2018-05-09 07:39:16 +03:00
|
|
|
# frozen_string_literal: true
|
|
|
|
|
|
|
|
require "forwardable"
|
|
|
|
|
|
|
|
class CSV
|
|
|
|
#
|
|
|
|
# A CSV::Table is a two-dimensional data structure for representing CSV
|
2019-10-12 08:03:21 +03:00
|
|
|
# documents. Tables allow you to work with the data by row or column,
|
2018-05-09 07:39:16 +03:00
|
|
|
# manipulate the data, and even convert the results back to CSV, if needed.
|
|
|
|
#
|
|
|
|
# All tables returned by CSV will be constructed from this class, if header
|
|
|
|
# row processing is activated.
|
|
|
|
#
|
|
|
|
class Table
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Constructs a new CSV::Table from +array_of_rows+, which are expected
|
|
|
|
# to be CSV::Row objects. All rows are assumed to have the same headers.
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2018-12-23 10:00:35 +03:00
|
|
|
# The optional +headers+ parameter can be set to Array of headers.
|
|
|
|
# If headers aren't set, headers are fetched from CSV::Row objects.
|
|
|
|
# Otherwise, headers() method will return headers being set in
|
2019-01-25 09:49:59 +03:00
|
|
|
# headers argument.
|
2018-12-23 10:00:35 +03:00
|
|
|
#
|
2018-05-09 07:39:16 +03:00
|
|
|
# A CSV::Table object supports the following Array methods through
|
|
|
|
# delegation:
|
|
|
|
#
|
|
|
|
# * empty?()
|
|
|
|
# * length()
|
|
|
|
# * size()
|
|
|
|
#
|
2018-12-23 10:00:35 +03:00
|
|
|
def initialize(array_of_rows, headers: nil)
|
2018-05-09 07:39:16 +03:00
|
|
|
@table = array_of_rows
|
2018-12-23 10:00:35 +03:00
|
|
|
@headers = headers
|
|
|
|
unless @headers
|
|
|
|
if @table.empty?
|
|
|
|
@headers = []
|
|
|
|
else
|
|
|
|
@headers = @table.first.headers
|
|
|
|
end
|
|
|
|
end
|
|
|
|
|
2018-05-09 07:39:16 +03:00
|
|
|
@mode = :col_or_row
|
|
|
|
end
|
|
|
|
|
|
|
|
# The current access mode for indexing and iteration.
|
|
|
|
attr_reader :mode
|
|
|
|
|
|
|
|
# Internal data format used to compare equality.
|
|
|
|
attr_reader :table
|
|
|
|
protected :table
|
|
|
|
|
|
|
|
### Array Delegation ###
|
|
|
|
|
|
|
|
extend Forwardable
|
|
|
|
def_delegators :@table, :empty?, :length, :size
|
|
|
|
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Returns a duplicate table object, in column mode. This is handy for
|
2018-05-09 07:39:16 +03:00
|
|
|
# chaining in a single call without changing the table mode, but be aware
|
|
|
|
# that this method can consume a fair amount of memory for bigger data sets.
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# This method returns the duplicate table for chaining. Don't chain
|
2018-05-09 07:39:16 +03:00
|
|
|
# destructive methods (like []=()) this way though, since you are working
|
|
|
|
# with a duplicate.
|
|
|
|
#
|
|
|
|
def by_col
|
|
|
|
self.class.new(@table.dup).by_col!
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Switches the mode of this table to column mode. All calls to indexing and
|
2018-05-09 07:39:16 +03:00
|
|
|
# iteration methods will work with columns until the mode is changed again.
|
|
|
|
#
|
|
|
|
# This method returns the table and is safe to chain.
|
|
|
|
#
|
|
|
|
def by_col!
|
|
|
|
@mode = :col
|
|
|
|
|
|
|
|
self
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Returns a duplicate table object, in mixed mode. This is handy for
|
2018-05-09 07:39:16 +03:00
|
|
|
# chaining in a single call without changing the table mode, but be aware
|
|
|
|
# that this method can consume a fair amount of memory for bigger data sets.
|
|
|
|
#
|
|
|
|
# This method returns the duplicate table for chaining. Don't chain
|
|
|
|
# destructive methods (like []=()) this way though, since you are working
|
|
|
|
# with a duplicate.
|
|
|
|
#
|
|
|
|
def by_col_or_row
|
|
|
|
self.class.new(@table.dup).by_col_or_row!
