = |record: &CsvRecord| {
+            let values: CsvRecord = record
+                .iter()
+                .enumerate()
+                .filter_map(|(i, h)| {
+                    if use_columns_set.contains(&i) {
+                        Some(h.trim().to_string())
+                    } else {
+                        None
+                    }
+                })
+                .collect();
+
+            Arc::new(DataBlockRecord::new(
+                values
+                    .iter()
+                    .enumerate()
+                    .filter_map(|(i, r)| {
+                        let record_val = Self::normalize_value(r);
+                        if !record_val.is_empty()
+                            && (sensitive_zeros_set.contains(&i) || record_val != "0")
+                        {
+                            Some(Arc::new(DataBlockValue::new(i, Arc::new(record_val))))
+                        } else {
+                            None
+                        }
+                    })
+                    .collect(),
+            ))
+        };
+
+        if record_limit > 0 {
+            records.iter().take(record_limit).map(map_result).collect()
+        } else {
+            records.iter().map(map_result).collect()
+        }
+    }
+
+    #[inline]
+    fn create(
+        input_res: Result<Self::InputType, Self::ErrorType>,
+        use_columns: &[String],
+        sensitive_zeros: &[String],
+        record_limit: usize,
+    ) -> Result<Arc<DataBlock>, Self::ErrorType> {
+        let mut input = input_res?;
+        let (headers, use_columns_set, sensitive_zeros_set) = Self::map_headers(
+            &mut Self::get_headers(&mut input)?,
+            use_columns,
+            sensitive_zeros,
+        );
+        let records = Self::map_records(
+            &Self::get_records(&mut input)?,
+            &use_columns_set,
+            &sensitive_zeros_set,
+            record_limit,
+        );
+
+        Ok(Arc::new(DataBlock::new(headers, records)))
+    }
+
+    /// Should be implemented to return the CsvRecords representing the headers
+    fn get_headers(input: &mut Self::InputType) -> Result<CsvRecord, Self::ErrorType>;
+
+    /// Should be implemented to return the vector of CsvRecords representing rows
+    fn get_records(input: &mut Self::InputType) -> Result<Vec<CsvRecord>, Self::ErrorType>;
+}

packages/core/src/data_block/mod.rs

@@ -1,10 +1,21 @@
 /// Module defining the structures that represent a data block
 pub mod block;
+
 /// Module to create data blocks from CSV files
 pub mod csv_block_creator;
+
+/// Defines io errors for handling csv files
+/// that are easier to bind to other languages
+pub mod csv_io_error;
+
+/// Module to create data blocks from different input types (trait definitions)
+pub mod data_block_creator;
+
 /// Module defining the structures that represent a data block record
 pub mod record;
+
 /// Type definitions related to data blocks
 pub mod typedefs;
+
 /// Module defining the structures that represent a data block value
 pub mod value;

packages/core/src/data_block/record.rs

@@ -1,10 +1,12 @@
 use super::value::DataBlockValue;
+use serde::{Deserialize, Serialize};
+use std::sync::Arc;
 
 /// Represents all the values of a given row in a data block
-#[derive(Debug)]
+#[derive(Debug, Serialize, Deserialize)]
 pub struct DataBlockRecord {
     /// Vector of data block values for a given row indexed by column
-    pub values: Vec<DataBlockValue>,
+    pub values: Vec<Arc<DataBlockValue>>,
 }
 
 impl DataBlockRecord {
@@ -12,7 +14,7 @@ impl DataBlockRecord {
     /// # Arguments
     /// * `values` - Vector of data block values for a given row indexed by column
     #[inline]
-    pub fn new(values: Vec<DataBlockValue>) -> DataBlockRecord {
+    pub fn new(values: Vec<Arc<DataBlockValue>>) -> DataBlockRecord {
         DataBlockRecord { values }
     }
 }

packages/core/src/data_block/typedefs.rs

@@ -1,5 +1,8 @@
 use super::{record::DataBlockRecord, value::DataBlockValue};
 use fnv::FnvHashMap;
+use std::sync::Arc;
+
+use crate::processing::evaluator::rare_combinations_comparison_data::CombinationComparison;
 
 /// Ordered vector of rows where a particular value combination is present
 pub type AttributeRows = Vec<usize>;
@@ -11,23 +14,38 @@ pub type AttributeRowsSlice = [usize];
 pub type CsvRecord = Vec<String>;
 
 /// The same as CsvRecord, but keeping a reference to the string value, so data does not have to be duplicated
-pub type CsvRecordRef<'data_block_value> = Vec<&'data_block_value str>;
+pub type CsvRecordRef = Vec<Arc<String>>;
 
 /// Slice of CsvRecord
 pub type CsvRecordSlice = [String];
 
-/// Vector of strings representhing the data block headers
-pub type DataBlockHeaders = CsvRecord;
+/// Slice of CsvRecord
+pub type CsvRecordRefSlice = [Arc<String>];
+
+/// Vector of strings representing the data block headers
+pub type DataBlockHeaders = CsvRecordRef;
 
 /// Slice of DataBlockHeaders
-pub type DataBlockHeadersSlice = CsvRecordSlice;
+pub type DataBlockHeadersSlice = CsvRecordRefSlice;
 
 /// Vector of data block records, where each record represents a row
-pub type DataBlockRecords = Vec<DataBlockRecord>;
-
-/// Slice of DataBlockRecords
-pub type DataBlockRecordsSlice = [DataBlockRecord];
+pub type DataBlockRecords = Vec<Arc<DataBlockRecord>>;
 
 /// HashMap with a data block value as key and all the attribute row indexes where it occurs as value
-pub type AttributeRowsMap<'data_block_value> =
-    FnvHashMap<&'data_block_value DataBlockValue, AttributeRows>;
+pub type AttributeRowsMap = FnvHashMap<Arc<DataBlockValue>, AttributeRows>;
+
+/// HashMap with a data block value as key and attribute row indexes as value
+pub type AttributeRowsRefMap = FnvHashMap<Arc<DataBlockValue>, Arc<AttributeRows>>;
+
+/// Maps the column index -> data block value -> rows where the value appear
+pub type AttributeRowsByColumnMap = FnvHashMap<usize, AttributeRowsMap>;
+
+/// Raw synthesized data (vector of csv record references to the original data block)
+pub type RawSyntheticData = Vec<CsvRecordRef>;
+
+/// A vector of combination comparisons
+/// (between sensitive and synthetic data)
+pub type CombinationsComparisons = Vec<CombinationComparison>;
+
+/// Maps a column name to the corresponding column index
+pub type ColumnIndexByName = FnvHashMap<String, usize>;

packages/core/src/data_block/value.rs

@@ -1,10 +1,16 @@
+use super::typedefs::DataBlockHeadersSlice;
+use serde::{Deserialize, Serialize};
+use std::{fmt::Display, str::FromStr, sync::Arc};
+
+const VALUE_DELIMITER: char = ':';
+
 /// Represents a value of a given data block for a particular row and column
-#[derive(Debug, PartialEq, Eq, Hash, PartialOrd, Ord)]
+#[derive(Debug, PartialEq, Eq, Hash, PartialOrd, Ord, Serialize, Deserialize)]
 pub struct DataBlockValue {
     /// Column index this value belongs to starting in '0'
     pub column_index: usize,
     /// Value stored on the CSV file for a given row at `column_index`
-    pub value: String,
+    pub value: Arc<String>,
 }
 
 impl DataBlockValue {
@@ -13,10 +19,72 @@ impl DataBlockValue {
     /// * `column_index` - Column index this value belongs to starting in '0'
     /// * `value` - Value stored on the CSV file for a given row at `column_index`
     #[inline]
-    pub fn new(column_index: usize, value: String) -> DataBlockValue {
+    pub fn new(column_index: usize, value: Arc<String>) -> DataBlockValue {
         DataBlockValue {
             column_index,
             value,
         }
     }
+
+    /// Formats a data block value as String using the
+    /// corresponding header name
+    /// The result is formatted as: `{header_name}:{block_value}`
+    /// # Arguments
+    /// * `headers` - data block headers
+    /// * `value` - value to be formatted
+    #[inline]
+    pub fn format_str_using_headers(&self, headers: &DataBlockHeadersSlice) -> String {
+        format!(
+            "{}{}{}",
+            headers[self.column_index], VALUE_DELIMITER, self.value
+        )
+    }
+}
+
+/// Error that can happen when parsing a data block from
+/// a string
+pub struct ParseDataBlockValueError {
+    error_message: String,
+}
+
+impl ParseDataBlockValueError {
+    #[inline]
+    /// Creates a new ParseDataBlockValueError with `error_message`
+    pub fn new(error_message: String) -> ParseDataBlockValueError {
+        ParseDataBlockValueError { error_message }
+    }
+}
+
+impl Display for ParseDataBlockValueError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}", self.error_message)
+    }
+}
+
+impl FromStr for DataBlockValue {
+    type Err = ParseDataBlockValueError;
+
+    /// Creates a new DataBlockValue by parsing `str_value`
+    fn from_str(str_value: &str) -> Result<Self, Self::Err> {
+        if let Some(pos) = str_value.find(VALUE_DELIMITER) {
+            Ok(DataBlockValue::new(
+                str_value[..pos]
+                    .parse::<usize>()
+                    .map_err(|err| ParseDataBlockValueError::new(err.to_string()))?,
+                Arc::new(str_value[pos + 1..].into()),
+            ))
+        } else {
+            Err(ParseDataBlockValueError::new(format!(
+                "data block value missing '{}'",
+                VALUE_DELIMITER
+            )))
+        }
+    }
+}
+
+impl Display for DataBlockValue {
+    /// Formats the DataBlockValue as a string
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        Ok(write!(f, "{}:{}", self.column_index, self.value)?)
+    }
 }

packages/core/src/processing/aggregator/aggregated_data.rs

@@ -0,0 +1,448 @@
+use super::{
+    privacy_risk_summary::PrivacyRiskSummary,
+    records_analysis_data::RecordsAnalysisData,
+    typedefs::{
+        AggregatedCountByLenMap, AggregatesCountMap, AggregatesCountStringMap, RecordsByLenMap,
+        RecordsSensitivity,
+    },
+};
+use itertools::Itertools;
+use log::info;
+use serde::{Deserialize, Serialize};
+use std::{
+    io::{BufReader, BufWriter, Error, Write},
+    sync::Arc,
+};
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+use crate::{
+    data_block::block::DataBlock,
+    processing::aggregator::typedefs::RecordsSet,
+    utils::{math::uround_down, time::ElapsedDurationLogger},
+};
+
+/// Aggregated data produced by the Aggregator
+#[cfg_attr(feature = "pyo3", pyclass)]
+#[derive(Serialize, Deserialize)]
+pub struct AggregatedData {
+    /// Data block from where this aggregated data was generated
+    pub data_block: Arc<DataBlock>,
+    /// Maps a value combination to its aggregated count
+    pub aggregates_count: AggregatesCountMap,
+    /// A vector of sensitivities for each record (the vector index is the record index)
+    pub records_sensitivity: RecordsSensitivity,
+    /// Maximum length used to compute attribute combinations
+    pub reporting_length: usize,
+}
+
+impl AggregatedData {
+    /// Creates a new AggregatedData struct with default values
+    #[inline]
+    pub fn default() -> AggregatedData {
+        AggregatedData {
+            data_block: Arc::new(DataBlock::default()),
+            aggregates_count: AggregatesCountMap::default(),
+            records_sensitivity: RecordsSensitivity::default(),
+            reporting_length: 0,
+        }
+    }
+
+    /// Creates a new AggregatedData struct
+    /// # Arguments:
+    /// * `data_block` - Data block with the original data
+    /// * `aggregates_count` - Computed aggregates count map
+    /// * `records_sensitivity` - Computed sensitivity for the records
+    /// * `reporting_length` - Maximum length used to compute attribute combinations
+    #[inline]
+    pub fn new(
+        data_block: Arc<DataBlock>,
+        aggregates_count: AggregatesCountMap,
+        records_sensitivity: RecordsSensitivity,
+        reporting_length: usize,
+    ) -> AggregatedData {
+        AggregatedData {
+            data_block,
+            aggregates_count,
+            records_sensitivity,
+            reporting_length,
+        }
+    }
+
+    #[inline]
+    /// Check if the records len map contains a value across all lengths
+    fn records_by_len_contains(records_by_len: &RecordsByLenMap, value: &usize) -> bool {
+        records_by_len
+            .values()
+            .any(|records| records.contains(value))
+    }
+
+    #[inline]
+    /// Whe first generated the RecordsByLenMap might contain
+    /// the same records appearing in different combination lengths.
+    /// This will keep only the record on the shortest length key
+    /// and remove the other occurrences
+    fn keep_records_only_on_shortest_len(records_by_len: &mut RecordsByLenMap) {
+        let lengths: Vec<usize> = records_by_len.keys().cloned().sorted().collect();
+        let max_len = lengths.last().copied().unwrap_or(0);
+
+        // make sure the record will be only present in the shortest len
+        // start on the shortest length
+        for l in lengths {
+            for r in records_by_len.get(&l).unwrap().clone() {
+                // search all lengths > l and remove the record from there
+                for n in l + 1..=max_len {
+                    if let Some(records) = records_by_len.get_mut(&n) {
+                        records.remove(&r);
+                    }
+                }
+            }
+        }
+        // retain only non-empty record lists
+        records_by_len.retain(|_, records| !records.is_empty());
+    }
+
+    #[inline]
+    fn _read_from_json(file_path: &str) -> Result<AggregatedData, Error> {
+        let _duration_logger = ElapsedDurationLogger::new("aggregated count json read");
+
+        Ok(serde_json::from_reader(BufReader::new(
+            std::fs::File::open(file_path)?,
+        ))?)
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl AggregatedData {
+    /// Builds a map from value combinations formatted as string to its aggregated count
+    /// This method will clone the data, so its recommended to have its result stored
+    /// in a local variable to avoid it being called multiple times
+    /// # Arguments:
+    /// * `combination_delimiter` - Delimiter used to join combinations
+    pub fn get_formatted_aggregates_count(
+        &self,
+        combination_delimiter: &str,
+    ) -> AggregatesCountStringMap {
+        self.aggregates_count
+            .iter()
+            .map(|(key, value)| {
+                (
+                    key.format_str_using_headers(&self.data_block.headers, combination_delimiter),
+                    value.clone(),
+                )
+            })
+            .collect()
+    }
+
+    #[cfg(feature = "pyo3")]
+    /// A vector of sensitivities for each record (the vector index is the record index)
+    /// This method will clone the data, so its recommended to have its result stored
+    /// in a local variable to avoid it being called multiple times
+    pub fn get_records_sensitivity(&self) -> RecordsSensitivity {
+        self.records_sensitivity.clone()
+    }
+
+    /// Round the aggregated counts down to the nearest multiple of resolution
+    /// # Arguments:
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn protect_aggregates_count(&mut self, resolution: usize) {
+        let _duration_logger = ElapsedDurationLogger::new("aggregates count protect");
+
+        info!(
+            "protecting aggregates counts with resolution {}",
+            resolution
+        );
+
+        for count in self.aggregates_count.values_mut() {
+            count.count = uround_down(count.count as f64, resolution as f64);
+        }
+        // remove 0 counts from response
+        self.aggregates_count.retain(|_, count| count.count > 0);
+    }
+
+    /// Calculates the records that contain rare combinations grouped by length.
+    /// This might contain duplicated records on different lengths if the record
+    /// contains more than one rare combination. Unique combinations are also contained
+    /// in this.
+    /// # Arguments:
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn calc_all_rare_combinations_records_by_len(&self, resolution: usize) -> RecordsByLenMap {
+        let _duration_logger =
+            ElapsedDurationLogger::new("all rare combinations records by len calculation");
+        let mut rare_records_by_len: RecordsByLenMap = RecordsByLenMap::default();
+
+        for (agg, count) in self.aggregates_count.iter() {
+            if count.count < resolution {
+                rare_records_by_len
+                    .entry(agg.len())
+                    .or_insert_with(RecordsSet::default)
+                    .extend(&count.contained_in_records);
+            }
+        }
+        rare_records_by_len
+    }
+
+    /// Calculates the records that contain unique combinations grouped by length.
+    /// This might contain duplicated records on different lengths if the record
+    /// contains more than one unique combination.
+    pub fn calc_all_unique_combinations_records_by_len(&self) -> RecordsByLenMap {
+        let _duration_logger =
+            ElapsedDurationLogger::new("all unique combinations records by len calculation");
+        let mut unique_records_by_len: RecordsByLenMap = RecordsByLenMap::default();
+
+        for (agg, count) in self.aggregates_count.iter() {
+            if count.count == 1 {
+                unique_records_by_len
+                    .entry(agg.len())
+                    .or_insert_with(RecordsSet::default)
+                    .extend(&count.contained_in_records);
+            }
+        }
+        unique_records_by_len
+    }
+
+    /// Calculate the records that contain unique and rare combinations grouped by length.
+    /// A tuple with the `(unique, rare)` is returned.
+    /// Both returned maps are ensured to only contain the records on the shortest length,
+    /// so each record will appear only on the shortest combination length that isolates it
+    /// within a rare group. Also, if the record has a unique combination, it will not
+    /// be present on the rare map, only on the unique one.
+    /// # Arguments:
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn calc_unique_rare_combinations_records_by_len(
+        &self,
+        resolution: usize,
+    ) -> (RecordsByLenMap, RecordsByLenMap) {
+        let _duration_logger =
+            ElapsedDurationLogger::new("unique/rare combinations records by len calculation");
+        let mut unique_records_by_len = self.calc_all_unique_combinations_records_by_len();
+        let mut rare_records_by_len = self.calc_all_rare_combinations_records_by_len(resolution);
+
+        AggregatedData::keep_records_only_on_shortest_len(&mut unique_records_by_len);
+        AggregatedData::keep_records_only_on_shortest_len(&mut rare_records_by_len);
+
+        // remove records with unique combinations from the rare map
+        rare_records_by_len.values_mut().for_each(|records| {
+            records.retain(|r| !AggregatedData::records_by_len_contains(&unique_records_by_len, r));
+        });
+
+        (unique_records_by_len, rare_records_by_len)
+    }
+
+    /// Perform the records analysis and returns the data containing
+    /// unique, rare and risky information grouped per length.
+    /// # Arguments:
+    /// * `resolution` - Reporting resolution used for data synthesis
+    /// * `protect` - Whether or not the counts should be rounded to the
+    /// nearest smallest multiple of resolution
+    pub fn calc_records_analysis_by_len(
+        &self,
+        resolution: usize,
+        protect: bool,
+    ) -> RecordsAnalysisData {
+        let _duration_logger = ElapsedDurationLogger::new("records analysis by len");
+        let (unique_records_by_len, rare_records_by_len) =
+            self.calc_unique_rare_combinations_records_by_len(resolution);
+
+        RecordsAnalysisData::from_unique_rare_combinations_records_by_len(
+            &unique_records_by_len,
+            &rare_records_by_len,
+            self.data_block.records.len(),
+            self.reporting_length,
+            resolution,
+            protect,
+        )
+    }
+
+    /// Calculates the number of rare combinations grouped by combination length
+    /// # Arguments:
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn calc_rare_combinations_count_by_len(
+        &self,
+        resolution: usize,
+    ) -> AggregatedCountByLenMap {
+        let _duration_logger =
+            ElapsedDurationLogger::new("rare combinations count by len calculation");
+        let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
+
+        info!(
+            "calculating rare combinations counts by length with resolution {}",
+            resolution
+        );
+
+        for (agg, count) in self.aggregates_count.iter() {
+            if count.count < resolution {
+                let curr_count = result.entry(agg.len()).or_insert(0);
+                *curr_count += 1;
+            }
+        }
+        result
+    }
+
+    /// Calculates the number of combinations grouped by combination length
+    pub fn calc_combinations_count_by_len(&self) -> AggregatedCountByLenMap {
+        let _duration_logger = ElapsedDurationLogger::new("combination count by len calculation");
+        let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
+
+        info!("calculating combination counts by length");
+
+        for agg in self.aggregates_count.keys() {
+            let curr_count = result.entry(agg.len()).or_insert(0);
+            *curr_count += 1;
+        }
+        result
+    }
+
+    /// Calculates the sum of all combination counts grouped by combination length
+    pub fn calc_combinations_sum_by_len(&self) -> AggregatedCountByLenMap {
+        let _duration_logger = ElapsedDurationLogger::new("combinations sum by len calculation");
+        let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
+
+        info!("calculating combination counts sums by length");
+
+        for (agg, count) in self.aggregates_count.iter() {
+            let curr_sum = result.entry(agg.len()).or_insert(0);
+            *curr_sum += count.count;
+        }
+        result
+    }
+
+    /// Calculates the privacy risk related with data block and the generated
+    /// aggregates counts
+    /// # Arguments:
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn calc_privacy_risk(&self, resolution: usize) -> PrivacyRiskSummary {
+        let _duration_logger = ElapsedDurationLogger::new("privacy risk calculation");
+
+        info!("calculating privacy risk...");
+
+        PrivacyRiskSummary::from_aggregates_count(
+            self.data_block.records.len(),
+            &self.aggregates_count,
+            resolution,
+        )
+    }
+
+    /// Writes the aggregates counts to the file system in a csv/tsv like format
+    /// # Arguments:
+    /// * `aggregates_path` - File path to be written
+    /// * `aggregates_delimiter` - Delimiter to use when writing to `aggregates_path`
+    /// * `combination_delimiter` - Delimiter used to join combinations and format then
+    /// as strings
+    /// * `resolution` - Reporting resolution used for data synthesis
+    /// * `protected` - Whether or not the counts were protected before calling this
+    pub fn write_aggregates_count(
+        &self,
+        aggregates_path: &str,
+        aggregates_delimiter: char,
+        combination_delimiter: &str,
+        resolution: usize,
+        protected: bool,
+    ) -> Result<(), Error> {
+        let _duration_logger = ElapsedDurationLogger::new("write aggregates count");
+
+        info!("writing file {}", aggregates_path);
+
+        let mut file = std::fs::File::create(aggregates_path)?;
+
+        file.write_all(
+            format!(
+                "selections{}{}\n",
+                aggregates_delimiter,
+                if protected {
+                    "protected_count"
+                } else {
+                    "count"
+                }
+            )
+            .as_bytes(),
+        )?;
+        file.write_all(
+            format!(
+                "selections{}{}\n",
+                aggregates_delimiter,
+                uround_down(self.data_block.records.len() as f64, resolution as f64)
+            )
+            .as_bytes(),
+        )?;
+        for aggregate in self.aggregates_count.keys() {
+            file.write_all(
+                format!(
+                    "{}{}{}\n",
+                    aggregate
+                        .format_str_using_headers(&self.data_block.headers, combination_delimiter),
+                    aggregates_delimiter,
+                    self.aggregates_count[aggregate].count
+                )
+                .as_bytes(),
+            )?
+        }
+        Ok(())
+    }
+
+    /// Writes the records sensitivity to the file system in a csv/tsv like format
+    /// # Arguments:
+    /// * `records_sensitivity_path` - File path to be written
+    /// * `records_sensitivity_delimiter` - Delimiter to use when writing to `records_sensitivity_path`
+    pub fn write_records_sensitivity(
+        &self,
+        records_sensitivity_path: &str,
+        records_sensitivity_delimiter: char,
+    ) -> Result<(), Error> {
+        let _duration_logger = ElapsedDurationLogger::new("write records sensitivity");
+
+        info!("writing file {}", records_sensitivity_path);
+
+        let mut file = std::fs::File::create(records_sensitivity_path)?;
+
+        file.write_all(
+            format!(
+                "record_index{}record_sensitivity\n",
+                records_sensitivity_delimiter
+            )
+            .as_bytes(),
+        )?;
+        for (i, sensitivity) in self.records_sensitivity.iter().enumerate() {
+            file.write_all(
+                format!("{}{}{}\n", i, records_sensitivity_delimiter, sensitivity).as_bytes(),
+            )?
+        }
+        Ok(())
+    }
+
+    /// Serializes the aggregated data to a json file
+    /// # Arguments:
+    /// * `file_path` - File path to be written
+    pub fn write_to_json(&self, file_path: &str) -> Result<(), Error> {
+        let _duration_logger = ElapsedDurationLogger::new("aggregated count json write");
+
+        Ok(serde_json::to_writer(
+            BufWriter::new(std::fs::File::create(file_path)?),
+            &self,
+        )?)
+    }
+
+    #[cfg(feature = "pyo3")]
+    #[staticmethod]
+    /// Deserializes the aggregated data from a json file
+    /// # Arguments:
+    /// * `file_path` - File path to read from
+    pub fn read_from_json(file_path: &str) -> Result<AggregatedData, Error> {
+        AggregatedData::_read_from_json(file_path)
+    }
+
+    #[cfg(not(feature = "pyo3"))]
+    /// Deserializes the aggregated data from a json file
+    /// # Arguments:
+    /// * `file_path` - File path to read from
+    pub fn read_from_json(file_path: &str) -> Result<AggregatedData, Error> {
+        AggregatedData::_read_from_json(file_path)
+    }
+}
+
+#[cfg(feature = "pyo3")]
+pub fn register(_py: Python, m: &PyModule) -> PyResult<()> {
+    m.add_class::<AggregatedData>()?;
+    Ok(())
+}

packages/core/src/processing/aggregator/data_aggregator.rs

@@ -0,0 +1,174 @@
+use super::aggregated_data::AggregatedData;
+use super::rows_aggregator::RowsAggregator;
+use super::typedefs::RecordsSet;
+use itertools::Itertools;
+use log::info;
+use serde::{Deserialize, Serialize};
+use std::sync::Arc;
+
+use crate::data_block::block::DataBlock;
+use crate::processing::aggregator::record_attrs_selector::RecordAttrsSelector;
+use crate::utils::math::calc_percentage;
+use crate::utils::reporting::ReportProgress;
+use crate::utils::threading::get_number_of_threads;
+use crate::utils::time::ElapsedDurationLogger;
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+/// Result of data aggregation for each combination
+#[cfg_attr(feature = "pyo3", pyclass)]
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct AggregatedCount {
+    /// How many times the combination appears on the records
+    pub count: usize,
+    /// Which records this combinations is part of
+    pub contained_in_records: RecordsSet,
+}
+
+#[cfg(feature = "pyo3")]
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl AggregatedCount {
+    /// How many times the combination appears on the records
+    #[getter]
+    fn count(&self) -> usize {
+        self.count
+    }
+
+    /// Which records this combinations is part of
+    /// This method will clone the data, so its recommended to have its result stored
+    /// in a local variable to avoid it being called multiple times
+    fn get_contained_in_records(&self) -> RecordsSet {
+        self.contained_in_records.clone()
+    }
+}
+
+impl Default for AggregatedCount {
+    fn default() -> AggregatedCount {
+        AggregatedCount {
+            count: 0,
+            contained_in_records: RecordsSet::default(),
+        }
+    }
+}
+
+/// Process a data block to produced aggregated data
+pub struct Aggregator {
+    data_block: Arc<DataBlock>,
+}
+
+impl Aggregator {
+    /// Returns a data aggregator for the given data block
+    /// # Arguments
+    /// * `data_block` - The data block to be processed
+    #[inline]
+    pub fn new(data_block: Arc<DataBlock>) -> Aggregator {
+        Aggregator { data_block }
+    }
+
+    /// Aggregates the data block and returns the aggregated data back
+    /// # Arguments
+    /// * `reporting_length` - Calculate combinations from 1 up to `reporting_length`
+    /// * `sensitivity_threshold` - Sensitivity threshold to filter record attributes
+    /// (0 means no suppression)
+    /// * `progress_reporter` - Will be used to report the processing
+    /// progress (`ReportProgress` trait). If `None`, nothing will be reported
+    pub fn aggregate<T>(
+        &mut self,
+        reporting_length: usize,
+        sensitivity_threshold: usize,
+        progress_reporter: &mut Option<T>,
+    ) -> AggregatedData
+    where
+        T: ReportProgress,
+    {
+        let _duration_logger = ElapsedDurationLogger::new("data aggregation");
+        let normalized_reporting_length =
+            self.data_block.normalize_reporting_length(reporting_length);
+        let length_range = (1..=normalized_reporting_length).collect::<Vec<usize>>();
+        let total_n_records = self.data_block.records.len();
+        let total_n_records_f64 = total_n_records as f64;
+
+        info!(
+            "aggregating data with reporting length = {}, sensitivity_threshold = {} and {} thread(s)",
+            normalized_reporting_length, sensitivity_threshold, get_number_of_threads()
+        );
+
+        let result = RowsAggregator::aggregate_all(
+            total_n_records,
+            &mut self.build_rows_aggregators(&length_range, sensitivity_threshold),
+            progress_reporter,
+        );
+
+        Aggregator::update_aggregate_progress(
+            progress_reporter,
+            total_n_records,
+            total_n_records_f64,
+        );
+
+        info!(
+            "data aggregated resulting in {} distinct combinations...",
+            result.aggregates_count.len()
+        );
+        info!(
+            "suppression ratio of aggregates is {:.2}%",
+            (1.0 - (result.selected_combs_count as f64 / result.all_combs_count as f64)) * 100.0
+        );
+
+        AggregatedData::new(
+            self.data_block.clone(),
+            result.aggregates_count,
+            result.records_sensitivity,
+            normalized_reporting_length,
+        )
+    }
+
+    #[inline]
+    fn build_rows_aggregators<'length_range>(
+        &self,
+        length_range: &'length_range [usize],
+        sensitivity_threshold: usize,
+    ) -> Vec<RowsAggregator<'length_range>> {
+        if self.data_block.records.is_empty() {
+            return Vec::default();
+        }
+
+        let chunk_size = ((self.data_block.records.len() as f64) / (get_number_of_threads() as f64))
+            .ceil() as usize;
+        let mut rows_aggregators: Vec<RowsAggregator> = Vec::default();
+        let attr_rows_map = Arc::new(self.data_block.calc_attr_rows());
+
+        for c in &self
+            .data_block
+            .records
+            .clone()
+            .drain(..)
+            .enumerate()
+            .chunks(chunk_size)
+        {
+            rows_aggregators.push(RowsAggregator::new(
+                self.data_block.clone(),
+                c.collect(),
+                RecordAttrsSelector::new(
+                    length_range,
+                    sensitivity_threshold,
+                    attr_rows_map.clone(),
+                ),
+            ))
+        }
+        rows_aggregators
+    }
+
+    #[inline]
+    fn update_aggregate_progress<T>(
+        progress_reporter: &mut Option<T>,
+        n_processed: usize,
+        total: f64,
+    ) where
+        T: ReportProgress,
+    {
+        if let Some(r) = progress_reporter {
+            r.report(calc_percentage(n_processed as f64, total));
+        }
+    }
+}

packages/core/src/processing/aggregator/mod.rs

@@ -1,501 +1,27 @@
+/// Module to represent aggregated data and provide
+/// some methods/utilities for information extracted from it
+pub mod aggregated_data;
+
+/// Dataset privacy risk definitions
+pub mod privacy_risk_summary;
+
 /// Module that can the used to suppress attributes from records to
 /// meet a certain sensitivity threshold
 pub mod record_attrs_selector;
 
+/// Defines structures related to records analysis (unique, rare and risky
+/// information)
+pub mod records_analysis_data;
+
 /// Type definitions related to the aggregation process
 pub mod typedefs;
 
-use self::typedefs::{
-    AggregatedCountByLenMap, AggregatesCountMap, RecordsSensitivity, RecordsSensitivitySlice,
-    RecordsSet, ValueCombinationSlice,
-};
-use instant::Duration;
-use itertools::Itertools;
-use log::Level::Trace;
-use log::{info, log_enabled, trace};
-use std::cmp::min;
-use std::io::{Error, Write};
+/// Defines structures to store value combinations generated
+/// during the aggregate process
+pub mod value_combination;
 
-use crate::data_block::block::DataBlock;
-use crate::data_block::typedefs::DataBlockHeadersSlice;
-use crate::data_block::value::DataBlockValue;
-use crate::measure_time;
-use crate::processing::aggregator::record_attrs_selector::RecordAttrsSelector;
-use crate::utils::math::{calc_n_combinations_range, calc_percentage, uround_down};
-use crate::utils::reporting::ReportProgress;
-use crate::utils::time::ElapsedDuration;
+mod data_aggregator;
 
-/// Represents the privacy risk information related to a data block
-#[derive(Debug)]
-pub struct PrivacyRiskSummary {
-    /// Total number of records on the data block
-    pub total_number_of_records: usize,
-    /// Total number of combinations aggregated (up to reporting length)
-    pub total_number_of_combinations: usize,
-    /// Number of records with unique combinations
-    pub records_with_unique_combinations_count: usize,
-    /// Number of records with rare combinations (combination count < resolution)
-    pub records_with_rare_combinations_count: usize,
-    /// Number of unique combinations
-    pub unique_combinations_count: usize,
-    /// Number of rare combinations
-    pub rare_combinations_count: usize,
-    /// Proportion of records containing unique combinations
-    pub records_with_unique_combinations_proportion: f64,
-    /// Proportion of records containing rare combinations
-    pub records_with_rare_combinations_proportion: f64,
-    /// Proportion of unique combinations
-    pub unique_combinations_proportion: f64,
-    /// Proportion of rare combinations
-    pub rare_combinations_proportion: f64,
-}
+pub use data_aggregator::{AggregatedCount, Aggregator};
 
-/// Result of data aggregation for each combination
-#[derive(Debug)]
-pub struct AggregatedCount {
-    /// How many times the combination appears on the records
-    pub count: usize,
-    /// Which records this combinations is part of
-    pub contained_in_records: RecordsSet,
-}
-
-impl Default for AggregatedCount {
-    fn default() -> AggregatedCount {
-        AggregatedCount {
-            count: 0,
-            contained_in_records: RecordsSet::default(),
-        }
-    }
-}
-
-#[derive(Debug)]
-struct AggregatorDurations {
-    aggregate: Duration,
-    calc_rare_combinations_count_by_len: Duration,
-    calc_combinations_count_by_len: Duration,
-    calc_combinations_sum_by_len: Duration,
-    calc_privacy_risk: Duration,
-    write_aggregates_count: Duration,
-}
-
-/// Aggregated data produced by the Aggregator
-pub struct AggregatedData<'data_block> {
-    /// Maps a value combination to its aggregated count
-    pub aggregates_count: AggregatesCountMap<'data_block>,
-    /// A vector of sensitivities for each record (the vector index is the record index)
-    pub records_sensitivity: RecordsSensitivity,
-}
-
-/// Process a data block to produced aggregated data
-pub struct Aggregator<'data_block> {
-    data_block: &'data_block DataBlock,
-    durations: AggregatorDurations,
-}
-
-impl<'data_block> Aggregator<'data_block> {
-    /// Returns a data aggregator for the given data block
-    /// # Arguments
-    /// * `data_block` - The data block to be processed
-    #[inline]
-    pub fn new(data_block: &'data_block DataBlock) -> Aggregator<'data_block> {
-        Aggregator {
-            data_block,
-            durations: AggregatorDurations {
-                aggregate: Duration::default(),
-                calc_rare_combinations_count_by_len: Duration::default(),
-                calc_combinations_count_by_len: Duration::default(),
-                calc_combinations_sum_by_len: Duration::default(),
-                calc_privacy_risk: Duration::default(),
-                write_aggregates_count: Duration::default(),
-            },
-        }
-    }
-
-    /// Formats a data block value as String.
-    /// The result is formatted as: `{header_name}:{block_value}`
-    /// # Arguments
-    /// * `headers`: data block headers
-    /// * `value`: value to be formatted
-    #[inline]
-    pub fn format_data_block_value_str(
-        headers: &DataBlockHeadersSlice,
-        value: &'data_block DataBlockValue,
-    ) -> String {
-        format!("{}:{}", headers[value.column_index], value.value)
-    }
-
-    /// Formats a value combination as String.
-    /// The result is formatted as:
-    /// `{header_name}:{block_value};{header_name}:{block_value}...`
-    /// # Arguments
-    /// * `headers`: data block headers
-    /// * `aggregate`: combinations to be formatted
-    #[inline]
-    pub fn format_aggregate_str(
-        headers: &DataBlockHeadersSlice,
-        aggregate: &ValueCombinationSlice<'data_block>,
-    ) -> String {
-        let mut str = String::default();
-        for comb in aggregate {
-            if !str.is_empty() {
-                str += ";";
-            }
-            str += Aggregator::format_data_block_value_str(headers, comb).as_str();
-        }
-        str
-    }
-
-    /// Aggregates the data block and returns the aggregated data back
-    /// # Arguments
-    /// * `reporting_length` - Calculate combinations from 1 up to `reporting_length`
-    /// * `sensitivity_threshold` - Sensitivity threshold to filter record attributes
-    /// (0 means no suppression)
-    /// * `progress_reporter` - Will be used to report the processing
-    /// progress (`ReportProgress` trait). If `None`, nothing will be reported
-    pub fn aggregate<T>(
-        &mut self,
-        reporting_length: usize,
-        sensitivity_threshold: usize,
-        progress_reporter: &mut Option<T>,
-    ) -> AggregatedData<'data_block>
-    where
-        T: ReportProgress,
-    {
-        measure_time!(
-            || {
-                let mut aggregates_count: AggregatesCountMap = AggregatesCountMap::default();
-                let max_len = if reporting_length == 0 {
-                    self.data_block.headers.len()
-                } else {
-                    min(reporting_length, self.data_block.headers.len())
-                };
-                let length_range = (1..=max_len).collect::<Vec<usize>>();
-                let mut record_attrs_selector = RecordAttrsSelector::new(
-                    &length_range,
-                    sensitivity_threshold,
-                    DataBlock::calc_attr_rows(&self.data_block.records),
-                );
-                let mut selected_combs_count = 0_u64;
-                let mut all_combs_count = 0_u64;
-                let mut records_sensitivity: RecordsSensitivity = RecordsSensitivity::default();
-                let total_n_records = self.data_block.records.len();
-                let total_n_records_f64 = total_n_records as f64;
-
-                records_sensitivity.resize(total_n_records, 0);
-
-                info!(
-                    "aggregating data with reporting length = {} and sensitivity_threshold = {}",
-                    max_len, sensitivity_threshold
-                );
-
-                for (record_index, record) in self.data_block.records.iter().enumerate() {
-                    let selected_attrs = record_attrs_selector.select_from_record(&record.values);
-
-                    all_combs_count +=
-                        calc_n_combinations_range(record.values.len(), &length_range);
-
-                    Aggregator::update_aggregate_progress(
-                        progress_reporter,
-                        record_index,
-                        total_n_records_f64,
-                    );
-
-                    for l in &length_range {
-                        for c in selected_attrs.iter().combinations(*l) {
-                            let curr_count = aggregates_count
-                                .entry(
-                                    c.iter()
-                                        .sorted_by_key(|k| {
-                                            Aggregator::format_data_block_value_str(
-                                                &self.data_block.headers,
-                                                k,
-                                            )
-                                        })
-                                        .cloned()
-                                        .copied()
-                                        .collect_vec(),
-                                )
-                                .or_insert_with(AggregatedCount::default);
-                            curr_count.count += 1;
-                            curr_count.contained_in_records.insert(record_index);
-                            records_sensitivity[record_index] += 1;
-                            selected_combs_count += 1;
-                        }
-                    }
-                }
-                Aggregator::update_aggregate_progress(
-                    progress_reporter,
-                    total_n_records,
-                    total_n_records_f64,
-                );
-
-                info!(
-                    "data aggregated resulting in {} distinct combinations...",
-                    aggregates_count.len()
-                );
-                info!(
-                    "suppression ratio of aggregates is {:.2}%",
-                    (1.0 - (selected_combs_count as f64 / all_combs_count as f64)) * 100.0
-                );
-                AggregatedData {
-                    aggregates_count,
-                    records_sensitivity,
-                }
-            },
-            (self.durations.aggregate)
-        )
-    }
-
-    /// Round the aggregated counts down to the nearest multiple of resolution
-    /// # Arguments:
-    /// * `aggregates_count` - Counts to be rounded in place
-    /// * `resolution` - Reporting resolution used for data synthesis
-    pub fn protect_aggregates_count(
-        aggregates_count: &mut AggregatesCountMap<'data_block>,
-        resolution: usize,
-    ) {
-        info!(
-            "protecting aggregates counts with resolution {}",
-            resolution
-        );
-        for count in aggregates_count.values_mut() {
-            count.count = uround_down(count.count as f64, resolution as f64);
-        }
-        // remove 0 counts from response
-        aggregates_count.retain(|_, count| count.count > 0);
-    }
-
-    /// Calculates the number of rare combinations grouped by combination length
-    /// # Arguments:
-    /// * `aggregates_count` - Calculated aggregates count map
-    /// * `resolution` - Reporting resolution used for data synthesis
-    pub fn calc_rare_combinations_count_by_len(
-        &mut self,
-        aggregates_count: &AggregatesCountMap<'data_block>,
-        resolution: usize,
-    ) -> AggregatedCountByLenMap {
-        info!(
-            "calculating rare combinations counts by length with resolution {}",
-            resolution
-        );
-        measure_time!(
-            || {
-                let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
-
-                for (agg, count) in aggregates_count.iter() {
-                    if count.count < resolution {
-                        let curr_count = result.entry(agg.len()).or_insert(0);
-                        *curr_count += 1;
-                    }
-                }
-                result
-            },
-            (self.durations.calc_rare_combinations_count_by_len)
-        )
-    }
-
-    /// Calculates the number of combinations grouped by combination length
-    /// # Arguments:
-    /// * `aggregates_count` - Calculated aggregates count map
-    pub fn calc_combinations_count_by_len(
-        &mut self,
-        aggregates_count: &AggregatesCountMap<'data_block>,
-    ) -> AggregatedCountByLenMap {
-        info!("calculating combination counts by length");
-        measure_time!(
-            || {
-                let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
-
-                for agg in aggregates_count.keys() {
-                    let curr_count = result.entry(agg.len()).or_insert(0);
-                    *curr_count += 1;
-                }
-                result
-            },
-            (self.durations.calc_combinations_count_by_len)
-        )
-    }
-
-    /// Calculates the sum of all combination counts grouped by combination length
-    /// # Arguments:
-    /// * `aggregates_count` - Calculated aggregates count map
-    pub fn calc_combinations_sum_by_len(
-        &mut self,
-        aggregates_count: &AggregatesCountMap<'data_block>,
-    ) -> AggregatedCountByLenMap {
-        info!("calculating combination counts sums by length");
-        measure_time!(
-            || {
-                let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
-
-                for (agg, count) in aggregates_count.iter() {
-                    let curr_sum = result.entry(agg.len()).or_insert(0);
-                    *curr_sum += count.count;
-                }
-                result
-            },
-            (self.durations.calc_combinations_sum_by_len)
-        )
-    }
-
-    /// Calculates the privacy risk related with data block and the generated
-    /// aggregates counts
-    /// # Arguments:
-    /// * `aggregates_count` - Calculated aggregates count map
-    /// * `resolution` - Reporting resolution used for data synthesis
-    pub fn calc_privacy_risk(
-        &mut self,
-        aggregates_count: &AggregatesCountMap<'data_block>,
-        resolution: usize,
-    ) -> PrivacyRiskSummary {
-        info!("calculating privacy risk...");
-        measure_time!(
-            || {
-                let mut records_with_unique_combinations = RecordsSet::default();
-                let mut records_with_rare_combinations = RecordsSet::default();
-                let mut unique_combinations_count: usize = 0;
-                let mut rare_combinations_count: usize = 0;
-                let total_number_of_records = self.data_block.records.len();
-                let total_number_of_combinations = aggregates_count.len();
-
-                for count in aggregates_count.values() {
-                    if count.count == 1 {
-                        unique_combinations_count += 1;
-                        count.contained_in_records.iter().for_each(|record_index| {
-                            records_with_unique_combinations.insert(*record_index);
-                        });
-                    }
-                    if count.count < resolution {
-                        rare_combinations_count += 1;
-                        count.contained_in_records.iter().for_each(|record_index| {
-                            records_with_rare_combinations.insert(*record_index);
-                        });
-                    }
-                }
-
-                PrivacyRiskSummary {
-                    total_number_of_records,
-                    total_number_of_combinations,
-                    records_with_unique_combinations_count: records_with_unique_combinations.len(),
-                    records_with_rare_combinations_count: records_with_rare_combinations.len(),
-                    unique_combinations_count,
-                    rare_combinations_count,
-                    records_with_unique_combinations_proportion: (records_with_unique_combinations
-                        .len()
-                        as f64)
-                        / (total_number_of_records as f64),
-                    records_with_rare_combinations_proportion: (records_with_rare_combinations.len()
-                        as f64)
-                        / (total_number_of_records as f64),
-                    unique_combinations_proportion: (unique_combinations_count as f64)
-                        / (total_number_of_combinations as f64),
-                    rare_combinations_proportion: (rare_combinations_count as f64)
-                        / (total_number_of_combinations as f64),
-                }
-            },
-            (self.durations.calc_privacy_risk)
-        )
-    }
-
-    /// Writes the aggregates counts to the file system in a csv/tsv like format
-    /// # Arguments:
-    /// * `aggregates_count` - Calculated aggregates count map
-    /// * `aggregates_path` - File path to be written
-    /// * `aggregates_delimiter` - Delimiter to use when writing to `aggregates_path`
-    /// * `resolution` - Reporting resolution used for data synthesis
-    /// * `protected` - Whether or not the counts were protected before calling this
-    pub fn write_aggregates_count(
-        &mut self,
-        aggregates_count: &AggregatesCountMap<'data_block>,
-        aggregates_path: &str,
-        aggregates_delimiter: char,
-        resolution: usize,
-        protected: bool,
-    ) -> Result<(), Error> {
-        info!("writing file {}", aggregates_path);
-
-        measure_time!(
-            || {
-                let mut file = std::fs::File::create(aggregates_path)?;
-
-                file.write_all(
-                    format!(
-                        "selections{}{}\n",
-                        aggregates_delimiter,
-                        if protected {
-                            "protected_count"
-                        } else {
-                            "count"
-                        }
-                    )
-                    .as_bytes(),
-                )?;
-                file.write_all(
-                    format!(
-                        "selections{}{}\n",
-                        aggregates_delimiter,
-                        uround_down(self.data_block.records.len() as f64, resolution as f64)
-                    )
-                    .as_bytes(),
-                )?;
-                for aggregate in aggregates_count.keys() {
-                    file.write_all(
-                        format!(
-                            "{}{}{}\n",
-                            Aggregator::format_aggregate_str(&self.data_block.headers, aggregate),
-                            aggregates_delimiter,
-                            aggregates_count[aggregate].count
-                        )
-                        .as_bytes(),
-                    )?
-                }
-                Ok(())
-            },
-            (self.durations.write_aggregates_count)
-        )
-    }
-
-    /// Writes the records sensitivity to the file system in a csv/tsv like format
-    /// # Arguments:
-    /// * `records_sensitivity` - Calculated records sensitivity
-    /// * `records_sensitivity_path` - File path to be written
-    pub fn write_records_sensitivity(
-        &mut self,
-        records_sensitivity: &RecordsSensitivitySlice,
-        records_sensitivity_path: &str,
-    ) -> Result<(), Error> {
-        info!("writing file {}", records_sensitivity_path);
-
-        measure_time!(
-            || {
-                let mut file = std::fs::File::create(records_sensitivity_path)?;
-
-                file.write_all("record_index\trecord_sensitivity\n".as_bytes())?;
-                for (i, sensitivity) in records_sensitivity.iter().enumerate() {
-                    file.write_all(format!("{}\t{}\n", i, sensitivity).as_bytes())?
-                }
-                Ok(())
-            },
-            (self.durations.write_aggregates_count)
-        )
-    }
-
-    #[inline]
-    fn update_aggregate_progress<T>(
-        progress_reporter: &mut Option<T>,
-        n_processed: usize,
-        total: f64,
-    ) where
-        T: ReportProgress,
-    {
-        if let Some(r) = progress_reporter {
-            r.report(calc_percentage(n_processed, total));
-        }
-    }
-}
-
-impl<'data_block> Drop for Aggregator<'data_block> {
-    fn drop(&mut self) {
-        trace!("aggregator durations: {:#?}", self.durations);
-    }
-}
+mod rows_aggregator;

