python_mozaggregator/mozaggregator/bigquery.py

139 строки
5.2 KiB
Python

import json
import gzip
from datetime import datetime, timedelta
from pyspark.sql import Row, SparkSession
class BigQueryDataset:
def __init__(self):
self.spark = SparkSession.builder.getOrCreate()
@staticmethod
def _date_add_days(date_ds, days):
dt = datetime.strptime(date_ds, "%Y%m%d")
return datetime.strftime(dt + timedelta(days), "%Y-%m-%d")
@staticmethod
def _extract_payload(row):
"""
The schema for the `payload_bytes_decoded` table is listed for reference.
root
|-- client_id: string (nullable = true)
|-- document_id: string (nullable = true)
|-- metadata: struct (nullable = true)
| |-- document_namespace: string (nullable = true)
| |-- document_type: string (nullable = true)
| |-- document_version: string (nullable = true)
| |-- geo: struct (nullable = true)
| | |-- city: string (nullable = true)
| | |-- country: string (nullable = true)
| | |-- db_version: string (nullable = true)
| | |-- subdivision1: string (nullable = true)
| | |-- subdivision2: string (nullable = true)
| |-- header: struct (nullable = true)
| | |-- date: string (nullable = true)
| | |-- dnt: string (nullable = true)
| | |-- x_debug_id: string (nullable = true)
| | |-- x_pingsender_version: string (nullable = true)
| |-- uri: struct (nullable = true)
| | |-- app_build_id: string (nullable = true)
| | |-- app_name: string (nullable = true)
| | |-- app_update_channel: string (nullable = true)
| | |-- app_version: string (nullable = true)
| |-- user_agent: struct (nullable = true)
| | |-- browser: string (nullable = true)
| | |-- os: string (nullable = true)
| | |-- version: string (nullable = true)
|-- normalized_app_name: string (nullable = true)
|-- normalized_channel: string (nullable = true)
|-- normalized_country_code: string (nullable = true)
|-- normalized_os: string (nullable = true)
|-- normalized_os_version: string (nullable = true)
|-- payload: binary (nullable = true)
|-- sample_id: long (nullable = true)
|-- submission_timestamp: timestamp (nullable = true)
"""
# Data is stored in payload_bytes_decoded as gzip.
data = json.loads(gzip.decompress(row.payload).decode("utf-8"))
# add `meta` fields for backwards compatibility
data["meta"] = {
"submissionDate": datetime.strftime(row.submission_timestamp, "%Y%m%d"),
"sampleId": row.sample_id,
# following 4 fields necessary for mobile_aggregates
"normalizedChannel": row.normalized_channel,
"appVersion": row.metadata.uri.app_version,
"appBuildId": row.metadata.uri.app_build_id,
"appName": row.metadata.uri.app_name,
}
return data
def load(
self,
project_id,
dataset_id,
doc_type,
submission_date,
channels=None,
filter_clause=None,
fraction=1,
doc_version="v4",
):
start = self._date_add_days(submission_date, 0)
end = self._date_add_days(submission_date, 1)
date_clause = (
f"submission_timestamp >= '{start}' AND submission_timestamp < '{end}'"
)
filters = [date_clause]
if channels:
# build up a clause like "(normalized_channel = 'nightly' OR normalized_channel = 'beta')"
clauses = [f"normalized_channel = '{channel}'" for channel in channels]
joined = f"({' OR '.join(clauses)})"
filters.append(joined)
if filter_clause:
filters.append(filter_clause)
df = (
self.spark.read.format("bigquery")
# Assumes the namespace is telemetry
.option(
"table",
f"{project_id}.{dataset_id}.telemetry_telemetry__{doc_type}_{doc_version}",
)
.option("filter", " AND ".join(filters))
.load()
)
# Size of the RDD sample is not deterministic
return df.rdd.map(self._extract_payload).sample(False, fraction)
def load_avro(
self,
prefix,
doc_type,
submission_date,
channels=None,
filter_clause=None,
doc_version="v4",
):
filters = []
if channels:
# build up a clause like "(normalized_channel = 'nightly' OR normalized_channel = 'beta')"
clauses = ' OR '.join([f"normalized_channel = '{channel}'" for channel in channels])
joined = f"({clauses})"
filters.append(joined)
if filter_clause:
filters.append(filter_clause)
df = self.spark.read.format("avro").load(
f"{prefix}/{submission_date}/{doc_type}_{doc_version}"
)
if filters:
df.where(" AND ".join(filters))
return df.rdd.map(self._extract_payload)