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Switches the mode of this table to mixed mode. All calls to indexing and
|
2018-05-09 07:39:16 +03:00
|
|
|
# iteration methods will use the default intelligent indexing system until
|
2019-10-12 08:03:21 +03:00
|
|
|
# the mode is changed again. In mixed mode an index is assumed to be a row
|
2018-05-09 07:39:16 +03:00
|
|
|
# reference while anything else is assumed to be column access by headers.
|
|
|
|
#
|
|
|
|
# This method returns the table and is safe to chain.
|
|
|
|
#
|
|
|
|
def by_col_or_row!
|
|
|
|
@mode = :col_or_row
|
|
|
|
|
|
|
|
self
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
|
|
|
# Returns a duplicate table object, in row mode. This is handy for chaining
|
|
|
|
# in a single call without changing the table mode, but be aware that this
|
|
|
|
# method can consume a fair amount of memory for bigger data sets.
|
|
|
|
#
|
|
|
|
# This method returns the duplicate table for chaining. Don't chain
|
|
|
|
# destructive methods (like []=()) this way though, since you are working
|
|
|
|
# with a duplicate.
|
|
|
|
#
|
|
|
|
def by_row
|
|
|
|
self.class.new(@table.dup).by_row!
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Switches the mode of this table to row mode. All calls to indexing and
|
2018-05-09 07:39:16 +03:00
|
|
|
# iteration methods will work with rows until the mode is changed again.
|
|
|
|
#
|
|
|
|
# This method returns the table and is safe to chain.
|
|
|
|
#
|
|
|
|
def by_row!
|
|
|
|
@mode = :row
|
|
|
|
|
|
|
|
self
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
|
|
|
# Returns the headers for the first row of this table (assumed to match all
|
2019-01-25 09:49:59 +03:00
|
|
|
# other rows). The headers Array passed to CSV::Table.new is returned for
|
|
|
|
# empty tables.
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
|
|
|
def headers
|
2019-01-25 09:49:59 +03:00
|
|
|
if @table.empty?
|
|
|
|
@headers.dup
|
|
|
|
else
|
|
|
|
@table.first.headers
|
|
|
|
end
|
2018-05-09 07:39:16 +03:00
|
|
|
end
|
|
|
|
|
2020-08-05 00:34:02 +03:00
|
|
|
# :call-seq:
|
|
|
|
# table[n] -> row
|
|
|
|
# table[range] -> array_of_rows
|
|
|
|
# table[header] -> array_of_fields
|
|
|
|
#
|
|
|
|
# Returns data from the table; does not modify the table.
|
|
|
|
#
|
|
|
|
# ---
|
|
|
|
#
|
|
|
|
# The expression <tt>table[n]</tt>, where +n+ is a non-negative \Integer,
|
|
|
|
# returns the +n+th row of the table, if that row exists,
|
|
|
|
# and if the access mode is <tt>:row</tt> or <tt>:col_or_row</tt>:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table.by_row! # => #<CSV::Table mode:row row_count:4>
|
|
|
|
# table[1] # => #<CSV::Row "Name":"bar" "Value":"1">
|
|
|
|
# table.by_col_or_row! # => #<CSV::Table mode:col_or_row row_count:4>
|
|
|
|
# table[1] # => #<CSV::Row "Name":"bar" "Value":"1">
|
|
|
|
#
|
|
|
|
# Counts backward from the last row if +n+ is negative:
|
|
|
|
# table[-1] # => #<CSV::Row "Name":"baz" "Value":"2">
|
|
|
|
#
|
|
|
|
# Returns +nil+ if +n+ is too large or too small:
|
|
|
|
# table[4] # => nil
|
|
|
|
# table[-4] => nil
|
|
|
|
#
|
|
|
|
# Raises an exception if the access mode is <tt>:row</tt>
|
|
|
|
# and +n+ is not an
|
|
|
|
# {Integer-convertible object}[https://docs.ruby-lang.org/en/master/implicit_conversion_rdoc.html#label-Integer-Convertible+Objects].