packages/core/src/processing/aggregator/privacy_risk_summary.rs

@@ -0,0 +1,144 @@
+use super::typedefs::{AggregatesCountMap, RecordsSet};
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+/// Represents the privacy risk information related to a data block
+#[cfg_attr(feature = "pyo3", pyclass)]
+#[derive(Debug)]
+pub struct PrivacyRiskSummary {
+    /// Total number of records on the data block
+    pub total_number_of_records: usize,
+    /// Total number of combinations aggregated (up to reporting length)
+    pub total_number_of_combinations: usize,
+    /// Number of records with unique combinations
+    pub records_with_unique_combinations_count: usize,
+    /// Number of records with rare combinations (combination count < resolution)
+    pub records_with_rare_combinations_count: usize,
+    /// Number of unique combinations
+    pub unique_combinations_count: usize,
+    /// Number of rare combinations
+    pub rare_combinations_count: usize,
+    /// Proportion of records containing unique combinations
+    pub records_with_unique_combinations_proportion: f64,
+    /// Proportion of records containing rare combinations
+    pub records_with_rare_combinations_proportion: f64,
+    /// Proportion of unique combinations
+    pub unique_combinations_proportion: f64,
+    /// Proportion of rare combinations
+    pub rare_combinations_proportion: f64,
+}
+
+#[cfg(feature = "pyo3")]
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl PrivacyRiskSummary {
+    /// Total number of records on the data block
+    #[getter]
+    fn total_number_of_records(&self) -> usize {
+        self.total_number_of_records
+    }
+
+    /// Total number of combinations aggregated (up to reporting length)
+    #[getter]
+    fn total_number_of_combinations(&self) -> usize {
+        self.total_number_of_combinations
+    }
+
+    /// Number of records with unique combinations
+    #[getter]
+    fn records_with_unique_combinations_count(&self) -> usize {
+        self.records_with_unique_combinations_count
+    }
+
+    /// Number of records with rare combinations (combination count < resolution)
+    #[getter]
+    fn records_with_rare_combinations_count(&self) -> usize {
+        self.records_with_rare_combinations_count
+    }
+
+    /// Number of unique combinations
+    #[getter]
+    fn unique_combinations_count(&self) -> usize {
+        self.unique_combinations_count
+    }
+
+    /// Number of rare combinations
+    #[getter]
+    fn rare_combinations_count(&self) -> usize {
+        self.rare_combinations_count
+    }
+
+    /// Proportion of records containing unique combinations
+    #[getter]
+    fn records_with_unique_combinations_proportion(&self) -> f64 {
+        self.records_with_unique_combinations_proportion
+    }
+
+    /// Proportion of records containing rare combinations
+    #[getter]
+    fn records_with_rare_combinations_proportion(&self) -> f64 {
+        self.records_with_rare_combinations_proportion
+    }
+
+    /// Proportion of unique combinations
+    #[getter]
+    fn unique_combinations_proportion(&self) -> f64 {
+        self.unique_combinations_proportion
+    }
+
+    /// Proportion of rare combinations
+    #[getter]
+    fn rare_combinations_proportion(&self) -> f64 {
+        self.rare_combinations_proportion
+    }
+}
+
+impl PrivacyRiskSummary {
+    #[inline]
+    /// Calculates ands returns the privacy risk related to the aggregates counts
+    /// # Arguments:
+    /// * `total_number_of_records` - Total number of records on the data block
+    /// * `aggregates_count` - Aggregates counts to compute the privacy risk for
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn from_aggregates_count(
+        total_number_of_records: usize,
+        aggregates_count: &AggregatesCountMap,
+        resolution: usize,
+    ) -> PrivacyRiskSummary {
+        let mut records_with_unique_combinations = RecordsSet::default();
+        let mut records_with_rare_combinations = RecordsSet::default();
+        let mut unique_combinations_count: usize = 0;
+        let mut rare_combinations_count: usize = 0;
+        let total_number_of_combinations = aggregates_count.len();
+
+        for count in aggregates_count.values() {
+            if count.count == 1 {
+                unique_combinations_count += 1;
+                records_with_unique_combinations.extend(&count.contained_in_records);
+            }
+            if count.count < resolution {
+                rare_combinations_count += 1;
+                records_with_rare_combinations.extend(&count.contained_in_records);
+            }
+        }
+
+        PrivacyRiskSummary {
+            total_number_of_records,
+            total_number_of_combinations,
+            records_with_unique_combinations_count: records_with_unique_combinations.len(),
+            records_with_rare_combinations_count: records_with_rare_combinations.len(),
+            unique_combinations_count,
+            rare_combinations_count,
+            records_with_unique_combinations_proportion: (records_with_unique_combinations.len()
+                as f64)
+                / (total_number_of_records as f64),
+            records_with_rare_combinations_proportion: (records_with_rare_combinations.len()
+                as f64)
+                / (total_number_of_records as f64),
+            unique_combinations_proportion: (unique_combinations_count as f64)
+                / (total_number_of_combinations as f64),
+            rare_combinations_proportion: (rare_combinations_count as f64)
+                / (total_number_of_combinations as f64),
+        }
+    }
+}

packages/core/src/processing/aggregator/record_attrs_selector.rs

@@ -1,5 +1,5 @@
 use fnv::FnvHashMap;
-use itertools::Itertools;
+use std::sync::Arc;
 
 use crate::{
     data_block::{typedefs::AttributeRowsMap, value::DataBlockValue},
@@ -13,19 +13,19 @@ use crate::{
 /// Sensitivity in this context means the number of combinations that
 /// can be generated based on the number of non-empty values on the record.
 /// So attributes will suppressed from the record until record sensitivity <= `sensitivity_threshold`
-pub struct RecordAttrsSelector<'length_range, 'data_block> {
+pub struct RecordAttrsSelector<'length_range> {
+    /// Range to compute the number of combinations [1...reporting_length]
+    pub length_range: &'length_range [usize],
     /// Cache for how many values should be suppressed based of the number
     /// of non-empty attributes on the record
     cache: FnvHashMap<usize, usize>,
-    /// Range to compute the number of combinations [1...reporting_length]
-    length_range: &'length_range [usize],
     /// Maximum sensitivity allowed for each record
     sensitivity_threshold: usize,
     /// Map with a data block value as key and all the attribute row indexes where it occurs as value
-    attr_rows_map: AttributeRowsMap<'data_block>,
+    attr_rows_map: Arc<AttributeRowsMap>,
 }
 
-impl<'length_range, 'data_block> RecordAttrsSelector<'length_range, 'data_block> {
+impl<'length_range> RecordAttrsSelector<'length_range> {
     /// Returns a new RecordAttrsSelector
     /// # Arguments
     /// * `length_range` - Range to compute the number of combinations [1...reporting_length]
@@ -36,8 +36,8 @@ impl<'length_range, 'data_block> RecordAttrsSelector<'length_range, 'data_block>
     pub fn new(
         length_range: &'length_range [usize],
         sensitivity_threshold: usize,
-        attr_rows_map: AttributeRowsMap<'data_block>,
-    ) -> RecordAttrsSelector<'length_range, 'data_block> {
+        attr_rows_map: Arc<AttributeRowsMap>,
+    ) -> RecordAttrsSelector<'length_range> {
         RecordAttrsSelector {
             cache: FnvHashMap::default(),
             length_range,
@@ -53,30 +53,30 @@ impl<'length_range, 'data_block> RecordAttrsSelector<'length_range, 'data_block>
     /// all records, higher the chance for it **not** to be suppressed
     ///
     /// # Arguments
-    /// * `record`: record to select attributes from
+    /// * `record` - record to select attributes from
     pub fn select_from_record(
         &mut self,
-        record: &'data_block [DataBlockValue],
-    ) -> Vec<&'data_block DataBlockValue> {
+        record: &[Arc<DataBlockValue>],
+    ) -> Vec<Arc<DataBlockValue>> {
         let n_attributes = record.len();
         let selected_attrs_count = n_attributes - self.get_suppressed_count(n_attributes);
 
         if selected_attrs_count < n_attributes {
             let mut attr_count_map: AttributeCountMap = record
                 .iter()
-                .map(|r| (r, self.attr_rows_map.get(r).unwrap().len()))
+                .map(|r| (r.clone(), self.attr_rows_map.get(r).unwrap().len()))
                 .collect();
-            let mut res: Vec<&DataBlockValue> = Vec::default();
+            let mut res: Vec<Arc<DataBlockValue>> = Vec::default();
 
             for _ in 0..selected_attrs_count {
                 if let Some(sample) = SynthesizerContext::sample_from_attr_counts(&attr_count_map) {
+                    attr_count_map.remove(&sample);
                     res.push(sample);
-                    attr_count_map.remove(sample);
                 }
             }
             res
         } else {
-            record.iter().collect_vec()
+            record.to_vec()
         }
     }
 

packages/core/src/processing/aggregator/records_analysis_data.rs

@@ -0,0 +1,247 @@
+use super::typedefs::{RecordsAnalysisByLenMap, RecordsByLenMap};
+use itertools::Itertools;
+use std::io::{Error, Write};
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+use crate::utils::math::{calc_percentage, uround_down};
+
+#[cfg_attr(feature = "pyo3", pyclass)]
+#[derive(Clone)]
+/// Analysis information related to a single record
+pub struct RecordsAnalysis {
+    /// Number of records containing unique combinations
+    pub unique_combinations_records_count: usize,
+    /// Percentage of records containing unique combinations
+    pub unique_combinations_records_percentage: f64,
+    /// Number of records containing rare combinations
+    /// (unique os not taken into account)
+    pub rare_combinations_records_count: usize,
+    /// Percentage of records containing rare combinations
+    /// (unique os not taken into account)
+    pub rare_combinations_records_percentage: f64,
+    /// Count of unique + rare
+    pub risky_combinations_records_count: usize,
+    /// Percentage of unique + rare
+    pub risky_combinations_records_percentage: f64,
+}
+
+impl RecordsAnalysis {
+    #[inline]
+    /// Created a new RecordsAnalysis with default values
+    pub fn default() -> RecordsAnalysis {
+        RecordsAnalysis {
+            unique_combinations_records_count: 0,
+            unique_combinations_records_percentage: 0.0,
+            rare_combinations_records_count: 0,
+            rare_combinations_records_percentage: 0.0,
+            risky_combinations_records_count: 0,
+            risky_combinations_records_percentage: 0.0,
+        }
+    }
+}
+
+#[cfg(feature = "pyo3")]
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl RecordsAnalysis {
+    #[getter]
+    /// Number of records containing unique combinations
+    fn unique_combinations_records_count(&self) -> usize {
+        self.unique_combinations_records_count
+    }
+
+    #[getter]
+    /// Percentage of records containing unique combinations
+    fn unique_combinations_records_percentage(&self) -> f64 {
+        self.unique_combinations_records_percentage
+    }
+
+    #[getter]
+    /// Number of records containing rare combinations
+    /// (unique os not taken into account)
+    fn rare_combinations_records_count(&self) -> usize {
+        self.rare_combinations_records_count
+    }
+
+    #[getter]
+    /// Percentage of records containing rare combinations
+    /// (unique os not taken into account)
+    fn rare_combinations_records_percentage(&self) -> f64 {
+        self.rare_combinations_records_percentage
+    }
+
+    #[getter]
+    /// Count of unique + rare
+    fn risky_combinations_records_count(&self) -> usize {
+        self.risky_combinations_records_count
+    }
+
+    #[getter]
+    /// Percentage of unique + rare
+    fn risky_combinations_records_percentage(&self) -> f64 {
+        self.risky_combinations_records_percentage
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pyclass)]
+/// Stores the records analysis for all records grouped by
+/// combination length
+pub struct RecordsAnalysisData {
+    /// Map of records analysis grouped by combination len
+    pub records_analysis_by_len: RecordsAnalysisByLenMap,
+}
+
+impl RecordsAnalysisData {
+    /// Computes the record analysis from the arguments.
+    /// # Arguments
+    /// * `unique_records_by_len` - unique records grouped by length
+    /// * `rare_records_by_len` - rare records grouped by length
+    /// * `total_number_of_records` - Total number of records on the data block
+    /// * `reporting_length` - Reporting length used for the data aggregation
+    /// * `resolution` - Reporting resolution used for data synthesis
+    /// * `protect` - Whether or not the counts should be rounded to the
+    /// nearest smallest multiple of resolution
+    #[inline]
+    pub fn from_unique_rare_combinations_records_by_len(
+        unique_records_by_len: &RecordsByLenMap,
+        rare_records_by_len: &RecordsByLenMap,
+        total_n_records: usize,
+        reporting_length: usize,
+        resolution: usize,
+        protect: bool,
+    ) -> RecordsAnalysisData {
+        let total_n_records_f64 = if protect {
+            uround_down(total_n_records as f64, resolution as f64)
+        } else {
+            total_n_records
+        } as f64;
+        let records_analysis_by_len: RecordsAnalysisByLenMap = (1..=reporting_length)
+            .map(|l| {
+                let mut ra = RecordsAnalysis::default();
+
+                ra.unique_combinations_records_count = unique_records_by_len
+                    .get(&l)
+                    .map_or(0, |records| records.len());
+                ra.rare_combinations_records_count = rare_records_by_len
+                    .get(&l)
+                    .map_or(0, |records| records.len());
+
+                if protect {
+                    ra.unique_combinations_records_count = uround_down(
+                        ra.unique_combinations_records_count as f64,
+                        resolution as f64,
+                    );
+                    ra.rare_combinations_records_count =
+                        uround_down(ra.rare_combinations_records_count as f64, resolution as f64)
+                }
+
+                ra.unique_combinations_records_percentage = calc_percentage(
+                    ra.unique_combinations_records_count as f64,
+                    total_n_records_f64,
+                );
+                ra.rare_combinations_records_percentage = calc_percentage(
+                    ra.rare_combinations_records_count as f64,
+                    total_n_records_f64,
+                );
+                ra.risky_combinations_records_count =
+                    ra.unique_combinations_records_count + ra.rare_combinations_records_count;
+                ra.risky_combinations_records_percentage = calc_percentage(
+                    ra.risky_combinations_records_count as f64,
+                    total_n_records_f64,
+                );
+
+                (l, ra)
+            })
+            .collect();
+        RecordsAnalysisData {
+            records_analysis_by_len,
+        }
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl RecordsAnalysisData {
+    /// Returns the records analysis map grouped by length.
+    /// This method will clone the data, so its recommended to have its result stored
+    /// in a local variable to avoid it being called multiple times
+    pub fn get_records_analysis_by_len(&self) -> RecordsAnalysisByLenMap {
+        self.records_analysis_by_len.clone()
+    }
+
+    /// Returns the total number of records containing unique combinations
+    /// for all lengths
+    pub fn get_total_unique(&self) -> usize {
+        self.records_analysis_by_len
+            .values()
+            .map(|v| v.unique_combinations_records_count)
+            .sum()
+    }
+
+    /// Returns the total number of records containing rare combinations
+    /// for all lengths
+    pub fn get_total_rare(&self) -> usize {
+        self.records_analysis_by_len
+            .values()
+            .map(|v| v.rare_combinations_records_count)
+            .sum()
+    }
+
+    /// Returns the total number of records containing risky (unique + rare)
+    /// combinations for all lengths
+    pub fn get_total_risky(&self) -> usize {
+        self.records_analysis_by_len
+            .values()
+            .map(|v| v.risky_combinations_records_count)
+            .sum()
+    }
+
+    /// Writes the records analysis to the file system in a csv/tsv like format
+    /// # Arguments:
+    /// * `records_analysis_path` - File path to be written
+    /// * `records_analysis_delimiter` - Delimiter to use when writing to `records_analysis_path`
+    pub fn write_records_analysis(
+        &mut self,
+        records_analysis_path: &str,
+        records_analysis_delimiter: char,
+    ) -> Result<(), Error> {
+        let mut file = std::fs::File::create(records_analysis_path)?;
+
+        file.write_all(
+            format!(
+                "combo_length{}sen_rare{}sen_rare_pct{}sen_unique{}sen_unique_pct{}sen_risky{}sen_risky_pct\n",
+                records_analysis_delimiter,
+                records_analysis_delimiter,
+                records_analysis_delimiter,
+                records_analysis_delimiter,
+                records_analysis_delimiter,
+                records_analysis_delimiter
+            )
+            .as_bytes(),
+        )?;
+        for l in self.records_analysis_by_len.keys().sorted() {
+            let ra = &self.records_analysis_by_len[l];
+
+            file.write_all(
+                format!(
+                    "{}{}{}{}{}{}{}{}{}{}{}{}{}\n",
+                    l,
+                    records_analysis_delimiter,
+                    ra.rare_combinations_records_count,
+                    records_analysis_delimiter,
+                    ra.rare_combinations_records_percentage,
+                    records_analysis_delimiter,
+                    ra.unique_combinations_records_count,
+                    records_analysis_delimiter,
+                    ra.unique_combinations_records_percentage,
+                    records_analysis_delimiter,
+                    ra.risky_combinations_records_count,
+                    records_analysis_delimiter,
+                    ra.risky_combinations_records_percentage,
+                )
+                .as_bytes(),
+            )?
+        }
+        Ok(())
+    }
+}

packages/core/src/processing/aggregator/rows_aggregator.rs

@@ -0,0 +1,195 @@
+use super::{
+    record_attrs_selector::RecordAttrsSelector,
+    typedefs::{AggregatesCountMap, EnumeratedDataBlockRecords, RecordsSensitivity},
+    value_combination::ValueCombination,
+    AggregatedCount,
+};
+use itertools::Itertools;
+use log::info;
+use std::sync::Arc;
+
+#[cfg(feature = "rayon")]
+use rayon::prelude::*;
+
+#[cfg(feature = "rayon")]
+use std::sync::Mutex;
+
+use crate::{
+    data_block::block::DataBlock,
+    utils::{
+        math::calc_n_combinations_range,
+        reporting::{ReportProgress, SendableProgressReporter, SendableProgressReporterRef},
+    },
+};
+
+pub struct RowsAggregatorResult {
+    pub all_combs_count: u64,
+    pub selected_combs_count: u64,
+    pub aggregates_count: AggregatesCountMap,
+    pub records_sensitivity: RecordsSensitivity,
+}
+
+impl RowsAggregatorResult {
+    #[inline]
+    pub fn new(total_n_records: usize) -> RowsAggregatorResult {
+        let mut records_sensitivity = RecordsSensitivity::default();
+
+        records_sensitivity.resize(total_n_records, 0);
+        RowsAggregatorResult {
+            all_combs_count: 0,
+            selected_combs_count: 0,
+            aggregates_count: AggregatesCountMap::default(),
+            records_sensitivity,
+        }
+    }
+}
+
+pub struct RowsAggregator<'length_range> {
+    data_block: Arc<DataBlock>,
+    enumerated_records: EnumeratedDataBlockRecords,
+    record_attrs_selector: RecordAttrsSelector<'length_range>,
+}
+
+impl<'length_range> RowsAggregator<'length_range> {
+    #[inline]
+    pub fn new(
+        data_block: Arc<DataBlock>,
+        enumerated_records: EnumeratedDataBlockRecords,
+        record_attrs_selector: RecordAttrsSelector<'length_range>,
+    ) -> RowsAggregator {
+        RowsAggregator {
+            data_block,
+            enumerated_records,
+            record_attrs_selector,
+        }
+    }
+
+    #[cfg(feature = "rayon")]
+    #[inline]
+    pub fn aggregate_all<T>(
+        total_n_records: usize,
+        rows_aggregators: &mut Vec<RowsAggregator>,
+        progress_reporter: &mut Option<T>,
+    ) -> RowsAggregatorResult
+    where
+        T: ReportProgress,
+    {
+        let sendable_pr =
+            Arc::new(Mutex::new(progress_reporter.as_mut().map(|r| {
+                SendableProgressReporter::new(total_n_records as f64, 1.0, r)
+            })));
+
+        RowsAggregator::join_partial_results(
+            total_n_records,
+            rows_aggregators
+                .par_iter_mut()
+                .map(|ra| ra.aggregate_rows(&mut sendable_pr.clone()))
+                .collect(),
+        )
+    }
+
+    #[cfg(not(feature = "rayon"))]
+    #[inline]
+    pub fn aggregate_all<T>(
+        total_n_records: usize,
+        rows_aggregators: &mut Vec<RowsAggregator>,
+        progress_reporter: &mut Option<T>,
+    ) -> RowsAggregatorResult
+    where
+        T: ReportProgress,
+    {
+        let mut sendable_pr = progress_reporter
+            .as_mut()
+            .map(|r| SendableProgressReporter::new(total_n_records as f64, 1.0, r));
+
+        RowsAggregator::join_partial_results(
+            total_n_records,
+            rows_aggregators
+                .iter_mut()
+                .map(|ra| ra.aggregate_rows(&mut sendable_pr))
+                .collect(),
+        )
+    }
+
+    #[inline]
+    fn join_partial_results(
+        total_n_records: usize,
+        mut partial_results: Vec<RowsAggregatorResult>,
+    ) -> RowsAggregatorResult {
+        info!("joining aggregated partial results...");
+
+        // take last element and the initial result
+        // or use the default one if there are no results
+        let mut final_result = partial_results
+            .pop()
+            .unwrap_or_else(|| RowsAggregatorResult::new(total_n_records));
+
+        // use drain instead of fold, so we do not duplicate memory
+        for mut partial_result in partial_results.drain(..) {
+            // join counts
+            final_result.all_combs_count += partial_result.all_combs_count;
+            final_result.selected_combs_count += partial_result.selected_combs_count;
+
+            // join aggregated counts
+            for (comb, value) in partial_result.aggregates_count.drain() {
+                let final_count = final_result
+                    .aggregates_count
+                    .entry(comb)
+                    .or_insert_with(AggregatedCount::default);
+                final_count.count += value.count;
+                final_count
+                    .contained_in_records
+                    .extend(value.contained_in_records);
+            }
+
+            // join records sensitivity
+            for (i, sensitivity) in partial_result.records_sensitivity.drain(..).enumerate() {
+                final_result.records_sensitivity[i] += sensitivity;
+            }
+        }
+        final_result
+    }
+
+    #[inline]
+    fn aggregate_rows<T>(
+        &mut self,
+        progress_reporter: &mut SendableProgressReporterRef<T>,
+    ) -> RowsAggregatorResult
+    where
+        T: ReportProgress,
+    {
+        let mut result = RowsAggregatorResult::new(self.data_block.records.len());
+
+        for (record_index, record) in self.enumerated_records.iter() {
+            let mut selected_attrs = self
+                .record_attrs_selector
+                .select_from_record(&record.values);
+
+            // sort the attributes here, so combinations will be already sorted
+            // and we do not need to sort entry by entry on the loop below
+            selected_attrs.sort_by_key(|k| k.format_str_using_headers(&self.data_block.headers));
+
+            result.all_combs_count += calc_n_combinations_range(
+                record.values.len(),
+                self.record_attrs_selector.length_range,
+            );
+
+            for l in self.record_attrs_selector.length_range {
+                for mut c in selected_attrs.iter().combinations(*l) {
+                    let current_count = result
+                        .aggregates_count
+                        .entry(ValueCombination::new(
+                            c.drain(..).map(|k| (*k).clone()).collect(),
+                        ))
+                        .or_insert_with(AggregatedCount::default);
+                    current_count.count += 1;
+                    current_count.contained_in_records.insert(*record_index);
+                    result.records_sensitivity[*record_index] += 1;
+                    result.selected_combs_count += 1;
+                }
+            }
+            SendableProgressReporter::update_progress(progress_reporter, 1.0);
+        }
+        result
+    }
+}

packages/core/src/processing/aggregator/typedefs.rs

@@ -1,26 +1,36 @@
-use super::AggregatedCount;
+use super::{
+    data_aggregator::AggregatedCount, records_analysis_data::RecordsAnalysis,
+    value_combination::ValueCombination,
+};
 use fnv::{FnvHashMap, FnvHashSet};
+use std::sync::Arc;
 
-use crate::data_block::value::DataBlockValue;
+use crate::data_block::record::DataBlockRecord;
 
 /// Set of records where the key is the record index starting in 0
 pub type RecordsSet = FnvHashSet<usize>;
 
-/// Vector of data block values representing a value combination (sorted by `{header_name}:{block_value}`)
-pub type ValueCombination<'data_block_value> = Vec<&'data_block_value DataBlockValue>;
-
-/// Slice of ValueCombination
-pub type ValueCombinationSlice<'data_block_value> = [&'data_block_value DataBlockValue];
-
 /// Maps a value combination to its aggregated count
-pub type AggregatesCountMap<'data_block_value> =
-    FnvHashMap<ValueCombination<'data_block_value>, AggregatedCount>;
+pub type AggregatesCountMap = FnvHashMap<ValueCombination, AggregatedCount>;
+
+/// Maps a value combination represented as a string to its aggregated count
+pub type AggregatesCountStringMap = FnvHashMap<String, AggregatedCount>;
 
 /// Maps a length (1,2,3... up to reporting length) to a determined count
 pub type AggregatedCountByLenMap = FnvHashMap<usize, usize>;
 
+/// Maps a length (1,2,3... up to reporting length) to a record set
+pub type RecordsByLenMap = FnvHashMap<usize, RecordsSet>;
+
 /// A vector of sensitivities for each record (the vector index is the record index)
 pub type RecordsSensitivity = Vec<usize>;
 
 /// Slice of RecordsSensitivity
 pub type RecordsSensitivitySlice = [usize];
+
+/// Vector of tuples:
+/// (index of the original record, reference to the original record)
+pub type EnumeratedDataBlockRecords = Vec<(usize, Arc<DataBlockRecord>)>;
+
+/// Map of records analysis grouped by combination len
+pub type RecordsAnalysisByLenMap = FnvHashMap<usize, RecordsAnalysis>;

packages/core/src/processing/aggregator/value_combination.rs

@@ -0,0 +1,152 @@
+use serde::{
+    de::{self, Visitor},
+    Deserialize, Serialize,
+};
+use std::{
+    fmt::Display,
+    marker::PhantomData,
+    ops::{Deref, DerefMut},
+    str::FromStr,
+    sync::Arc,
+};
+
+use crate::data_block::{
+    typedefs::DataBlockHeadersSlice,
+    value::{DataBlockValue, ParseDataBlockValueError},
+};
+
+const COMBINATIONS_DELIMITER: char = ';';
+
+#[derive(Eq, PartialEq, Hash)]
+/// Wraps a vector of data block values representing a value
+/// combination (sorted by `{header_name}:{block_value}`)
+pub struct ValueCombination {
+    combination: Vec<Arc<DataBlockValue>>,
+}
+
+impl ValueCombination {
+    #[inline]
+    /// Creates a new ValueCombination with default values
+    pub fn default() -> ValueCombination {
+        ValueCombination::new(Vec::default())
+    }
+
+    #[inline]
+    /// Creates a new ValueCombination
+    /// # Arguments
+    /// * `combination` - raw vector of value combinations
+    /// sorted by `{header_name}:{block_value}`
+    pub fn new(combination: Vec<Arc<DataBlockValue>>) -> ValueCombination {
+        ValueCombination { combination }
+    }
+
+    /// Formats a value combination as String using the headers.
+    /// The result is formatted as:
+    /// `{header_name}:{block_value};{header_name}:{block_value}...`
+    /// # Arguments
+    /// * `headers` - Data block headers
+    /// * `combination_delimiter` - Delimiter used to join combinations
+    #[inline]
+    pub fn format_str_using_headers(
+        &self,
+        headers: &DataBlockHeadersSlice,
+        combination_delimiter: &str,
+    ) -> String {
+        let mut str = String::default();
+
+        if let Some(comb) = self.combination.get(0) {
+            str.push_str(&comb.format_str_using_headers(headers));
+        }
+        for comb in self.combination.iter().skip(1) {
+            str += combination_delimiter;
+            str.push_str(&comb.format_str_using_headers(headers));
+        }
+        str
+    }
+}
+
+impl Display for ValueCombination {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        if let Some(comb) = self.combination.get(0) {
+            write!(f, "{}", comb)?;
+        }
+        for comb in self.combination.iter().skip(1) {
+            write!(f, "{}{}", COMBINATIONS_DELIMITER, comb)?;
+        }
+        Ok(())
+    }
+}
+
+impl FromStr for ValueCombination {
+    type Err = ParseDataBlockValueError;
+
+    /// Creates a new ValueCombination by parsing `str_value`
+    fn from_str(str_value: &str) -> Result<Self, Self::Err> {
+        Ok(ValueCombination::new(
+            str_value
+                .split(COMBINATIONS_DELIMITER)
+                .map(|v| Ok(Arc::new(DataBlockValue::from_str(v)?)))
+                .collect::<Result<Vec<Arc<DataBlockValue>>, Self::Err>>()?,
+        ))
+    }
+}
+
+impl Deref for ValueCombination {
+    type Target = Vec<Arc<DataBlockValue>>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.combination
+    }
+}
+
+impl DerefMut for ValueCombination {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.combination
+    }
+}
+
+// Implementing a custom serializer, so this can
+// be properly written to a json file
+impl Serialize for ValueCombination {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::Serializer,
+    {
+        serializer.serialize_str(&format!("{}", self))
+    }
+}
+
+struct ValueCombinationVisitor {
+    marker: PhantomData<fn() -> ValueCombination>,
+}
+
+impl ValueCombinationVisitor {
+    fn new() -> Self {
+        ValueCombinationVisitor {
+            marker: PhantomData,
+        }
+    }
+}
+
+impl<'de> Visitor<'de> for ValueCombinationVisitor {
+    type Value = ValueCombination;
+
+    fn expecting(&self, formatter: &mut std::fmt::Formatter) -> std::fmt::Result {
+        formatter.write_str("a string representing the value combination")
+    }
+
+    fn visit_str<E: de::Error>(self, v: &str) -> Result<Self::Value, E> {
+        ValueCombination::from_str(v).map_err(E::custom)
+    }
+}
+
+// Implementing a custom deserializer, so this can
+// be properly read from a json file
+impl<'de> Deserialize<'de> for ValueCombination {
+    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
+    where
+        D: serde::Deserializer<'de>,
+    {
+        deserializer.deserialize_string(ValueCombinationVisitor::new())
+    }
+}

packages/core/src/processing/evaluator/bucket.rs

@@ -1,86 +0,0 @@
-use std::ops::{Deref, DerefMut};
-
-const INITIAL_BIN: usize = 10;
-const BIN_RATIO: usize = 2;
-
-/// Stores the preservation information related to a particular bucket.
-/// A bucket stores count in a certain range value.
-/// For example: (0, 10], (10, 20], (20, 40]...
-#[derive(Debug)]
-pub struct PreservationByCountBucket {
-    /// How many elements are stored in the bucket
-    pub size: usize,
-    /// Preservation sum of all elements in the bucket
-    pub preservation_sum: f64,
-    /// Combination length sum of all elements in the bucket
-    pub length_sum: usize,
-}
-
-impl PreservationByCountBucket {
-    /// Return a new PreservationByCountBucket with default values
-    pub fn default() -> PreservationByCountBucket {
-        PreservationByCountBucket {
-            size: 0,
-            preservation_sum: 0.0,
-            length_sum: 0,
-        }
-    }
-
-    /// Adds a new value to the bucket
-    /// # Arguments
-    /// * `preservation` - Preservation related to the value
-    /// * `length` - Combination length related to the value
-    pub fn add(&mut self, preservation: f64, length: usize) {
-        self.size += 1;
-        self.preservation_sum += preservation;
-        self.length_sum += length;
-    }
-}
-
-/// Stores the bins for each bucket. For example:
-/// [10, 20, 40, 80, 160...]
-#[derive(Debug)]
-pub struct PreservationByCountBucketBins {
-    /// Bins where the index of the vector is the bin index,
-    /// and the value is the max value count allowed on the bin
-    bins: Vec<usize>,
-}
-
-impl PreservationByCountBucketBins {
-    /// Generates a new PreservationByCountBucketBins with bins up to a `max_val`
-    /// # Arguments
-    /// * `max_val` - Max value allowed on the last bucket (inclusive)
-    pub fn new(max_val: usize) -> PreservationByCountBucketBins {
-        let mut bins: Vec<usize> = vec![INITIAL_BIN];
-        loop {
-            let last = *bins.last().unwrap();
-            if last >= max_val {
-                break;
-            }
-            bins.push(last * BIN_RATIO);
-        }
-        PreservationByCountBucketBins { bins }
-    }
-
-    /// Find the first `bucket_max_val` where `val >= bucket_max_val`
-    /// # Arguments
-    /// * `val` - Count to look in which bucket it will fit
-    pub fn find_bucket_max_val(&self, val: usize) -> usize {
-        // find first element x where x >= val
-        self.bins[self.bins.partition_point(|x| *x < val)]
-    }
-}
-
-impl Deref for PreservationByCountBucketBins {
-    type Target = Vec<usize>;
-
-    fn deref(&self) -> &Self::Target {
-        &self.bins
-    }
-}
-
-impl DerefMut for PreservationByCountBucketBins {
-    fn deref_mut(&mut self) -> &mut Self::Target {
-        &mut self.bins
-    }
-}

packages/core/src/processing/evaluator/data_evaluator.rs

@@ -0,0 +1,252 @@
+use super::preservation_by_count::{PreservationByCountBucketBins, PreservationByCountBuckets};
+use super::rare_combinations_comparison_data::RareCombinationsComparisonData;
+use fnv::FnvHashSet;
+use log::info;
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+use crate::processing::aggregator::aggregated_data::AggregatedData;
+use crate::processing::aggregator::typedefs::AggregatedCountByLenMap;
+use crate::processing::aggregator::value_combination::ValueCombination;
+use crate::processing::evaluator::preservation_bucket::PreservationBucket;
+use crate::processing::evaluator::preservation_by_length::PreservationByLengthBuckets;
+use crate::utils::time::ElapsedDurationLogger;
+
+#[cfg_attr(feature = "pyo3", pyclass)]
+/// Evaluates aggregated, sensitive and synthesized data
+pub struct Evaluator {}
+
+impl Evaluator {
+    /// Returns a new Evaluator
+    pub fn default() -> Evaluator {
+        Evaluator {}
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl Evaluator {
+    /// Returns a new Evaluator
+    #[cfg(feature = "pyo3")]
+    #[new]
+    pub fn default_pyo3() -> Evaluator {
+        Evaluator::default()
+    }
+
+    /// Calculates the leakage counts grouped by combination length
+    /// (how many attribute combinations exist on the sensitive data and
+    /// appear on the synthetic data with `count < resolution`).
+    /// By design this should be `0`
+    /// # Arguments
+    /// * `sensitive_aggregated_data` - Calculated aggregated data for the sensitive data
+    /// * `synthetic_aggregated_data` - Calculated aggregated data for the synthetic data
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn calc_leakage_count(
+        &self,
+        sensitive_aggregated_data: &AggregatedData,
+        synthetic_aggregated_data: &AggregatedData,
+        resolution: usize,
+    ) -> AggregatedCountByLenMap {
+        let _duration_logger = ElapsedDurationLogger::new("leakage count calculation");
+        let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
+
+        info!("calculating rare sensitive combination leakages by length");
+
+        for (sensitive_agg, sensitive_count) in sensitive_aggregated_data.aggregates_count.iter() {
+            if sensitive_count.count < resolution {
+                if let Some(synthetic_count) = synthetic_aggregated_data
+                    .aggregates_count
+                    .get(sensitive_agg)
+                {
+                    if synthetic_count.count < resolution {
+                        let leaks = result.entry(sensitive_agg.len()).or_insert(0);
+                        *leaks += 1;
+                    }
+                }
+            }
+        }
+        result
+    }
+
+    /// Calculates the fabricated counts grouped by combination length
+    /// (how many attribute combinations exist on the synthetic data that do not
+    /// exist on the sensitive data).
+    /// By design this should be `0`
+    /// # Arguments
+    /// * `sensitive_aggregated_data` - Calculated aggregated data for the sensitive data
+    /// * `synthetic_aggregated_data` - Calculated aggregated data for the synthetic data
+    pub fn calc_fabricated_count(
+        &self,
+        sensitive_aggregated_data: &AggregatedData,
+        synthetic_aggregated_data: &AggregatedData,
+    ) -> AggregatedCountByLenMap {
+        let _duration_logger = ElapsedDurationLogger::new("fabricated count calculation");
+        let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
+
+        info!("calculating fabricated synthetic combinations by length");
+
+        for synthetic_agg in synthetic_aggregated_data.aggregates_count.keys() {
+            if sensitive_aggregated_data
+                .aggregates_count
+                .get(synthetic_agg)
+                .is_none()
+            {
+                let fabricated = result.entry(synthetic_agg.len()).or_insert(0);
+                *fabricated += 1;
+            }
+        }
+        result
+    }
+
+    /// Calculates the preservation information grouped by combination count.
+    /// An example output might be a map like:
+    /// `{ 10 -> PreservationBucket, 20 -> PreservationBucket, ...}`
+    /// # Arguments
+    /// * `sensitive_aggregated_data` - Calculated aggregated data for the sensitive data
+    /// * `synthetic_aggregated_data` - Calculated aggregated data for the synthetic data
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn calc_preservation_by_count(
+        &self,
+        sensitive_aggregated_data: &AggregatedData,
+        synthetic_aggregated_data: &AggregatedData,
+        resolution: usize,
+    ) -> PreservationByCountBuckets {
+        let _duration_logger = ElapsedDurationLogger::new("preservation by count calculation");
+
+        info!(
+            "calculating preservation by count with resolution: {}",
+            resolution
+        );
+
+        let max_syn_count = *synthetic_aggregated_data
+            .aggregates_count
+            .values()
+            .map(|a| &a.count)
+            .max()
+            .unwrap_or(&0);
+        let bins: PreservationByCountBucketBins = PreservationByCountBucketBins::new(max_syn_count);
+        let mut buckets: PreservationByCountBuckets = PreservationByCountBuckets::default();
+        let mut processed_combs: FnvHashSet<&ValueCombination> = FnvHashSet::default();
+
+        for (comb, count) in sensitive_aggregated_data.aggregates_count.iter() {
+            // exclude sensitive rare combinations
+            if count.count >= resolution && !processed_combs.contains(comb) {
+                buckets.populate(
+                    &bins,
+                    comb,
+                    &sensitive_aggregated_data.aggregates_count,
+                    &synthetic_aggregated_data.aggregates_count,
+                );
+                processed_combs.insert(comb);
+            }
+        }
+        for comb in synthetic_aggregated_data.aggregates_count.keys() {
+            if !processed_combs.contains(comb) {
+                buckets.populate(
+                    &bins,
+                    comb,
+                    &sensitive_aggregated_data.aggregates_count,
+                    &synthetic_aggregated_data.aggregates_count,
+                );
+                processed_combs.insert(comb);
+            }
+        }
+
+        // fill empty buckets with default value
+        for bin in bins.iter() {
+            buckets
+                .entry(*bin)
+                .or_insert_with(PreservationBucket::default);
+        }
+
+        buckets
+    }
+
+    /// Calculates the preservation information grouped by combination length.
+    /// An example output might be a map like:
+    /// `{ 1 -> PreservationBucket, 2 -> PreservationBucket, ...}`
+    /// # Arguments
+    /// * `sensitive_aggregated_data` - Calculated aggregated data for the sensitive data
+    /// * `synthetic_aggregated_data` - Calculated aggregated data for the synthetic data
+    /// * `resolution` - Reporting resolution used for data synthesis
+    pub fn calc_preservation_by_length(
+        &self,
+        sensitive_aggregated_data: &AggregatedData,
+        synthetic_aggregated_data: &AggregatedData,
+        resolution: usize,
+    ) -> PreservationByLengthBuckets {
+        let _duration_logger = ElapsedDurationLogger::new("preservation by length calculation");
+
+        info!(
+            "calculating preservation by length with resolution: {}",
+            resolution
+        );
+
+        let mut buckets: PreservationByLengthBuckets = PreservationByLengthBuckets::default();
+        let mut processed_combs: FnvHashSet<&ValueCombination> = FnvHashSet::default();
+
+        for (comb, count) in sensitive_aggregated_data.aggregates_count.iter() {
+            // exclude sensitive rare combinations
+            if count.count >= resolution && !processed_combs.contains(comb) {
+                buckets.populate(
+                    comb,
+                    &sensitive_aggregated_data.aggregates_count,
+                    &synthetic_aggregated_data.aggregates_count,
+                );
+                processed_combs.insert(comb);
+            }
+        }
+        for comb in synthetic_aggregated_data.aggregates_count.keys() {
+            if !processed_combs.contains(comb) {
+                buckets.populate(
+                    comb,
+                    &sensitive_aggregated_data.aggregates_count,
+                    &synthetic_aggregated_data.aggregates_count,
+                );
+                processed_combs.insert(comb);
+            }
+        }
+
+        // fill empty buckets with default value
+        for l in 1..=synthetic_aggregated_data.reporting_length {
+            buckets.entry(l).or_insert_with(PreservationBucket::default);
+        }
+
+        buckets
+    }
+
+    //// Compares the rare combinations on the synthetic data with
+    /// the sensitive data counts
+    /// # Arguments
+    /// * `sensitive_aggregated_data` - Calculated aggregated data for the sensitive data
+    /// * `synthetic_aggregated_data` - Calculated aggregated data for the synthetic data
+    /// * `resolution` - Reporting resolution used for data synthesis
+    /// * `combination_delimiter` - Delimiter used to join combinations and format then
+    /// as strings
+    /// * `protect` - Whether or not the sensitive counts should be rounded to the
+    /// nearest smallest multiple of resolution
+    pub fn compare_synthetic_and_sensitive_rare(
+        &self,
+        synthetic_aggregated_data: &AggregatedData,
+        sensitive_aggregated_data: &AggregatedData,
+        resolution: usize,
+        combination_delimiter: &str,
+        protect: bool,
+    ) -> RareCombinationsComparisonData {
+        let _duration_logger =
+            ElapsedDurationLogger::new("synthetic and sensitive rare comparison");
+        RareCombinationsComparisonData::from_synthetic_and_sensitive_aggregated_data(
+            synthetic_aggregated_data,
+            sensitive_aggregated_data,
+            resolution,
+            combination_delimiter,
+            protect,
+        )
+    }
+}
+
+#[cfg(feature = "pyo3")]
+pub fn register(_py: Python, m: &PyModule) -> PyResult<()> {
+    m.add_class::<Evaluator>()?;
+    Ok(())
+}

packages/core/src/processing/evaluator/mod.rs

@@ -1,227 +1,23 @@
-pub mod bucket;
+/// Module defining routines to calculate preservation by count
+pub mod preservation_by_count;
+
+/// Module defining routines to calculate preservation by length
+pub mod preservation_by_length;
+
+/// Module defining a bucket used to calculate preservation by count
+/// and length
+pub mod preservation_bucket;
+
+/// Defines structures related to rare combination comparisons
+/// between the synthetic and sensitive datasets
+pub mod rare_combinations_comparison_data;
+
+/// Type definitions related to the evaluation process
 pub mod typedefs;
 
-use self::bucket::{PreservationByCountBucket, PreservationByCountBucketBins};
-use self::typedefs::PreservationByCountBuckets;
-use fnv::FnvHashSet;
-use instant::Duration;
-use log::Level::Trace;
-use log::{info, log_enabled, trace};
+mod data_evaluator;
 
-use crate::measure_time;
-use crate::processing::aggregator::typedefs::ValueCombination;
-use crate::utils::time::ElapsedDuration;
+pub use data_evaluator::Evaluator;
 
-use super::aggregator::typedefs::{
-    AggregatedCountByLenMap, AggregatesCountMap, ValueCombinationSlice,
-};
-
-#[derive(Debug)]
-struct EvaluatorDurations {
-    calc_leakage_count: Duration,
-    calc_fabricated_count: Duration,
-    calc_preservation_by_count: Duration,
-    calc_combination_loss: Duration,
-}
-
-/// Evaluates aggregated, sensitive and synthesized data
-pub struct Evaluator {
-    durations: EvaluatorDurations,
-}
-
-impl Evaluator {
-    /// Returns a new Evaluator
-    #[inline]
-    pub fn default() -> Evaluator {
-        Evaluator {
-            durations: EvaluatorDurations {
-                calc_leakage_count: Duration::default(),
-                calc_fabricated_count: Duration::default(),
-                calc_preservation_by_count: Duration::default(),
-                calc_combination_loss: Duration::default(),
-            },
-        }
-    }
-
-    /// Calculates the leakage counts grouped by combination length
-    /// (how many attribute combinations exist on the sensitive data and
-    /// appear on the synthetic data with `count < resolution`).
-    /// By design this should be `0`
-    /// # Arguments
-    /// - `sensitive_aggregates` - Calculated aggregates counts for the sensitive data
-    /// - `synthetic_aggregates` - Calculated aggregates counts for the synthetic data
-    pub fn calc_leakage_count(
-        &mut self,
-        sensitive_aggregates: &AggregatesCountMap,
-        synthetic_aggregates: &AggregatesCountMap,
-        resolution: usize,
-    ) -> AggregatedCountByLenMap {
-        info!("calculating rare sensitive combination leakages by length");
-        measure_time!(
-            || {
-                let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
-
-                for (sensitive_agg, sensitive_count) in sensitive_aggregates.iter() {
-                    if sensitive_count.count < resolution {
-                        if let Some(synthetic_count) = synthetic_aggregates.get(sensitive_agg) {
-                            if synthetic_count.count < resolution {
-                                let leaks = result.entry(sensitive_agg.len()).or_insert(0);
-                                *leaks += 1;
-                            }
-                        }
-                    }
-                }
-                result
-            },
-            (self.durations.calc_leakage_count)
-        )
-    }
-
-    /// Calculates the fabricated counts grouped by combination length
-    /// (how many attribute combinations exist on the synthetic data that do not
-    /// exist on the sensitive data).
-    /// By design this should be `0`
-    /// # Arguments
-    /// - `sensitive_aggregates` - Calculated aggregates counts for the sensitive data
-    /// - `synthetic_aggregates` - Calculated aggregates counts for the synthetic data
-    pub fn calc_fabricated_count(
-        &mut self,
-        sensitive_aggregates: &AggregatesCountMap,
-        synthetic_aggregates: &AggregatesCountMap,
-    ) -> AggregatedCountByLenMap {
-        info!("calculating fabricated synthetic combinations by length");
-        measure_time!(
-            || {
-                let mut result: AggregatedCountByLenMap = AggregatedCountByLenMap::default();
-
-                for synthetic_agg in synthetic_aggregates.keys() {
-                    if sensitive_aggregates.get(synthetic_agg).is_none() {
-                        let fabricated = result.entry(synthetic_agg.len()).or_insert(0);
-                        *fabricated += 1;
-                    }
-                }
-                result
-            },
-            (self.durations.calc_fabricated_count)
-        )
-    }
-
-    /// Calculates the preservation information by bucket.
-    /// An example output might be a map like:
-    /// `{ 10 -> PreservationByCountBucket, 20 -> PreservationByCountBucket, ...}`
-    /// # Arguments
-    /// - `sensitive_aggregates` - Calculated aggregates counts for the sensitive data
-    /// - `synthetic_aggregates` - Calculated aggregates counts for the synthetic data
-    /// * `resolution` - Reporting resolution used for data synthesis
-    pub fn calc_preservation_by_count(
-        &mut self,
-        sensitive_aggregates: &AggregatesCountMap,
-        synthetic_aggregates: &AggregatesCountMap,
-        resolution: usize,
-    ) -> PreservationByCountBuckets {
-        info!(
-            "calculating preservation by count with resolution: {}",
-            resolution
-        );
-        measure_time!(
-            || {
-                let max_syn_count = *synthetic_aggregates
-                    .values()
-                    .map(|a| &a.count)
-                    .max()
-                    .unwrap_or(&0);
-                let bins: PreservationByCountBucketBins =
-                    PreservationByCountBucketBins::new(max_syn_count);
-                let mut buckets: PreservationByCountBuckets = PreservationByCountBuckets::default();
-                let mut processed_combs: FnvHashSet<&ValueCombination> = FnvHashSet::default();
-
-                for (comb, count) in sensitive_aggregates.iter() {
-                    if count.count >= resolution && !processed_combs.contains(comb) {
-                        Evaluator::populate_buckets(
-                            &mut buckets,
-                            &bins,
-                            comb,
-                            sensitive_aggregates,
-                            synthetic_aggregates,
-                        );
-                        processed_combs.insert(comb);
-                    }
-                }
-                for comb in synthetic_aggregates.keys() {
-                    if !processed_combs.contains(comb) {
-                        Evaluator::populate_buckets(
-                            &mut buckets,
-                            &bins,
-                            comb,
-                            sensitive_aggregates,
-                            synthetic_aggregates,
-                        );
-                        processed_combs.insert(comb);
-                    }
-                }
-
-                // fill empty buckets with default value
-                for bin in bins.iter() {
-                    if !buckets.contains_key(bin) {
-                        buckets.insert(*bin, PreservationByCountBucket::default());
-                    }
-                }
-
-                buckets
-            },
-            (self.durations.calc_preservation_by_count)
-        )
-    }
-
-    /// Calculates the combination loss for the calculated bucket preservation information
-    /// (`combination_loss = avg(1 - bucket_preservation_sum / bucket_size) for all buckets`)
-    /// # Arguments
-    /// - `buckets` - Calculated preservation buckets
-    pub fn calc_combination_loss(&mut self, buckets: &PreservationByCountBuckets) -> f64 {
-        measure_time!(
-            || {
-                buckets
-                    .values()
-                    .map(|b| 1.0 - (b.preservation_sum / (b.size as f64)))
-                    .sum::<f64>()
-                    / (buckets.len() as f64)
-            },
-            (self.durations.calc_combination_loss)
-        )
-    }
-
-    #[inline]
-    fn populate_buckets(
-        buckets: &mut PreservationByCountBuckets,
-        bins: &PreservationByCountBucketBins,
-        comb: &ValueCombinationSlice,
-        sensitive_aggregates: &AggregatesCountMap,
-        synthetic_aggregates: &AggregatesCountMap,
-    ) {
-        let sen_count = match sensitive_aggregates.get(comb) {
-            Some(count) => count.count,
-            _ => 0,
-        };
-        let syn_count = match synthetic_aggregates.get(comb) {
-            Some(count) => count.count,
-            _ => 0,
-        };
-        let preservation = if sen_count > 0 {
-            // max value is 100%, so use min
-            f64::min((syn_count as f64) / (sen_count as f64), 1.0)
-        } else {
-            0.0
-        };
-
-        buckets
-            .entry(bins.find_bucket_max_val(syn_count))
-            .or_insert_with(PreservationByCountBucket::default)
-            .add(preservation, comb.len());
-    }
-}
-
-impl Drop for Evaluator {
-    fn drop(&mut self) {
-        trace!("evaluator durations: {:#?}", self.durations);
-    }
-}
+#[cfg(feature = "pyo3")]
+pub use data_evaluator::register;

packages/core/src/processing/evaluator/preservation_bucket.rs

@@ -0,0 +1,58 @@
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+/// Bucket to store preservation information
+#[cfg_attr(feature = "pyo3", pyclass)]
+#[derive(Debug, Clone)]
+pub struct PreservationBucket {
+    /// How many elements are stored in the bucket
+    pub size: usize,
+    /// Preservation sum of all elements in the bucket
+    pub preservation_sum: f64,
+    /// Combination length sum of all elements in the bucket
+    pub length_sum: usize,
+    /// Combination count sum
+    pub combination_count_sum: usize,
+}
+
+impl PreservationBucket {
+    /// Return a new PreservationBucket with default values
+    pub fn default() -> PreservationBucket {
+        PreservationBucket {
+            size: 0,
+            preservation_sum: 0.0,
+            length_sum: 0,
+            combination_count_sum: 0,
+        }
+    }
+
+    /// Adds a new value to the bucket
+    /// # Arguments
+    /// * `preservation` - Preservation related to the value
+    /// * `length` - Combination length related to the value
+    /// * `combination_count` - Combination count related to the value
+    pub fn add(&mut self, preservation: f64, length: usize, combination_count: usize) {
+        self.size += 1;
+        self.preservation_sum += preservation;
+        self.length_sum += length;
+        self.combination_count_sum += combination_count;
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl PreservationBucket {
+    /// Gets the mean preservation for the values in this bucket
+    pub fn get_mean_preservation(&self) -> f64 {
+        self.preservation_sum / (self.size as f64)
+    }
+
+    /// Gets the mean combination length for the values in this bucket
+    pub fn get_mean_combination_length(&self) -> f64 {
+        (self.length_sum as f64) / (self.size as f64)
+    }
+
+    /// Gets the mean combination count for the values in this bucket
+    pub fn get_mean_combination_count(&self) -> f64 {
+        (self.combination_count_sum as f64) / (self.size as f64)
+    }
+}

packages/core/src/processing/evaluator/preservation_by_count.rs

@@ -0,0 +1,191 @@
+use super::{preservation_bucket::PreservationBucket, typedefs::PreservationBucketsMap};
+use itertools::Itertools;
+use std::{
+    io::{Error, Write},
+    ops::{Deref, DerefMut},
+};
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+use crate::{
+    processing::aggregator::{typedefs::AggregatesCountMap, value_combination::ValueCombination},
+    utils::time::ElapsedDurationLogger,
+};
+
+const INITIAL_BIN: usize = 10;
+const BIN_RATIO: usize = 2;
+
+#[cfg_attr(feature = "pyo3", pyclass)]
+/// Wrapping struct mapping the max value allowed in the bucket
+/// to its correspondent PreservationBucket.
+/// In this context a PreservationBucket stores the preservation information
+/// related to a particular bucket.
+/// A bucket stores counts in a certain range value.
+/// For example: (0, 10], (10, 20], (20, 40]...
+pub struct PreservationByCountBuckets {
+    buckets_map: PreservationBucketsMap,
+}
+
+impl PreservationByCountBuckets {
+    /// Returns a new default PreservationByCountBuckets
+    #[inline]
+    pub fn default() -> PreservationByCountBuckets {
+        PreservationByCountBuckets {
+            buckets_map: PreservationBucketsMap::default(),
+        }
+    }
+
+    #[inline]
+    pub(super) fn populate(
+        &mut self,
+        bins: &PreservationByCountBucketBins,
+        comb: &ValueCombination,
+        sensitive_aggregates: &AggregatesCountMap,
+        synthetic_aggregates: &AggregatesCountMap,
+    ) {
+        let sen_count = match sensitive_aggregates.get(comb) {
+            Some(count) => count.count,
+            _ => 0,
+        };
+        let syn_count = match synthetic_aggregates.get(comb) {
+            Some(count) => count.count,
+            _ => 0,
+        };
+        let preservation = if sen_count > 0 {
+            // max value is 100%, so use min
+            f64::min((syn_count as f64) / (sen_count as f64), 1.0)
+        } else {
+            0.0
+        };
+
+        self.buckets_map
+            .entry(bins.find_bucket_max_val(syn_count))
+            .or_insert_with(PreservationBucket::default)
+            .add(preservation, comb.len(), syn_count);
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl PreservationByCountBuckets {
+    /// Returns the actual buckets grouped by its max value.
+    /// This method will clone the data, so its recommended to have its result stored
+    /// in a local variable to avoid it being called multiple times
+    pub fn get_buckets(&self) -> PreservationBucketsMap {
+        self.buckets_map.clone()
+    }
+
+    /// Calculates the combination loss for the calculated bucket preservation information
+    /// (`combination_loss = avg(1 - bucket_preservation_sum / bucket_size) for all buckets`)
+    pub fn calc_combination_loss(&self) -> f64 {
+        let _duration_logger = ElapsedDurationLogger::new("combination loss calculation");
+
+        self.buckets_map
+            .values()
+            .map(|b| 1.0 - (b.preservation_sum / (b.size as f64)))
+            .sum::<f64>()
+            / (self.buckets_map.len() as f64)
+    }
+
+    /// Writes the preservation grouped by counts to the file system in a csv/tsv like format
+    /// # Arguments:
+    /// * `preservation_by_count_path` - File path to be written
+    /// * `preservation_by_count_delimiter` - Delimiter to use when writing to `preservation_by_count_path`
+    pub fn write_preservation_by_count(
+        &self,
+        preservation_by_count_path: &str,
+        preservation_by_count_delimiter: char,
+    ) -> Result<(), Error> {
+        let mut file = std::fs::File::create(preservation_by_count_path)?;
+
+        file.write_all(
+            format!(
+                "syn_count_bucket{}mean_combo_count{}mean_combo_length{}count_preservation\n",
+                preservation_by_count_delimiter,
+                preservation_by_count_delimiter,
+                preservation_by_count_delimiter,
+            )
+            .as_bytes(),
+        )?;
+        for max_val in self.buckets_map.keys().sorted().rev() {
+            let b = &self.buckets_map[max_val];
+
+            file.write_all(
+                format!(
+                    "{}{}{}{}{}{}{}\n",
+                    max_val,
+                    preservation_by_count_delimiter,
+                    b.get_mean_combination_count(),
+                    preservation_by_count_delimiter,
+                    b.get_mean_combination_length(),
+                    preservation_by_count_delimiter,
+                    b.get_mean_preservation(),
+                )
+                .as_bytes(),
+            )?
+        }
+        Ok(())
+    }
+}
+
+impl Deref for PreservationByCountBuckets {
+    type Target = PreservationBucketsMap;
+
+    fn deref(&self) -> &Self::Target {
+        &self.buckets_map
+    }
+}
+
+impl DerefMut for PreservationByCountBuckets {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.buckets_map
+    }
+}
+
+/// Stores the bins for each bucket. For example:
+/// [10, 20, 40, 80, 160...]
+#[derive(Debug)]
+pub struct PreservationByCountBucketBins {
+    /// Bins where the index of the vector is the bin index,
+    /// and the value is the max value count allowed on the bin
+    bins: Vec<usize>,
+}
+
+impl PreservationByCountBucketBins {
+    /// Generates a new PreservationByCountBucketBins with bins up to a `max_val`
+    /// # Arguments
+    /// * `max_val` - Max value allowed on the last bucket (inclusive)
+    pub fn new(max_val: usize) -> PreservationByCountBucketBins {
+        let mut bins: Vec<usize> = vec![INITIAL_BIN];
+        loop {
+            let last = *bins.last().unwrap();
+            if last >= max_val {
+                break;
+            }
+            bins.push(last * BIN_RATIO);
+        }
+        PreservationByCountBucketBins { bins }
+    }
+
+    /// Find the first `bucket_max_val` where `val >= bucket_max_val`
+    /// # Arguments
+    /// * `val` - Count to look in which bucket it will fit
+    pub fn find_bucket_max_val(&self, val: usize) -> usize {
+        // find first element x where x >= val
+        self.bins[self.bins.partition_point(|x| *x < val)]
+    }
+}
+
+impl Deref for PreservationByCountBucketBins {
+    type Target = Vec<usize>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.bins
+    }
+}
+
+impl DerefMut for PreservationByCountBucketBins {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.bins
+    }
+}

packages/core/src/processing/evaluator/preservation_by_length.rs

@@ -0,0 +1,118 @@
+use super::{preservation_bucket::PreservationBucket, typedefs::PreservationBucketsMap};
+use itertools::Itertools;
+use std::{
+    io::{Error, Write},
+    ops::{Deref, DerefMut},
+};
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+use crate::processing::aggregator::{
+    typedefs::AggregatesCountMap, value_combination::ValueCombination,
+};
+
+#[cfg_attr(feature = "pyo3", pyclass)]
+/// Wrapping to store the preservation buckets grouped
+/// by length
+pub struct PreservationByLengthBuckets {
+    buckets_map: PreservationBucketsMap,
+}
+
+impl PreservationByLengthBuckets {
+    /// Returns a new default PreservationByLengthBuckets
+    #[inline]
+    pub fn default() -> PreservationByLengthBuckets {
+        PreservationByLengthBuckets {
+            buckets_map: PreservationBucketsMap::default(),
+        }
+    }
+
+    #[inline]
+    pub(super) fn populate(
+        &mut self,
+        comb: &ValueCombination,
+        sensitive_aggregates: &AggregatesCountMap,
+        synthetic_aggregates: &AggregatesCountMap,
+    ) {
+        let sen_count = match sensitive_aggregates.get(comb) {
+            Some(count) => count.count,
+            _ => 0,
+        };
+        let syn_count = match synthetic_aggregates.get(comb) {
+            Some(count) => count.count,
+            _ => 0,
+        };
+        let preservation = if sen_count > 0 {
+            // max value is 100%, so use min as syn count might be > sen_count
+            f64::min((syn_count as f64) / (sen_count as f64), 1.0)
+        } else {
+            0.0
+        };
+
+        self.buckets_map
+            .entry(comb.len())
+            .or_insert_with(PreservationBucket::default)
+            .add(preservation, comb.len(), syn_count);
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl PreservationByLengthBuckets {
+    /// Returns the actual buckets grouped by length.
+    /// This method will clone the data, so its recommended to have its result stored
+    /// in a local variable to avoid it being called multiple times
+    pub fn get_buckets(&self) -> PreservationBucketsMap {
+        self.buckets_map.clone()
+    }
+
+    /// Writes the preservation grouped by length to the file system in a csv/tsv like format
+    /// # Arguments:
+    /// * `preservation_by_length_path` - File path to be written
+    /// * `preservation_by_length_delimiter` - Delimiter to use when writing to `preservation_by_length_path`
+    pub fn write_preservation_by_length(
+        &mut self,
+        preservation_by_length_path: &str,
+        preservation_by_length_delimiter: char,
+    ) -> Result<(), Error> {
+        let mut file = std::fs::File::create(preservation_by_length_path)?;
+
+        file.write_all(
+            format!(
+                "syn_combo_length{}mean_combo_count{}count_preservation\n",
+                preservation_by_length_delimiter, preservation_by_length_delimiter,
+            )
+            .as_bytes(),
+        )?;
+        for length in self.buckets_map.keys().sorted() {
+            let b = &self.buckets_map[length];
+
+            file.write_all(
+                format!(
+                    "{}{}{}{}{}\n",
+                    length,
+                    preservation_by_length_delimiter,
+                    b.get_mean_combination_count(),
+                    preservation_by_length_delimiter,
+                    b.get_mean_preservation(),
+                )
+                .as_bytes(),
+            )?
+        }
+        Ok(())
+    }
+}
+
+impl Deref for PreservationByLengthBuckets {
+    type Target = PreservationBucketsMap;
+
+    fn deref(&self) -> &Self::Target {
+        &self.buckets_map
+    }
+}
+
+impl DerefMut for PreservationByLengthBuckets {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.buckets_map
+    }
+}

packages/core/src/processing/evaluator/rare_combinations_comparison_data.rs

@@ -0,0 +1,197 @@
+use std::io::{Error, Write};
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+use crate::{
+    data_block::typedefs::CombinationsComparisons,
+    processing::aggregator::aggregated_data::AggregatedData, utils::math::uround_down,
+};
+
+#[cfg_attr(feature = "pyo3", pyclass)]
+#[derive(Clone)]
+/// Represents a single combination comparison
+pub struct CombinationComparison {
+    /// Length of the combination
+    pub combination_length: usize,
+    /// Combination formatted as string
+    pub combination: String,
+    /// Index of the record where this combination occur
+    pub record_index: usize,
+    /// Number of times this combination occurs on the synthetic data
+    pub synthetic_count: usize,
+    /// Number of times this combination occurs on the sensitive data
+    pub sensitive_count: usize,
+}
+
+impl CombinationComparison {
+    #[inline]
+    /// Creates a new CombinationComparison
+    pub fn new(
+        combination_length: usize,
+        combination: String,
+        record_index: usize,
+        synthetic_count: usize,
+        sensitive_count: usize,
+    ) -> CombinationComparison {
+        CombinationComparison {
+            combination_length,
+            combination,
+            record_index,
+            synthetic_count,
+            sensitive_count,
+        }
+    }
+}
+
+#[cfg(feature = "pyo3")]
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl CombinationComparison {
+    #[getter]
+    /// Length of the combination
+    fn combination_length(&self) -> usize {
+        self.combination_length
+    }
+
+    #[getter]
+    /// Combination formatted as string
+    fn combination(&self) -> String {
+        self.combination.clone()
+    }
+
+    #[getter]
+    /// Index of the record where this combination occur
+    fn record_index(&self) -> usize {
+        self.record_index
+    }
+
+    #[getter]
+    /// Number of times this combination occurs on the synthetic data
+    fn synthetic_count(&self) -> usize {
+        self.synthetic_count
+    }
+
+    #[getter]
+    /// Number of times this combination occurs on the sensitive data
+    fn sensitive_count(&self) -> usize {
+        self.sensitive_count
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pyclass)]
+/// Computed rare combination comparisons between the synthetic
+/// and sensitive datasets for all records
+pub struct RareCombinationsComparisonData {
+    /// Rare combination comparison for all records
+    /// (compares rare combinations on the synthetic dataset with
+    /// the sensitive counts)
+    pub rare_combinations: CombinationsComparisons,
+}
+
+impl RareCombinationsComparisonData {
+    //// Build a new comparison between the rare combinations on the synthetic
+    /// data and the sensitive data counts
+    /// # Arguments
+    /// * `sensitive_aggregated_data` - Calculated aggregated data for the sensitive data
+    /// * `synthetic_aggregated_data` - Calculated aggregated data for the synthetic data
+    /// * `resolution` - Reporting resolution used for data synthesis
+    /// * `combination_delimiter` - Delimiter used to join combinations and format then
+    /// as strings
+    /// * `protect` - Whether or not the sensitive counts should be rounded to the
+    /// nearest smallest multiple of resolution
+    #[inline]
+    pub fn from_synthetic_and_sensitive_aggregated_data(
+        synthetic_aggregated_data: &AggregatedData,
+        sensitive_aggregated_data: &AggregatedData,
+        resolution: usize,
+        combination_delimiter: &str,
+        protect: bool,
+    ) -> RareCombinationsComparisonData {
+        let mut rare_combinations: CombinationsComparisons = CombinationsComparisons::default();
+        let resolution_f64 = resolution as f64;
+
+        for (agg, count) in synthetic_aggregated_data.aggregates_count.iter() {
+            if count.count < resolution {
+                let combination_str = agg.format_str_using_headers(
+                    &synthetic_aggregated_data.data_block.headers,
+                    combination_delimiter,
+                );
+                let mut sensitive_count = sensitive_aggregated_data
+                    .aggregates_count
+                    .get(agg)
+                    .map_or(0, |c| c.count);
+
+                if protect {
+                    sensitive_count = uround_down(sensitive_count as f64, resolution_f64)
+                }
+
+                for record_index in count.contained_in_records.iter() {
+                    rare_combinations.push(CombinationComparison::new(
+                        agg.len(),
+                        combination_str.clone(),
+                        *record_index,
+                        count.count,
+                        sensitive_count,
+                    ));
+                }
+            }
+        }
+
+        // sort result by combination length
+        rare_combinations.sort_by_key(|c| c.combination_length);
+
+        RareCombinationsComparisonData { rare_combinations }
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl RareCombinationsComparisonData {
+    /// Returns the rare combination comparisons between the synthetic
+    /// and sensitive datasets for all records.
+    /// This method will clone the data, so its recommended to have its result stored
+    /// in a local variable to avoid it being called multiple times
+    pub fn get_rare_combinations(&self) -> CombinationsComparisons {
+        self.rare_combinations.clone()
+    }
+
+    /// Writes the rare combinations to the file system in a csv/tsv like format
+    /// # Arguments:
+    /// * `rare_combinations_path` - File path to be written
+    /// * `rare_combinations_delimiter` - Delimiter to use when writing to `rare_combinations_path`
+    pub fn write_rare_combinations(
+        &mut self,
+        rare_combinations_path: &str,
+        rare_combinations_delimiter: char,
+    ) -> Result<(), Error> {
+        let mut file = std::fs::File::create(rare_combinations_path)?;
+
+        file.write_all(
+            format!(
+                "combo_length{}combo{}record_id{}syn_count{}sen_count\n",
+                rare_combinations_delimiter,
+                rare_combinations_delimiter,
+                rare_combinations_delimiter,
+                rare_combinations_delimiter,
+            )
+            .as_bytes(),
+        )?;
+        for ra in self.rare_combinations.iter() {
+            file.write_all(
+                format!(
+                    "{}{}{}{}{}{}{}{}{}\n",
+                    ra.combination_length,
+                    rare_combinations_delimiter,
+                    ra.combination,
+                    rare_combinations_delimiter,
+                    ra.record_index,
+                    rare_combinations_delimiter,
+                    ra.synthetic_count,
+                    rare_combinations_delimiter,
+                    ra.sensitive_count,
+                )
+                .as_bytes(),
+            )?
+        }
+        Ok(())
+    }
+}

packages/core/src/processing/evaluator/typedefs.rs

@@ -1,6 +1,5 @@
-use super::bucket::PreservationByCountBucket;
+use super::preservation_bucket::PreservationBucket;
 use fnv::FnvHashMap;
 
-/// Maps the max value allowed in the bucket to its
-/// correspondent PreservationByCountBucket
-pub type PreservationByCountBuckets = FnvHashMap<usize, PreservationByCountBucket>;
+/// Maps a value to its correspondent PreservationBucket
+pub type PreservationBucketsMap = FnvHashMap<usize, PreservationBucket>;

packages/core/src/processing/generator/data_generator.rs

@@ -0,0 +1,143 @@
+use super::generated_data::GeneratedData;
+use super::synthesizer::typedefs::SynthesizedRecords;
+use super::synthesizer::unseeded::UnseededSynthesizer;
+use super::SynthesisMode;
+use log::info;
+use std::sync::Arc;
+
+use crate::data_block::block::DataBlock;
+use crate::data_block::typedefs::RawSyntheticData;
+use crate::processing::generator::synthesizer::cache::SynthesizerCacheKey;
+use crate::processing::generator::synthesizer::seeded::SeededSynthesizer;
+use crate::utils::reporting::ReportProgress;
+use crate::utils::time::ElapsedDurationLogger;
+
+/// Process a data block and generates new synthetic data
+pub struct Generator {
+    data_block: Arc<DataBlock>,
+}
+
+impl Generator {
+    /// Returns a new Generator
+    /// # Arguments
+    /// * `data_block` - Sensitive data to be synthesized
+    #[inline]
+    pub fn new(data_block: Arc<DataBlock>) -> Generator {
+        Generator { data_block }
+    }
+
+    /// Generates new synthetic data based on sensitive data
+    /// # Arguments
+    /// * `resolution` - Reporting resolution used for data synthesis
+    /// * `cache_max_size` - Maximum cache size allowed
+    /// * `empty_value` - Empty values on the synthetic data will be represented by this
+    /// * `mode` - Which mode to perform the data synthesis
+    /// * `progress_reporter` - Will be used to report the processing
+    /// progress (`ReportProgress` trait). If `None`, nothing will be reported
+    pub fn generate<T>(
+        &mut self,
+        resolution: usize,
+        cache_max_size: usize,
+        empty_value: String,
+        mode: SynthesisMode,
+        progress_reporter: &mut Option<T>,
+    ) -> GeneratedData
+    where
+        T: ReportProgress,
+    {
+        let _duration_logger = ElapsedDurationLogger::new("data generation");
+        let empty_value_arc = Arc::new(empty_value);
+
+        info!("starting {} generation...", mode);
+
+        let synthesized_records = match mode {
+            SynthesisMode::Seeded => {
+                self.seeded_synthesis(resolution, cache_max_size, progress_reporter)
+            }
+            SynthesisMode::Unseeded => self.unseeded_synthesis(
+                resolution,
+                cache_max_size,
+                &empty_value_arc,
+                progress_reporter,
+            ),
+        };
+
+        self.build_generated_data(synthesized_records, empty_value_arc)
+    }
+
+    #[inline]
+    fn build_generated_data(
+        &self,
+        mut synthesized_records: SynthesizedRecords,
+        empty_value: Arc<String>,
+    ) -> GeneratedData {
+        let mut result: RawSyntheticData = RawSyntheticData::default();
+        let mut records: RawSyntheticData = synthesized_records
+            .drain(..)
+            .map(|r| {
+                SynthesizerCacheKey::new(self.data_block.headers.len(), &r)
+                    .format_record(&empty_value)
+            })
+            .collect();
+
+        // sort by number of defined attributes
+        records.sort();
+        records.sort_by_key(|r| {
+            -r.iter()
+                .map(|s| if s.is_empty() { 0 } else { 1 })
+                .sum::<isize>()
+        });
+
+        result.push(self.data_block.headers.to_vec());
+        result.extend(records);
+
+        let expansion_ratio = (result.len() - 1) as f64 / self.data_block.records.len() as f64;
+
+        info!("expansion ratio: {:.4?}", expansion_ratio);
+
+        GeneratedData::new(result, expansion_ratio)
+    }
+
+    #[inline]
+    fn seeded_synthesis<T>(
+        &self,
+        resolution: usize,
+        cache_max_size: usize,
+        progress_reporter: &mut Option<T>,
+    ) -> SynthesizedRecords
+    where
+        T: ReportProgress,
+    {
+        let attr_rows_map = Arc::new(self.data_block.calc_attr_rows());
+        let mut synth = SeededSynthesizer::new(
+            self.data_block.clone(),
+            attr_rows_map,
+            resolution,
+            cache_max_size,
+        );
+        synth.run(progress_reporter)
+    }
+
+    #[inline]
+    fn unseeded_synthesis<T>(
+        &self,
+        resolution: usize,
+        cache_max_size: usize,
+        empty_value: &Arc<String>,
+        progress_reporter: &mut Option<T>,
+    ) -> SynthesizedRecords
+    where
+        T: ReportProgress,
+    {
+        let attr_rows_map_by_column =
+            Arc::new(self.data_block.calc_attr_rows_by_column(empty_value));
+        let mut synth = UnseededSynthesizer::new(
+            self.data_block.clone(),
+            attr_rows_map_by_column,
+            resolution,
+            cache_max_size,
+            empty_value.clone(),
+        );
+        synth.run(progress_reporter)
+    }
+}

packages/core/src/processing/generator/generated_data.rs

@@ -0,0 +1,93 @@
+use csv::WriterBuilder;
+use log::info;
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+#[cfg(feature = "pyo3")]
+use crate::data_block::typedefs::CsvRecord;
+
+use crate::{
+    data_block::{csv_io_error::CsvIOError, typedefs::RawSyntheticData},
+    utils::time::ElapsedDurationLogger,
+};
+
+#[cfg_attr(feature = "pyo3", pyclass)]
+/// Synthetic data generated by the Generator
+pub struct GeneratedData {
+    /// Synthesized data - headers (index 0) and records indexes 1...
+    pub synthetic_data: RawSyntheticData,
+    /// `Synthetic data length / Sensitive data length` (header not included)
+    pub expansion_ratio: f64,
+}
+
+impl GeneratedData {
+    /// Returns a new GeneratedData struct with default values
+    #[inline]
+    pub fn default() -> GeneratedData {
+        GeneratedData {
+            synthetic_data: RawSyntheticData::default(),
+            expansion_ratio: 0.0,
+        }
+    }
+
+    /// Returns a new GeneratedData struct
+    /// # Arguments
+    /// * `synthetic_data` - Synthesized data - headers (index 0) and records indexes 1...
+    /// * `expansion_ratio` - `Synthetic data length / Sensitive data length` (header not included)
+    #[inline]
+    pub fn new(synthetic_data: RawSyntheticData, expansion_ratio: f64) -> GeneratedData {
+        GeneratedData {
+            synthetic_data,
+            expansion_ratio,
+        }
+    }
+}
+
+#[cfg_attr(feature = "pyo3", pymethods)]
+impl GeneratedData {
+    #[cfg(feature = "pyo3")]
+    /// Synthesized data - headers (index 0) and records indexes 1...
+    /// This method will clone the data, so its recommended to have its result stored
+    /// in a local variable to avoid it being called multiple times
+    fn get_synthetic_data(&self) -> Vec<CsvRecord> {
+        self.synthetic_data
+            .iter()
+            .map(|row| row.iter().map(|value| (**value).clone()).collect())
+            .collect()
+    }
+
+    #[cfg(feature = "pyo3")]
+    #[getter]
+    /// `Synthetic data length / Sensitive data length` (header not included)
+    fn expansion_ratio(&self) -> f64 {
+        self.expansion_ratio
+    }
+
+    /// Writes the synthesized data to the file system
+    /// # Arguments
+    /// * `path` - File path to be written
+    /// * `delimiter` - Delimiter to use when writing to `path`
+    pub fn write_synthetic_data(&self, path: &str, delimiter: char) -> Result<(), CsvIOError> {
+        let _duration_logger = ElapsedDurationLogger::new("write synthetic data");
+
+        info!("writing file {}", path);
+
+        let mut wtr = match WriterBuilder::new()
+            .delimiter(delimiter as u8)
+            .from_path(&path)
+        {
+            Ok(writer) => writer,
+            Err(err) => return Err(CsvIOError::new(err)),
+        };
+
+        // write header and records
+        for r in self.synthetic_data.iter() {
+            match wtr.write_record(r.iter().map(|v| v.as_str())) {
+                Ok(_) => {}
+                Err(err) => return Err(CsvIOError::new(err)),
+            };
+        }
+        Ok(())
+    }
+}

packages/core/src/processing/generator/mod.rs

@@ -1,156 +1,13 @@
 /// Module defining the data synthesis process
 pub mod synthesizer;
 
-use csv::{Error, WriterBuilder};
-use log::Level::Trace;
-use log::{info, log_enabled, trace};
-use std::time::Duration;
+mod synthesis_mode;
 
-use crate::data_block::block::DataBlock;
-use crate::data_block::typedefs::CsvRecordRef;
-use crate::measure_time;
-use crate::processing::generator::synthesizer::cache::SynthesizerCacheKey;
-use crate::processing::generator::synthesizer::seeded::SeededSynthesizer;
-use crate::utils::reporting::ReportProgress;
-use crate::utils::time::ElapsedDuration;
+pub use synthesis_mode::SynthesisMode;
 
-use self::synthesizer::typedefs::SynthesizedRecords;
+mod data_generator;
 
-#[derive(Debug)]
-struct GeneratorDurations {
-    generate: Duration,
-    write_records: Duration,
-}
+pub use data_generator::Generator;
 