|
|
|
|
# table.by_row! # => #<CSV::Table mode:row row_count:4>
|
|
|
|
# # Raises TypeError (no implicit conversion of String into Integer):
|
|
|
|
# table['Name']
|
|
|
|
#
|
|
|
|
# ---
|
|
|
|
#
|
|
|
|
# The expression <tt>table[range]</tt>, where +range+ is a Range object,
|
|
|
|
# returns rows from the table, beginning at row <tt>range.first</tt>,
|
|
|
|
# if those rows exist, and if the access mode is <tt>:row</tt> or <tt>:col_or_row</tt>:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table.by_row! # => #<CSV::Table mode:row row_count:4>
|
|
|
|
# rows = table[1..2] # => #<CSV::Row "Name":"bar" "Value":"1">
|
|
|
|
# rows # => [#<CSV::Row "Name":"bar" "Value":"1">, #<CSV::Row "Name":"baz" "Value":"2">]
|
|
|
|
# table.by_col_or_row! # => #<CSV::Table mode:col_or_row row_count:4>
|
|
|
|
# rows = table[1..2] # => #<CSV::Row "Name":"bar" "Value":"1">
|
|
|
|
# rows # => [#<CSV::Row "Name":"bar" "Value":"1">, #<CSV::Row "Name":"baz" "Value":"2">]
|
|
|
|
#
|
|
|
|
# If there are too few rows, returns all from <tt>range.first</tt> to the end:
|
|
|
|
# rows = table[1..50] # => #<CSV::Row "Name":"bar" "Value":"1">
|
|
|
|
# rows # => [#<CSV::Row "Name":"bar" "Value":"1">, #<CSV::Row "Name":"baz" "Value":"2">]
|
|
|
|
#
|
|
|
|
# Special case: if <tt>range.start == table.size</tt>, returns an empty \Array:
|
|
|
|
# table[table.size..50] # => []
|
|
|
|
#
|
|
|
|
# If <tt>range.end</tt> is negative, calculates the ending index from the end:
|
|
|
|
# rows = table[0..-1]
|
|
|
|
# rows # => [#<CSV::Row "Name":"foo" "Value":"0">, #<CSV::Row "Name":"bar" "Value":"1">, #<CSV::Row "Name":"baz" "Value":"2">]
|
|
|
|
#
|
|
|
|
# If <tt>range.start</tt> is negative, calculates the starting index from the end:
|
|
|
|
# rows = table[-1..2]
|
|
|
|
# rows # => [#<CSV::Row "Name":"baz" "Value":"2">]
|
|
|
|
#
|
|
|
|
# If <tt>range.start</tt> is larger than <tt>table.size</tt>, returns +nil+:
|
|
|
|
# table[4..4] # => nil
|
|
|
|
#
|
|
|
|
# ---
|
|
|
|
#
|
|
|
|
# The expression <tt>table[header]</tt>, where +header+ is a \String,
|
|
|
|
# returns column values (\Array of \Strings) if the column exists
|
|
|
|
# and if the access mode is <tt>:col</tt> or <tt>:col_or_row</tt>:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table.by_col! # => #<CSV::Table mode:col row_count:4>
|
|
|
|
# table['Name'] # => ["foo", "bar", "baz"]
|
|
|
|
# table.by_col_or_row! # => #<CSV::Table mode:col_or_row row_count:4>
|
|
|
|
# col = table['Name']
|
|
|
|
# col # => ["foo", "bar", "baz"]
|
|
|
|
#
|
|
|
|