-/// Process a data block and generates new synthetic data
-pub struct Generator<'data_block> {
-    data_block: &'data_block DataBlock,
-    durations: GeneratorDurations,
-}
-
-/// Synthetic data generated by the Generator
-pub struct GeneratedData<'data_block> {
-    /// Synthesized data - headers (index 0) and records indexes 1...
-    pub synthetic_data: Vec<CsvRecordRef<'data_block>>,
-    /// `Synthetic data length / Sensitive data length` (header not included)
-    pub expansion_ratio: f64,
-}
-
-impl<'data_block> Generator<'data_block> {
-    /// Returns a new Generator
-    /// # Arguments
-    /// * `data_block` - Sensitive data to be synthesized
-    #[inline]
-    pub fn new(data_block: &'data_block DataBlock) -> Generator<'data_block> {
-        Generator {
-            data_block,
-            durations: GeneratorDurations {
-                generate: Duration::default(),
-                write_records: Duration::default(),
-            },
-        }
-    }
-
-    /// Generates new synthetic data based on sensitive data
-    /// # Arguments
-    /// * `resolution` - Reporting resolution used for data synthesis
-    /// * `cache_max_size` - Maximum cache size allowed
-    /// * `empty_value` - Empty values on the synthetic data will be represented by this
-    /// * `progress_reporter` - Will be used to report the processing
-    /// progress (`ReportProgress` trait). If `None`, nothing will be reported
-    pub fn generate<T>(
-        &mut self,
-        resolution: usize,
-        cache_max_size: usize,
-        empty_value: &'data_block str,
-        progress_reporter: &mut Option<T>,
-    ) -> GeneratedData<'data_block>
-    where
-        T: ReportProgress,
-    {
-        info!("starting generation...");
-
-        measure_time!(
-            || {
-                let mut result: Vec<CsvRecordRef<'data_block>> = Vec::default();
-                let mut records: Vec<CsvRecordRef<'data_block>> = self
-                    .seeded_synthesis(resolution, cache_max_size, progress_reporter)
-                    .iter()
-                    .map(|r| {
-                        SynthesizerCacheKey::new(self.data_block.headers.len(), r)
-                            .format_record(empty_value)
-                    })
-                    .collect();
-
-                // sort by number of defined attributes
-                records.sort();
-                records.sort_by_key(|r| {
-                    -r.iter()
-                        .map(|s| if s.is_empty() { 0 } else { 1 })
-                        .sum::<isize>()
-                });
-
-                let expansion_ratio = records.len() as f64 / self.data_block.records.len() as f64;
-
-                info!("expansion ratio: {:.4?}", expansion_ratio);
-
-                result.push(self.data_block.headers.iter().map(|h| h.as_str()).collect());
-                result.extend(records);
-
-                GeneratedData {
-                    synthetic_data: result,
-                    expansion_ratio,
-                }
-            },
-            (self.durations.generate)
-        )
-    }
-
-    /// Writes the synthesized data to the file system
-    /// # Arguments
-    /// * `result` - Synthetic data to write (headers included)
-    /// * `path` - File path to be written
-    /// * `delimiter` - Delimiter to use when writing to `path`
-    pub fn write_records(
-        &mut self,
-        result: &[CsvRecordRef<'data_block>],
-        path: &str,
-        delimiter: u8,
-    ) -> Result<(), Error> {
-        info!("writing file {}", path);
-
-        measure_time!(
-            || {
-                let mut wtr = WriterBuilder::new().delimiter(delimiter).from_path(&path)?;
-
-                // write header and records
-                for r in result.iter() {
-                    wtr.write_record(r)?;
-                }
-                Ok(())
-            },
-            (self.durations.write_records)
-        )
-    }
-
-    fn seeded_synthesis<T>(
-        &mut self,
-        resolution: usize,
-        cache_max_size: usize,
-        progress_reporter: &mut Option<T>,
-    ) -> SynthesizedRecords<'data_block>
-    where
-        T: ReportProgress,
-    {
-        let attr_rows_map = DataBlock::calc_attr_rows(&self.data_block.records);
-        let mut synth =
-            SeededSynthesizer::new(self.data_block, &attr_rows_map, resolution, cache_max_size);
-        synth.run(progress_reporter)
-    }
-}
-
-impl<'data_block> Drop for Generator<'data_block> {
-    fn drop(&mut self) {
-        trace!("generator durations: {:#?}", self.durations);
-    }
-}
+/// Module to describe the synthetic data generated by the Generator
+pub mod generated_data;

packages/core/src/processing/generator/synthesis_mode.rs

@@ -0,0 +1,29 @@
+use std::{fmt::Display, str::FromStr};
+
+#[derive(Debug)]
+/// Modes to execute the data generation/synthesis
+pub enum SynthesisMode {
+    Seeded,
+    Unseeded,
+}
+
+impl FromStr for SynthesisMode {
+    type Err = &'static str;
+
+    fn from_str(mode: &str) -> Result<Self, Self::Err> {
+        match mode.to_lowercase().as_str() {
+            "seeded" => Ok(SynthesisMode::Seeded),
+            "unseeded" => Ok(SynthesisMode::Unseeded),
+            _ => Err("invalid mode, should be seeded or unseeded"),
+        }
+    }
+}
+
+impl Display for SynthesisMode {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            SynthesisMode::Seeded => Ok(write!(f, "seeded")?),
+            SynthesisMode::Unseeded => Ok(write!(f, "unseeded")?),
+        }
+    }
+}

packages/core/src/processing/generator/synthesizer/cache.rs

@@ -1,6 +1,7 @@
 use super::typedefs::SynthesizedRecord;
 use fnv::FnvBuildHasher;
 use lru::LruCache;
+use std::sync::Arc;
 
 use crate::data_block::{typedefs::CsvRecordRef, value::DataBlockValue};
 
@@ -8,29 +9,26 @@ use crate::data_block::{typedefs::CsvRecordRef, value::DataBlockValue};
 /// (`columns[{column_index}] = `value` if value exists,
 /// or `columns[{column_index}] = None` otherwise)
 #[derive(Debug, PartialEq, Eq, Hash, Clone)]
-pub struct SynthesizerCacheKey<'value> {
+pub struct SynthesizerCacheKey {
     /// Values for a given synthesized record are indexed by `column_index`
     /// on this vector
-    columns: Vec<Option<&'value String>>,
+    columns: Vec<Option<Arc<String>>>,
 }
 
-impl<'value> SynthesizerCacheKey<'value> {
+impl SynthesizerCacheKey {
     /// Returns a new SynthesizerCacheKey
     /// # Arguments
     /// * `num_columns` - Number of columns in the data block
     /// * `records` - Synthesized record to build the key for
     #[inline]
-    pub fn new(
-        num_columns: usize,
-        values: &SynthesizedRecord<'value>,
-    ) -> SynthesizerCacheKey<'value> {
+    pub fn new(num_columns: usize, values: &SynthesizedRecord) -> SynthesizerCacheKey {
         let mut key = SynthesizerCacheKey {
             columns: Vec::with_capacity(num_columns),
         };
         key.columns.resize_with(num_columns, || None);
 
         for v in values.iter() {
-            key.columns[v.column_index] = Some(&v.value);
+            key.columns[v.column_index] = Some(v.value.clone());
         }
         key
     }
@@ -40,9 +38,9 @@ impl<'value> SynthesizerCacheKey<'value> {
     /// # Arguments
     /// * `value` - New data block value to be added in the new key
     #[inline]
-    pub fn new_with_value(&self, value: &'value DataBlockValue) -> SynthesizerCacheKey<'value> {
+    pub fn new_with_value(&self, value: &Arc<DataBlockValue>) -> SynthesizerCacheKey {
         let mut new_key = self.clone();
-        new_key.columns[value.column_index] = Some(&value.value);
+        new_key.columns[value.column_index] = Some(value.value.clone());
         new_key
     }
 
@@ -59,29 +57,29 @@ impl<'value> SynthesizerCacheKey<'value> {
     /// # Arguments
     /// * `empty_value` - String to be used in case the `columns[column_index] == None`
     #[inline]
-    pub fn format_record(&self, empty_value: &'value str) -> CsvRecordRef<'value> {
+    pub fn format_record(&self, empty_value: &Arc<String>) -> CsvRecordRef {
         self.columns
             .iter()
             .map(|c_opt| match c_opt {
-                Some(c) => c,
-                None => empty_value,
+                Some(c) => (*c).clone(),
+                None => empty_value.clone(),
             })
             .collect()
     }
 }
 
 /// Cache to store keys-values used during the synthesis process
-pub struct SynthesizerCache<'value, T> {
+pub struct SynthesizerCache<T> {
     /// LruCache to store the keys mapping to a generic type T
-    cache: LruCache<SynthesizerCacheKey<'value>, T, FnvBuildHasher>,
+    cache: LruCache<SynthesizerCacheKey, T, FnvBuildHasher>,
 }
 
-impl<'value, T> SynthesizerCache<'value, T> {
+impl<T> SynthesizerCache<T> {
     /// Returns a new SynthesizerCache
     /// # Arguments
     /// * `cache_max_size` - Maximum cache size allowed for the LRU
     #[inline]
-    pub fn new(cache_max_size: usize) -> SynthesizerCache<'value, T> {
+    pub fn new(cache_max_size: usize) -> SynthesizerCache<T> {
         SynthesizerCache {
             cache: LruCache::with_hasher(cache_max_size, FnvBuildHasher::default()),
         }
@@ -93,7 +91,7 @@ impl<'value, T> SynthesizerCache<'value, T> {
     /// # Arguments
     /// * `key` - Key to look for the value
     #[inline]
-    pub fn get(&mut self, key: &SynthesizerCacheKey<'value>) -> Option<&T> {
+    pub fn get(&mut self, key: &SynthesizerCacheKey) -> Option<&T> {
         self.cache.get(key)
     }
 
@@ -104,7 +102,7 @@ impl<'value, T> SynthesizerCache<'value, T> {
     /// * `key` - Key to be inserted
     /// * `value` - Value to associated with the key
     #[inline]
-    pub fn insert(&mut self, key: SynthesizerCacheKey<'value>, value: T) -> Option<T> {
+    pub fn insert(&mut self, key: SynthesizerCacheKey, value: T) -> Option<T> {
         self.cache.put(key, value)
     }
 }

packages/core/src/processing/generator/synthesizer/context.rs

@@ -2,93 +2,76 @@ use super::cache::{SynthesizerCache, SynthesizerCacheKey};
 use super::typedefs::{
     AttributeCountMap, NotAllowedAttrSet, SynthesizedRecord, SynthesizerSeedSlice,
 };
+use itertools::Itertools;
 use rand::Rng;
+use std::sync::Arc;
 
-use crate::data_block::typedefs::{AttributeRows, AttributeRowsMap, AttributeRowsSlice};
+use crate::data_block::typedefs::{
+    AttributeRows, AttributeRowsByColumnMap, AttributeRowsMap, AttributeRowsRefMap,
+    AttributeRowsSlice,
+};
 use crate::data_block::value::DataBlockValue;
+use crate::processing::aggregator::typedefs::RecordsSet;
 use crate::utils::collections::ordered_vec_intersection;
 
 /// Represents a synthesis context, containing the information
 /// required to synthesize new records
 /// (common to seeded and unseeded synthesis)
-pub struct SynthesizerContext<'data_block, 'attr_rows_map> {
+pub struct SynthesizerContext {
     /// Number of headers in the data block
-    headers_len: usize,
+    pub headers_len: usize,
     /// Number of records in the data block
-    records_len: usize,
-    /// Maps a data block value to all the rows where it occurs
-    attr_rows_map: &'attr_rows_map AttributeRowsMap<'data_block>,
+    pub records_len: usize,
     /// Cache used to store the rows where value combinations occurs
-    cache: SynthesizerCache<'data_block, AttributeRows>,
+    cache: SynthesizerCache<Arc<AttributeRows>>,
     /// Reporting resolution used for data synthesis
     resolution: usize,
 }
 
-impl<'data_block, 'attr_rows_map> SynthesizerContext<'data_block, 'attr_rows_map> {
+impl SynthesizerContext {
     /// Returns a new SynthesizerContext
     /// # Arguments
     /// * `headers_len` - Number of headers in the data block
     /// * `records_len` - Number of records in the data block
-    /// * `attr_rows_map` - Maps a data block value to all the rows where it occurs
     /// * `resolution` - Reporting resolution used for data synthesis
     /// * `cache_max_size` - Maximum cache size allowed
     #[inline]
     pub fn new(
         headers_len: usize,
         records_len: usize,
-        attr_rows_map: &'attr_rows_map AttributeRowsMap<'data_block>,
         resolution: usize,
         cache_max_size: usize,
-    ) -> SynthesizerContext<'data_block, 'attr_rows_map> {
+    ) -> SynthesizerContext {
         SynthesizerContext {
             headers_len,
             records_len,
-            attr_rows_map,
             cache: SynthesizerCache::new(cache_max_size),
             resolution,
         }
     }
 
-    /// Returns the configured mapping of a data block value to all
-    /// the rows where it occurs
-    #[inline]
-    pub fn get_attr_rows_map(&self) -> &AttributeRowsMap<'data_block> {
-        self.attr_rows_map
-    }
-
-    /// Returns the configured reporting resolution
-    #[inline]
-    pub fn get_resolution(&self) -> usize {
-        self.resolution
-    }
-
     /// Samples an attribute based on its count
     /// (the higher the count the greater the chance for the
     /// attribute to be selected).
     /// Returns `None` if all the counts are 0 or the map is empty
     /// # Arguments
     /// * `counts` - Maps an attribute to its count for sampling
-    pub fn sample_from_attr_counts(
-        counts: &AttributeCountMap<'data_block>,
-    ) -> Option<&'data_block DataBlockValue> {
-        let mut res: Option<&'data_block DataBlockValue> = None;
+    pub fn sample_from_attr_counts(counts: &AttributeCountMap) -> Option<Arc<DataBlockValue>> {
+        let mut res: Option<Arc<DataBlockValue>> = None;
         let total: usize = counts.values().sum();
 
         if total != 0 {
-            let random = if total == 1 {
-                1
-            } else {
-                rand::thread_rng().gen_range(1..total)
-            };
-            let mut current_sim: usize = 0;
+            let random = rand::thread_rng().gen_range(1..=total);
+            let mut current_sum: usize = 0;
 
-            for (value, count) in counts.iter() {
-                current_sim += count;
-                if current_sim < random {
-                    res = Some(value);
-                    break;
+            for (value, count) in counts.iter().sorted_by_key(|(_, c)| **c) {
+                if *count > 0 {
+                    current_sum += count;
+                    res = Some(value.clone());
+                    if current_sum >= random {
+                        break;
+                    }
                 }
-                res = Some(value)
             }
         }
         res
@@ -100,52 +83,130 @@ impl<'data_block, 'attr_rows_map> SynthesizerContext<'data_block, 'attr_rows_map
     /// * `synthesized_record` - Record synthesized so far
     /// * `current_seed` - Current seed/record used for sampling
     /// * `not_allowed_attr_set` - Attributes not allowed to be sampled
-    pub fn sample_next_attr(
+    /// * `attr_rows_map` - Maps a data block value to all the rows where it occurs
+    pub fn sample_next_attr_from_seed(
         &mut self,
-        synthesized_record: &SynthesizedRecord<'data_block>,
-        current_seed: &SynthesizerSeedSlice<'data_block>,
-        not_allowed_attr_set: &NotAllowedAttrSet<'data_block>,
-    ) -> Option<&'data_block DataBlockValue> {
-        let counts =
-            self.calc_next_attr_count(synthesized_record, current_seed, not_allowed_attr_set);
+        synthesized_record: &SynthesizedRecord,
+        current_seed: &SynthesizerSeedSlice,
+        not_allowed_attr_set: &NotAllowedAttrSet,
+        attr_rows_map: &AttributeRowsMap,
+    ) -> Option<Arc<DataBlockValue>> {
+        let counts = self.calc_next_attr_count(
+            synthesized_record,
+            current_seed,
+            not_allowed_attr_set,
+            attr_rows_map,
+        );
         SynthesizerContext::sample_from_attr_counts(&counts)
     }
 
+    /// Samples the next attribute from the column at `column_index`.
+    /// Returns a tuple with
+    /// (intersection of `current_attrs_rows` and rows that sampled value appear, sampled value)
+    /// # Arguments
+    /// * `synthesized_record` - Record synthesized so far
+    /// * `column_index` - Index of the column to sample from
+    /// * `attr_rows_map_by_column` - Maps the column index -> data block value -> rows where the value appear
+    /// * `current_attrs_rows` - Rows where the so far sampled combination appear
+    /// * `empty_value` - Empty values on the synthetic data will be represented by this
+    pub fn sample_next_attr_from_column(
+        &mut self,
+        synthesized_record: &SynthesizedRecord,
+        column_index: usize,
+        attr_rows_map_by_column: &AttributeRowsByColumnMap,
+        current_attrs_rows: &AttributeRowsSlice,
+        empty_value: &Arc<String>,
+    ) -> Option<(Arc<AttributeRows>, Arc<DataBlockValue>)> {
+        let cache_key = SynthesizerCacheKey::new(self.headers_len, synthesized_record);
+        let empty_block_value = Arc::new(DataBlockValue::new(column_index, empty_value.clone()));
+        let mut values_to_sample: AttributeRowsRefMap = AttributeRowsRefMap::default();
+        let mut counts = AttributeCountMap::default();
+
+        // calculate the row intersections for each value in the column
+        for (value, rows) in attr_rows_map_by_column[&column_index].iter() {
+            let new_cache_key = cache_key.new_with_value(value);
+            let rows_intersection = match self.cache.get(&new_cache_key) {
+                Some(cached_value) => cached_value.clone(),
+                None => {
+                    let intersection = Arc::new(ordered_vec_intersection(current_attrs_rows, rows));
+                    self.cache.insert(new_cache_key, intersection.clone());
+                    intersection
+                }
+            };
+            values_to_sample.insert(value.clone(), rows_intersection);
+        }
+
+        // store the rows with empty values
+        let mut rows_with_empty_values: RecordsSet = values_to_sample[&empty_block_value]
+            .iter()
+            .cloned()
+            .collect();
+
+        for (value, rows) in values_to_sample.iter() {
+            if rows.len() < self.resolution {
+                // if the combination containing the attribute appears in less
+                // than resolution rows, we can't use it so we tag it as an empty value
+                rows_with_empty_values.extend(rows.iter());
+            } else if **value != *empty_block_value {
+                // if we can use the combination containing the attribute
+                // gather its count for sampling
+                counts.insert(value.clone(), rows.len());
+            }
+        }
+
+        // if there are empty values that can be sampled, add them for sampling
+        if !rows_with_empty_values.is_empty() {
+            counts.insert(empty_block_value, rows_with_empty_values.len());
+        }
+
+        Self::sample_from_attr_counts(&counts).map(|sampled_value| {
+            (
+                values_to_sample.remove(&sampled_value).unwrap(),
+                sampled_value,
+            )
+        })
+    }
+
     fn calc_current_attrs_rows(
         &mut self,
-        cache_key: &SynthesizerCacheKey<'data_block>,
-        synthesized_record: &SynthesizedRecord<'data_block>,
-    ) -> AttributeRows {
+        cache_key: &SynthesizerCacheKey,
+        synthesized_record: &SynthesizedRecord,
+        attr_rows_map: &AttributeRowsMap,
+    ) -> Arc<AttributeRows> {
         match self.cache.get(cache_key) {
             Some(cache_value) => cache_value.clone(),
             None => {
                 let mut current_attrs_rows: AttributeRows = AttributeRows::new();
+                let current_attrs_rows_arc: Arc<AttributeRows>;
 
                 if !synthesized_record.is_empty() {
                     for sr in synthesized_record.iter() {
                         current_attrs_rows = if current_attrs_rows.is_empty() {
-                            self.attr_rows_map[sr].clone()
+                            attr_rows_map[sr].clone()
                         } else {
-                            ordered_vec_intersection(&current_attrs_rows, &self.attr_rows_map[sr])
+                            ordered_vec_intersection(&current_attrs_rows, &attr_rows_map[sr])
                         }
                     }
                 } else {
                     current_attrs_rows = (0..self.records_len).collect();
                 }
+                current_attrs_rows_arc = Arc::new(current_attrs_rows);
+
                 self.cache
-                    .insert(cache_key.clone(), current_attrs_rows.clone());
-                current_attrs_rows
+                    .insert(cache_key.clone(), current_attrs_rows_arc.clone());
+                current_attrs_rows_arc
             }
         }
     }
 
     fn gen_attr_count_map(
         &mut self,
-        cache_key: &mut SynthesizerCacheKey<'data_block>,
-        current_seed: &SynthesizerSeedSlice<'data_block>,
+        cache_key: &SynthesizerCacheKey,
+        current_seed: &SynthesizerSeedSlice,
         current_attrs_rows: &AttributeRowsSlice,
-        not_allowed_attr_set: &NotAllowedAttrSet<'data_block>,
-    ) -> AttributeCountMap<'data_block> {
+        not_allowed_attr_set: &NotAllowedAttrSet,
+        attr_rows_map: &AttributeRowsMap,
+    ) -> AttributeCountMap {
         let mut attr_count_map: AttributeCountMap = AttributeCountMap::default();
 
         for value in current_seed.iter() {
@@ -156,10 +217,10 @@ impl<'data_block, 'attr_rows_map> SynthesizerContext<'data_block, 'attr_rows_map
                 let count = match self.cache.get(&new_cache_key) {
                     Some(cached_value) => cached_value.len(),
                     None => {
-                        let intersection = ordered_vec_intersection(
+                        let intersection = Arc::new(ordered_vec_intersection(
                             current_attrs_rows,
-                            self.attr_rows_map.get(value).unwrap(),
-                        );
+                            attr_rows_map.get(value).unwrap(),
+                        ));
                         let count = intersection.len();
 
                         self.cache.insert(new_cache_key, intersection);
@@ -168,7 +229,7 @@ impl<'data_block, 'attr_rows_map> SynthesizerContext<'data_block, 'attr_rows_map
                 };
 
                 if count >= self.resolution {
-                    attr_count_map.insert(value, count);
+                    attr_count_map.insert(value.clone(), count);
                 }
             }
         }
@@ -177,18 +238,21 @@ impl<'data_block, 'attr_rows_map> SynthesizerContext<'data_block, 'attr_rows_map
 
     fn calc_next_attr_count(
         &mut self,
-        synthesized_record: &SynthesizedRecord<'data_block>,
-        current_seed: &SynthesizerSeedSlice<'data_block>,
-        not_allowed_attr_set: &NotAllowedAttrSet<'data_block>,
-    ) -> AttributeCountMap<'data_block> {
-        let mut cache_key = SynthesizerCacheKey::new(self.headers_len, synthesized_record);
-        let current_attrs_rows = self.calc_current_attrs_rows(&cache_key, synthesized_record);
+        synthesized_record: &SynthesizedRecord,
+        current_seed: &SynthesizerSeedSlice,
+        not_allowed_attr_set: &NotAllowedAttrSet,
+        attr_rows_map: &AttributeRowsMap,
+    ) -> AttributeCountMap {
+        let cache_key = SynthesizerCacheKey::new(self.headers_len, synthesized_record);
+        let current_attrs_rows =
+            self.calc_current_attrs_rows(&cache_key, synthesized_record, attr_rows_map);
 
         self.gen_attr_count_map(
-            &mut cache_key,
+            &cache_key,
             current_seed,
             &current_attrs_rows,
             not_allowed_attr_set,
+            attr_rows_map,
         )
     }
 }

packages/core/src/processing/generator/synthesizer/mod.rs

@@ -1,8 +1,18 @@
 /// Module defining the cache used during the synthesis process
 pub mod cache;
+
 /// Module defining the context used for data synthesis
 pub mod context;
+
 /// Module defining the seeded synthesis process
 pub mod seeded;
+
+/// Module defining the unseeded synthesis process
+pub mod unseeded;
+
 /// Type definitions related to the synthesis process
 pub mod typedefs;
+
+mod seeded_rows_synthesizer;
+
+mod unseeded_rows_synthesizer;

packages/core/src/processing/generator/synthesizer/seeded.rs

@@ -1,56 +1,46 @@
 use super::{
     context::SynthesizerContext,
+    seeded_rows_synthesizer::SeededRowsSynthesizer,
     typedefs::{
         AvailableAttrsMap, NotAllowedAttrSet, SynthesizedRecord, SynthesizedRecords,
         SynthesizedRecordsSlice, SynthesizerSeed, SynthesizerSeedSlice,
     },
 };
 use fnv::FnvHashMap;
-use itertools::izip;
-use log::Level::Trace;
-use log::{info, log_enabled, trace};
+use itertools::{izip, Itertools};
+use log::info;
 use rand::{prelude::SliceRandom, thread_rng};
-use std::time::Duration;
+use std::sync::Arc;
 
 use crate::{
-    data_block::{
-        block::DataBlock,
-        record::DataBlockRecord,
-        typedefs::{AttributeRowsMap, DataBlockRecords},
-        value::DataBlockValue,
-    },
-    measure_time,
+    data_block::{block::DataBlock, typedefs::AttributeRowsMap, value::DataBlockValue},
     utils::{
         math::{calc_percentage, iround_down},
         reporting::ReportProgress,
-        time::ElapsedDuration,
+        threading::get_number_of_threads,
+        time::ElapsedDurationLogger,
     },
 };
 
-#[derive(Debug)]
-struct SeededSynthesizerDurations {
-    consolidate: Duration,
-    suppress: Duration,
-    synthesize_rows: Duration,
-}
-
 /// Represents all the information required to perform the seeded data synthesis
-pub struct SeededSynthesizer<'data_block, 'attr_rows_map> {
-    /// Records to be synthesized
-    pub records: &'data_block DataBlockRecords,
-    /// Context used for the synthesis process
-    pub context: SynthesizerContext<'data_block, 'attr_rows_map>,
+pub struct SeededSynthesizer {
+    /// Reference to the original data block
+    data_block: Arc<DataBlock>,
+    /// Maps a data block value to all the rows where it occurs
+    attr_rows_map: Arc<AttributeRowsMap>,
+    /// Reporting resolution used for data synthesis
+    resolution: usize,
+    /// Maximum cache size allowed
+    cache_max_size: usize,
     /// Percentage already completed on the row synthesis step
     synthesize_percentage: f64,
     /// Percentage already completed on the consolidation step
     consolidate_percentage: f64,
     /// Percentage already completed on the suppression step
     suppress_percentage: f64,
-    /// Elapsed durations on each step for benchmarking
-    durations: SeededSynthesizerDurations,
 }
 
-impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map> {
+impl SeededSynthesizer {
     /// Returns a new SeededSynthesizer
     /// # Arguments
     /// * `data_block` - Sensitive data to be synthesized
@@ -59,28 +49,19 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
     /// * `cache_max_size` - Maximum cache size allowed
     #[inline]
     pub fn new(
-        data_block: &'data_block DataBlock,
-        attr_rows_map: &'attr_rows_map AttributeRowsMap<'data_block>,
+        data_block: Arc<DataBlock>,
+        attr_rows_map: Arc<AttributeRowsMap>,
         resolution: usize,
         cache_max_size: usize,
-    ) -> SeededSynthesizer<'data_block, 'attr_rows_map> {
+    ) -> SeededSynthesizer {
         SeededSynthesizer {
-            records: &data_block.records,
-            context: SynthesizerContext::new(
-                data_block.headers.len(),
-                data_block.records.len(),
-                attr_rows_map,
-                resolution,
-                cache_max_size,
-            ),
+            data_block,
+            attr_rows_map,
+            resolution,
+            cache_max_size,
             synthesize_percentage: 0.0,
             consolidate_percentage: 0.0,
             suppress_percentage: 0.0,
-            durations: SeededSynthesizerDurations {
-                consolidate: Duration::default(),
-                suppress: Duration::default(),
-                synthesize_rows: Duration::default(),
-            },
         }
     }
 
@@ -89,64 +70,71 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
     /// # Arguments
     /// * `progress_reporter` - Will be used to report the processing
     /// progress (`ReportProgress` trait). If `None`, nothing will be reported
-    pub fn run<T>(&mut self, progress_reporter: &mut Option<T>) -> SynthesizedRecords<'data_block>
+    pub fn run<T>(&mut self, progress_reporter: &mut Option<T>) -> SynthesizedRecords
     where
         T: ReportProgress,
     {
-        let mut synthesized_records: SynthesizedRecords<'data_block> = SynthesizedRecords::new();
+        let mut synthesized_records: SynthesizedRecords = SynthesizedRecords::new();
 
-        self.synthesize_percentage = 0.0;
-        self.consolidate_percentage = 0.0;
-        self.suppress_percentage = 0.0;
+        if !self.data_block.records.is_empty() {
+            let mut rows_synthesizers: Vec<SeededRowsSynthesizer> = self.build_rows_synthesizers();
 
-        self.synthesize_rows(&mut synthesized_records, progress_reporter);
-        self.consolidate(&mut synthesized_records, progress_reporter);
-        self.suppress(&mut synthesized_records, progress_reporter);
+            self.synthesize_percentage = 0.0;
+            self.consolidate_percentage = 0.0;
+            self.suppress_percentage = 0.0;
 
+            self.synthesize_rows(
+                &mut synthesized_records,
+                &mut rows_synthesizers,
+                progress_reporter,
+            );
+            self.consolidate(
+                &mut synthesized_records,
+                progress_reporter,
+                // use the first context to leverage already cached intersections
+                &mut rows_synthesizers[0].context,
+            );
+            self.suppress(&mut synthesized_records, progress_reporter);
+        }
         synthesized_records
     }
 
     #[inline]
-    fn synthesize_row(
-        &mut self,
-        seed: &'data_block DataBlockRecord,
-    ) -> SynthesizedRecord<'data_block> {
-        let current_seed: SynthesizerSeed = seed.values.iter().collect();
-        let mut synthesized_record = SynthesizedRecord::default();
-        let not_allowed_attr_set = NotAllowedAttrSet::default();
+    fn build_rows_synthesizers(&self) -> Vec<SeededRowsSynthesizer> {
+        let chunk_size = ((self.data_block.records.len() as f64) / (get_number_of_threads() as f64))
+            .ceil() as usize;
+        let mut rows_synthesizers: Vec<SeededRowsSynthesizer> = Vec::default();
 
-        loop {
-            let next = self.context.sample_next_attr(
-                &synthesized_record,
-                &current_seed,
-                &not_allowed_attr_set,
-            );
-
-            match next {
-                Some(value) => {
-                    synthesized_record.insert(value);
-                }
-                None => break,
-            }
+        for c in &self.data_block.records.iter().chunks(chunk_size) {
+            rows_synthesizers.push(SeededRowsSynthesizer::new(
+                SynthesizerContext::new(
+                    self.data_block.headers.len(),
+                    self.data_block.records.len(),
+                    self.resolution,
+                    self.cache_max_size,
+                ),
+                c.cloned().collect(),
+                self.attr_rows_map.clone(),
+            ));
         }
-        synthesized_record
+        rows_synthesizers
     }
 
     #[inline]
     fn count_not_used_attrs(
         &mut self,
-        synthesized_records: &SynthesizedRecordsSlice<'data_block>,
-    ) -> AvailableAttrsMap<'data_block> {
+        synthesized_records: &SynthesizedRecordsSlice,
+    ) -> AvailableAttrsMap {
         let mut available_attrs: AvailableAttrsMap = AvailableAttrsMap::default();
 
         // go through the pairs (original_record, synthesized_record) and count how many
         // attributes were not used
         for (original_record, synthesized_record) in
-            izip!(self.records.iter(), synthesized_records.iter())
+            izip!(self.data_block.records.iter(), synthesized_records.iter())
         {
             for d in original_record.values.iter() {
                 if !synthesized_record.contains(d) {
-                    let attr = available_attrs.entry(d).or_insert(0);
+                    let attr = available_attrs.entry(d.clone()).or_insert(0);
                     *attr += 1;
                 }
             }
@@ -156,16 +144,16 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
 
     fn calc_available_attrs(
         &mut self,
-        synthesized_records: &SynthesizedRecordsSlice<'data_block>,
-    ) -> AvailableAttrsMap<'data_block> {
+        synthesized_records: &SynthesizedRecordsSlice,
+    ) -> AvailableAttrsMap {
         let mut available_attrs = self.count_not_used_attrs(synthesized_records);
-        let resolution_f64 = self.context.get_resolution() as f64;
+        let resolution_f64 = self.resolution as f64;
 
         // add attributes for consolidation
-        for (attr, value) in self.context.get_attr_rows_map().iter() {
+        for (attr, value) in self.attr_rows_map.iter() {
             let n_rows = value.len();
 
-            if n_rows >= self.context.get_resolution() {
+            if n_rows >= self.resolution {
                 let target_attr_count = iround_down(n_rows as f64, resolution_f64)
                     - (n_rows as isize)
                     + match available_attrs.get(attr) {
@@ -175,7 +163,7 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
 
                 if target_attr_count > 0 {
                     // insert/update the final target count
-                    available_attrs.insert(attr, target_attr_count);
+                    available_attrs.insert(attr.clone(), target_attr_count);
                 } else {
                     // remove negative and zero values
                     available_attrs.remove(attr);
@@ -192,18 +180,17 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
     #[inline]
     fn calc_not_allowed_attrs(
         &mut self,
-        available_attrs: &mut AvailableAttrsMap<'data_block>,
-    ) -> NotAllowedAttrSet<'data_block> {
-        self.context
-            .get_attr_rows_map()
+        available_attrs: &mut AvailableAttrsMap,
+    ) -> NotAllowedAttrSet {
+        self.attr_rows_map
             .keys()
             .filter_map(|attr| match available_attrs.get(attr) {
                 // not on available attributes
-                None => Some(*attr),
+                None => Some(attr.clone()),
                 Some(at) => {
                     if *at <= 0 {
                         // on available attributes, but count <= 0
-                        Some(*attr)
+                        Some(attr.clone())
                     } else {
                         None
                     }
@@ -215,18 +202,20 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
     #[inline]
     fn consolidate_record(
         &mut self,
-        available_attrs: &mut AvailableAttrsMap<'data_block>,
-        current_seed: &SynthesizerSeedSlice<'data_block>,
-    ) -> SynthesizedRecord<'data_block> {
+        available_attrs: &mut AvailableAttrsMap,
+        current_seed: &SynthesizerSeedSlice,
+        context: &mut SynthesizerContext,
+    ) -> SynthesizedRecord {
         let mut not_allowed_attr_set: NotAllowedAttrSet =
             self.calc_not_allowed_attrs(available_attrs);
         let mut synthesized_record = SynthesizedRecord::default();
 
         loop {
-            let next = self.context.sample_next_attr(
+            let next = context.sample_next_attr_from_seed(
                 &synthesized_record,
                 current_seed,
                 &not_allowed_attr_set,
+                &self.attr_rows_map,
             );
 
             match next {
@@ -236,9 +225,9 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
 
                     if next_count <= 1 {
                         available_attrs.remove(&value);
-                        not_allowed_attr_set.insert(value);
+                        not_allowed_attr_set.insert(value.clone());
                     } else {
-                        available_attrs.insert(value, next_count - 1);
+                        available_attrs.insert(value.clone(), next_count - 1);
                     }
                     synthesized_record.insert(value);
                 }
@@ -249,44 +238,43 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
 
     fn consolidate<T>(
         &mut self,
-        synthesized_records: &mut SynthesizedRecords<'data_block>,
+        synthesized_records: &mut SynthesizedRecords,
         progress_reporter: &mut Option<T>,
+        context: &mut SynthesizerContext,
     ) where
         T: ReportProgress,
     {
-        measure_time!(
-            || {
-                info!("consolidating...");
+        let _duration_logger = ElapsedDurationLogger::new("consolidation");
+        info!("consolidating...");
 
-                let mut available_attrs = self.calc_available_attrs(synthesized_records);
-                let current_seed: SynthesizerSeed = available_attrs.keys().cloned().collect();
-                let total = available_attrs.len();
-                let total_f64 = total as f64;
-                let mut n_processed = 0;
+        let mut available_attrs = self.calc_available_attrs(synthesized_records);
+        let current_seed: SynthesizerSeed = available_attrs.keys().cloned().collect();
+        let total = available_attrs.len();
+        let total_f64 = total as f64;
+        let mut n_processed = 0;
 
-                while !available_attrs.is_empty() {
-                    self.update_consolidate_progress(n_processed, total_f64, progress_reporter);
-                    synthesized_records
-                        .push(self.consolidate_record(&mut available_attrs, &current_seed));
-                    n_processed = total - available_attrs.len();
-                }
-                self.update_consolidate_progress(n_processed, total_f64, progress_reporter);
-            },
-            (self.durations.consolidate)
-        );
+        while !available_attrs.is_empty() {
+            self.update_consolidate_progress(n_processed, total_f64, progress_reporter);
+            synthesized_records.push(self.consolidate_record(
+                &mut available_attrs,
+                &current_seed,
+                context,
+            ));
+            n_processed = total - available_attrs.len();
+        }
+        self.update_consolidate_progress(n_processed, total_f64, progress_reporter);
     }
 
     #[inline]
     fn count_synthesized_records_attrs(
         &mut self,
-        synthesized_records: &mut SynthesizedRecords<'data_block>,
-    ) -> FnvHashMap<&'data_block DataBlockValue, isize> {
-        let mut current_counts: FnvHashMap<&'data_block DataBlockValue, isize> =
-            FnvHashMap::default();
+        synthesized_records: &mut SynthesizedRecords,
+    ) -> FnvHashMap<Arc<DataBlockValue>, isize> {
+        let mut current_counts: FnvHashMap<Arc<DataBlockValue>, isize> = FnvHashMap::default();
 
         for r in synthesized_records.iter() {
             for v in r.iter() {
-                let count = current_counts.entry(v).or_insert(0);
+                let count = current_counts.entry(v.clone()).or_insert(0);
                 *count += 1;
             }
         }
@@ -296,16 +284,16 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
     #[inline]
     fn calc_exceeded_count_attrs(
         &mut self,
-        current_counts: &FnvHashMap<&'data_block DataBlockValue, isize>,
-    ) -> FnvHashMap<&'data_block DataBlockValue, isize> {
-        let mut targets: FnvHashMap<&'data_block DataBlockValue, isize> = FnvHashMap::default();
+        current_counts: &FnvHashMap<Arc<DataBlockValue>, isize>,
+    ) -> FnvHashMap<Arc<DataBlockValue>, isize> {
+        let mut targets: FnvHashMap<Arc<DataBlockValue>, isize> = FnvHashMap::default();
 
-        for (attr, rows) in self.context.get_attr_rows_map().iter() {
-            if rows.len() >= self.context.get_resolution() {
-                let t = current_counts[attr]
-                    - iround_down(rows.len() as f64, self.context.get_resolution() as f64);
+        for (attr, rows) in self.attr_rows_map.iter() {
+            if rows.len() >= self.resolution {
+                let t =
+                    current_counts[attr] - iround_down(rows.len() as f64, self.resolution as f64);
                 if t > 0 {
-                    targets.insert(attr, t);
+                    targets.insert(attr.clone(), t);
                 }
             }
         }
@@ -314,76 +302,73 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
 
     fn suppress<T>(
         &mut self,
-        synthesized_records: &mut SynthesizedRecords<'data_block>,
+        synthesized_records: &mut SynthesizedRecords,
         progress_reporter: &mut Option<T>,
     ) where
         T: ReportProgress,
     {
-        measure_time!(
-            || {
-                info!("suppressing...");
+        let _duration_logger = ElapsedDurationLogger::new("suppression");
+        info!("suppressing...");
 
-                let current_counts: FnvHashMap<&'data_block DataBlockValue, isize> =
-                    self.count_synthesized_records_attrs(synthesized_records);
-                let mut targets: FnvHashMap<&'data_block DataBlockValue, isize> =
-                    self.calc_exceeded_count_attrs(&current_counts);
-                let total = synthesized_records.len() as f64;
-                let mut n_processed = 0;
+        let current_counts: FnvHashMap<Arc<DataBlockValue>, isize> =
+            self.count_synthesized_records_attrs(synthesized_records);
+        let mut targets: FnvHashMap<Arc<DataBlockValue>, isize> =
+            self.calc_exceeded_count_attrs(&current_counts);
+        let total = synthesized_records.len() as f64;
+        let mut n_processed = 0;
 
-                synthesized_records.shuffle(&mut thread_rng());
+        synthesized_records.shuffle(&mut thread_rng());
 
-                for r in synthesized_records.iter_mut() {
-                    let mut new_record = SynthesizedRecord::default();
+        for r in synthesized_records.iter_mut() {
+            let mut new_record = SynthesizedRecord::default();
 
-                    self.update_suppress_progress(n_processed, total, progress_reporter);
-                    n_processed += 1;
-                    for attr in r.iter() {
-                        match targets.get(attr).cloned() {
-                            None => {
-                                new_record.insert(attr);
-                            }
-                            Some(attr_count) => {
-                                if attr_count == 1 {
-                                    targets.remove(attr);
-                                } else {
-                                    targets.insert(attr, attr_count - 1);
-                                }
-                            }
+            self.update_suppress_progress(n_processed, total, progress_reporter);
+            n_processed += 1;
+            for attr in r.iter() {
+                match targets.get(attr).cloned() {
+                    None => {
+                        new_record.insert(attr.clone());
+                    }
+                    Some(attr_count) => {
+                        if attr_count == 1 {
+                            targets.remove(attr);
+                        } else {
+                            targets.insert(attr.clone(), attr_count - 1);
                         }
                     }
-                    *r = new_record;
                 }
-                synthesized_records.retain(|r| !r.is_empty());
-                self.update_suppress_progress(n_processed, total, progress_reporter);
-            },
-            (self.durations.suppress)
-        );
+            }
+            *r = new_record;
+        }
+        synthesized_records.retain(|r| !r.is_empty());
+        self.update_suppress_progress(n_processed, total, progress_reporter);
     }
 
     fn synthesize_rows<T>(
         &mut self,
-        synthesized_records: &mut SynthesizedRecords<'data_block>,
+        synthesized_records: &mut SynthesizedRecords,
+        rows_synthesizers: &mut Vec<SeededRowsSynthesizer>,
         progress_reporter: &mut Option<T>,
     ) where
         T: ReportProgress,
     {
-        measure_time!(
-            || {
-                info!("synthesizing rows...");
+        let _duration_logger = ElapsedDurationLogger::new("rows synthesis");
 
-                let mut n_processed = 0;
-                let records = self.records;
-                let total = records.len() as f64;
-
-                for seed in records.iter() {
-                    self.update_synthesize_progress(n_processed, total, progress_reporter);
-                    n_processed += 1;
-                    synthesized_records.push(self.synthesize_row(seed));
-                }
-                self.update_synthesize_progress(n_processed, total, progress_reporter);
-            },
-            (self.durations.synthesize_rows)
+        info!(
+            "synthesizing rows with {} thread(s)...",
+            get_number_of_threads()
         );
+
+        let total = self.data_block.records.len() as f64;
+
+        SeededRowsSynthesizer::synthesize_all(
+            total,
+            synthesized_records,
+            rows_synthesizers,
+            progress_reporter,
+        );
+
+        self.update_synthesize_progress(self.data_block.records.len(), total, progress_reporter);
     }
 
     #[inline]
@@ -396,7 +381,7 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
         T: ReportProgress,
     {
         if let Some(r) = progress_reporter {
-            self.synthesize_percentage = calc_percentage(n_processed, total);
+            self.synthesize_percentage = calc_percentage(n_processed as f64, total);
             r.report(self.calc_overall_progress());
         }
     }
@@ -411,7 +396,7 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
         T: ReportProgress,
     {
         if let Some(r) = progress_reporter {
-            self.consolidate_percentage = calc_percentage(n_processed, total);
+            self.consolidate_percentage = calc_percentage(n_processed as f64, total);
             r.report(self.calc_overall_progress());
         }
     }
@@ -426,7 +411,7 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
         T: ReportProgress,
     {
         if let Some(r) = progress_reporter {
-            self.suppress_percentage = calc_percentage(n_processed, total);
+            self.suppress_percentage = calc_percentage(n_processed as f64, total);
             r.report(self.calc_overall_progress());
         }
     }
@@ -438,9 +423,3 @@ impl<'data_block, 'attr_rows_map> SeededSynthesizer<'data_block, 'attr_rows_map>
             + self.suppress_percentage * 0.2
     }
 }
-
-impl<'data_block, 'attr_rows_map> Drop for SeededSynthesizer<'data_block, 'attr_rows_map> {
-    fn drop(&mut self) {
-        trace!("seed synthesizer durations: {:#?}", self.durations);
-    }
-}

packages/core/src/processing/generator/synthesizer/seeded_rows_synthesizer.rs

@@ -0,0 +1,127 @@
+use super::{
+    context::SynthesizerContext,
+    typedefs::{NotAllowedAttrSet, SynthesizedRecord, SynthesizedRecords, SynthesizerSeed},
+};
+use crate::{
+    data_block::typedefs::AttributeRowsMap,
+    utils::reporting::{SendableProgressReporter, SendableProgressReporterRef},
+};
+use std::sync::Arc;
+
+#[cfg(feature = "rayon")]
+use rayon::prelude::*;
+
+#[cfg(feature = "rayon")]
+use std::sync::Mutex;
+
+use crate::{
+    data_block::{record::DataBlockRecord, typedefs::DataBlockRecords},
+    utils::reporting::ReportProgress,
+};
+
+pub struct SeededRowsSynthesizer {
+    pub context: SynthesizerContext,
+    pub records: DataBlockRecords,
+    pub attr_rows_map: Arc<AttributeRowsMap>,
+}
+
+impl SeededRowsSynthesizer {
+    #[inline]
+    pub fn new(
+        context: SynthesizerContext,
+        records: DataBlockRecords,
+        attr_rows_map: Arc<AttributeRowsMap>,
+    ) -> SeededRowsSynthesizer {
+        SeededRowsSynthesizer {
+            context,
+            records,
+            attr_rows_map,
+        }
+    }
+
+    #[cfg(feature = "rayon")]
+    #[inline]
+    pub fn synthesize_all<T>(
+        total: f64,
+        synthesized_records: &mut SynthesizedRecords,
+        rows_synthesizers: &mut Vec<SeededRowsSynthesizer>,
+        progress_reporter: &mut Option<T>,
+    ) where
+        T: ReportProgress,
+    {
+        let sendable_pr = Arc::new(Mutex::new(
+            progress_reporter
+                .as_mut()
+                .map(|r| SendableProgressReporter::new(total, 0.5, r)),
+        ));
+
+        synthesized_records.par_extend(
+            rows_synthesizers
+                .par_iter_mut()
+                .flat_map(|rs| rs.synthesize_rows(&mut sendable_pr.clone())),
+        );
+    }
+
+    #[cfg(not(feature = "rayon"))]
+    #[inline]
+    pub fn synthesize_all<T>(
+        total: f64,
+        synthesized_records: &mut SynthesizedRecords,
+        rows_synthesizers: &mut Vec<SeededRowsSynthesizer>,
+        progress_reporter: &mut Option<T>,
+    ) where
+        T: ReportProgress,
+    {
+        let mut sendable_pr = progress_reporter
+            .as_mut()
+            .map(|r| SendableProgressReporter::new(total, 0.5, r));
+
+        synthesized_records.extend(
+            rows_synthesizers
+                .iter_mut()
+                .flat_map(|rs| rs.synthesize_rows(&mut sendable_pr)),
+        );
+    }
+
+    #[inline]
+    fn synthesize_rows<T>(
+        &mut self,
+        progress_reporter: &mut SendableProgressReporterRef<T>,
+    ) -> SynthesizedRecords
+    where
+        T: ReportProgress,
+    {
+        let mut synthesized_records = SynthesizedRecords::default();
+        let records = self.records.clone();
+
+        for seed in records.iter() {
+            synthesized_records.push(self.synthesize_row(seed));
+            SendableProgressReporter::update_progress(progress_reporter, 1.0);
+        }
+        synthesized_records
+    }
+
+    #[inline]
+    fn synthesize_row(&mut self, seed: &DataBlockRecord) -> SynthesizedRecord {
+        let current_seed: &SynthesizerSeed = &seed.values;
+        let mut synthesized_record = SynthesizedRecord::default();
+        let not_allowed_attr_set = NotAllowedAttrSet::default();
+
+        loop {
+            let next = self.context.sample_next_attr_from_seed(
+                &synthesized_record,
+                current_seed,
+                &not_allowed_attr_set,
+                &self.attr_rows_map,
+            );
+
+            match next {
+                Some(value) => {
+                    synthesized_record.insert(value);
+                }
+                None => break,
+            }
+        }
+        synthesized_record
+    }
+}

packages/core/src/processing/generator/synthesizer/typedefs.rs

@@ -1,33 +1,31 @@
 use fnv::{FnvHashMap, FnvHashSet};
-use std::collections::BTreeSet;
+use std::{collections::BTreeSet, sync::Arc};
 
 use crate::data_block::value::DataBlockValue;
 
 /// If the data block value were added to the synthesized record, this maps to the
 /// number of rows the resulting attribute combination will be part of
-pub type AttributeCountMap<'data_block_value> =
-    FnvHashMap<&'data_block_value DataBlockValue, usize>;
+pub type AttributeCountMap = FnvHashMap<Arc<DataBlockValue>, usize>;
 
 /// The seeds used for the current synthesis step (aka current record being processed)
-pub type SynthesizerSeed<'data_block_value> = Vec<&'data_block_value DataBlockValue>;
+pub type SynthesizerSeed = Vec<Arc<DataBlockValue>>;
 
 /// Slice of SynthesizerSeed
-pub type SynthesizerSeedSlice<'data_block_value> = [&'data_block_value DataBlockValue];
+pub type SynthesizerSeedSlice = [Arc<DataBlockValue>];
 
 /// Attributes not allowed to be sampled in a particular data synthesis stage
-pub type NotAllowedAttrSet<'data_block_value> = FnvHashSet<&'data_block_value DataBlockValue>;
+pub type NotAllowedAttrSet = FnvHashSet<Arc<DataBlockValue>>;
 
 /// Record synthesized at a particular stage
-pub type SynthesizedRecord<'data_block_value> = BTreeSet<&'data_block_value DataBlockValue>;
+pub type SynthesizedRecord = BTreeSet<Arc<DataBlockValue>>;
 
 /// Vector of synthesized records
-pub type SynthesizedRecords<'data_block_value> = Vec<SynthesizedRecord<'data_block_value>>;
+pub type SynthesizedRecords = Vec<SynthesizedRecord>;
 
 /// Slice of SynthesizedRecords
-pub type SynthesizedRecordsSlice<'data_block_value> = [SynthesizedRecord<'data_block_value>];
+pub type SynthesizedRecordsSlice = [SynthesizedRecord];
 
 /// How many attributes (of the ones available) should be added during the
 /// consolidation step for the count to match the a multiple of reporting resolution
 /// (rounded down)
-pub type AvailableAttrsMap<'data_block_value> =
-    FnvHashMap<&'data_block_value DataBlockValue, isize>;
+pub type AvailableAttrsMap = FnvHashMap<Arc<DataBlockValue>, isize>;

packages/core/src/processing/generator/synthesizer/unseeded.rs

@@ -0,0 +1,163 @@
+use super::{
+    context::SynthesizerContext, typedefs::SynthesizedRecords,
+    unseeded_rows_synthesizer::UnseededRowsSynthesizer,
+};
+use log::info;
+use std::sync::Arc;
+
+use crate::{
+    data_block::{block::DataBlock, typedefs::AttributeRowsByColumnMap},
+    utils::{
+        math::calc_percentage, reporting::ReportProgress, threading::get_number_of_threads,
+        time::ElapsedDurationLogger,
+    },
+};
+
+/// Represents all the information required to perform the unseeded data synthesis
+pub struct UnseededSynthesizer {
+    /// Reference to the original data block
+    data_block: Arc<DataBlock>,
+    /// Maps a data block value to all the rows where it occurs grouped by column
+    attr_rows_map_by_column: Arc<AttributeRowsByColumnMap>,
+    /// Reporting resolution used for data synthesis
+    resolution: usize,
+    /// Maximum cache size allowed
+    cache_max_size: usize,
+    /// Empty values on the synthetic data will be represented by this
+    empty_value: Arc<String>,
+    /// Percentage already completed on the row synthesis step
+    synthesize_percentage: f64,
+}
+
+impl UnseededSynthesizer {
+    /// Returns a new UnseededSynthesizer
+    /// # Arguments
+    /// * `data_block` - Sensitive data to be synthesized
+    /// * `attr_rows_map_by_column` - Maps a data block value to all the rows where it occurs grouped by column
+    /// * `resolution` - Reporting resolution used for data synthesis
+    /// * `cache_max_size` - Maximum cache size allowed
+    /// * `empty_value` - Empty values on the synthetic data will be represented by this
+    #[inline]
+    pub fn new(
+        data_block: Arc<DataBlock>,
+        attr_rows_map_by_column: Arc<AttributeRowsByColumnMap>,
+        resolution: usize,
+        cache_max_size: usize,
+        empty_value: Arc<String>,
+    ) -> UnseededSynthesizer {
+        UnseededSynthesizer {
+            data_block,
+            attr_rows_map_by_column,
+            resolution,
+            cache_max_size,
+            empty_value,
+            synthesize_percentage: 0.0,
+        }
+    }
+
+    /// Performs the row synthesis
+    /// Returns the synthesized records
+    /// # Arguments
+    /// * `progress_reporter` - Will be used to report the processing
+    /// progress (`ReportProgress` trait). If `None`, nothing will be reported
+    pub fn run<T>(&mut self, progress_reporter: &mut Option<T>) -> SynthesizedRecords
+    where
+        T: ReportProgress,
+    {
+        let mut synthesized_records: SynthesizedRecords = SynthesizedRecords::new();
+
+        if !self.data_block.records.is_empty() {
+            let mut rows_synthesizers: Vec<UnseededRowsSynthesizer> =
+                self.build_rows_synthesizers();
+
+            self.synthesize_percentage = 0.0;
+
+            self.synthesize_rows(
+                &mut synthesized_records,
+                &mut rows_synthesizers,
+                progress_reporter,
+            );
+        }
+        synthesized_records
+    }
+
+    #[inline]
+    fn build_rows_synthesizers(&self) -> Vec<UnseededRowsSynthesizer> {
+        let mut total_size = self.data_block.records.len();
+        let chunk_size = ((total_size as f64) / (get_number_of_threads() as f64)).ceil() as usize;
+        let mut rows_synthesizers: Vec<UnseededRowsSynthesizer> = Vec::default();
+
+        loop {
+            if total_size > chunk_size {
+                rows_synthesizers.push(UnseededRowsSynthesizer::new(
+                    SynthesizerContext::new(
+                        self.data_block.headers.len(),
+                        self.data_block.records.len(),
+                        self.resolution,
+                        self.cache_max_size,
+                    ),
+                    chunk_size,
+                    self.attr_rows_map_by_column.clone(),
+                    self.empty_value.clone(),
+                ));
+                total_size -= chunk_size;
+            } else {
+                rows_synthesizers.push(UnseededRowsSynthesizer::new(
+                    SynthesizerContext::new(
+                        self.data_block.headers.len(),
+                        self.data_block.records.len(),
+                        self.resolution,
+                        self.cache_max_size,
+                    ),
+                    total_size,
+                    self.attr_rows_map_by_column.clone(),
+                    self.empty_value.clone(),
+                ));
+                break;
+            }
+        }
+        rows_synthesizers
+    }
+
+    fn synthesize_rows<T>(
+        &mut self,
+        synthesized_records: &mut SynthesizedRecords,
+        rows_synthesizers: &mut Vec<UnseededRowsSynthesizer>,
+        progress_reporter: &mut Option<T>,
+    ) where
+        T: ReportProgress,
+    {
+        let _duration_logger = ElapsedDurationLogger::new("rows synthesis");
+
+        info!(
+            "synthesizing rows with {} thread(s)...",
+            get_number_of_threads()
+        );
+
+        let total = self.data_block.records.len() as f64;
+
+        UnseededRowsSynthesizer::synthesize_all(
+            total,
+            synthesized_records,
+            rows_synthesizers,
+            progress_reporter,
+        );
+
+        self.update_synthesize_progress(self.data_block.records.len(), total, progress_reporter);
+    }
+
+    #[inline]
+    fn update_synthesize_progress<T>(
+        &mut self,
+        n_processed: usize,
+        total: f64,
+        progress_reporter: &mut Option<T>,
+    ) where
+        T: ReportProgress,
+    {
+        if let Some(r) = progress_reporter {
+            self.synthesize_percentage = calc_percentage(n_processed as f64, total);
+            r.report(self.synthesize_percentage);
+        }
+    }
+}

packages/core/src/processing/generator/synthesizer/unseeded_rows_synthesizer.rs

@@ -0,0 +1,128 @@
+use super::{
+    context::SynthesizerContext,
+    typedefs::{SynthesizedRecord, SynthesizedRecords},
+};
+use crate::{
+    data_block::typedefs::{AttributeRows, AttributeRowsByColumnMap},
+    utils::reporting::{SendableProgressReporter, SendableProgressReporterRef},
+};
+use rand::{prelude::SliceRandom, thread_rng};
+use std::sync::Arc;
+
+#[cfg(feature = "rayon")]
+use rayon::prelude::*;
+
+#[cfg(feature = "rayon")]
+use std::sync::Mutex;
+
+use crate::utils::reporting::ReportProgress;
+
+pub struct UnseededRowsSynthesizer {
+    context: SynthesizerContext,
+    chunk_size: usize,
+    column_indexes: Vec<usize>,
+    attr_rows_map_by_column: Arc<AttributeRowsByColumnMap>,
+    empty_value: Arc<String>,
+}
+
+impl UnseededRowsSynthesizer {
+    #[inline]
+    pub fn new(
+        context: SynthesizerContext,
+        chunk_size: usize,
+        attr_rows_map_by_column: Arc<AttributeRowsByColumnMap>,
+        empty_value: Arc<String>,
+    ) -> UnseededRowsSynthesizer {
+        UnseededRowsSynthesizer {
+            context,
+            chunk_size,
+            column_indexes: attr_rows_map_by_column.keys().cloned().collect(),
+            attr_rows_map_by_column,
+            empty_value,
+        }
+    }
+
+    #[cfg(feature = "rayon")]
+    #[inline]
+    pub fn synthesize_all<T>(
+        total: f64,
+        synthesized_records: &mut SynthesizedRecords,
+        rows_synthesizers: &mut Vec<UnseededRowsSynthesizer>,
+        progress_reporter: &mut Option<T>,
+    ) where
+        T: ReportProgress,
+    {
+        let sendable_pr = Arc::new(Mutex::new(
+            progress_reporter
+                .as_mut()
+                .map(|r| SendableProgressReporter::new(total, 1.0, r)),
+        ));
+
+        synthesized_records.par_extend(
+            rows_synthesizers
+                .par_iter_mut()
+                .flat_map(|rs| rs.synthesize_rows(&mut sendable_pr.clone())),
+        );
+    }
+
+    #[cfg(not(feature = "rayon"))]
+    #[inline]
+    pub fn synthesize_all<T>(
+        total: f64,
+        synthesized_records: &mut SynthesizedRecords,
+        rows_synthesizers: &mut Vec<UnseededRowsSynthesizer>,
+        progress_reporter: &mut Option<T>,
+    ) where
+        T: ReportProgress,
+    {
+        let mut sendable_pr = progress_reporter
+            .as_mut()
+            .map(|r| SendableProgressReporter::new(total, 1.0, r));
+
+        synthesized_records.extend(
+            rows_synthesizers
+                .iter_mut()
+                .flat_map(|rs| rs.synthesize_rows(&mut sendable_pr)),
+        );
+    }
+
+    #[inline]
+    fn synthesize_rows<T>(
+        &mut self,
+        progress_reporter: &mut SendableProgressReporterRef<T>,
+    ) -> SynthesizedRecords
+    where
+        T: ReportProgress,
+    {
+        let mut synthesized_records = SynthesizedRecords::default();
+
+        for _ in 0..self.chunk_size {
+            synthesized_records.push(self.synthesize_row());
+            SendableProgressReporter::update_progress(progress_reporter, 1.0);
+        }
+        synthesized_records
+    }
+
+    #[inline]
+    fn synthesize_row(&mut self) -> SynthesizedRecord {
+        let mut synthesized_record = SynthesizedRecord::default();
+        let mut current_attrs_rows: Arc<AttributeRows> =
+            Arc::new((0..self.context.records_len).collect());
+
+        self.column_indexes.shuffle(&mut thread_rng());
+
+        for column_index in self.column_indexes.iter() {
+            if let Some((next_attrs_rows, sample)) = self.context.sample_next_attr_from_column(
+                &synthesized_record,
+                *column_index,
+                &self.attr_rows_map_by_column,
+                &current_attrs_rows,
+                &self.empty_value,
+            ) {
+                current_attrs_rows = next_attrs_rows;
+                synthesized_record.insert(sample);
+            }
+        }
+        synthesized_record
+    }
+}

packages/core/src/utils/collections.rs

@@ -4,8 +4,8 @@ use std::hash::Hash;
 /// Given two sorted vectors, calculates the intersection between them.
 /// Returns the result intersection as a new vector
 /// # Arguments
-/// - `a`: first sorted vector
-/// - `b`: second sorted vector
+/// * `a` - first sorted vector
+/// * `b` - second sorted vector
 #[inline]
 pub fn ordered_vec_intersection<T>(a: &[T], b: &[T]) -> Vec<T>
 where

packages/core/src/utils/math.rs

@@ -54,6 +54,6 @@ pub fn calc_n_combinations_range(n: usize, range: &[usize]) -> u64 {
 
 /// Calculates the percentage of processed elements up to a total
 #[inline]
-pub fn calc_percentage(n_processed: usize, total: f64) -> f64 {
-    (n_processed as f64) * 100.0 / total
+pub fn calc_percentage(n_processed: f64, total: f64) -> f64 {
+    n_processed * 100.0 / total
 }

packages/core/src/utils/mod.rs

@@ -1,8 +1,14 @@
 /// Module for collection utilities
 pub mod collections;
+
 /// Module for math utilities
 pub mod math;
+
 /// Module for reporting utilities
 pub mod reporting;
+
+/// Module for threading utilities
+pub mod threading;
+
 /// Module for time utilities
 pub mod time;

packages/core/src/utils/reporting/logger_progress_reporter.rs

@@ -1,11 +1,6 @@
+use super::ReportProgress;
 use log::{log, Level};
 
-/// Implement this trait to inform progress
-pub trait ReportProgress {
-    /// Receives the updated progress
-    fn report(&mut self, new_progress: f64);
-}
-
 /// Simple progress reporter using the default logger.
 /// * It will log progress using the configured `log_level`
 /// * It will only log at every 1% completed
@@ -16,8 +11,8 @@ pub struct LoggerProgressReporter {
 
 impl LoggerProgressReporter {
     /// Returns a new LoggerProgressReporter
-    /// # Arguments:
-    /// * `log_level`: which log level use to log progress
+    /// # Arguments
+    /// * `log_level - which log level use to log progress
     pub fn new(log_level: Level) -> LoggerProgressReporter {
         LoggerProgressReporter {
             progress: 0.0,

packages/core/src/utils/reporting/mod.rs

@@ -0,0 +1,11 @@
+mod report_progress;
+
+mod logger_progress_reporter;
+
+mod sendable_progress_reporter;
+
+pub use report_progress::ReportProgress;
+
+pub use logger_progress_reporter::LoggerProgressReporter;
+
+pub use sendable_progress_reporter::{SendableProgressReporter, SendableProgressReporterRef};

packages/core/src/utils/reporting/report_progress.rs

@@ -0,0 +1,5 @@
+/// Implement this trait to inform progress
+pub trait ReportProgress {
+    /// Receives the updated progress
+    fn report(&mut self, new_progress: f64);
+}

packages/core/src/utils/reporting/sendable_progress_reporter.rs

@@ -0,0 +1,92 @@
+#[cfg(feature = "rayon")]
+use std::sync::{Arc, Mutex};
+
+use crate::utils::{math::calc_percentage, reporting::ReportProgress};
+
+/// Progress reporter to be used in parallel environments
+/// (should be wrapped with `Arc` and `Mutex` if multi-threading is enabled).
+/// It will calculate `proportion * (n_processed * 100.0 / total)` and report
+/// it to the main reporter
+pub struct SendableProgressReporter<'main_reporter, T>
+where
+    T: ReportProgress,
+{
+    total: f64,
+    n_processed: f64,
+    proportion: f64,
+    main_reporter: &'main_reporter mut T,
+}
+
+impl<'main_reporter, T> SendableProgressReporter<'main_reporter, T>
+where
+    T: ReportProgress,
+{
+    /// Creates a new SendableProgressReporter
+    /// # Arguments
+    /// * `total` - Total number of steps to be reported
+    /// * `proportion` - Value to multiply the percentage before reporting to
+    /// main_reporter
+    /// * `main_reporter` - Main reporter to which this should report
+    pub fn new(
+        total: f64,
+        proportion: f64,
+        main_reporter: &'main_reporter mut T,
+    ) -> SendableProgressReporter<T> {
+        SendableProgressReporter {
+            total,
+            proportion,
+            n_processed: 0.0,
+            main_reporter,
+        }
+    }
+
+    /// Updates `reporter` by adding `value_to_add` and reporting
+    /// to the main reporter in a thread safe way
+    #[inline]
+    pub fn update_progress(reporter: &mut SendableProgressReporterRef<T>, value_to_add: f64)
+    where
+        T: ReportProgress,
+    {
+        #[cfg(feature = "rayon")]
+        if let Ok(guard) = &mut reporter.lock() {
+            if let Some(r) = guard.as_mut() {
+                r.report(value_to_add);
+            }
+        }
+        #[cfg(not(feature = "rayon"))]
+        if let Some(r) = reporter {
+            r.report(value_to_add);
+        }
+    }
+}
+
+impl<'main_reporter, T> ReportProgress for SendableProgressReporter<'main_reporter, T>
+where
+    T: ReportProgress,
+{
+    /// Will add `value_to_add` to `n_processed` and call the main reporter with
+    /// `proportion * (n_processed * 100.0 / total)`
+    fn report(&mut self, value_to_add: f64) {
+        self.n_processed += value_to_add;
+        self.main_reporter
+            .report(self.proportion * calc_percentage(self.n_processed, self.total))
+    }
+}
+
+#[cfg(feature = "rayon")]
+///  People using this should correctly handle race conditions with a `Mutex`
+/// (see `SendableProgressReporterRef`), so we inform the compiler this struct is thread safe
+unsafe impl<'main_reporter, T> Send for SendableProgressReporter<'main_reporter, T> where
+    T: ReportProgress
+{
+}
+
+#[cfg(feature = "rayon")]
+/// Use this to refer to SendableProgressReporter if multi-threading is enabled
+pub type SendableProgressReporterRef<'main_reporter, T> =
+    Arc<Mutex<Option<SendableProgressReporter<'main_reporter, T>>>>;
+
+#[cfg(not(feature = "rayon"))]
+/// Use this to refer to SendableProgressReporter if multi-threading is disabled
+pub type SendableProgressReporterRef<'main_reporter, T> =
+    Option<SendableProgressReporter<'main_reporter, T>>;

packages/core/src/utils/threading.rs

@@ -0,0 +1,34 @@
+#[cfg(feature = "rayon")]
+use rayon;
+
+#[cfg(feature = "pyo3")]
+use pyo3::prelude::*;
+
+pub fn get_number_of_threads() -> usize {
+    #[cfg(feature = "rayon")]
+    {
+        rayon::current_num_threads()
+    }
+    #[cfg(not(feature = "rayon"))]
+    {
+        1
+    }
+}
+
+#[cfg(feature = "rayon")]
+#[cfg_attr(feature = "pyo3", pyfunction)]
+#[allow(unused_must_use)]
+/// Sets the number of threads used for parallel processing
+/// # Arguments
+/// * `n` - number of threads
+pub fn set_number_of_threads(n: usize) {
+    rayon::ThreadPoolBuilder::new()
+        .num_threads(n)
+        .build_global();
+}
+
+#[cfg(feature = "pyo3")]
+pub fn register(_py: Python, m: &PyModule) -> PyResult<()> {
+    m.add_function(wrap_pyfunction!(set_number_of_threads, m)?)?;
+    Ok(())
+}

packages/core/src/utils/time.rs

@@ -16,7 +16,7 @@ pub struct ElapsedDuration<'lifetime> {
 impl<'lifetime> ElapsedDuration<'lifetime> {
     /// Returns a new ElapsedDuration
     /// # Arguments
-    /// - `result`: Duration where to sum the elapsed duration when the `ElapsedDuration`
+    /// * `result` - Duration where to sum the elapsed duration when the `ElapsedDuration`
     /// instance goes out of scope
     pub fn new(result: &mut Duration) -> ElapsedDuration {
         ElapsedDuration {
@@ -41,10 +41,10 @@ pub struct ElapsedDurationLogger {
 }
 
 impl ElapsedDurationLogger {
-    pub fn new(message: String) -> ElapsedDurationLogger {
+    pub fn new<S: Into<String>>(message: S) -> ElapsedDurationLogger {
         ElapsedDurationLogger {
             _start: Instant::now(),
-            _message: message,
+            _message: message.into(),
         }
     }
 }

packages/lib-python/Cargo.toml

@@ -14,4 +14,5 @@ crate-type = ["cdylib"]
 log = { version = "0.4", features = ["std"] }
 csv = { version = "1.1" }
 pyo3 = { version = "0.15", features = ["extension-module"] }
-sds-core = { path = "../core", features = ["pyo3"] }
+sds-core = { path = "../core", features = ["pyo3", "rayon"] }
+env_logger = { version = "0.9" }

packages/lib-python/src/data_processor.rs

@@ -0,0 +1,133 @@
+use csv::ReaderBuilder;
+use pyo3::prelude::*;
+use sds_core::{
+    data_block::{
+        block::DataBlock, csv_block_creator::CsvDataBlockCreator, csv_io_error::CsvIOError,
+        data_block_creator::DataBlockCreator,
+    },
+    processing::{
+        aggregator::{aggregated_data::AggregatedData, Aggregator},
+        generator::{generated_data::GeneratedData, Generator, SynthesisMode},
+    },
+    utils::reporting::LoggerProgressReporter,
+};
+use std::sync::Arc;
+
+#[pyclass]
+/// Processor exposing the main features
+pub struct SDSProcessor {
+    data_block: Arc<DataBlock>,
+}
+
+#[pymethods]
+impl SDSProcessor {
+    /// Creates a new processor by reading the content of a given file
+    /// # Arguments
+    /// * `path` - File to be read to build the data block
+    /// * `delimiter` - CSV/TSV separator for the content on `path`
+    /// * `use_columns` - Column names to be used (if `[]` use all columns)
+    /// * `sensitive_zeros` - Column names containing sensitive zeros
+    /// (if `[]` no columns are considered to have sensitive zeros)
+    /// * `record_limit` - Use only the first `record_limit` records (if `0` use all records)
+    #[new]
+    pub fn new(
+        path: &str,
+        delimiter: char,
+        use_columns: Vec<String>,
+        sensitive_zeros: Vec<String>,
+        record_limit: usize,
+    ) -> Result<SDSProcessor, CsvIOError> {
+        match CsvDataBlockCreator::create(
+            ReaderBuilder::new()
+                .delimiter(delimiter as u8)
+                .from_path(path),
+            &use_columns,
+            &sensitive_zeros,
+            record_limit,
+        ) {
+            Ok(data_block) => Ok(SDSProcessor { data_block }),
+            Err(err) => Err(CsvIOError::new(err)),
+        }
+    }
+
+    #[staticmethod]
+    /// Load the SDS Processor for the data block linked to `aggregated_data`
+    pub fn from_aggregated_data(aggregated_data: &AggregatedData) -> SDSProcessor {
+        SDSProcessor {
+            data_block: aggregated_data.data_block.clone(),
+        }
+    }
+
+    #[inline]
+    /// Returns the number of records on the data block
+    pub fn number_of_records(&self) -> usize {
+        self.data_block.number_of_records()
+    }
+
+    #[inline]
+    /// Returns the number of records on the data block protected by `resolution`
+    pub fn protected_number_of_records(&self, resolution: usize) -> usize {
+        self.data_block.protected_number_of_records(resolution)
+    }
+
+    #[inline]
+    /// Normalizes the reporting length based on the number of selected headers.
+    /// Returns the normalized value
+    /// # Arguments
+    /// * `reporting_length` - Reporting length to be normalized (0 means use all columns)
+    pub fn normalize_reporting_length(&self, reporting_length: usize) -> usize {
+        self.data_block.normalize_reporting_length(reporting_length)
+    }
+
+    /// Builds the aggregated data for the content
+    /// using the specified `reporting_length` and `sensitivity_threshold`.
+    /// The result is written to `sensitive_aggregates_path` and `reportable_aggregates_path`.
+    /// # Arguments
+    /// * `reporting_length` - Maximum length to compute attribute combinations
+    /// * `sensitivity_threshold` - Sensitivity threshold to filter record attributes
+    /// (0 means no suppression)
+    pub fn aggregate(
+        &self,
+        reporting_length: usize,
+        sensitivity_threshold: usize,
+    ) -> AggregatedData {
+        let mut progress_reporter: Option<LoggerProgressReporter> = None;
+        let mut aggregator = Aggregator::new(self.data_block.clone());
+        aggregator.aggregate(
+            reporting_length,
+            sensitivity_threshold,
+            &mut progress_reporter,
+        )
+    }
+
+    /// Synthesizes the content using the specified `resolution` and
+    /// returns the generated data
+    /// # Arguments
+    /// * `cache_max_size` - Maximum cache size used during the synthesis process
+    /// * `resolution` - Reporting resolution to be used
+    /// * `empty_value` - Empty values on the synthetic data will be represented by this
+    /// * `seeded` - True for seeded synthesis, False for unseeded
+    pub fn generate(
+        &self,
+        cache_max_size: usize,
+        resolution: usize,
+        empty_value: String,
+        seeded: bool,
+    ) -> GeneratedData {
+        let mut progress_reporter: Option<LoggerProgressReporter> = None;
+        let mut generator = Generator::new(self.data_block.clone());
+        let mode = if seeded {
+            SynthesisMode::Seeded
+        } else {
+            SynthesisMode::Unseeded
+        };
+
+        generator.generate(
+            resolution,
+            cache_max_size,
+            empty_value,
+            mode,
+            &mut progress_reporter,
+        )
+    }
+}

packages/lib-python/src/lib.rs

@@ -1,160 +1,24 @@
 //! This crate will generate python bindings for the main features
 //! of the `sds_core` library.
-
-use csv::ReaderBuilder;
-use pyo3::{exceptions::PyIOError, prelude::*};
+use data_processor::SDSProcessor;
+use pyo3::prelude::*;
 use sds_core::{
-    data_block::{
-        block::{DataBlock, DataBlockCreator},
-        csv_block_creator::CsvDataBlockCreator,
-    },
-    processing::{
-        aggregator::{typedefs::AggregatedCountByLenMap, Aggregator},
-        generator::Generator,
-    },
-    utils::reporting::LoggerProgressReporter,
+    processing::{aggregator::aggregated_data, evaluator},
+    utils::threading,
 };
 
-/// Creates a data block by reading the content of a given file
-/// # Arguments
-/// * `path` - File to be read to build the data block
-/// * `delimiter` - CSV/TSV separator for the content on `path`
-/// * `use_columns` - Column names to be used (if `[]` use all columns)
-/// * `sensitive_zeros` - Column names containing sensitive zeros
-/// (if `[]` no columns are considered to have sensitive zeros)
-/// * `record_limit` - Use only the first `record_limit` records (if `0` use all records)
-#[pyfunction]
-pub fn create_data_block_from_file(
-    path: &str,
-    delimiter: char,
-    use_columns: Vec<String>,
-    sensitive_zeros: Vec<String>,
-    record_limit: usize,
-) -> PyResult<DataBlock> {
-    match CsvDataBlockCreator::create(
-        ReaderBuilder::new()
-            .delimiter(delimiter as u8)
-            .from_path(path),
-        &use_columns,
-        &sensitive_zeros,
-        record_limit,
-    ) {
-        Ok(data_block) => Ok(data_block),
-        Err(err) => Err(PyIOError::new_err(format!(
-            "error reading data block: {}",
-            err
-        ))),
-    }
-}
-
-/// Builds the protected and aggregated data for the content in `data_block`
-/// using the specified `reporting_length` and `resolution`.
-/// The result is written to `sensitive_aggregates_path` and `reportable_aggregates_path`.
-///
-/// The function returns a tuple (combo_counts, rare_counts).
-///
-/// * `combo_counts` - computed number of distinct combinations grouped by combination length (not protected)
-/// * `rare_counts` - computed number of rare combinations (`count < resolution`) grouped by combination length (not protected)
-///
-/// # Arguments
-/// * `data_block` - Data block with the content to be aggregated
-/// * `sensitive_aggregates_path` - Path to write the sensitive aggregates
-/// * `reportable_aggregates_path` - Path to write the aggregates with protected count
-/// (rounded down to the nearest multiple of `resolution`)
-/// * `delimiter` - CSV/TSV separator for the content be written in
-/// `sensitive_aggregates_path` and `reportable_aggregates_path`
-/// * `reporting_length` - Maximum length to compute attribute combinations
-/// * `resolution` - Reporting resolution to be used
-#[pyfunction]
-pub fn aggregate_and_write(
-    data_block: &DataBlock,
-    sensitive_aggregates_path: &str,
-    reportable_aggregates_path: &str,
-    delimiter: char,
-    reporting_length: usize,
-    resolution: usize,
-) -> PyResult<(AggregatedCountByLenMap, AggregatedCountByLenMap)> {
-    let mut progress_reporter: Option<LoggerProgressReporter> = None;
-    let mut aggregator = Aggregator::new(data_block);
-    let mut aggregated_data = aggregator.aggregate(reporting_length, 0, &mut progress_reporter);
-    let combo_count = aggregator.calc_combinations_count_by_len(&aggregated_data.aggregates_count);
-    let rare_count = aggregator
-        .calc_rare_combinations_count_by_len(&aggregated_data.aggregates_count, resolution);
-
-    if let Err(err) = aggregator.write_aggregates_count(
-        &aggregated_data.aggregates_count,
-        sensitive_aggregates_path,
-        delimiter,
-        resolution,
-        false,
-    ) {
-        return Err(PyIOError::new_err(format!(
-            "error writing sensitive aggregates: {}",
-            err
-        )));
-    }
-
-    Aggregator::protect_aggregates_count(&mut aggregated_data.aggregates_count, resolution);
-
-    match aggregator.write_aggregates_count(
-        &aggregated_data.aggregates_count,
-        reportable_aggregates_path,
-        delimiter,
-        resolution,
-        true,
-    ) {
-        Ok(_) => Ok((combo_count, rare_count)),
-        Err(err) => Err(PyIOError::new_err(format!(
-            "error writing reportable aggregates: {}",
-            err
-        ))),
-    }
-}
-
-/// Synthesizes the `data_block` using the specified `resolution`.
-/// The result is written to `synthetic_path`.
-///
-/// The function returns the expansion ratio:
-/// `Synthetic data length / Sensitive data length` (header not included)
-///
-/// # Arguments
-/// * `data_block` - Data block with the content to be synthesized
-/// * `synthetic_path` - Path to write the synthetic data
-/// * `synthetic_delimiter` - CSV/TSV separator for the content be written in `synthetic_path`
-/// * `cache_max_size` - Maximum cache size used during the synthesis process
-/// * `resolution` - Reporting resolution to be used
-#[pyfunction]
-pub fn generate_seeded_and_write(
-    data_block: &DataBlock,
-    synthetic_path: &str,
-    synthetic_delimiter: char,
-    cache_max_size: usize,
-    resolution: usize,
-) -> PyResult<f64> {
-    let mut progress_reporter: Option<LoggerProgressReporter> = None;
-    let mut generator = Generator::new(data_block);
-    let generated_data = generator.generate(resolution, cache_max_size, "", &mut progress_reporter);
-
-    match generator.write_records(
-        &generated_data.synthetic_data,
-        synthetic_path,
-        synthetic_delimiter as u8,
-    ) {
-        Ok(_) => Ok(generated_data.expansion_ratio),
-        Err(err) => Err(PyIOError::new_err(format!(
-            "error writing synthetic data: {}",
-            err
-        ))),
-    }
-}
+/// Module that exposes the main processor
+pub mod data_processor;
 
 /// A Python module implemented in Rust. The name of this function must match
 /// the `lib.name` setting in the `Cargo.toml`, else Python will not be able to
 /// import the module.
 #[pymodule]
-fn sds(_py: Python, m: &PyModule) -> PyResult<()> {
-    m.add_function(wrap_pyfunction!(create_data_block_from_file, m)?)?;
-    m.add_function(wrap_pyfunction!(aggregate_and_write, m)?)?;
-    m.add_function(wrap_pyfunction!(generate_seeded_and_write, m)?)?;
+fn sds(py: Python, m: &PyModule) -> PyResult<()> {
+    env_logger::init();
+    m.add_class::<SDSProcessor>()?;
+    threading::register(py, m)?;
+    aggregated_data::register(py, m)?;
+    evaluator::register(py, m)?;
     Ok(())
 }

packages/lib-wasm/Cargo.toml

@@ -19,7 +19,7 @@ csv = { version = "1.1" }
 web-sys = { version = "0.3", features = [ "console" ]}
 sds-core = { path = "../core" }
 js-sys = { version = "0.3" }
-serde = { version = "1.0", features = [ "derive" ] }
+serde = { version = "1.0", features = [ "derive", "rc" ] }
 