# Modifying the returned column values does not modify the table:
|
|
|
|
# col[0] = 'bat'
|
|
|
|
# col # => ["bat", "bar", "baz"]
|
|
|
|
# table['Name'] # => ["foo", "bar", "baz"]
|
|
|
|
#
|
|
|
|
# Returns an \Array of +nil+ values if there is no such column:
|
|
|
|
# table['Nosuch'] # => [nil, nil, nil]
|
2018-05-09 07:39:16 +03:00
|
|
|
def [](index_or_header)
|
|
|
|
if @mode == :row or # by index
|
|
|
|
(@mode == :col_or_row and (index_or_header.is_a?(Integer) or index_or_header.is_a?(Range)))
|
|
|
|
@table[index_or_header]
|
|
|
|
else # by header
|
|
|
|
@table.map { |row| row[index_or_header] }
|
|
|
|
end
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
|
|
|
# In the default mixed mode, this method assigns rows for index access and
|
2019-10-12 08:03:21 +03:00
|
|
|
# columns for header access. You can force the index association by first
|
2018-05-09 07:39:16 +03:00
|
|
|
# calling by_col!() or by_row!().
|
|
|
|
#
|
|
|
|
# Rows may be set to an Array of values (which will inherit the table's
|
|
|
|
# headers()) or a CSV::Row.
|
|
|
|
#
|
|
|
|
# Columns may be set to a single value, which is copied to each row of the
|
2019-10-12 08:03:21 +03:00
|
|
|
# column, or an Array of values. Arrays of values are assigned to rows top
|
|
|
|
# to bottom in row major order. Excess values are ignored and if the Array
|
2018-05-09 07:39:16 +03:00
|
|
|
# does not have a value for each row the extra rows will receive a +nil+.
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Assigning to an existing column or row clobbers the data. Assigning to
|
2018-05-09 07:39:16 +03:00
|
|
|
# new columns creates them at the right end of the table.
|
|
|
|
#
|
|
|
|
def []=(index_or_header, value)
|
|
|
|
if @mode == :row or # by index
|
|
|
|
(@mode == :col_or_row and index_or_header.is_a? Integer)
|
|
|
|
if value.is_a? Array
|
|
|
|
@table[index_or_header] = Row.new(headers, value)
|
|
|
|
else
|
|
|
|
@table[index_or_header] = value
|
|
|
|
end
|
|
|
|
else # set column
|
2018-12-23 10:00:35 +03:00
|
|
|
unless index_or_header.is_a? Integer
|
|
|
|
index = @headers.index(index_or_header) || @headers.size
|
|
|
|
@headers[index] = index_or_header
|
|
|
|
end
|
2018-05-09 07:39:16 +03:00
|
|
|
if value.is_a? Array # multiple values
|
|
|
|
@table.each_with_index do |row, i|
|
|
|
|
if row.header_row?
|
|
|
|
row[index_or_header] = index_or_header
|
|
|
|
else
|
|
|
|
row[index_or_header] = value[i]
|
|
|
|
end
|
|
|
|
end
|
|
|
|
else # repeated value
|
|
|
|
@table.each do |row|
|
|
|
|
if row.header_row?