 # The `console_error_panic_hook` crate provides better debugging of panics by
 # logging them with `console.error`. This is great for development, but requires

packages/lib-wasm/src/aggregator/mod.rs

@@ -1,116 +0,0 @@
-use js_sys::Object;
-use js_sys::Reflect::set;
-use log::error;
-use sds_core::data_block::block::DataBlock;
-use sds_core::data_block::typedefs::DataBlockHeaders;
-use sds_core::processing::aggregator::typedefs::{AggregatedCountByLenMap, AggregatesCountMap};
-use sds_core::processing::aggregator::{Aggregator, PrivacyRiskSummary};
-use sds_core::utils::reporting::ReportProgress;
-use sds_core::utils::time::ElapsedDurationLogger;
-use serde::Serialize;
-use wasm_bindgen::JsValue;
-
-use crate::utils::js::serializers::{serialize_aggregates_count, serialize_privacy_risk};
-use crate::{match_or_return_undefined, set_or_return_undefined};
-
-#[derive(Serialize)]
-pub struct AggregatedCombination {
-    combination_key: String,
-    count: usize,
-    length: usize,
-}
-
-impl AggregatedCombination {
-    #[inline]
-    pub fn new(combination_key: String, count: usize, length: usize) -> AggregatedCombination {
-        AggregatedCombination {
-            combination_key,
-            count,
-            length,
-        }
-    }
-}
-
-impl From<AggregatedCombination> for JsValue {
-    #[inline]
-    fn from(aggregated_combination: AggregatedCombination) -> Self {
-        match_or_return_undefined!(JsValue::from_serde(&aggregated_combination))
-    }
-}
-
-pub struct AggregatedResult<'data_block> {
-    pub headers: &'data_block DataBlockHeaders,
-    pub aggregates_count: AggregatesCountMap<'data_block>,
-    pub rare_combinations_count_by_len: AggregatedCountByLenMap,
-    pub combinations_count_by_len: AggregatedCountByLenMap,
-    pub combinations_sum_by_len: AggregatedCountByLenMap,
-    pub privacy_risk: PrivacyRiskSummary,
-}
-
-impl<'data_block> From<AggregatedResult<'data_block>> for JsValue {
-    #[inline]
-    fn from(aggregated_result: AggregatedResult) -> Self {
-        let _duration_logger =
-            ElapsedDurationLogger::new(String::from("AggregatedResult conversion to JsValue"));
-        let result = Object::new();
-
-        set_or_return_undefined!(
-            &result,
-            &"aggregatedCombinations".into(),
-            &serialize_aggregates_count(
-                &aggregated_result.aggregates_count,
-                aggregated_result.headers
-            ),
-        );
-        set_or_return_undefined!(
-            &result,
-            &"rareCombinationsCountByLen".into(),
-            &match_or_return_undefined!(JsValue::from_serde(
-                &aggregated_result.rare_combinations_count_by_len
-            ))
-        );
-        set_or_return_undefined!(
-            &result,
-            &"combinationsCountByLen".into(),
-            &match_or_return_undefined!(JsValue::from_serde(
-                &aggregated_result.combinations_count_by_len
-            ))
-        );
-        set_or_return_undefined!(
-            &result,
-            &"combinationsSumByLen".into(),
-            &match_or_return_undefined!(JsValue::from_serde(
-                &aggregated_result.combinations_sum_by_len
-            ))
-        );
-        set_or_return_undefined!(
-            &result,
-            &"privacyRisk".into(),
-            &serialize_privacy_risk(&aggregated_result.privacy_risk)
-        );
-
-        result.into()
-    }
-}
-
-pub fn aggregate<'data_block, T: ReportProgress>(
-    data_block: &'data_block DataBlock,
-    reporting_length: usize,
-    resolution: usize,
-    progress_reporter: &mut Option<T>,
-) -> Result<AggregatedResult<'data_block>, String> {
-    let mut aggregator = Aggregator::new(data_block);
-    let aggregates_data = aggregator.aggregate(reporting_length, 0, progress_reporter);
-
-    Ok(AggregatedResult {
-        headers: &data_block.headers,
-        rare_combinations_count_by_len: aggregator
-            .calc_rare_combinations_count_by_len(&aggregates_data.aggregates_count, resolution),
-        combinations_count_by_len: aggregator
-            .calc_combinations_count_by_len(&aggregates_data.aggregates_count),
-        combinations_sum_by_len: aggregator
-            .calc_combinations_sum_by_len(&aggregates_data.aggregates_count),
-        privacy_risk: aggregator.calc_privacy_risk(&aggregates_data.aggregates_count, resolution),
-        aggregates_count: aggregates_data.aggregates_count,
-    })
-}

packages/lib-wasm/src/evaluator/mod.rs

@@ -1,170 +0,0 @@
-use js_sys::Reflect::set;
-use js_sys::{Array, Function, Object};
-use log::{error, info};
-use sds_core::processing::aggregator::typedefs::AggregatedCountByLenMap;
-use sds_core::processing::aggregator::Aggregator;
-use sds_core::processing::evaluator::typedefs::PreservationByCountBuckets;
-use sds_core::processing::evaluator::Evaluator;
-use sds_core::utils::reporting::ReportProgress;
-use sds_core::utils::time::ElapsedDurationLogger;
-use wasm_bindgen::prelude::*;
-use wasm_bindgen::JsValue;
-
-use crate::aggregator::{aggregate, AggregatedResult};
-use crate::utils::js::deserializers::{
-    deserialize_csv_data, deserialize_sensitive_zeros, deserialize_use_columns,
-};
-use crate::utils::js::js_progress_reporter::JsProgressReporter;
-use crate::utils::js::serializers::serialize_buckets;
-use crate::{match_or_return_undefined, set_or_return_undefined};
-
-struct EvaluatedResult<'data_block> {
-    sensitive_aggregated_result: AggregatedResult<'data_block>,
-    synthetic_aggregated_result: AggregatedResult<'data_block>,
-    leakage_count_by_len: AggregatedCountByLenMap,
-    fabricated_count_by_len: AggregatedCountByLenMap,
-    preservation_by_count_buckets: PreservationByCountBuckets,
-    combination_loss: f64,
-    record_expansion: f64,
-}
-
-impl<'data_block> From<EvaluatedResult<'data_block>> for JsValue {
-    #[inline]
-    fn from(evaluated_result: EvaluatedResult) -> Self {
-        let _duration_logger =
-            ElapsedDurationLogger::new(String::from("EvaluatedResult conversion to JsValue"));
-        let result = Object::new();
-
-        set_or_return_undefined!(
-            &result,
-            &"sensitiveAggregatedResult".into(),
-            &evaluated_result.sensitive_aggregated_result.into(),
-        );
-        set_or_return_undefined!(
-            &result,
-            &"syntheticAggregatedResult".into(),
-            &evaluated_result.synthetic_aggregated_result.into(),
-        );
-        set_or_return_undefined!(
-            &result,
-            &"leakageCountByLen".into(),
-            &match_or_return_undefined!(JsValue::from_serde(
-                &evaluated_result.leakage_count_by_len
-            ))
-        );
-        set_or_return_undefined!(
-            &result,
-            &"fabricatedCountByLen".into(),
-            &match_or_return_undefined!(JsValue::from_serde(
-                &evaluated_result.fabricated_count_by_len
-            ))
-        );
-        set_or_return_undefined!(
-            &result,
-            &"preservationByCountBuckets".into(),
-            &serialize_buckets(&evaluated_result.preservation_by_count_buckets)
-        );
-        set_or_return_undefined!(
-            &result,
-            &"combinationLoss".into(),
-            &match_or_return_undefined!(JsValue::from_serde(&evaluated_result.combination_loss))
-        );
-        set_or_return_undefined!(
-            &result,
-            &"recordExpansion".into(),
-            &match_or_return_undefined!(JsValue::from_serde(&evaluated_result.record_expansion))
-        );
-
-        result.into()
-    }
-}
-
-#[wasm_bindgen]
-#[allow(clippy::too_many_arguments)]
-pub fn evaluate(
-    sensitive_csv_data: Array,
-    synthetic_csv_data: Array,
-    use_columns: Array,
-    sensitive_zeros: Array,
-    record_limit: usize,
-    reporting_length: usize,
-    resolution: usize,
-    progress_callback: Function,
-) -> JsValue {
-    let _duration_logger = ElapsedDurationLogger::new(String::from("evaluation process"));
-    let use_columns_vec = match_or_return_undefined!(deserialize_use_columns(use_columns));
-    let sensitive_zeros_vec =
-        match_or_return_undefined!(deserialize_sensitive_zeros(sensitive_zeros));
-
-    info!("aggregating sensitive data...");
-    let sensitive_data_block = match_or_return_undefined!(deserialize_csv_data(
-        sensitive_csv_data,
-        &use_columns_vec,
-        &sensitive_zeros_vec,
-        record_limit,
-    ));
-    let mut sensitive_aggregated_result = match_or_return_undefined!(aggregate(
-        &sensitive_data_block,
-        reporting_length,
-        resolution,
-        &mut Some(JsProgressReporter::new(&progress_callback, &|p| p * 0.40))
-    ));
-
-    info!("aggregating synthetic data...");
-    let synthetic_data_block = match_or_return_undefined!(deserialize_csv_data(
-        synthetic_csv_data,
-        &use_columns_vec,
-        &sensitive_zeros_vec,
-        record_limit,
-    ));
-    let synthetic_aggregated_result = match_or_return_undefined!(aggregate(
-        &synthetic_data_block,
-        reporting_length,
-        resolution,
-        &mut Some(JsProgressReporter::new(&progress_callback, &|p| {
-            40.0 + (p * 0.40)
-        }))
-    ));
-    let mut evaluator_instance = Evaluator::default();
-
-    info!("evaluating synthetic data based on sensitive data...");
-
-    let buckets = evaluator_instance.calc_preservation_by_count(
-        &sensitive_aggregated_result.aggregates_count,
-        &synthetic_aggregated_result.aggregates_count,
-        resolution,
-    );
-    let leakage_count_by_len = evaluator_instance.calc_leakage_count(
-        &sensitive_aggregated_result.aggregates_count,
-        &synthetic_aggregated_result.aggregates_count,
-        resolution,
-    );
-    let fabricated_count_by_len = evaluator_instance.calc_fabricated_count(
-        &sensitive_aggregated_result.aggregates_count,
-        &synthetic_aggregated_result.aggregates_count,
-    );
-    let combination_loss = evaluator_instance.calc_combination_loss(&buckets);
-
-    Aggregator::protect_aggregates_count(
-        &mut sensitive_aggregated_result.aggregates_count,
-        resolution,
-    );
-
-    JsProgressReporter::new(&progress_callback, &|p| p).report(100.0);
-
-    EvaluatedResult {
-        leakage_count_by_len,
-        fabricated_count_by_len,
-        combination_loss,
-        preservation_by_count_buckets: buckets,
-        record_expansion: (synthetic_aggregated_result
-            .privacy_risk
-            .total_number_of_records as f64)
-            / (sensitive_aggregated_result
-                .privacy_risk
-                .total_number_of_records as f64),
-        sensitive_aggregated_result,
-        synthetic_aggregated_result,
-    }
-    .into()
-}

packages/lib-wasm/src/generator/mod.rs

@@ -1,45 +0,0 @@
-use js_sys::{Array, Function};
-use log::error;
-use sds_core::{processing::generator::Generator, utils::time::ElapsedDurationLogger};
-use wasm_bindgen::prelude::*;
-
-use crate::{
-    match_or_return_undefined,
-    utils::js::{
-        deserializers::{
-            deserialize_csv_data, deserialize_sensitive_zeros, deserialize_use_columns,
-        },
-        js_progress_reporter::JsProgressReporter,
-    },
-};
-
-#[wasm_bindgen]
-pub fn generate(
-    csv_data: Array,
-    use_columns: Array,
-    sensitive_zeros: Array,
-    record_limit: usize,
-    resolution: usize,
-    cache_size: usize,
-    progress_callback: Function,
-) -> JsValue {
-    let _duration_logger = ElapsedDurationLogger::new(String::from("generation process"));
-    let data_block = match_or_return_undefined!(deserialize_csv_data(
-        csv_data,
-        &match_or_return_undefined!(deserialize_use_columns(use_columns)),
-        &match_or_return_undefined!(deserialize_sensitive_zeros(sensitive_zeros)),
-        record_limit,
-    ));
-    let mut generator = Generator::new(&data_block);
-
-    match_or_return_undefined!(JsValue::from_serde(
-        &generator
-            .generate(
-                resolution,
-                cache_size,
-                "",
-                &mut Some(JsProgressReporter::new(&progress_callback, &|p| p))
-            )
-            .synthetic_data
-    ))
-}

packages/lib-wasm/src/lib.rs

@@ -1,141 +1,3 @@
-//! This crate will generate wasm bindings for the main features
-//! of the `sds_core` library:
-//!
-//! # Init Logger
-//! ```typescript
-//! init_logger(
-//!     level_str: string
-//! ): boolean
-//! ```
-//!
-//! Initializes logging using a particular log level.
-//! Returns `true` if successfully initialized, `false` otherwise
-//!
-//! ## Arguments:
-//! * `level_str` - String containing a valid log level
-//! (`off`, `error`, `warn`, `info`, `debug` or `trace`)
-//!
-//! # Generate
-//! ```typescript
-//! generate(
-//!     csv_data: string[][],
-//!     use_columns: string[],
-//!     sensitive_zeros: string[],
-//!     record_limit: number,
-//!     resolution: number,
-//!     cache_size: number,
-//!     progress_callback: (progress: value) => void
-//! ): string[][]
-//! ```
-//!
-//! Synthesizes the `csv_data` using the configured parameters and returns it
-//!
-//! ## Arguments
-//! * `csv_data` - Data to be synthesized
-//!     * `csv_data[0]` - Should be the headers
-//!     * `csv_data[1...]` - Should be the records
-//! * `use_columns` - Column names to be used (if `[]` use all columns)
-//! * `sensitive_zeros` - Column names containing sensitive zeros
-//! (if `[]` no columns are considered to have sensitive zeros)
-//! * `record_limit` - Use only the first `record_limit` records (if `0` use all records)
-//! * `resolution` - Reporting resolution to be used
-//! * `cache_size` - Maximum cache size used during the synthesis process
-//! * `progress_callback` - Callback that informs the processing percentage (0.0 % - 100.0 %)
-//!
-//! # Evaluate
-//! ```typescript
-//! interface IAggregatedCombination {
-//!     combination_key: number
-//!     count: number
-//!     length: number
-//! }
-//!
-//! interface IAggregatedCombinations {
-//!     [name: string]: IAggregatedCombination
-//! }
-//!
-//! interface IAggregatedCountByLen {
-//!     [length: number]: number
-//! }
-//!
-//! interface IPrivacyRiskSummary {
-//!     totalNumberOfRecords: number
-//!     totalNumberOfCombinations: number
-//!     recordsWithUniqueCombinationsCount: number
-//!     recordsWithRareCombinationsCount: number
-//!     uniqueCombinationsCount: number
-//!     rareCombinationsCount: number
-//!     recordsWithUniqueCombinationsProportion: number
-//!     recordsWithRareCombinationsProportion: number
-//!     uniqueCombinationsProportion: number
-//!     rareCombinationsProportion: number
-//! }
-//!
-//! interface IAggregatedResult {
-//!     aggregatedCombinations?: IAggregatedCombinations
-//!     rareCombinationsCountByLen?: IAggregatedCountByLen
-//!     combinationsCountByLen?: IAggregatedCountByLen
-//!     combinationsSumByLen?: IAggregatedCountByLen
-//!     privacyRisk?: IPrivacyRiskSummary
-//! }
-//!
-//! interface IPreservationByCountBucket {
-//!     size: number
-//!     preservationSum: number
-//!     lengthSum: number
-//! }
-//!
-//! interface IPreservationByCountBuckets {
-//!     [bucket_index: number]: IPreservationByCountBucket
-//! }
-//!
-//! interface IEvaluatedResult {
-//!     sensitiveAggregatedResult?: IAggregatedResult
-//!     syntheticAggregatedResult?: IAggregatedResult
-//!     leakageCountByLen?: IAggregatedCountByLen
-//!     fabricatedCountByLen?: IAggregatedCountByLen
-//!     preservationByCountBuckets?: IPreservationByCountBuckets
-//!     combinationLoss?: number
-//!     recordExpansion?: number
-//! }
-//!
-//! evaluate(
-//!     sensitive_csv_data: string[][],
-//!     synthetic_csv_data: string[][],
-//!     use_columns: string[],
-//!     sensitive_zeros: string[],
-//!     record_limit: number,
-//!     reporting_length: number,
-//!     resolution: number,
-//!     progress_callback: (progress: value) => void
-//! ): IEvaluatedResult
-//! ```
-//! Evaluates the synthetic data based on the sensitive data and produces a `IEvaluatedResult`
-//!
-//! ## Arguments
-//! * `sensitive_csv_data` - Sensitive data to be evaluated
-//!     * `sensitive_csv_data[0]` - Should be the headers
-//!     * `sensitive_csv_data[1...]` - Should be the records
-//! * `synthetic_csv_data` - Synthetic data produced from the synthetic data
-//!     * `synthetic_csv_data[0]` - Should be the headers
-//!     * `synthetic_csv_data[1...]` - Should be the records
-//! * `use_columns` - Column names to be used (if `[]` use all columns)
-//! * `sensitive_zeros` - Column names containing sensitive zeros
-//! (if `[]` no columns are considered to have sensitive zeros)
-//! * `record_limit` - Use only the first `record_limit` records (if `0` use all records)
-//! * `reporting_length` - Maximum length to compute attribute combinations
-//! for analysis
-//! * `resolution` - Reporting resolution to be used
-//! * `progress_callback` - Callback that informs the processing percentage (0.0 % - 100.0 %)
+pub mod processing;
 
-#[doc(hidden)]
-pub mod aggregator;
-
-#[doc(hidden)]
-pub mod evaluator;
-
-#[doc(hidden)]
-pub mod generator;
-
-#[doc(hidden)]
 pub mod utils;

packages/lib-wasm/src/processing/aggregator/aggregate_count_and_length.rs

@@ -0,0 +1,31 @@
+use serde::Serialize;
+use std::convert::TryFrom;
+use wasm_bindgen::{JsCast, JsValue};
+
+use crate::utils::js::ts_definitions::JsAggregateCountAndLength;
+
+#[derive(Serialize)]
+pub struct WasmAggregateCountAndLength {
+    count: usize,
+    length: usize,
+}
+
+impl WasmAggregateCountAndLength {
+    #[inline]
+    pub fn new(count: usize, length: usize) -> WasmAggregateCountAndLength {
+        WasmAggregateCountAndLength { count, length }
+    }
+}
+
+impl TryFrom<WasmAggregateCountAndLength> for JsAggregateCountAndLength {
+    type Error = JsValue;
+
+    #[inline]
+    fn try_from(
+        aggregate_count_and_length: WasmAggregateCountAndLength,
+    ) -> Result<Self, Self::Error> {
+        JsValue::from_serde(&aggregate_count_and_length)
+            .map_err(|err| JsValue::from(err.to_string()))
+            .map(|c| c.unchecked_into())
+    }
+}

packages/lib-wasm/src/processing/aggregator/aggregate_result.rs

@@ -0,0 +1,228 @@
+use super::aggregate_count_and_length::WasmAggregateCountAndLength;
+use js_sys::{Object, Reflect::set};
+use sds_core::{
+    processing::aggregator::aggregated_data::AggregatedData, utils::time::ElapsedDurationLogger,
+};
+use std::{
+    convert::TryFrom,
+    ops::{Deref, DerefMut},
+};
+use wasm_bindgen::{prelude::*, JsCast};
+
+use crate::utils::js::ts_definitions::{
+    JsAggregateCountAndLength, JsAggregateCountByLen, JsAggregateResult, JsAggregatesCount,
+    JsPrivacyRiskSummary, JsResult,
+};
+
+#[wasm_bindgen]
+pub struct WasmAggregateResult {
+    aggregated_data: AggregatedData,
+}
+
+impl WasmAggregateResult {
+    #[inline]
+    pub fn default() -> WasmAggregateResult {
+        WasmAggregateResult {
+            aggregated_data: AggregatedData::default(),
+        }
+    }
+
+    #[inline]
+    pub fn new(aggregated_data: AggregatedData) -> WasmAggregateResult {
+        WasmAggregateResult { aggregated_data }
+    }
+}
+
+#[wasm_bindgen]
+impl WasmAggregateResult {
+    #[wasm_bindgen(getter)]
+    #[wasm_bindgen(js_name = "reportingLength")]
+    pub fn reporting_length(&self) -> usize {
+        self.aggregated_data.reporting_length
+    }
+
+    #[wasm_bindgen(js_name = "aggregatesCountToJs")]
+    pub fn aggregates_count_to_js(
+        &self,
+        combination_delimiter: &str,
+    ) -> JsResult<JsAggregatesCount> {
+        let result = Object::new();
+
+        for (agg, count) in self.aggregated_data.aggregates_count.iter() {
+            set(
+                &result,
+                &agg.format_str_using_headers(
+                    &self.aggregated_data.data_block.headers,
+                    combination_delimiter,
+                )
+                .into(),
+                &JsAggregateCountAndLength::try_from(WasmAggregateCountAndLength::new(
+                    count.count,
+                    agg.len(),
+                ))?
+                .into(),
+            )?;
+        }
+
+        Ok(result.unchecked_into::<JsAggregatesCount>())
+    }
+
+    #[wasm_bindgen(js_name = "rareCombinationsCountByLenToJs")]
+    pub fn rare_combinations_count_by_len_to_js(
+        &self,
+        resolution: usize,
+    ) -> JsResult<JsAggregateCountByLen> {
+        let count = self
+            .aggregated_data
+            .calc_rare_combinations_count_by_len(resolution);
+
+        Ok(JsValue::from_serde(&count)
+            .map_err(|err| JsValue::from(err.to_string()))?
+            .unchecked_into::<JsAggregateCountByLen>())
+    }
+
+    #[wasm_bindgen(js_name = "combinationsCountByLenToJs")]
+    pub fn combinations_count_by_len_to_js(&self) -> JsResult<JsAggregateCountByLen> {
+        let count = self.aggregated_data.calc_combinations_count_by_len();
+
+        Ok(JsValue::from_serde(&count)
+            .map_err(|err| JsValue::from(err.to_string()))?
+            .unchecked_into::<JsAggregateCountByLen>())
+    }
+
+    #[wasm_bindgen(js_name = "combinationsSumByLenToJs")]
+    pub fn combinations_sum_by_len_to_js(&self) -> JsResult<JsAggregateCountByLen> {
+        let count = self.aggregated_data.calc_combinations_sum_by_len();
+
+        Ok(JsValue::from_serde(&count)
+            .map_err(|err| JsValue::from(err.to_string()))?
+            .unchecked_into::<JsAggregateCountByLen>())
+    }
+
+    #[wasm_bindgen(js_name = "privacyRiskToJs")]
+    pub fn privacy_risk_to_js(&self, resolution: usize) -> JsResult<JsPrivacyRiskSummary> {
+        let pr = self.aggregated_data.calc_privacy_risk(resolution);
+        let result = Object::new();
+
+        set(
+            &result,
+            &"totalNumberOfRecords".into(),
+            &pr.total_number_of_records.into(),
+        )?;
+        set(
+            &result,
+            &"totalNumberOfCombinations".into(),
+            &pr.total_number_of_combinations.into(),
+        )?;
+        set(
+            &result,
+            &"recordsWithUniqueCombinationsCount".into(),
+            &pr.records_with_unique_combinations_count.into(),
+        )?;
+        set(
+            &result,
+            &"recordsWithRareCombinationsCount".into(),
+            &pr.records_with_rare_combinations_count.into(),
+        )?;
+        set(
+            &result,
+            &"uniqueCombinationsCount".into(),
+            &pr.unique_combinations_count.into(),
+        )?;
+        set(
+            &result,
+            &"rareCombinationsCount".into(),
+            &pr.rare_combinations_count.into(),
+        )?;
+        set(
+            &result,
+            &"recordsWithUniqueCombinationsProportion".into(),
+            &pr.records_with_unique_combinations_proportion.into(),
+        )?;
+        set(
+            &result,
+            &"recordsWithRareCombinationsProportion".into(),
+            &pr.records_with_rare_combinations_proportion.into(),
+        )?;
+        set(
+            &result,
+            &"uniqueCombinationsProportion".into(),
+            &pr.unique_combinations_proportion.into(),
+        )?;
+        set(
+            &result,
+            &"rareCombinationsProportion".into(),
+            &pr.rare_combinations_proportion.into(),
+        )?;
+
+        Ok(result.unchecked_into::<JsPrivacyRiskSummary>())
+    }
+
+    #[wasm_bindgen(js_name = "toJs")]
+    pub fn to_js(
+        &self,
+        combination_delimiter: &str,
+        resolution: usize,
+        include_aggregates_count: bool,
+    ) -> JsResult<JsAggregateResult> {
+        let _duration_logger =
+            ElapsedDurationLogger::new(String::from("aggregate result serialization"));
+        let result = Object::new();
+
+        set(
+            &result,
+            &"reportingLength".into(),
+            &self.reporting_length().into(),
+        )?;
+        if include_aggregates_count {
+            set(
+                &result,
+                &"aggregatesCount".into(),
+                &self.aggregates_count_to_js(combination_delimiter)?.into(),
+            )?;
+        }
+        set(
+            &result,
+            &"rareCombinationsCountByLen".into(),
+            &self
+                .rare_combinations_count_by_len_to_js(resolution)?
+                .into(),
+        )?;
+        set(
+            &result,
+            &"combinationsCountByLen".into(),
+            &self.combinations_count_by_len_to_js()?.into(),
+        )?;
+        set(
+            &result,
+            &"combinationsSumByLen".into(),
+            &self.combinations_sum_by_len_to_js()?.into(),
+        )?;
+        set(
+            &result,
+            &"privacyRisk".into(),
+            &self.privacy_risk_to_js(resolution)?.into(),
+        )?;
+
+        Ok(JsValue::from(result).unchecked_into::<JsAggregateResult>())
+    }
+
+    #[wasm_bindgen(js_name = "protectAggregatesCount")]
+    pub fn protect_aggregates_count(&mut self, resolution: usize) {
+        self.aggregated_data.protect_aggregates_count(resolution)
+    }
+}
+
+impl Deref for WasmAggregateResult {
+    type Target = AggregatedData;
+
+    fn deref(&self) -> &Self::Target {
+        &self.aggregated_data
+    }
+}
+
+impl DerefMut for WasmAggregateResult {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.aggregated_data
+    }
+}

packages/lib-wasm/src/processing/aggregator/mod.rs

@@ -0,0 +1,3 @@
+pub mod aggregate_count_and_length;
+
+pub mod aggregate_result;

packages/lib-wasm/src/processing/evaluator/evaluate_result.rs

@@ -0,0 +1,184 @@
+use super::preservation_by_count::WasmPreservationByCount;
+use js_sys::{Object, Reflect::set};
+use sds_core::{processing::evaluator::Evaluator, utils::time::ElapsedDurationLogger};
+use wasm_bindgen::{prelude::*, JsCast};
+
+use crate::{
+    processing::aggregator::aggregate_result::WasmAggregateResult,
+    utils::js::ts_definitions::{
+        JsAggregateCountByLen, JsAggregateResult, JsEvaluateResult, JsResult,
+    },
+};
+
+#[wasm_bindgen]
+pub struct WasmEvaluateResult {
+    pub(crate) sensitive_aggregate_result: WasmAggregateResult,
+    pub(crate) synthetic_aggregate_result: WasmAggregateResult,
+    evaluator: Evaluator,
+}
+
+impl WasmEvaluateResult {
+    #[inline]
+    pub fn default() -> WasmEvaluateResult {
+        WasmEvaluateResult {
+            sensitive_aggregate_result: WasmAggregateResult::default(),
+            synthetic_aggregate_result: WasmAggregateResult::default(),
+            evaluator: Evaluator::default(),
+        }
+    }
+
+    #[inline]
+    pub fn new(
+        sensitive_aggregate_result: WasmAggregateResult,
+        synthetic_aggregate_result: WasmAggregateResult,
+    ) -> WasmEvaluateResult {
+        WasmEvaluateResult {
+            sensitive_aggregate_result,
+            synthetic_aggregate_result,
+            evaluator: Evaluator::default(),
+        }
+    }
+}
+
+#[wasm_bindgen]
+impl WasmEvaluateResult {
+    #[wasm_bindgen(constructor)]
+    pub fn from_aggregate_results(
+        sensitive_aggregate_result: WasmAggregateResult,
+        synthetic_aggregate_result: WasmAggregateResult,
+    ) -> WasmEvaluateResult {
+        WasmEvaluateResult::new(sensitive_aggregate_result, synthetic_aggregate_result)
+    }
+
+    #[wasm_bindgen(js_name = "sensitiveAggregateResultToJs")]
+    pub fn sensitive_aggregate_result_to_js(
+        &self,
+        combination_delimiter: &str,
+        resolution: usize,
+        include_aggregates_count: bool,
+    ) -> JsResult<JsAggregateResult> {
+        self.sensitive_aggregate_result.to_js(
+            combination_delimiter,
+            resolution,
+            include_aggregates_count,
+        )
+    }
+
+    #[wasm_bindgen(js_name = "syntheticAggregateResultToJs")]
+    pub fn synthetic_aggregate_result_to_js(
+        &self,
+        combination_delimiter: &str,
+        resolution: usize,
+        include_aggregates_count: bool,
+    ) -> JsResult<JsAggregateResult> {
+        self.synthetic_aggregate_result.to_js(
+            combination_delimiter,
+            resolution,
+            include_aggregates_count,
+        )
+    }
+
+    #[wasm_bindgen(js_name = "leakageCountByLenToJs")]
+    pub fn leakage_count_by_len_to_js(&self, resolution: usize) -> JsResult<JsAggregateCountByLen> {
+        let count = self.evaluator.calc_leakage_count(
+            &self.sensitive_aggregate_result,
+            &self.synthetic_aggregate_result,
+            resolution,
+        );
+
+        Ok(JsValue::from_serde(&count)
+            .map_err(|err| JsValue::from(err.to_string()))?
+            .unchecked_into::<JsAggregateCountByLen>())
+    }
+
+    #[wasm_bindgen(js_name = "fabricatedCountByLenToJs")]
+    pub fn fabricated_count_by_len_to_js(&self) -> JsResult<JsAggregateCountByLen> {
+        let count = self.evaluator.calc_fabricated_count(
+            &self.sensitive_aggregate_result,
+            &self.synthetic_aggregate_result,
+        );
+
+        Ok(JsValue::from_serde(&count)
+            .map_err(|err| JsValue::from(err.to_string()))?
+            .unchecked_into::<JsAggregateCountByLen>())
+    }
+
+    #[wasm_bindgen(js_name = "preservationByCount")]
+    pub fn preservation_by_count(&self, resolution: usize) -> WasmPreservationByCount {
+        WasmPreservationByCount::new(self.evaluator.calc_preservation_by_count(
+            &self.sensitive_aggregate_result,
+            &self.synthetic_aggregate_result,
+            resolution,
+        ))
+    }
+
+    #[wasm_bindgen(js_name = "recordExpansion")]
+    pub fn record_expansion(&self) -> f64 {
+        (self
+            .synthetic_aggregate_result
+            .data_block
+            .number_of_records() as f64)
+            / (self
+                .sensitive_aggregate_result
+                .data_block
+                .number_of_records() as f64)
+    }
+
+    #[wasm_bindgen(js_name = "toJs")]
+    pub fn to_js(
+        &self,
+        combination_delimiter: &str,
+        resolution: usize,
+        include_aggregates_count: bool,
+    ) -> JsResult<JsEvaluateResult> {
+        let _duration_logger =
+            ElapsedDurationLogger::new(String::from("evaluate result serialization"));
+        let result = Object::new();
+
+        set(
+            &result,
+            &"sensitiveAggregateResult".into(),
+            &self
+                .sensitive_aggregate_result_to_js(
+                    combination_delimiter,
+                    resolution,
+                    include_aggregates_count,
+                )?
+                .into(),
+        )?;
+        set(
+            &result,
+            &"syntheticAggregateResult".into(),
+            &self
+                .synthetic_aggregate_result_to_js(
+                    combination_delimiter,
+                    resolution,
+                    include_aggregates_count,
+                )?
+                .into(),
+        )?;
+
+        set(
+            &result,
+            &"leakageCountByLen".into(),
+            &self.leakage_count_by_len_to_js(resolution)?.into(),
+        )?;
+        set(
+            &result,
+            &"fabricatedCountByLen".into(),
+            &self.fabricated_count_by_len_to_js()?.into(),
+        )?;
+        set(
+            &result,
+            &"preservationByCount".into(),
+            &self.preservation_by_count(resolution).to_js()?.into(),
+        )?;
+        set(
+            &result,
+            &"recordExpansion".into(),
+            &self.record_expansion().into(),
+        )?;
+
+        Ok(JsValue::from(result).unchecked_into::<JsEvaluateResult>())
+    }
+}

packages/lib-wasm/src/processing/evaluator/mod.rs

@@ -0,0 +1,3 @@
+pub mod preservation_by_count;
+
+pub mod evaluate_result;

packages/lib-wasm/src/processing/evaluator/preservation_by_count.rs

@@ -0,0 +1,93 @@
+use js_sys::{Object, Reflect::set};
+use sds_core::{
+    processing::evaluator::preservation_by_count::PreservationByCountBuckets,
+    utils::time::ElapsedDurationLogger,
+};
+use std::ops::{Deref, DerefMut};
+use wasm_bindgen::{prelude::*, JsCast};
+
+use crate::utils::js::ts_definitions::{
+    JsPreservationByCount, JsPreservationByCountBuckets, JsResult,
+};
+
+#[wasm_bindgen]
+pub struct WasmPreservationByCount {
+    preservation_by_count_buckets: PreservationByCountBuckets,
+}
+
+impl WasmPreservationByCount {
+    #[inline]
+    pub fn new(
+        preservation_by_count_buckets: PreservationByCountBuckets,
+    ) -> WasmPreservationByCount {
+        WasmPreservationByCount {
+            preservation_by_count_buckets,
+        }
+    }
+}
+
+#[wasm_bindgen]
+impl WasmPreservationByCount {
+    #[wasm_bindgen(js_name = "bucketsToJs")]
+    pub fn buckets_to_js(&self) -> JsResult<JsPreservationByCountBuckets> {
+        let result = Object::new();
+
+        for (bucket_index, b) in self.preservation_by_count_buckets.iter() {
+            let serialized_bucket = Object::new();
+
+            set(&serialized_bucket, &"size".into(), &b.size.into())?;
+            set(
+                &serialized_bucket,
+                &"preservationSum".into(),
+                &b.preservation_sum.into(),
+            )?;
+            set(
+                &serialized_bucket,
+                &"lengthSum".into(),
+                &b.length_sum.into(),
+            )?;
+            set(
+                &serialized_bucket,
+                &"combinationCountSum".into(),
+                &b.combination_count_sum.into(),
+            )?;
+
+            set(&result, &(*bucket_index).into(), &serialized_bucket.into())?;
+        }
+        Ok(result.unchecked_into::<JsPreservationByCountBuckets>())
+    }
+
+    #[wasm_bindgen(js_name = "combinationLoss")]
+    pub fn combination_loss(&self) -> f64 {
+        self.preservation_by_count_buckets.calc_combination_loss()
+    }
+
+    #[wasm_bindgen(js_name = "toJs")]
+    pub fn to_js(&self) -> JsResult<JsPreservationByCount> {
+        let _duration_logger =
+            ElapsedDurationLogger::new(String::from("preservation by count serialization"));
+        let result = Object::new();
+
+        set(&result, &"buckets".into(), &self.buckets_to_js()?.into())?;
+        set(
+            &result,
+            &"combinationLoss".into(),
+            &self.combination_loss().into(),
+        )?;
+        Ok(JsValue::from(result).unchecked_into::<JsPreservationByCount>())
+    }
+}
+
+impl Deref for WasmPreservationByCount {
+    type Target = PreservationByCountBuckets;
+
+    fn deref(&self) -> &Self::Target {
+        &self.preservation_by_count_buckets
+    }
+}
+
+impl DerefMut for WasmPreservationByCount {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.preservation_by_count_buckets
+    }
+}

packages/lib-wasm/src/processing/generator/generate_result.rs

@@ -0,0 +1,77 @@
+use js_sys::{Object, Reflect::set};
+use sds_core::{
+    processing::generator::generated_data::GeneratedData, utils::time::ElapsedDurationLogger,
+};
+use std::ops::{Deref, DerefMut};
+use wasm_bindgen::{prelude::*, JsCast};
+
+use crate::utils::js::ts_definitions::{JsCsvData, JsGenerateResult, JsResult};
+
+#[wasm_bindgen]
+pub struct WasmGenerateResult {
+    generated_data: GeneratedData,
+}
+
+impl WasmGenerateResult {
+    #[inline]
+    pub fn default() -> WasmGenerateResult {
+        WasmGenerateResult {
+            generated_data: GeneratedData::default(),
+        }
+    }
+
+    #[inline]
+    pub fn new(generated_data: GeneratedData) -> WasmGenerateResult {
+        WasmGenerateResult { generated_data }
+    }
+}
+
+#[wasm_bindgen]
+impl WasmGenerateResult {
+    #[wasm_bindgen(getter)]
+    #[wasm_bindgen(js_name = "expansionRatio")]
+    pub fn expansion_ratio(&self) -> f64 {
+        self.generated_data.expansion_ratio
+    }
+
+    #[wasm_bindgen(js_name = "syntheticDataToJs")]
+    pub fn synthetic_data_to_js(&self) -> JsResult<JsCsvData> {
+        Ok(JsValue::from_serde(&self.generated_data.synthetic_data)
+            .map_err(|err| JsValue::from(err.to_string()))?
+            .unchecked_into::<JsCsvData>())
+    }
+
+    #[wasm_bindgen(js_name = "toJs")]
+    pub fn to_js(&self) -> JsResult<JsGenerateResult> {
+        let _duration_logger =
+            ElapsedDurationLogger::new(String::from("generate result serialization"));
+        let result = Object::new();
+
+        set(
+            &result,
+            &"expansionRatio".into(),
+            &self.expansion_ratio().into(),
+        )?;
+        set(
+            &result,
+            &"syntheticData".into(),
+            &self.synthetic_data_to_js()?.into(),
+        )?;
+
+        Ok(JsValue::from(result).unchecked_into::<JsGenerateResult>())
+    }
+}
+
+impl Deref for WasmGenerateResult {
+    type Target = GeneratedData;
+
+    fn deref(&self) -> &Self::Target {
+        &self.generated_data
+    }
+}
+
+impl DerefMut for WasmGenerateResult {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.generated_data
+    }
+}

packages/lib-wasm/src/processing/generator/mod.rs

@@ -0,0 +1 @@
+pub mod generate_result;

packages/lib-wasm/src/processing/mod.rs

@@ -0,0 +1,11 @@
+pub mod aggregator;
+
+pub mod evaluator;
+
+pub mod generator;
+
+pub mod navigator;
+
+pub mod sds_processor;
+
+pub mod sds_context;

packages/lib-wasm/src/processing/navigator/attributes_intersection.rs

@@ -0,0 +1,75 @@
+use js_sys::{Array, Object, Reflect::set};
+use std::{collections::HashMap, convert::TryFrom, sync::Arc};
+use wasm_bindgen::{JsCast, JsValue};
+
+use crate::utils::js::ts_definitions::{
+    JsAttributesIntersection, JsAttributesIntersectionByColumn,
+};
+
+pub struct WasmAttributesIntersection {
+    pub(crate) value: Arc<String>,
+    pub(crate) estimated_count: usize,
+    pub(crate) actual_count: Option<usize>,
+}
+
+impl WasmAttributesIntersection {
+    #[inline]
+    pub fn new(
+        value: Arc<String>,
+        estimated_count: usize,
+        actual_count: Option<usize>,
+    ) -> WasmAttributesIntersection {
+        WasmAttributesIntersection {
+            value,
+            estimated_count,
+            actual_count,
+        }
+    }
+}
+
+impl TryFrom<WasmAttributesIntersection> for JsAttributesIntersection {
+    type Error = JsValue;
+
+    fn try_from(attributes_intersection: WasmAttributesIntersection) -> Result<Self, Self::Error> {
+        let result = Object::new();
+
+        set(
+            &result,
+            &"value".into(),
+            &attributes_intersection.value.as_str().into(),
+        )?;
+        set(
+            &result,
+            &"estimatedCount".into(),
+            &attributes_intersection.estimated_count.into(),
+        )?;
+        if let Some(actual_count) = attributes_intersection.actual_count {
+            set(&result, &"actualCount".into(), &actual_count.into())?;
+        }
+
+        Ok(result.unchecked_into::<JsAttributesIntersection>())
+    }
+}
+
+pub type WasmAttributesIntersectionByColumn = HashMap<usize, Vec<WasmAttributesIntersection>>;
+
+impl TryFrom<WasmAttributesIntersectionByColumn> for JsAttributesIntersectionByColumn {
+    type Error = JsValue;
+
+    fn try_from(
+        mut intersections_by_column: WasmAttributesIntersectionByColumn,
+    ) -> Result<Self, Self::Error> {
+        let result = Object::new();
+
+        for (column_index, mut attrs_intersections) in intersections_by_column.drain() {
+            let intersections: Array = Array::default();
+
+            for intersection in attrs_intersections.drain(..) {
+                intersections.push(&JsAttributesIntersection::try_from(intersection)?.into());
+            }
+            set(&result, &column_index.into(), &intersections)?;
+        }
+
+        Ok(result.unchecked_into::<JsAttributesIntersectionByColumn>())
+    }
+}

packages/lib-wasm/src/processing/navigator/mod.rs

@@ -0,0 +1,5 @@
+pub mod attributes_intersection;
+
+pub mod navigate_result;
+
+pub mod selected_attributes;

packages/lib-wasm/src/processing/navigator/navigate_result.rs

@@ -0,0 +1,229 @@
+use super::{
+    attributes_intersection::{WasmAttributesIntersection, WasmAttributesIntersectionByColumn},
+    selected_attributes::{WasmSelectedAttributes, WasmSelectedAttributesByColumn},
+};
+use sds_core::{
+    data_block::{
+        block::DataBlock,
+        typedefs::{
+            AttributeRows, AttributeRowsByColumnMap, AttributeRowsMap, ColumnIndexByName, CsvRecord,
+        },
+        value::DataBlockValue,
+    },
+    processing::aggregator::value_combination::ValueCombination,
+    utils::collections::ordered_vec_intersection,
+};
+use std::{cmp::Reverse, convert::TryFrom, sync::Arc};
+use wasm_bindgen::prelude::*;
+
+use crate::{
+    processing::{aggregator::aggregate_result::WasmAggregateResult, sds_processor::SDSProcessor},
+    utils::js::ts_definitions::{
+        JsAttributesIntersectionByColumn, JsHeaderNames, JsResult, JsSelectedAttributesByColumn,
+    },
+};
+
+#[wasm_bindgen]
+pub struct WasmNavigateResult {
+    synthetic_data_block: Arc<DataBlock>,
+    attr_rows_by_column: AttributeRowsByColumnMap,
+    selected_attributes: WasmSelectedAttributesByColumn,
+    selected_attr_rows: AttributeRows,
+    all_attr_rows: AttributeRows,
+    column_index_by_name: ColumnIndexByName,
+}
+
+impl WasmNavigateResult {
+    #[inline]
+    pub fn default() -> WasmNavigateResult {
+        WasmNavigateResult::new(
+            Arc::new(DataBlock::default()),
+            AttributeRowsByColumnMap::default(),
+            WasmSelectedAttributesByColumn::default(),
+            AttributeRows::default(),
+            AttributeRows::default(),
+            ColumnIndexByName::default(),
+        )
+    }
+
+    #[inline]
+    pub fn new(
+        synthetic_data_block: Arc<DataBlock>,
+        attr_rows_by_column: AttributeRowsByColumnMap,
+        selected_attributes: WasmSelectedAttributesByColumn,
+        selected_attr_rows: AttributeRows,
+        all_attr_rows: AttributeRows,
+        column_index_by_name: ColumnIndexByName,
+    ) -> WasmNavigateResult {
+        WasmNavigateResult {
+            synthetic_data_block,
+            attr_rows_by_column,
+            selected_attributes,
+            selected_attr_rows,
+            all_attr_rows,
+            column_index_by_name,
+        }
+    }
+
+    #[inline]
+    fn intersect_attributes(&self, attributes: &WasmSelectedAttributesByColumn) -> AttributeRows {
+        let default_empty_attr_rows = AttributeRowsMap::default();
+        let default_empty_rows = AttributeRows::default();
+        let mut result = self.all_attr_rows.clone();
+
+        for (column_index, values) in attributes.iter() {
+            let attr_rows = self
+                .attr_rows_by_column
+                .get(column_index)
+                .unwrap_or(&default_empty_attr_rows);
+            for value in values.iter() {
+                result = ordered_vec_intersection(
+                    &result,
+                    attr_rows.get(value).unwrap_or(&default_empty_rows),
+                );
+            }
+        }
+        result
+    }
+
+    #[inline]
+    fn get_sensitive_aggregates_count(
+        &self,
+        attributes: &WasmSelectedAttributesByColumn,
+        sensitive_aggregate_result: &WasmAggregateResult,
+    ) -> Option<usize> {
+        let mut combination: Vec<Arc<DataBlockValue>> = attributes
+            .values()
+            .flat_map(|values| values.iter().cloned())
+            .collect();
+
+        combination.sort_by_key(|k| k.format_str_using_headers(&self.synthetic_data_block.headers));
+
+        sensitive_aggregate_result
+            .aggregates_count
+            .get(&ValueCombination::new(combination))
+            .map(|c| c.count)
+    }
+
+    #[inline]
+    fn get_selected_but_current_column(
+        &self,
+        column_index: usize,
+    ) -> (WasmSelectedAttributesByColumn, AttributeRows) {
+        let mut selected_attributes_but_current_column = self.selected_attributes.clone();
+
+        selected_attributes_but_current_column.remove(&column_index);
+
+        let selected_attr_rows_but_current_column =
+            self.intersect_attributes(&selected_attributes_but_current_column);
+
+        (
+            selected_attributes_but_current_column,
+            selected_attr_rows_but_current_column,
+        )
+    }
+
+    #[inline]
+    fn selected_attributes_contains_value(
+        &self,
+        column_index: usize,
+        value: &Arc<DataBlockValue>,
+    ) -> bool {
+        self.selected_attributes
+            .get(&column_index)
+            .map(|values| values.contains(value))
+            .unwrap_or(false)
+    }
+}
+
+#[wasm_bindgen]
+impl WasmNavigateResult {
+    #[wasm_bindgen(constructor)]
+    pub fn from_synthetic_processor(synthetic_processor: &SDSProcessor) -> WasmNavigateResult {
+        WasmNavigateResult::new(
+            synthetic_processor.data_block.clone(),
+            synthetic_processor
+                .data_block
+                .calc_attr_rows_with_no_empty_values(),
+            WasmSelectedAttributesByColumn::default(),
+            AttributeRows::default(),
+            (0..synthetic_processor.data_block.number_of_records()).collect(),
+            synthetic_processor.data_block.calc_column_index_by_name(),
+        )
+    }
+
+    #[wasm_bindgen(js_name = "selectAttributes")]
+    pub fn select_attributes(&mut self, attributes: JsSelectedAttributesByColumn) -> JsResult<()> {
+        self.selected_attributes = WasmSelectedAttributesByColumn::try_from(attributes)?;
+        self.selected_attr_rows = self.intersect_attributes(&self.selected_attributes);
+        Ok(())
+    }
+
+    #[wasm_bindgen(js_name = "attributesIntersectionsByColumn")]
+    pub fn attributes_intersections_by_column(
+        &mut self,
+        columns: JsHeaderNames,
+        sensitive_aggregate_result: &WasmAggregateResult,
+    ) -> JsResult<JsAttributesIntersectionByColumn> {
+        let column_indexes: AttributeRows = CsvRecord::try_from(columns)?
+            .iter()
+            .filter_map(|header_name| self.column_index_by_name.get(header_name))
+            .cloned()
+            .collect();
+        let mut result = WasmAttributesIntersectionByColumn::default();
+
+        for column_index in column_indexes {
+            let result_column_entry = result.entry(column_index).or_insert_with(Vec::default);
+            if let Some(attr_rows) = self.attr_rows_by_column.get(&column_index) {
+                let (selected_attributes_but_current_column, selected_attr_rows_but_current_column) =
+                    self.get_selected_but_current_column(column_index);
+
+                for (value, rows) in attr_rows {
+                    let current_selected_attr_rows;
+                    let current_selected_attributes;
+
+                    if self.selected_attributes_contains_value(column_index, value) {
+                        // if the selected attributes contain the current value, we
+                        // have already the selected attr rows cached
+                        current_selected_attributes = &self.selected_attributes;
+                        current_selected_attr_rows = &self.selected_attr_rows;
+                    } else {
+                        // if the selected attributes do not contain the current value
+                        // use the intersection between all selected values, but the
+                        // ones in the current column
+                        current_selected_attributes = &selected_attributes_but_current_column;
+                        current_selected_attr_rows = &selected_attr_rows_but_current_column;
+                    };
+
+                    let estimated_count =
+                        ordered_vec_intersection(current_selected_attr_rows, rows).len();
+
+                    if estimated_count > 0 {
+                        let mut attributes = current_selected_attributes.clone();
+                        let attributes_entry = attributes
+                            .entry(column_index)
+                            .or_insert_with(WasmSelectedAttributes::default);
+
+                        attributes_entry.insert(value.clone());
+
+                        result_column_entry.push(WasmAttributesIntersection::new(
+                            value.value.clone(),
+                            estimated_count,
+                            self.get_sensitive_aggregates_count(
+                                &attributes,
+                                sensitive_aggregate_result,
+                            ),
+                        ));
+                    }
+                }
+            }
+        }
+
+        // sort by estimated count in descending order
+        for intersections in result.values_mut() {
+            intersections.sort_by_key(|intersection| Reverse(intersection.estimated_count))
+        }
+
+        JsAttributesIntersectionByColumn::try_from(result)
+    }
+}

packages/lib-wasm/src/processing/navigator/selected_attributes.rs

@@ -0,0 +1,48 @@
+use js_sys::{Array, Object, Set};
+use sds_core::data_block::value::DataBlockValue;
+use std::{
+    collections::{HashMap, HashSet},
+    convert::TryFrom,
+    sync::Arc,
+};
+use wasm_bindgen::{JsCast, JsValue};
+
+use crate::utils::js::ts_definitions::JsSelectedAttributesByColumn;
+
+pub type WasmSelectedAttributes = HashSet<Arc<DataBlockValue>>;
+
+pub type WasmSelectedAttributesByColumn = HashMap<usize, WasmSelectedAttributes>;
+
+impl TryFrom<JsSelectedAttributesByColumn> for WasmSelectedAttributesByColumn {
+    type Error = JsValue;
+
+    fn try_from(js_selected_attributes: JsSelectedAttributesByColumn) -> Result<Self, Self::Error> {
+        let mut result = WasmSelectedAttributesByColumn::default();
+
+        for entry_res in Object::entries(&js_selected_attributes.dyn_into::<Object>()?).values() {
+            let entry = entry_res?.dyn_into::<Array>()?;
+            let column_index = entry
+                .get(0)
+                .as_string()
+                .ok_or_else(|| {
+                    JsValue::from("invalid column index on selected attributes by column")
+                })?
+                .parse::<usize>()
+                .map_err(|err| JsValue::from(err.to_string()))?;
+            let result_entry = result.entry(column_index).or_insert_with(HashSet::default);
+
+            for value_res in entry.get(1).dyn_into::<Set>()?.keys() {
+                let value = value_res?;
+                let value_str = value.as_string().ok_or_else(|| {
+                    JsValue::from("invalid value on selected attributes by column")
+                })?;
+
+                result_entry.insert(Arc::new(DataBlockValue::new(
+                    column_index,
+                    Arc::new(value_str),
+                )));
+            }
+        }
+        Ok(result)
+    }
+}

packages/lib-wasm/src/processing/sds_context.rs

@@ -0,0 +1,209 @@
+use super::{
+    evaluator::evaluate_result::WasmEvaluateResult, generator::generate_result::WasmGenerateResult,
+    navigator::navigate_result::WasmNavigateResult, sds_processor::SDSProcessor,
+};
+use log::debug;
+use wasm_bindgen::{prelude::*, JsCast};
+
+use crate::utils::js::ts_definitions::{
+    JsAttributesIntersectionByColumn, JsCsvData, JsEvaluateResult, JsGenerateResult, JsHeaderNames,
+    JsReportProgressCallback, JsResult, JsSelectedAttributesByColumn,
+};
+
+#[wasm_bindgen]
+pub struct SDSContext {
+    use_columns: JsHeaderNames,
+    sensitive_zeros: JsHeaderNames,
+    record_limit: usize,
+    sensitive_processor: SDSProcessor,
+    generate_result: WasmGenerateResult,
+    resolution: usize,
+    synthetic_processor: SDSProcessor,
+    evaluate_result: WasmEvaluateResult,
+    navigate_result: WasmNavigateResult,
+}
+
+#[wasm_bindgen]
+impl SDSContext {
+    #[wasm_bindgen(constructor)]
+    pub fn default() -> SDSContext {
+        SDSContext {
+            use_columns: JsHeaderNames::default(),
+            sensitive_zeros: JsHeaderNames::default(),
+            record_limit: 0,
+            sensitive_processor: SDSProcessor::default(),
+            generate_result: WasmGenerateResult::default(),
+            resolution: 0,
+            synthetic_processor: SDSProcessor::default(),
+            evaluate_result: WasmEvaluateResult::default(),
+            navigate_result: WasmNavigateResult::default(),
+        }
+    }
+
+    #[wasm_bindgen(js_name = "clearSensitiveData")]
+    pub fn clear_sensitive_data(&mut self) {
+        self.use_columns = JsHeaderNames::default();
+        self.sensitive_zeros = JsHeaderNames::default();
+        self.record_limit = 0;
+        self.sensitive_processor = SDSProcessor::default();
+        self.clear_generate();
+    }
+
+    #[wasm_bindgen(js_name = "clearGenerate")]
+    pub fn clear_generate(&mut self) {
+        self.generate_result = WasmGenerateResult::default();
+        self.resolution = 0;
+        self.synthetic_processor = SDSProcessor::default();
+        self.clear_evaluate()
+    }
+
+    #[wasm_bindgen(js_name = "clearEvaluate")]
+    pub fn clear_evaluate(&mut self) {
+        self.evaluate_result = WasmEvaluateResult::default();
+        self.clear_navigate()
+    }
+
+    #[wasm_bindgen(js_name = "clearNavigate")]
+    pub fn clear_navigate(&mut self) {
+        self.navigate_result = WasmNavigateResult::default();
+    }
+
+    #[wasm_bindgen(js_name = "setSensitiveData")]
+    pub fn set_sensitive_data(
+        &mut self,
+        csv_data: JsCsvData,
+        use_columns: JsHeaderNames,
+        sensitive_zeros: JsHeaderNames,
+        record_limit: usize,
+    ) -> JsResult<()> {
+        debug!("setting sensitive data...");
+
+        self.use_columns = use_columns;
+        self.sensitive_zeros = sensitive_zeros;
+        self.record_limit = record_limit;
+        self.sensitive_processor = SDSProcessor::new(
+            csv_data,
+            self.use_columns.clone().unchecked_into(),
+            self.sensitive_zeros.clone().unchecked_into(),
+            self.record_limit,
+        )?;
+        self.clear_generate();
+        Ok(())
+    }
+
+    pub fn generate(
+        &mut self,
+        cache_max_size: usize,
+        resolution: usize,
+        empty_value: String,
+        seeded: bool,
+        progress_callback: JsReportProgressCallback,
+    ) -> JsResult<()> {
+        debug!("generating synthetic data...");
+
+        self.generate_result = self.sensitive_processor.generate(
+            cache_max_size,
+            resolution,
+            empty_value,
+            seeded,
+            progress_callback,
+        )?;
+        self.resolution = resolution;
+
+        debug!("creating synthetic data processor...");
+
+        self.synthetic_processor = SDSProcessor::new(
+            self.generate_result.synthetic_data_to_js()?,
+            self.use_columns.clone().unchecked_into(),
+            self.sensitive_zeros.clone().unchecked_into(),
+            self.record_limit,
+        )?;
+        self.clear_evaluate();
+        Ok(())
+    }
+
+    pub fn evaluate(
+        &mut self,
+        reporting_length: usize,
+        sensitivity_threshold: usize,
+        sensitive_progress_callback: JsReportProgressCallback,
+        synthetic_progress_callback: JsReportProgressCallback,
+    ) -> JsResult<()> {
+        debug!("aggregating sensitive data...");
+
+        let sensitive_aggregate_result = self.sensitive_processor.aggregate(
+            reporting_length,
+            sensitivity_threshold,
+            sensitive_progress_callback,
+        )?;
+
+        debug!("aggregating synthetic data...");
+
+        let synthetic_aggregate_result = self.synthetic_processor.aggregate(
+            reporting_length,
+            sensitivity_threshold,
+            synthetic_progress_callback,
+        )?;
+
+        debug!("evaluating synthetic data based on sensitive data...");
+
+        self.evaluate_result = WasmEvaluateResult::from_aggregate_results(
+            sensitive_aggregate_result,
+            synthetic_aggregate_result,
+        );
+
+        self.clear_navigate();
+
+        Ok(())
+    }
+
+    #[wasm_bindgen(js_name = "protectSensitiveAggregatesCount")]
+    pub fn protect_sensitive_aggregates_count(&mut self) {
+        debug!("protecting sensitive aggregates count...");
+
+        self.evaluate_result
+            .sensitive_aggregate_result
+            .protect_aggregates_count(self.resolution);
+    }
+
+    pub fn navigate(&mut self) {
+        debug!("creating navigate result...");
+
+        self.navigate_result =
+            WasmNavigateResult::from_synthetic_processor(&self.synthetic_processor);
+    }
+
+    #[wasm_bindgen(js_name = "selectAttributes")]
+    pub fn select_attributes(&mut self, attributes: JsSelectedAttributesByColumn) -> JsResult<()> {
+        self.navigate_result.select_attributes(attributes)
+    }
+
+    #[wasm_bindgen(js_name = "attributesIntersectionsByColumn")]
+    pub fn attributes_intersections_by_column(
+        &mut self,
+        columns: JsHeaderNames,
+    ) -> JsResult<JsAttributesIntersectionByColumn> {
+        self.navigate_result.attributes_intersections_by_column(
+            columns,
+            &self.evaluate_result.sensitive_aggregate_result,
+        )
+    }
+
+    #[wasm_bindgen(js_name = "generateResultToJs")]
+    pub fn generate_result_to_js(&self) -> JsResult<JsGenerateResult> {
+        self.generate_result.to_js()
+    }
+
+    #[wasm_bindgen(js_name = "evaluateResultToJs")]
+    pub fn evaluate_result_to_js(
+        &self,
+        combination_delimiter: &str,
+        include_aggregates_count: bool,
+    ) -> JsResult<JsEvaluateResult> {
+        self.evaluate_result.to_js(
+            combination_delimiter,
+            self.resolution,
+            include_aggregates_count,
+        )
+    }
+}

packages/lib-wasm/src/processing/sds_processor/header_names.rs

@@ -0,0 +1,23 @@
+use js_sys::Array;
+use sds_core::data_block::typedefs::CsvRecord;
+use std::convert::TryFrom;
+use wasm_bindgen::{JsCast, JsValue};
+
+use crate::utils::js::ts_definitions::JsHeaderNames;
+
+impl JsHeaderNames {
+    #[inline]
+    pub fn default() -> JsHeaderNames {
+        Array::default().unchecked_into()
+    }
+}
+
+impl TryFrom<JsHeaderNames> for CsvRecord {
+    type Error = JsValue;
+
+    fn try_from(js_csv_record: JsHeaderNames) -> Result<Self, Self::Error> {
+        js_csv_record
+            .into_serde::<CsvRecord>()
+            .map_err(|err| JsValue::from(err.to_string()))
+    }
+}

packages/lib-wasm/src/processing/sds_processor/mod.rs

@@ -0,0 +1,5 @@
+pub mod header_names;
+
+mod processor;
+
+pub use processor::SDSProcessor;

packages/lib-wasm/src/processing/sds_processor/processor.rs

@@ -0,0 +1,120 @@
+use js_sys::Function;
+use sds_core::{
+    data_block::{block::DataBlock, data_block_creator::DataBlockCreator, typedefs::CsvRecord},
+    processing::{
+        aggregator::Aggregator,
+        generator::{Generator, SynthesisMode},
+    },
+    utils::time::ElapsedDurationLogger,
+};
+use std::convert::TryFrom;
+use std::sync::Arc;
+use wasm_bindgen::{prelude::*, JsCast};
+
+use crate::{
+    utils::js::js_progress_reporter::JsProgressReporter,
+    {
+        processing::{
+            aggregator::aggregate_result::WasmAggregateResult,
+            generator::generate_result::WasmGenerateResult,
+        },
+        utils::js::{
+            js_block_creator::JsDataBlockCreator,
+            ts_definitions::{JsCsvData, JsHeaderNames, JsReportProgressCallback},
+        },
+    },
+};
+
+#[wasm_bindgen]
+pub struct SDSProcessor {
+    pub(crate) data_block: Arc<DataBlock>,
+}
+
+#[wasm_bindgen]
+impl SDSProcessor {
+    pub fn default() -> SDSProcessor {
+        SDSProcessor {
+            data_block: Arc::new(DataBlock::default()),
+        }
+    }
+}
+
+#[wasm_bindgen]
+impl SDSProcessor {
+    #[wasm_bindgen(constructor)]
+    pub fn new(
+        csv_data: JsCsvData,
+        use_columns: JsHeaderNames,
+        sensitive_zeros: JsHeaderNames,
+        record_limit: usize,
+    ) -> Result<SDSProcessor, JsValue> {
+        let _duration_logger = ElapsedDurationLogger::new(String::from("sds processor creation"));
+        let data_block = JsDataBlockCreator::create(
+            Ok(csv_data.dyn_into()?),
+            &CsvRecord::try_from(use_columns)?,
+            &CsvRecord::try_from(sensitive_zeros)?,
+            record_limit,
+        )?;
+
+        Ok(SDSProcessor { data_block })
+    }
+
+    #[wasm_bindgen(js_name = "numberOfRecords")]
+    pub fn number_of_records(&self) -> usize {
+        self.data_block.number_of_records()
+    }
+
+    #[wasm_bindgen(js_name = "protectedNumberOfRecords")]
+    pub fn protected_number_of_records(&self, resolution: usize) -> usize {
+        self.data_block.protected_number_of_records(resolution)
+    }
+
+    #[wasm_bindgen(js_name = "normalizeReportingLength")]
+    pub fn normalize_reporting_length(&self, reporting_length: usize) -> usize {
+        self.data_block.normalize_reporting_length(reporting_length)
+    }
+
+    pub fn aggregate(
+        &self,
+        reporting_length: usize,
+        sensitivity_threshold: usize,
+        progress_callback: JsReportProgressCallback,
+    ) -> Result<WasmAggregateResult, JsValue> {
+        let _duration_logger =
+            ElapsedDurationLogger::new(String::from("sds processor aggregation"));
+        let js_callback: Function = progress_callback.dyn_into()?;
+        let mut aggregator = Aggregator::new(self.data_block.clone());
+
+        Ok(WasmAggregateResult::new(aggregator.aggregate(
+            reporting_length,
+            sensitivity_threshold,
+            &mut Some(JsProgressReporter::new(&js_callback, &|p| p)),
+        )))
+    }
+
+    pub fn generate(
+        &self,
+        cache_max_size: usize,
+        resolution: usize,
+        empty_value: String,
+        seeded: bool,
+        progress_callback: JsReportProgressCallback,
+    ) -> Result<WasmGenerateResult, JsValue> {
+        let _duration_logger = ElapsedDurationLogger::new(String::from("sds processor generation"));
+        let js_callback: Function = progress_callback.dyn_into()?;
+        let mut generator = Generator::new(self.data_block.clone());
+        let mode = if seeded {
+            SynthesisMode::Seeded
+        } else {
+            SynthesisMode::Unseeded
+        };
+
+        Ok(WasmGenerateResult::new(generator.generate(
+            resolution,
+            cache_max_size,
+            empty_value,
+            mode,
+            &mut Some(JsProgressReporter::new(&js_callback, &|p| p)),
+        )))
+    }
+}

packages/lib-wasm/src/utils/js/deserializers.rs

@@ -1,41 +0,0 @@
-use super::js_block_creator::JsDataBlockCreator;
-use js_sys::Array;
-use sds_core::{
-    data_block::{
-        block::{DataBlock, DataBlockCreator},
-        typedefs::CsvRecord,
-    },
-    utils::time::ElapsedDurationLogger,
-};
-
-#[inline]
-pub fn deserialize_use_columns(use_columns: Array) -> Result<CsvRecord, String> {
-    let _duration_logger = ElapsedDurationLogger::new(String::from("deserialize_use_columns"));
-
-    match use_columns.into_serde::<CsvRecord>() {
-        Ok(v) => Ok(v),
-        _ => Err(String::from("use_columns should be an Array<string>")),
-    }
-}
-
-#[inline]
-pub fn deserialize_sensitive_zeros(sensitive_zeros: Array) -> Result<CsvRecord, String> {
-    let _duration_logger = ElapsedDurationLogger::new(String::from("deserialize_sensitive_zeros"));
-
-    match sensitive_zeros.into_serde::<CsvRecord>() {
-        Ok(v) => Ok(v),
-        _ => Err(String::from("sensitive_zeros should be an Array<string>")),
-    }
-}
-
-#[inline]
-pub fn deserialize_csv_data(
-    csv_data: Array,
-    use_columns: &[String],
-    sensitive_zeros: &[String],
-    record_limit: usize,
-) -> Result<DataBlock, String> {
-    let _duration_logger = ElapsedDurationLogger::new(String::from("deserialize_csv_data"));
-
-    JsDataBlockCreator::create(Ok(csv_data), use_columns, sensitive_zeros, record_limit)
-}

packages/lib-wasm/src/utils/js/js_block_creator.rs

@@ -1,33 +1,34 @@
 use js_sys::Array;
-use sds_core::data_block::{block::DataBlockCreator, typedefs::CsvRecord};
+use sds_core::data_block::{data_block_creator::DataBlockCreator, typedefs::CsvRecord};
+use wasm_bindgen::JsValue;
 
 pub struct JsDataBlockCreator;
 
 impl DataBlockCreator for JsDataBlockCreator {
     type InputType = Array;
-    type ErrorType = String;
+    type ErrorType = JsValue;
 
-    fn get_headers(csv_data: &mut Array) -> Result<CsvRecord, String> {
+    fn get_headers(csv_data: &mut Self::InputType) -> Result<CsvRecord, Self::ErrorType> {
         let headers = csv_data.get(0);
 
         if headers.is_undefined() {
-            Err(String::from("missing headers"))
+            Err(JsValue::from("csv data missing headers"))
         } else {
-            match headers.into_serde::<Vec<String>>() {
-                Ok(h) => Ok(h),
-                Err(_) => Err(String::from("headers should an Array<string>")),
-            }
+            headers
+                .into_serde::<CsvRecord>()
+                .map_err(|err| JsValue::from(err.to_string()))
         }
     }
 
-    fn get_records(csv_data: &mut Self::InputType) -> Result<Vec<CsvRecord>, String> {
+    fn get_records(csv_data: &mut Self::InputType) -> Result<Vec<CsvRecord>, Self::ErrorType> {
         csv_data
             .slice(1, csv_data.length())
             .iter()
-            .map(|record| match record.into_serde::<Vec<String>>() {
-                Ok(h) => Ok(h),
-                Err(_) => Err(String::from("records should an Array<string>")),
+            .map(|record| {
+                record
+                    .into_serde::<CsvRecord>()
+                    .map_err(|err| JsValue::from(err.to_string()))
             })
-            .collect::<Result<Vec<CsvRecord>, String>>()
+            .collect::<Result<Vec<CsvRecord>, Self::ErrorType>>()
     }
 }

packages/lib-wasm/src/utils/js/macros.rs

@@ -1,32 +0,0 @@
-#[doc(hidden)]
-#[macro_export]
-macro_rules! match_or_return_undefined {
-    ($result_to_match: expr) => {
-        match $result_to_match {
-            Ok(result) => result,
-            Err(err) => {
-                error!("{}", err);
-                return JsValue::undefined();
-            }
-        }
-    };
-}
-
-#[doc(hidden)]
-#[macro_export]
-macro_rules! set_or_return_undefined {
-    ($($arg:tt)*) => {
-        match set($($arg)*) {
-            Ok(has_been_set) => {
-                if !has_been_set {
-                    error!("unable to set object value from wasm");
-                    return JsValue::undefined();
-                }
-            }
-            _ => {
-                error!("unable to set object value from wasm");
-                return JsValue::undefined();
-            }
-        };
-    };
-}

packages/lib-wasm/src/utils/js/mod.rs

@@ -1,16 +1,9 @@
-pub fn set_panic_hook() {
-    // When the `console_error_panic_hook` feature is enabled, we can call the
-    // `set_panic_hook` function at least once during initialization, and then
-    // we will get better error messages if our code ever panics.
-    //
-    // For more details see
-    // https://github.com/rustwasm/console_error_panic_hook#readme
-    #[cfg(feature = "console_error_panic_hook")]
-    console_error_panic_hook::set_once();
-}
-
-pub mod deserializers;
 pub mod js_block_creator;
+
 pub mod js_progress_reporter;
-pub mod macros;
-pub mod serializers;
+
+mod set_panic_hook;
+
+pub use set_panic_hook::set_panic_hook;
+
+pub mod ts_definitions;

packages/lib-wasm/src/utils/js/serializers.rs

@@ -1,124 +0,0 @@
-use js_sys::{Object, Reflect::set};
-use log::error;
-use sds_core::{
-    data_block::typedefs::DataBlockHeadersSlice,
-    processing::{
-        aggregator::{typedefs::AggregatesCountMap, Aggregator, PrivacyRiskSummary},
-        evaluator::typedefs::PreservationByCountBuckets,
-    },
-    utils::time::ElapsedDurationLogger,
-};
-use wasm_bindgen::JsValue;
-
-use crate::{aggregator::AggregatedCombination, set_or_return_undefined};
-
-#[inline]
-pub fn serialize_aggregates_count(
-    aggregates_count: &AggregatesCountMap,
-    headers: &DataBlockHeadersSlice,
-) -> JsValue {
-    let _duration_logger = ElapsedDurationLogger::new(String::from("serialize_aggregates_count"));
-    let aggregated_values = Object::new();
-
-    for (agg, count) in aggregates_count.iter() {
-        let combination_key = Aggregator::format_aggregate_str(headers, agg);
-
-        set_or_return_undefined!(
-            &aggregated_values,
-            &combination_key.clone().into(),
-            &AggregatedCombination::new(combination_key, count.count, agg.len()).into()
-        );
-    }
-    aggregated_values.into()
-}
-
-#[inline]
-pub fn serialize_buckets(buckets: &PreservationByCountBuckets) -> JsValue {
-    let _duration_logger = ElapsedDurationLogger::new(String::from("serialize_buckets"));
-    let serialized_buckets = Object::new();
-
-    for (bucket_index, b) in buckets.iter() {
-        let serialized_bucket = Object::new();
-
-        set_or_return_undefined!(&serialized_bucket, &"size".into(), &b.size.into());
-        set_or_return_undefined!(
-            &serialized_bucket,
-            &"preservationSum".into(),
-            &b.preservation_sum.into()
-        );
-        set_or_return_undefined!(
-            &serialized_bucket,
-            &"lengthSum".into(),
-            &b.length_sum.into()
-        );
-
-        set_or_return_undefined!(
-            &serialized_buckets,
-            &(*bucket_index).into(),
-            &serialized_bucket.into()
-        );
-    }
-    serialized_buckets.into()
-}
-
-#[inline]
-pub fn serialize_privacy_risk(privacy_risk: &PrivacyRiskSummary) -> JsValue {
-    let _duration_logger = ElapsedDurationLogger::new(String::from("serialize_privacy_risk"));
-    let serialized_privacy_risk = Object::new();
-
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"totalNumberOfRecords".into(),
-        &privacy_risk.total_number_of_records.into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"totalNumberOfCombinations".into(),
-        &privacy_risk.total_number_of_combinations.into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"recordsWithUniqueCombinationsCount".into(),
-        &privacy_risk.records_with_unique_combinations_count.into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"recordsWithRareCombinationsCount".into(),
-        &privacy_risk.records_with_rare_combinations_count.into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"uniqueCombinationsCount".into(),
-        &privacy_risk.unique_combinations_count.into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"rareCombinationsCount".into(),
-        &privacy_risk.rare_combinations_count.into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"recordsWithUniqueCombinationsProportion".into(),
-        &privacy_risk
-            .records_with_unique_combinations_proportion
-            .into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"recordsWithRareCombinationsProportion".into(),
-        &privacy_risk
-            .records_with_rare_combinations_proportion
-            .into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"uniqueCombinationsProportion".into(),
-        &privacy_risk.unique_combinations_proportion.into()
-    );
-    set_or_return_undefined!(
-        &serialized_privacy_risk,
-        &"rareCombinationsProportion".into(),
-        &privacy_risk.rare_combinations_proportion.into()
-    );
-    serialized_privacy_risk.into()
-}

packages/lib-wasm/src/utils/js/set_panic_hook.rs

@@ -0,0 +1,10 @@
+pub fn set_panic_hook() {
+    // When the `console_error_panic_hook` feature is enabled, we can call the
+    // `set_panic_hook` function at least once during initialization, and then
+    // we will get better error messages if our code ever panics.
+    //
+    // For more details see
+    // https://github.com/rustwasm/console_error_panic_hook#readme
+    #[cfg(feature = "console_error_panic_hook")]
+    console_error_panic_hook::set_once();
+}

packages/lib-wasm/src/utils/js/ts_definitions.rs

@@ -0,0 +1,146 @@
+use wasm_bindgen::prelude::*;
+
+#[wasm_bindgen(typescript_custom_section)]
+const TS_APPEND_CONTENT: &'static str = r#"
+export type ReportProgressCallback = (progress: number) => void;
+
+export type HeaderNames = string[];
+
+export type CsvRecord = string[];
+
+export type CsvData = CsvRecord[];
+
+export interface IGenerateResult {
+  expansionRatio: number;
+  syntheticData: CsvData;
+}
+
+export interface IPrivacyRiskSummary {
+  totalNumberOfRecords: number
+  totalNumberOfCombinations: number
+  recordsWithUniqueCombinationsCount: number
+  recordsWithRareCombinationsCount: number
+  uniqueCombinationsCount: number
+  rareCombinationsCount: number
+  recordsWithUniqueCombinationsProportion: number
+  recordsWithRareCombinationsProportion: number
+  uniqueCombinationsProportion: number
+  rareCombinationsProportion: number
+}
+
+export interface IAggregateCountByLen {
+  [length: number]: number
+}
+
+export interface IAggregateCountAndLength {
+  count: number;
+  length: number;
+}
+
+export interface IAggregatesCount {
+  [name: string]: IAggregateCountAndLength;
+}
+
+export interface IAggregateResult {
+  reportingLength: number;
+  aggregatesCount?: IAggregatesCount;
+  rareCombinationsCountByLen: IAggregateCountByLen;
+  combinationsCountByLen: IAggregateCountByLen;
+  combinationsSumByLen: IAggregateCountByLen;
+  privacyRisk: IPrivacyRiskSummary;
+}
+
+export interface IPreservationByCountBucket {
+  size: number;
+  preservationSum: number;
+  lengthSum: number;
+  combinationCountSum: number;
+}
+
+export interface IPreservationByCountBuckets {
+  [bucket_index: number]: IPreservationByCountBucket;
+}
+
+export interface IPreservationByCount {
+  buckets: IPreservationByCountBuckets;
+  combinationLoss: number;
+}
+
+export interface IEvaluateResult {
+  sensitiveAggregateResult: IAggregateResult;
+  syntheticAggregateResult: IAggregateResult;
+  leakageCountByLen: IAggregateCountByLen;
+  fabricatedCountByLen: IAggregateCountByLen;
+  preservationByCount: IPreservationByCount;
+  recordExpansion: number;
+}
+
+export interface ISelectedAttributesByColumn {
+  [columnIndex: number]: Set<string>;
+}
+
+export interface IAttributesIntersection {
+  value: string;
+  estimatedCount: number;
+  actualCount?: number;
+}
+
+export interface IAttributesIntersectionByColumn {
+  [columnIndex: number]: IAttributesIntersection[];
+}"#;
+
+#[wasm_bindgen]
+extern "C" {
+    #[wasm_bindgen(typescript_type = "ReportProgressCallback")]
+    pub type JsReportProgressCallback;
+
+    #[wasm_bindgen(typescript_type = "HeaderNames")]
+    pub type JsHeaderNames;
+
+    #[wasm_bindgen(typescript_type = "CsvRecord")]
+    pub type JsCsvRecord;
+
+    #[wasm_bindgen(typescript_type = "CsvData")]
+    pub type JsCsvData;
+
+    #[wasm_bindgen(typescript_type = "IGenerateResult")]
+    pub type JsGenerateResult;
+
+    #[wasm_bindgen(typescript_type = "IPrivacyRiskSummary")]
+    pub type JsPrivacyRiskSummary;
+
+    #[wasm_bindgen(typescript_type = "IAggregateCountByLen")]
+    pub type JsAggregateCountByLen;
+
+    #[wasm_bindgen(typescript_type = "IAggregateCountAndLength")]
+    pub type JsAggregateCountAndLength;
+
+    #[wasm_bindgen(typescript_type = "IAggregatesCount")]
+    pub type JsAggregatesCount;
+
+    #[wasm_bindgen(typescript_type = "IAggregateResult")]
+    pub type JsAggregateResult;
+
+    #[wasm_bindgen(typescript_type = "IPreservationByCountBucket")]
+    pub type JsPreservationByCountBucket;
+
+    #[wasm_bindgen(typescript_type = "IPreservationByCountBuckets")]
+    pub type JsPreservationByCountBuckets;
+
+    #[wasm_bindgen(typescript_type = "IPreservationByCount")]
+    pub type JsPreservationByCount;
+
+    #[wasm_bindgen(typescript_type = "IEvaluateResult")]
+    pub type JsEvaluateResult;
+
+    #[wasm_bindgen(typescript_type = "ISelectedAttributesByColumn")]
+    pub type JsSelectedAttributesByColumn;
+
+    #[wasm_bindgen(typescript_type = "IAttributesIntersection")]
+    pub type JsAttributesIntersection;
+
+    #[wasm_bindgen(typescript_type = "IAttributesIntersectionByColumn")]
+    pub type JsAttributesIntersectionByColumn;
+}
+
+pub type JsResult<T> = Result<T, JsValue>;

packages/lib-wasm/src/utils/mod.rs

@@ -1,2 +1,3 @@
 pub mod js;
+
 pub mod logger;

packages/python-pipeline/src/aggregator.py

@@ -30,11 +30,13 @@ def aggregate(config):
     sensitive_zeros = config['sensitive_zeros']
     output_dir = config['output_dir']
     prefix = config['prefix']
+    sensitive_aggregated_data_json = path.join(
+        output_dir, f'{prefix}_sensitive_aggregated_data.json')
 
     logging.info(f'Aggregate {sensitive_microdata_path}')
     start_time = time.time()
 
-    data_block = sds.create_data_block_from_file(
+    sds_processor = sds.SDSProcessor(
         sensitive_microdata_path,
         sensitive_microdata_delimiter,
         use_columns,
@@ -42,13 +44,32 @@ def aggregate(config):
         max(record_limit, 0)
     )
 
-    len_to_combo_count, len_to_rare_count = sds.aggregate_and_write(
-        data_block,
+    aggregated_data = sds_processor.aggregate(
+        reporting_length,
+        0
+    )
+    len_to_combo_count = aggregated_data.calc_combinations_count_by_len()
+    len_to_rare_count = aggregated_data.calc_rare_combinations_count_by_len(
+        reporting_resolution)
+
+    aggregated_data.write_aggregates_count(
         sensitive_aggregates_path,
+        '\t',
+        ';',
+        reporting_resolution,
+        False
+    )
+
+    aggregated_data.write_to_json(sensitive_aggregated_data_json)
+
+    aggregated_data.protect_aggregates_count(reporting_resolution)
+
+    aggregated_data.write_aggregates_count(
         reportable_aggregates_path,
         '\t',
-        reporting_length,
-        reporting_resolution
+        ';',
+        reporting_resolution,
+        True
     )
 
     leakage_tsv = path.join(

packages/python-pipeline/src/evaluator.py

@@ -2,8 +2,247 @@ import time
 import datetime
 import logging
 from os import path
-from collections import defaultdict
 import util as util
+import sds
+
+
+class Evaluator:
+    def __init__(self, config):
+        self.sds_evaluator = sds.Evaluator()
+        self.use_columns = config['use_columns']
+        self.record_limit = max(config['record_limit'], 0)
+        self.reporting_length = max(config['reporting_length'], 0)
+        self.reporting_resolution = config['reporting_resolution']
+        self.sensitive_microdata_path = config['sensitive_microdata_path']
+        self.sensitive_microdata_delimiter = config['sensitive_microdata_delimiter']
+        self.synthetic_microdata_path = config['synthetic_microdata_path']
+        self.sensitive_zeros = config['sensitive_zeros']
+        self.output_dir = config['output_dir']
+        self.prefix = config['prefix']
+        self.sensitive_aggregated_data_json = path.join(
+            self.output_dir, f'{self.prefix}_sensitive_aggregated_data.json'
+        )
+        self.record_analysis_tsv = path.join(
+            self.output_dir, f'{self.prefix}_sensitive_analysis_by_length.tsv'
+        )
+        self.synthetic_rare_combos_tsv = path.join(
+            self.output_dir, f'{self.prefix}_synthetic_rare_combos_by_length.tsv'
+        )
+        self.parameters_tsv = path.join(
+            self.output_dir, f'{self.prefix}_parameters.tsv'
+        )
+        self.leakage_tsv = path.join(
+            self.output_dir, f'{self.prefix}_synthetic_leakage_by_length.tsv'
+        )
+        self.leakage_svg = path.join(
+            self.output_dir, f'{self.prefix}_synthetic_leakage_by_length.svg'
+        )
+        self.preservation_by_count_tsv = path.join(
+            self.output_dir, f'{self.prefix}_preservation_by_count.tsv'
+        )
+        self.preservation_by_count_svg = path.join(
+            self.output_dir, f'{self.prefix}_preservation_by_count.svg'
+        )
+        self.preservation_by_length_tsv = path.join(
+            self.output_dir, f'{self.prefix}_preservation_by_length.tsv'
+        )
+        self.preservation_by_length_svg = path.join(
+            self.output_dir, f'{self.prefix}_preservation_by_length.svg'
+        )
+
+    def _load_sensitive_aggregates(self):
+        if not path.exists(self.sensitive_aggregated_data_json):
+            logging.info('Computing sensitive aggregates...')
+            self.sen_sds_processor = sds.SDSProcessor(
+                self.sensitive_microdata_path,
+                self.sensitive_microdata_delimiter,
+                self.use_columns,
+                self.sensitive_zeros,
+                self.record_limit
+            )
+            self.sen_aggregated_data = self.sen_sds_processor.aggregate(
+                self.reporting_length,
+                0
+            )
+        else:
+            logging.info('Loading sensitive aggregates...')
+            self.sen_aggregated_data = sds.AggregatedData.read_from_json(
+                self.sensitive_aggregated_data_json
+            )
+            self.sen_sds_processor = sds.SDSProcessor.from_aggregated_data(
+                self.sen_aggregated_data
+            )
+
+        self.reporting_length = self.sen_sds_processor.normalize_reporting_length(
+            self.reporting_length
+        )
+
+    def _load_synthetic_aggregates(self):
+        logging.info('Computing synthetic aggregates...')
+        self.syn_sds_processor = sds.SDSProcessor(
+            self.synthetic_microdata_path,
+            "\t",
+            self.use_columns,
+            self.sensitive_zeros,
+            self.record_limit
+        )
+        self.syn_aggregated_data = self.syn_sds_processor.aggregate(
+            self.reporting_length,
+            0
+        )
+
+    def _do_records_analysis(self):
+        logging.info('Performing records analysis on sensitive aggregates...')
+        self.records_analysis_data = self.sen_aggregated_data.calc_records_analysis_by_len(
+            self.reporting_resolution, True
+        )
+        self.records_analysis_data.write_records_analysis(
+            self.record_analysis_tsv, '\t'
+        )
+
+    def _compare_synthetic_and_sensitive_rare(self):
+        logging.info(
+            'Comparing synthetic and sensitive aggregates rare combinations...')
+        rare_combinations_data = self.sds_evaluator.compare_synthetic_and_sensitive_rare(
+            self.syn_aggregated_data, self.sen_aggregated_data, self.reporting_resolution, ' AND ', True
+        )
+        rare_combinations_data.write_rare_combinations(
+            self.synthetic_rare_combos_tsv, '\t'
+        )
+
+    def _write_parameters(self):
+        logging.info('Writing evaluation parameters...')
+        total_sen = self.sen_sds_processor.protected_number_of_records(
+            self.reporting_resolution
+        )
+        with open(self.parameters_tsv, 'w') as f:
+            f.write('\t'.join(['parameter', 'value'])+'\n')
+            f.write(
+                '\t'.join(['resolution', str(self.reporting_resolution)])+'\n'
+            )
+            f.write('\t'.join(['limit', str(self.reporting_length)])+'\n')
+            f.write(
+                '\t'.join(['total_sensitive_records', str(total_sen)])+'\n'
+            )
+            f.write('\t'.join(['unique_identifiable', str(
+                self.records_analysis_data.get_total_unique())])+'\n'
+            )
+            f.write('\t'.join(['rare_identifiable', str(
+                self.records_analysis_data.get_total_rare())])+'\n'
+            )
+            f.write('\t'.join(['risky_identifiable', str(
+                self.records_analysis_data.get_total_risky())])+'\n'
+            )
+            f.write(
+                '\t'.join(['risky_identifiable_pct', str(
+                    100*self.records_analysis_data.get_total_risky()/total_sen)])+'\n'
+            )
+
+    def _find_leakages(self):
+        logging.info(
+            'Looking for leakages from the sensitive dataset on the synthetic dataset...')
+        comb_counts = self.syn_aggregated_data.calc_combinations_count_by_len()
+        leakage_counts = self.sds_evaluator.calc_leakage_count(
+            self.sen_aggregated_data, self.syn_aggregated_data, self.reporting_resolution
+        )
+
+        with open(self.leakage_tsv, 'w') as f:
+            f.write('\t'.join(
+                ['syn_combo_length', 'combo_count', 'leak_count', 'leak_proportion'])+'\n'
+            )
+            for length in range(1, self.reporting_length + 1):
+                combo_count = comb_counts.get(length, 0)
+                leak_count = leakage_counts.get(length, 0)
+                # by design there should be no leakage
+                assert leak_count == 0
+                leak_prop = leak_count/combo_count if combo_count > 0 else 0
+                f.write('\t'.join(
+                    [str(length), str(combo_count), str(leak_count), str(leak_prop)])+'\n'
+                )
+
+        util.plotStats(
+            x_axis='syn_combo_length',
+            x_axis_title='Length of Synthetic Combination',
+            y_bar='combo_count',
+            y_bar_title='Count of Combinations',
+            y_line='leak_proportion',
+            y_line_title=f'Proportion of Leaked (<{self.reporting_resolution}) Combinations',
+            color='violet',
+            darker_color='darkviolet',
+            stats_tsv=self.leakage_tsv,
+            stats_svg=self.leakage_svg,
+            delimiter='\t',
+            style='whitegrid',
+            palette='magma'
+        )
+
+    def _calculate_preservation_by_count(self):
+        logging.info('Calculating preservation by count...')
+        preservation_by_count = self.sds_evaluator.calc_preservation_by_count(
+            self.sen_aggregated_data, self.syn_aggregated_data, self.reporting_resolution
+        )
+        preservation_by_count.write_preservation_by_count(
+            self.preservation_by_count_tsv, '\t'
+        )
+
+        util.plotStats(
+            x_axis='syn_count_bucket',
+            x_axis_title='Count of Filtered Synthetic Records',
+            y_bar='mean_combo_length',
+            y_bar_title='Mean Length of Combinations',
+            y_line='count_preservation',
+            y_line_title='Count Preservation (Synthetic/Sensitive)',
+            color='lightgreen',
+            darker_color='green',
+            stats_tsv=self.preservation_by_count_tsv,
+            stats_svg=self.preservation_by_count_svg,
+            delimiter='\t',
+            style='whitegrid',
+            palette='magma'
+        )
+
+    def _calculate_preservation_by_length(self):
+        logging.info('Calculating preservation by length...')
+        preservation_by_length = self.sds_evaluator.calc_preservation_by_length(
+            self.sen_aggregated_data, self.syn_aggregated_data, self.reporting_resolution
+        )
+        preservation_by_length.write_preservation_by_length(
+            self.preservation_by_length_tsv, '\t'
+        )
+
+        util.plotStats(
+            x_axis='syn_combo_length',
+            x_axis_title='Length of Combination',
+            y_bar='mean_combo_count',
+            y_bar_title='Mean Synthetic Count',
+            y_line='count_preservation',
+            y_line_title='Count Preservation (Synthetic/Sensitive)',
+            color='cornflowerblue',
+            darker_color='mediumblue',
+            stats_tsv=self.preservation_by_length_tsv,
+            stats_svg=self.preservation_by_length_svg,
+            delimiter='\t',
+            style='whitegrid',
+            palette='magma'
+        )
+
+    def run(self):
+        logging.info(
+            f'Evaluate {self.synthetic_microdata_path} vs {self.sensitive_microdata_path}'
+        )
+        start_time = time.time()
+
+        self._load_sensitive_aggregates()
+        self._load_synthetic_aggregates()
+        self._do_records_analysis()
+        self._compare_synthetic_and_sensitive_rare()
+        self._write_parameters()
+        self._find_leakages()
+        self._calculate_preservation_by_count()
+        self._calculate_preservation_by_length()
+
+        logging.info(
+            f'Evaluated {self.synthetic_microdata_path} vs {self.sensitive_microdata_path}, took {datetime.timedelta(seconds = time.time() - start_time)}s')
 
 
 def evaluate(config):
@@ -15,309 +254,4 @@ def evaluate(config):
     Args:
         config: options from the json config file, else default values.
     """
-
-    use_columns = config['use_columns']
-    record_limit = config['record_limit']
-    reporting_length = config['reporting_length']
-    reporting_resolution = config['reporting_resolution']
-    sensitive_microdata_path = config['sensitive_microdata_path']
-    sensitive_microdata_delimiter = config['sensitive_microdata_delimiter']
-    synthetic_microdata_path = config['synthetic_microdata_path']
-    sensitive_zeros = config['sensitive_zeros']
-    parallel_jobs = config['parallel_jobs']
-    output_dir = config['output_dir']
-    prefix = config['prefix']
-
-    logging.info(
-        f'Evaluate {synthetic_microdata_path} vs {sensitive_microdata_path}')
-    start_time = time.time()
-
-    sen_counts = None
-    sen_records = None
-    sen_df = util.loadMicrodata(
-        path=sensitive_microdata_path, delimiter=sensitive_microdata_delimiter, record_limit=record_limit,
-        use_columns=use_columns)
-    sen_records = util.genRowList(sen_df, sensitive_zeros)
-    if not path.exists(config['sensitive_aggregates_path']):
-        logging.info('Computing sensitive aggregates...')
-        if reporting_length == -1:
-            reporting_length = max([len(row) for row in sen_records])
-        sen_counts = util.countAllCombos(
-            sen_records, reporting_length, parallel_jobs)
-    else:
-        logging.info('Loading sensitive aggregates...')
-        sen_counts = util.loadSavedAggregates(
-            config['sensitive_aggregates_path'])
-        if reporting_length == -1:
-            reporting_length = max(sen_counts.keys())
-
-    if use_columns != []:
-        reporting_length = min(reporting_length, len(use_columns))
-
-    filtered_sen_counts = {length: {combo: count for combo, count in combo_to_counts.items(
-    ) if count >= reporting_resolution} for length, combo_to_counts in sen_counts.items()}
-    syn_df = util.loadMicrodata(path=synthetic_microdata_path,
-                                delimiter='\t', record_limit=-1, use_columns=use_columns)
-    syn_records = util.genRowList(syn_df, sensitive_zeros)
-    syn_counts = util.countAllCombos(
-        syn_records, reporting_length, parallel_jobs)
-
-    len_to_syn_count = {length: len(combo_to_count)
-                        for length, combo_to_count in syn_counts.items()}
-    len_to_sen_rare = {length: {combo: count for combo, count in combo_to_count.items() if count < reporting_resolution}
-                       for length, combo_to_count in sen_counts.items()}
-    len_to_syn_rare = {length: {combo: count for combo, count in combo_to_count.items() if count < reporting_resolution}
-                       for length, combo_to_count in syn_counts.items()}
-    len_to_syn_leak = {length: len([1 for rare in rares if rare in syn_counts.get(length, {}).keys()])
-                       for length, rares in len_to_sen_rare.items()}
-
-    sen_unique_to_records, sen_rare_to_records, _ = util.mapShortestUniqueRareComboLengthToRecords(
-        sen_records, len_to_sen_rare)
-    sen_rare_to_sen_count = {length: util.protect(len(records), reporting_resolution)
-                             for length, records in sen_rare_to_records.items()}
-    sen_unique_to_sen_count = {length: util.protect(len(records), reporting_resolution)
-                               for length, records in sen_unique_to_records.items()}
-
-    total_sen = util.protect(len(sen_records), reporting_resolution)
-    unique_total = sum(
-        [v for k, v in sen_unique_to_sen_count.items() if k > 0])
-    rare_total = sum([v for k, v in sen_rare_to_sen_count.items() if k > 0])
-    risky_total = unique_total + rare_total
-    risky_total_pct = 100*risky_total/total_sen
-
-    record_analysis_tsv = path.join(
-        output_dir, f'{prefix}_sensitive_analysis_by_length.tsv')
-    with open(record_analysis_tsv, 'w') as f:
-        f.write('\t'.join(['combo_length', 'sen_rare', 'sen_rare_pct',
-                'sen_unique', 'sen_unique_pct', 'sen_risky', 'sen_risky_pct'])+'\n')
-        for length in sen_counts.keys():
-            sen_rare = sen_rare_to_sen_count.get(length, 0)
-            sen_rare_pct = 100*sen_rare / total_sen if total_sen > 0 else 0
-            sen_unique = sen_unique_to_sen_count.get(length, 0)
-            sen_unique_pct = 100*sen_unique / total_sen if total_sen > 0 else 0
-            sen_risky = sen_rare + sen_unique
-            sen_risky_pct = 100*sen_risky / total_sen if total_sen > 0 else 0
-            f.write(
-                '\t'.join(
-                    [str(length),
-                     str(sen_rare),
-                     str(sen_rare_pct),
-                     str(sen_unique),
-                     str(sen_unique_pct),
-                     str(sen_risky),
-                     str(sen_risky_pct)]) + '\n')
-
-    _, _, syn_length_to_combo_to_rare = util.mapShortestUniqueRareComboLengthToRecords(
-        syn_records, len_to_syn_rare)
-    combos_tsv = path.join(
-        output_dir, f'{prefix}_synthetic_rare_combos_by_length.tsv')
-    with open(combos_tsv, 'w') as f:
-        f.write('\t'.join(['combo_length', 'combo',
-                'record_id', 'syn_count', 'sen_count'])+'\n')
-        for length, combo_to_rare in syn_length_to_combo_to_rare.items():
-            for combo, rare_ids in combo_to_rare.items():
-                syn_count = len(rare_ids)
-                for rare_id in rare_ids:
-                    sen_count = util.protect(sen_counts.get(length, {})[
-                                             combo], reporting_resolution)
-                    f.write(
-                        '\t'.join(
-                            [str(length),
-                             util.comboToString(combo).replace(';', ' AND '),
-                             str(rare_id),
-                             str(syn_count),
-                             str(sen_count)]) + '\n')
-
-    parameters_tsv = path.join(output_dir, f'{prefix}_parameters.tsv')
-
-    with open(parameters_tsv, 'w') as f:
-        f.write('\t'.join(['parameter', 'value'])+'\n')
-        f.write('\t'.join(['resolution', str(reporting_resolution)])+'\n')
-        f.write('\t'.join(['limit', str(reporting_length)])+'\n')
-        f.write('\t'.join(['total_sensitive_records', str(total_sen)])+'\n')
-        f.write('\t'.join(['unique_identifiable', str(unique_total)])+'\n')
-        f.write('\t'.join(['rare_identifiable', str(rare_total)])+'\n')
-        f.write('\t'.join(['risky_identifiable', str(risky_total)])+'\n')
-        f.write(
-            '\t'.join(['risky_identifiable_pct', str(risky_total_pct)])+'\n')
-
-    leakage_tsv = path.join(
-        output_dir, f'{prefix}_synthetic_leakage_by_length.tsv')
-    leakage_svg = path.join(
-        output_dir, f'{prefix}_synthetic_leakage_by_length.svg')
-    with open(leakage_tsv, 'w') as f:
-        f.write('\t'.join(['syn_combo_length', 'combo_count',
-                'leak_count', 'leak_proportion'])+'\n')
-        for length, leak_count in len_to_syn_leak.items():
-            combo_count = len_to_syn_count.get(length, 0)
-            leak_prop = leak_count/combo_count if combo_count > 0 else 0
-            f.write('\t'.join([str(length), str(combo_count),
-                    str(leak_count), str(leak_prop)])+'\n')
-
-    util.plotStats(
-        x_axis='syn_combo_length',
-        x_axis_title='Length of Synthetic Combination',
-        y_bar='combo_count',
-        y_bar_title='Count of Combinations',
-        y_line='leak_proportion',
-        y_line_title=f'Proportion of Leaked (<{reporting_resolution}) Combinations',
-        color='violet',
-        darker_color='darkviolet',
-        stats_tsv=leakage_tsv,
-        stats_svg=leakage_svg,
-        delimiter='\t',
-        style='whitegrid',
-        palette='magma')
-
-    compareDatasets(filtered_sen_counts, syn_counts, output_dir, prefix)
-
-    logging.info(
-        f'Evaluated {synthetic_microdata_path} vs {sensitive_microdata_path}, took {datetime.timedelta(seconds = time.time() - start_time)}s')
-
-
-def compareDatasets(sensitive_length_to_combo_to_count, synthetic_length_to_combo_to_count, output_dir, prefix):
-    """Evaluates the error in the synthetic microdata with respect to the sensitive microdata.
-
-    Produces output statistics and graphics binned by attribute count.
-
-    Args:
-        sensitive_length_to_combo_to_count: counts from sensitive microdata.
-        synthetic_length_to_combo_to_count: counts from synthetic microdata.
-        output_dir: where to save output statistics and graphics.
-        prefix: prefix to add to output files.
-    """
-
-    all_count_length_preservation = []
-    max_syn_count = 0
-
-    all_combos = set()
-    for length, combo_to_count in sensitive_length_to_combo_to_count.items():
-        for combo in combo_to_count.keys():
-            all_combos.add((length, combo))
-    for length, combo_to_count in synthetic_length_to_combo_to_count.items():
-        for combo in combo_to_count.keys():
-            all_combos.add((length, combo))
-    tot = len(all_combos)
-    for i, (length, combo) in enumerate(all_combos):
-        if i % 10000 == 0:
-            logging.info(f'{100*i/tot:.1f}% through comparisons')
-        sen_count = sensitive_length_to_combo_to_count.get(
-            length, {}).get(combo, 0)
-        syn_count = synthetic_length_to_combo_to_count.get(
-            length, {}).get(combo, 0)
-        max_syn_count = max(syn_count, max_syn_count)
-        preservation = 0
-        preservation = syn_count / sen_count if sen_count > 0 else 0
-        if sen_count == 0:
-            logging.error(
-                f'Error: For {combo}, synthetic count is {syn_count} but no sensitive count')
-        all_count_length_preservation.append((syn_count, length, preservation))
-        max_syn_count = max(syn_count, max_syn_count)
-
-    generateStatsAndGraphics(output_dir, max_syn_count,
-                             all_count_length_preservation, prefix)
-
-
-def generateStatsAndGraphics(output_dir, max_syn_count, count_length_preservation, prefix):
-    """Generates count error statistics for buckets of user-observed counts (post-filtering).
-
-    Outputs the files preservation_by_length and preservation_by_count tsv and svg files.
-
-    Args:
-        output_dir: the folder in which to output summary files.
-        max_syn_count: the maximum observed count of synthetic records matching a single attribute value.
-        count_length_preservation: list of (count, length, preservation) tuples for observed preservation of sensitive counts after filtering by a combo of length.
-    """
-    sorted_counts = sorted(count_length_preservation,
-                           key=lambda x: x[0], reverse=False)
-    buckets = []
-    next_bucket = 10
-    while next_bucket < max_syn_count:
-        buckets.append(next_bucket)
-        next_bucket *= 2
-    buckets.append(next_bucket)
-    bucket_to_preservations = defaultdict(list)
-    bucket_to_counts = defaultdict(list)
-    bucket_to_lengths = defaultdict(list)
-    length_to_preservations = defaultdict(list)
-    length_to_counts = defaultdict(list)
-    for (count, length, preservation) in sorted_counts:
-        bucket = buckets[0]
-        for bi in range(len(buckets)-1, -1, -1):
-            if count > buckets[bi]:
-                bucket = buckets[bi+1]
-                break
-        bucket_to_preservations[bucket].append(preservation)
-        bucket_to_lengths[bucket].append(length)
-        bucket_to_counts[bucket].append(count)
-        length_to_counts[length].append(count)
-        length_to_preservations[length].append(preservation)
-
-    bucket_to_mean_count = {bucket: sum(counts)/len(counts) if len(counts) >
-                            0 else 0 for bucket, counts in bucket_to_counts.items()}
-    bucket_to_mean_preservation = {bucket: sum(preservations) / len(preservations)
-                                   if len(preservations) > 0 else 0 for bucket,
-                                   preservations in bucket_to_preservations.items()}
-    bucket_to_mean_length = {bucket: sum(lengths)/len(lengths) if len(lengths) >
-                             0 else 0 for bucket, lengths in bucket_to_lengths.items()}
-
-    counts_tsv = path.join(output_dir, f'{prefix}_preservation_by_count.tsv')
-    counts_svg = path.join(output_dir, f'{prefix}_preservation_by_count.svg')
-    with open(counts_tsv, 'w') as f:
-        f.write('\t'.join(['syn_count_bucket', 'mean_combo_count',
-                'mean_combo_length', 'count_preservation'])+'\n')
-        for bucket in reversed(buckets):
-            f.write(
-                '\t'.join(
-                    [str(bucket),
-                     str(bucket_to_mean_count.get(bucket, 0)),
-                     str(bucket_to_mean_length.get(bucket, 0)),
-                     str(bucket_to_mean_preservation.get(bucket, 0))]) + '\n')
-
-    util.plotStats(
-        x_axis='syn_count_bucket',
-        x_axis_title='Count of Filtered Synthetic Records',
-        y_bar='mean_combo_length',
-        y_bar_title='Mean Length of Combinations',
-        y_line='count_preservation',
-        y_line_title='Count Preservation (Synthetic/Sensitive)',
-        color='lightgreen',
-        darker_color='green',
-        stats_tsv=counts_tsv,
-        stats_svg=counts_svg,
-        delimiter='\t',
-        style='whitegrid',
-        palette='magma')
-
-    length_to_mean_preservation = {length: sum(preservations) / len(preservations)
-                                   if len(preservations) > 0 else 0 for length,
-                                   preservations in length_to_preservations.items()}
-    length_to_mean_count = {length: sum(counts)/len(counts) if len(counts) >
-                            0 else 0 for length, counts in length_to_counts.items()}
-
-    lengths_tsv = path.join(output_dir, f'{prefix}_preservation_by_length.tsv')
-    lengths_svg = path.join(output_dir, f'{prefix}_preservation_by_length.svg')
-    with open(lengths_tsv, 'w') as f:
-        f.write(
-            '\t'.join(['syn_combo_length', 'mean_combo_count', 'count_preservation'])+'\n')
-        for length in sorted(length_to_preservations.keys()):
-            f.write(
-                '\t'.join(
-                    [str(length),
-                     str(length_to_mean_count.get(length, 0)),
-                     str(length_to_mean_preservation.get(length, 0))]) + '\n')
-
-    util.plotStats(
-        x_axis='syn_combo_length',
-        x_axis_title='Length of Combination',
-        y_bar='mean_combo_count',
-        y_bar_title='Mean Synthetic Count',
-        y_line='count_preservation',
-        y_line_title='Count Preservation (Synthetic/Sensitive)',
-        color='cornflowerblue',
-        darker_color='mediumblue',
-        stats_tsv=lengths_tsv,
-        stats_svg=lengths_svg,
-        delimiter='\t',
-        style='whitegrid',
-        palette='magma')
+    Evaluator(config).run()

packages/python-pipeline/src/generator.py

@@ -1,12 +1,6 @@
 import time
 import datetime
 import logging
-import joblib
-import psutil
-import pandas as pd
-from math import ceil
-from random import random, shuffle
-import util as util
 import sds
 
 
@@ -20,14 +14,12 @@ def generate(config):
     """
 
     use_columns = config['use_columns']
-    parallel_jobs = config['parallel_jobs']
     record_limit = config['record_limit']
     sensitive_microdata_path = config['sensitive_microdata_path']
     sensitive_microdata_delimiter = config['sensitive_microdata_delimiter']
     synthetic_microdata_path = config['synthetic_microdata_path']
     sensitive_zeros = config['sensitive_zeros']
     resolution = config['reporting_resolution']
-    memory_limit = config['memory_limit_pct']
     cache_max_size = config['cache_max_size']
     seeded = config['seeded']
 
@@ -36,215 +28,26 @@ def generate(config):
 
     if seeded:
         logging.info(f'Generating from seeds')
-        data_block = sds.create_data_block_from_file(
-            sensitive_microdata_path,
-            sensitive_microdata_delimiter,
-            use_columns,
-            sensitive_zeros,
-            max(record_limit, 0)
-        )
-        syn_ratio = sds.generate_seeded_and_write(
-            data_block,
-            synthetic_microdata_path,
-            '\t',
-            cache_max_size,
-            resolution
-        )
-
     else:
-        df = util.loadMicrodata(path=sensitive_microdata_path, delimiter=sensitive_microdata_delimiter,
-                                record_limit=record_limit, use_columns=use_columns)
-        columns = df.columns.values
-        num = len(df)
-        logging.info(f'Prepared data')
-        records = []
-
-        chunk = ceil(num/(parallel_jobs))
-
         logging.info(f'Generating unseeded')
-        chunks = [chunk for i in range(parallel_jobs)]
-        col_val_ids = util.genColValIdsDict(df, sensitive_zeros)
-        res = joblib.Parallel(
-            n_jobs=parallel_jobs, backend='loky', verbose=1)(
-            joblib.delayed(synthesizeRowsUnseeded)(
-                chunk, num, columns, col_val_ids, resolution, memory_limit)
-            for chunk in chunks)
-        for rows in res:
-            records.extend(rows)
-        # trim any overgenerated records because of uniform chunk size
-        records = records[:num]
 
-        records.sort()
-        records.sort(key=lambda x: len(
-            [y for y in x if y != '']), reverse=True)
-
-        sdf = pd.DataFrame(records, columns=df.columns)
-        syn_ratio = len(sdf) / len(df)
-        sdf.to_csv(synthetic_microdata_path, sep='\t', index=False)
+    sds_processor = sds.SDSProcessor(
+        sensitive_microdata_path,
+        sensitive_microdata_delimiter,
+        use_columns,
+        sensitive_zeros,
+        max(record_limit, 0)
+    )
+    generated_data = sds_processor.generate(
+        cache_max_size,
+        resolution,
+        "",
+        seeded
+    )
+    generated_data.write_synthetic_data(synthetic_microdata_path, '\t')
+    syn_ratio = generated_data.expansion_ratio
 
     config['expansion_ratio'] = syn_ratio
 
     logging.info(
         f'Generated {synthetic_microdata_path} from {sensitive_microdata_path} with synthesis ratio {syn_ratio}, took {datetime.timedelta(seconds = time.time() - start_time)}s')
-
-
-def synthesizeRowsUnseeded(chunk, num_rows, columns, col_val_ids, resolution, memory_limit):
-    """Create synthetic records through unconstrained sampling of attribute distributions.
-
-    Args:
-        chunk: how many records to synthesize.
-        num_rows: how many rows/records in the sensitive dataset.
-        columns: the columns of the sensitive dataset.
-        atts_to_ids: a dict mapping attributes to row ids.
-        resolution: the minimum count of sensitive attribute combinations for inclusion in synthetic records.
-        memory_limit: the percentage memory use not to exceed.
-
-    Returns:
-        rows: synthetic records created through unconstrained sampling of attribute distributions.
-        overall_cache_hits: the number of times the cache was hit.
-        overall_cache_misses: the number of times the cache was missed.
-    """
-    rows = []
-    filter_cache = {}
-    overall_cache_hits = 0
-    overall_cache_misses = 0
-
-    for i in range(chunk):
-        if i % 100 == 99:
-            logging.info(
-                f'{(100*i/chunk):.1f}% through row synthesis, cache utilization {100*overall_cache_hits/(overall_cache_hits + overall_cache_misses):.1f}%')
-        filters, cache_hits, cache_misses = synthesizeRowUnseeded(
-            filter_cache, num_rows, columns, col_val_ids, resolution, memory_limit)
-        overall_cache_hits += cache_hits
-        overall_cache_misses += cache_misses
-        row = normalize(columns, filters)
-        rows.append(row)
-
-    return rows
-
-
-def synthesizeRowUnseeded(filter_cache, num_rows, columns, col_val_ids, resolution, memory_limit):
-    """Creates a synthetic record through unconstrained sampling of attribute distributions.
-
-    Args:
-        num_rows: how many rows/records in the sensitive dataset.
-        columns: the columns of the sensitive dataset.
-        atts_to_ids: a dict mapping attributes to row ids.
-        resolution: the minimum count of sensitive attribute combinations for inclusion in synthetic records.
-        memory_limit: the percentage memory use not to exceed.
-
-    Returns:
-        row: a synthetic record created through unconstrained sampling of attribute distributions.
-        row_cache_hits: the number of times the cache was hit.
-        row_cache_misses: the number of times the cache was missed.
-    """
-    shuffled_columns = list(columns)
-    shuffle(shuffled_columns)
-    output_atts = []
-    residual_ids = set(range(num_rows))
-    row_cache_hits = 0
-    row_cache_misses = 0
-    # do not add to cache once memory limit is reached
-    use_cache = psutil.virtual_memory()[2] <= memory_limit
-    for col in shuffled_columns:
-        vals_to_sample = {}
-        for val, ids in col_val_ids[col].items():
-            next_filters = tuple(
-                sorted((*output_atts, (col, val)), key=lambda x: f'{x[0]}:{x[1]}'.lower()))
-            if next_filters in filter_cache.keys():
-                vals_to_sample[val] = set(filter_cache[next_filters])
-                row_cache_hits += 1
-            else:
-                row_cache_misses += 1
-                res = set(ids).intersection(residual_ids)
-                vals_to_sample[val] = res
-                if use_cache:
-                    filter_cache[next_filters] = res
-        if '' not in vals_to_sample.keys():
-            vals_to_sample[''] = set()
-        vals_to_remove = {k: v for k,
-                          v in vals_to_sample.items() if len(v) < resolution}
-        for val, ids in vals_to_remove.items():
-            vals_to_sample[''].update(ids)
-            if val != '':
-                del vals_to_sample[val]
-        val_to_counts = {k: len(v) for k, v in vals_to_sample.items()}
-        sampled_val = sampleFromCounts(val_to_counts, True)
-        if sampled_val != None:
-            output_atts.append((col, sampled_val))
-            residual_ids = set(vals_to_sample[sampled_val])
-
-    filters = tuple(
-        sorted(output_atts, key=lambda x: f'{x[0]}:{x[1]}'.lower()))
-    return filters, row_cache_hits, row_cache_misses
-
-
-def sampleFromCounts(counts, preferNotNone):
-    """Samples from a dict of counts based on their relative frequency
-
-    Args:
-        counts: the dict of counts.
-        preferNotNone: whether to avoid sampling None if possible
-
-    Returns:
-        sampled_value: the sampled value
-    """
-    dist = convertCountsToCumulativeDistribution(counts)
-    sampled_value = None
-    r = random()
-    keys = sorted(dist.keys())
-    for p in keys:
-        v = dist[p]
-        if r < p:
-            if preferNotNone and v == None:
-                continue
-            else:
-                sampled_value = v
-                break
-        sampled_value = v
-    return sampled_value
-
-
-def convertCountsToCumulativeDistribution(att_to_count):
-    """Converts a dict of counts to a cumulative probability distribution for sampling.
-
-    Args:
-        att_to_count: a dict of attribute counts.
-
-    Returns:
-        dist a cumulative probability distribution mapping cumulative probabilities [0,1] to attributes. 
-    """
-    dist = {}
-    total = sum(att_to_count.values())
-    if total == 0:
-        return dist
-    cumulative = 0
-    for att, count in att_to_count.items():
-        if count > 0:
-            p = count/total
-            cumulative += p
-            dist[cumulative] = att
-    return dist
-
-
-def normalize(columns, atts):
-    """Creates an output record according to the columns schema from a set of attributes.
-
-    Args:
-        columns: the columns of the output record.
-        atts: the attribute values of the output record.
-
-    Returns:
-        row: a normalized row ready for dataframe integration.
-    """
-    row = []
-    for c in columns:
-        added = False
-        for a in atts:
-            if a[0] == c:
-                row.append(a[1])
-                added = True
-                break
-        if not added:
-            row.append('')
-    return row

packages/python-pipeline/src/showcase.py

@@ -1,6 +1,7 @@
 import json
 import logging
 import argparse
+import sds
 from os import path, mkdir
 import aggregator as aggregator
 import generator as generator
@@ -49,7 +50,7 @@ def main():
             config['evaluate'] = True
             config['navigate'] = True
 
-         # set based on the number of cores/memory available
+        # set based on the number of cores/memory available
         config['parallel_jobs'] = config.get('parallel_jobs', 1)
         config['memory_limit_pct'] = config.get('memory_limit_pct', 80)
         config['cache_max_size'] = config.get('cache_max_size', 100000)
@@ -102,6 +103,9 @@ def runPipeline(config):
         config: options from the json config file, else default values.
     """
 
+    # set number of threads to be used for processing
+    sds.set_number_of_threads(config['parallel_jobs'])
+
     if config['aggregate']:
         aggregator.aggregate(config)
 

packages/python-pipeline/src/util.py

@@ -1,14 +1,10 @@
 
 import matplotlib.ticker as ticker
 import matplotlib.pyplot as plt
-import logging
 import numpy as np
 import pandas as pd
-import joblib
-from itertools import combinations
-from collections import defaultdict
 import seaborn as sns
-from math import ceil, floor
+from math import ceil
 import matplotlib
 # fixes matplotlib + joblib bug "RuntimeError: main thread is not in main loop Tcl_AsyncDelete: async handler deleted by the wrong thread"
 matplotlib.use('Agg')
@@ -36,219 +32,6 @@ def loadMicrodata(path, delimiter, record_limit, use_columns):
     return df
 
 
-def genRowList(df, sensitive_zeros):
-    """Converts a dataframe to a list of rows.
-
-    Args:
-        df: the dataframe.
-        sensitive_zeros: columns for which negative values should be controlled.
-
-    Returns:
-        row_list: the list of rows.
-    """
-    row_list = []
-    for _, row in df.iterrows():
-        res = []
-        for c in df.columns:
-            val = str(row[c])
-            if val != '' and (c in sensitive_zeros or val != '0'):
-                res.append((c, val))
-        row2 = sorted(res, key=lambda x: f'{x[0]}:{x[1]}'.lower())
-        row_list.append(row2)
-    return row_list
-
-
-def genColValIdsDict(df, sensitive_zeros):
-    """Converts a dataframe to a dict of col->val->ids.
-
-    Args:
-        df: the dataframe
-
-    Returns:
-        colValIds: the dict of col->val->ids.
-    """
-    colValIds = {}
-    for rid, row in df.iterrows():
-        for c in df.columns:
-            val = str(row[c])
-            if c not in colValIds.keys():
-                colValIds[c] = defaultdict(list)
-            if val == '0':
-                if c in sensitive_zeros:
-                    colValIds[c][val].append(rid)
-                else:
-                    colValIds[c][''].append(rid)
-            else:
-                colValIds[c][val].append(rid)
-
-    return colValIds
-
-
-def countAllCombos(row_list, length_limit, parallel_jobs):
-    """Counts all combinations in the given rows up to a limit.
-
-    Args:
-        row_list: a list of rows.
-        length_limit: the maximum length to compute counts for (all lengths if -1).
-        parallel_jobs: the number of processor cores to use for parallelized counting.
-
-    Returns:
-        length_to_combo_to_count: a dict mapping combination lengths to dicts mapping combinations to counts. 
-    """
-    length_to_combo_to_count = {}
-    if length_limit == -1:
-        length_limit = max([len(x) for x in row_list])
-    for length in range(1, length_limit+1):
-        logging.info(f'counting combos of length {length}')
-        res = joblib.Parallel(n_jobs=parallel_jobs, backend='loky', verbose=1)(
-            joblib.delayed(genAllCombos)(row, length) for row in row_list)
-        length_to_combo_to_count[length] = defaultdict(int)
-        for combos in res:
-            for combo in combos:
-                length_to_combo_to_count.get(length, {})[combo] += 1
-
-    return length_to_combo_to_count
-
-
-def genAllCombos(row, length):
-    """Generates all combos from row up to and including size length.
-
-    Args:
-        row: the row to extract combinations from.
-        length: the maximum combination length to extract.
-
-    Returns:
-        combos: list of combinations extracted from row.
-    """
-    res = []
-    if len(row) == length:
-        canonical_combo = tuple(
-            sorted(list(row), key=lambda x: f'{x[0]}:{x[1]}'.lower()))
-        res.append(canonical_combo)
-    else:
-        for combo in combinations(row, length):
-            canonical_combo = tuple(
-                sorted(list(combo), key=lambda x: f'{x[0]}:{x[1]}'.lower()))
-            res.append(canonical_combo)
-    return res
-
-
-def mapShortestUniqueRareComboLengthToRecords(records, length_to_rare):
-    """
-    Maps each record to the shortest combination length that isolates it within a rare group (i.e., below resolution).
-
-    Args:
-        records: the input records.
-        length_to_rare: a dict of length to rare combo to count.
-
-    Returns:
-        rare_to_records: dict of rare combination lengths mapped to record lists
-    """
-    rare_to_records = defaultdict(set)
-    unique_to_records = defaultdict(set)
-    length_to_combo_to_rare = {length: defaultdict(
-        set) for length in length_to_rare.keys()}
-    for i, record in enumerate(records):
-        matchedRare = False
-        matchedUnique = False
-        for length in sorted(length_to_rare.keys()):
-            if matchedUnique:
-                break
-            for combo in combinations(record, length):
-                canonical_combo = tuple(
-                    sorted(list(combo), key=lambda x: f'{x[0]}:{x[1]}'.lower()))
-                if canonical_combo in length_to_rare.get(length, {}).keys():
-                    # unique
-                    if length_to_rare.get(length, {})[canonical_combo] == 1:
-                        unique_to_records[length].add(i)
-                        matchedUnique = True
-                        length_to_combo_to_rare.get(
-                            length, {})[canonical_combo].add(i)
-                        if i in rare_to_records[length]:
-                            rare_to_records[length].remove(i)
-
-                    else:
-                        if i not in unique_to_records[length]:
-                            rare_to_records[length].add(i)
-                        matchedRare = True
-                        length_to_combo_to_rare.get(
-                            length, {})[canonical_combo].add(i)
-
-            if matchedUnique:
-                break
-        if not matchedRare and not matchedUnique:
-            rare_to_records[0].add(i)
-        if not matchedUnique:
-            unique_to_records[0].add(i)
-    return unique_to_records, rare_to_records, length_to_combo_to_rare
-
-
-def protect(value, resolution):
-    """Protects a value from a privacy perspective by rounding down to the closest multiple of the supplied resolution.
-
-    Args:
-        value: the value to protect.
-        resolution: round values down to the closest multiple of this.
-    """
-    rounded = floor(value / resolution) * resolution
-    return rounded
-
-
-def comboToString(combo_tuple):
-    """Creates a string from a tuple of (attribute, value) tuples.
-
-    Args:
-        combo_tuple: tuple of (attribute, value) tuples.
-
-    Returns:
-        combo_string: a ;-delimited string of ':'-concatenated attribute-value pairs.
-    """
-    combo_string = ''
-    for (col, val) in combo_tuple:
-        combo_string += f'{col}:{val};'
-    return combo_string[:-1]
-
-
-def loadSavedAggregates(path):
-    """Loads sensitive aggregates from file to speed up evaluation.
-
-    Args:
-        path: location of the saved sensitive aggregates.
-
-    Returns:
-        length_to_combo_to_count: a dict mapping combination lengths to dicts mapping combinations to counts
-    """
-    length_to_combo_to_count = defaultdict(dict)
-    with open(path, 'r') as f:
-        for i, line in enumerate(f):
-            if i == 0:
-                continue  # skip header
-            parts = [x.strip() for x in line.split('\t')]
-            if len(parts[0]) > 0:
-                length, combo = stringToLengthAndCombo(parts[0])
-                length_to_combo_to_count[length][combo] = int(parts[1])
-    return length_to_combo_to_count
-
-
-def stringToLengthAndCombo(combo_string):
-    """Creates a tuple of (attribute, value) tuples from a given string.
-
-    Args:
-        combo_string: string representation of (attribute, value) tuples.
-
-    Returns:
-        combo_tuple: tuple of (attribute, value) tuples.
-    """
-    length = len(combo_string.split(';'))
-    combo_list = []
-    for col_vals in combo_string.split(';'):
-        parts = col_vals.split(':')
-        if len(parts) == 2:
-            combo_list.append((parts[0], parts[1]))
-    combo_tuple = tuple(combo_list)
-    return length, combo_tuple
-
-
 def plotStats(
         x_axis, x_axis_title, y_bar, y_bar_title, y_line, y_line_title, color, darker_color, stats_tsv, stats_svg,
         delimiter, style='whitegrid', palette='magma'):

packages/webapp/src/components/AttributeSelector/ColumnAttributeSelector.tsx

@@ -2,31 +2,24 @@
  * Copyright (c) Microsoft. All rights reserved.
  * Licensed under the MIT license. See LICENSE file in the project.
  */
-import { Stack, Spinner, Label } from '@fluentui/react'
-import _ from 'lodash'
+import { Stack, Label, Spinner } from '@fluentui/react'
 import { memo, useCallback, useEffect, useRef, useState } from 'react'
-import { useOnAttributeSelection } from './hooks'
+import { IAttributesIntersection } from 'sds-wasm'
+import { useMaxCount } from './hooks'
 import { AttributeIntersectionValueChart } from '~components/Charts/AttributeIntersectionValueChart'
 import { useStopPropagation } from '~components/Charts/hooks'
-import { CsvRecord, IAttributesIntersectionValue } from '~models'
-import {
-	useEvaluatedResultValue,
-	useNavigateResultValue,
-	useWasmWorkerValue,
-} from '~states'
-import {
-	useSelectedAttributeRows,
-	useSelectedAttributes,
-} from '~states/dataShowcaseContext'
+import { SetSelectedAttributesCallback } from '~components/Pages/DataShowcasePage/DataNavigation'
+import { useWasmWorkerValue } from '~states'
 
 export interface ColumnAttributeSelectorProps {
-	headers: CsvRecord
 	headerName: string
 	columnIndex: number
 	height: string | number
-	width: number | number
-	chartHeight: number
+	width: string | number
+	chartBarHeight: number
 	minHeight?: string | number
+	selectedAttributes: Set<string>
+	onSetSelectedAttributes: SetSelectedAttributesCallback
 }
 
 // fixed value for rough axis height so charts don't squish if selected down to 1
@@ -34,87 +27,49 @@ const AXIS_HEIGHT = 16
 
 export const ColumnAttributeSelector: React.FC<ColumnAttributeSelectorProps> =
 	memo(function ColumnAttributeSelector({
-		headers,
 		headerName,
 		columnIndex,
 		height,
 		width,
-		chartHeight,
+		chartBarHeight,
 		minHeight,
+		selectedAttributes,
+		onSetSelectedAttributes,
 	}: ColumnAttributeSelectorProps) {
-		const [selectedAttributeRows, setSelectedAttributeRows] =
-			useSelectedAttributeRows()
-		const [selectedAttributes, setSelectedAttributes] = useSelectedAttributes()
-		const navigateResult = useNavigateResultValue()
-		const [items, setItems] = useState<IAttributesIntersectionValue[]>([])
+		const [items, setItems] = useState<IAttributesIntersection[]>([])
 		const [isLoading, setIsLoading] = useState(false)
-		const isMounted = useRef(true)
-		const evaluatedResult = useEvaluatedResultValue()
 		const worker = useWasmWorkerValue()
-		const selectedAttribute = selectedAttributes[columnIndex]
-		const maxCount =
-			_.max([
-				_.maxBy(items, item => item.estimatedCount)?.estimatedCount,
-				_.maxBy(items, item => item.actualCount ?? 0)?.actualCount,
-			]) ?? 1
-
-		const onAttributeSelection = useOnAttributeSelection(
-			setIsLoading,
-			selectedAttributes,
-			worker,
-			navigateResult,
-			isMounted,
-			setSelectedAttributes,
-			setSelectedAttributeRows,
-		)
+		const maxCount = useMaxCount(items)
+		const isMounted = useRef(true)
+		const stopPropagation = useStopPropagation()
 
 		const handleSelection = useCallback(
-			(item: IAttributesIntersectionValue | undefined) => {
+			async (item: IAttributesIntersection | undefined) => {
 				const newValue = item?.value
 				// toggle off with re-click
-				if (newValue === selectedAttribute) {
-					onAttributeSelection(columnIndex, undefined)
+				if (newValue === undefined || selectedAttributes.has(newValue)) {
+					await onSetSelectedAttributes(columnIndex, undefined)
 				} else {
-					onAttributeSelection(columnIndex, newValue)
+					await onSetSelectedAttributes(columnIndex, item)
 				}
 			},
-			[selectedAttribute, columnIndex, onAttributeSelection],
+			[selectedAttributes, onSetSelectedAttributes, columnIndex],
 		)
 
-		const stopPropagation = useStopPropagation()
-
 		useEffect(() => {
 			if (worker) {
 				setIsLoading(true)
 				worker
-					.intersectAttributesInColumnsWith(
-						headers,
-						navigateResult.allRows,
-						navigateResult.attrsInColumnsMap[columnIndex] ?? new Set(),
-						selectedAttributeRows,
-						selectedAttributes,
-						navigateResult.attrRowsMap,
-						columnIndex,
-						evaluatedResult.sensitiveAggregatedResult?.aggregatedCombinations,
-					)
-					.then(newItems => {
-						if (isMounted.current) {
-							setItems(newItems ?? [])
-							setIsLoading(false)
+					.attributesIntersectionsByColumn([headerName])
+					.then(intersections => {
+						if (!isMounted.current || !intersections) {
+							return
 						}
+						setItems(intersections[columnIndex] ?? [])
+						setIsLoading(false)
 					})
 			}
-		}, [
-			worker,
-			navigateResult,
-			headers,
-			selectedAttributeRows,
-			selectedAttributes,
-			columnIndex,
-			evaluatedResult,
-			setItems,
-			setIsLoading,
-		])
+		}, [worker, setIsLoading, setItems, headerName, columnIndex])
 
 		useEffect(() => {
 			return () => {
@@ -151,8 +106,8 @@ export const ColumnAttributeSelector: React.FC<ColumnAttributeSelectorProps> =
 							items={items}
 							onClick={handleSelection}
 							maxCount={maxCount}
-							height={chartHeight * Math.max(items.length, 1) + AXIS_HEIGHT}
-							selectedValue={selectedAttribute}
+							height={chartBarHeight * Math.max(items.length, 1) + AXIS_HEIGHT}
+							selectedAttributes={selectedAttributes}
 						/>
 					</Stack.Item>
 				)}