|
|
|
|
row[index_or_header] = index_or_header
|
|
|
|
else
|
|
|
|
row[index_or_header] = value
|
|
|
|
end
|
|
|
|
end
|
|
|
|
end
|
|
|
|
end
|
|
|
|
end
|
|
|
|
|
2020-08-12 03:48:00 +03:00
|
|
|
# :call-seq:
|
|
|
|
# table.values_at(*indexes) -> array_of_rows
|
|
|
|
# table.values_at(*headers) -> array_of_columns_data
|
|
|
|
#
|
|
|
|
# If the access mode is <tt>:row</tt> or <tt>:col_or_row</tt>,
|
|
|
|
# and each argument is either an \Integer or a \Range,
|
|
|
|
# returns rows.
|
|
|
|
# Otherwise, returns columns data.
|
|
|
|
#
|
|
|
|
# In either case, the returned values are in the order
|
|
|
|
# specified by the arguments. Arguments may be repeated.
|
|
|
|
#
|
|
|
|
# ---
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-12 03:48:00 +03:00
|
|
|
# Returns rows as an \Array of \CSV::Row objects.
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-12 03:48:00 +03:00
|
|
|
# No argument:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table.values_at # => []
|
|
|
|
#
|
|
|
|
# One index:
|
|
|
|
# values = table.values_at(0)
|
|
|
|
# values # => [#<CSV::Row "Name":"foo" "Value":"0">]
|
|
|
|
#
|
|
|
|
# Two indexes:
|
|
|
|
# values = table.values_at(2, 0)
|
|
|
|
# values # => [#<CSV::Row "Name":"baz" "Value":"2">, #<CSV::Row "Name":"foo" "Value":"0">]
|
|
|
|
#
|
|
|
|
# One \Range:
|
|
|
|
# values = table.values_at(1..2)
|
|
|
|
# values # => [#<CSV::Row "Name":"bar" "Value":"1">, #<CSV::Row "Name":"baz" "Value":"2">]
|
|
|
|
#
|
|
|
|
# \Ranges and indexes:
|
|
|
|
# values = table.values_at(0..1, 1..2, 0, 2)
|
|
|
|
# pp values
|
|
|
|
# Output:
|
|
|
|
# [#<CSV::Row "Name":"foo" "Value":"0">,
|
|
|
|
# #<CSV::Row "Name":"bar" "Value":"1">,
|
|
|
|
# #<CSV::Row "Name":"bar" "Value":"1">,
|
|
|
|
# #<CSV::Row "Name":"baz" "Value":"2">,
|
|
|
|
# #<CSV::Row "Name":"foo" "Value":"0">,
|
|
|
|
# #<CSV::Row "Name":"baz" "Value":"2">]
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-12 03:48:00 +03:00
|
|
|
# ---
|
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# Returns columns data as row Arrays,
|
2020-08-12 03:48:00 +03:00
|
|
|
# each consisting of the specified columns data for that row:
|
|
|
|
# values = table.values_at('Name')
|
|
|
|
# values # => [["foo"], ["bar"], ["baz"]]
|
|
|
|
# values = table.values_at('Value', 'Name')
|
|
|
|
# values # => [["0", "foo"], ["1", "bar"], ["2", "baz"]]
|
2018-05-09 07:39:16 +03:00
|
|
|
def values_at(*indices_or_headers)
|
|
|
|
if @mode == :row or # by indices
|
|
|
|
( @mode == :col_or_row and indices_or_headers.all? do |index|
|
|
|
|
index.is_a?(Integer) or
|
|
|
|
( index.is_a?(Range) and
|
|
|
|
index.first.is_a?(Integer) and
|
|
|
|
index.last.is_a?(Integer) )
|
|
|
|
end )
|
|
|
|
@table.values_at(*indices_or_headers)
|
|
|
|
else # by headers
|
|
|
|
@table.map { |row| row.values_at(*indices_or_headers) }
|
|
|
|
end
|
|
|
|
end
|
|
|
|
|
2020-08-12 03:48:00 +03:00
|
|
|
# :call-seq:
|
|
|
|
# table << row_or_array -> self
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-12 03:48:00 +03:00
|
|
|
# If +row_or_array+ is a \CSV::Row object,
|
|
|
|
# it is appended to the table:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table << CSV::Row.new(table.headers, ['bat', 3])
|
|
|
|
# table[3] # => #<CSV::Row "Name":"bat" "Value":3>
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-12 03:48:00 +03:00
|
|
|
# If +row_or_array+ is an \Array, it is used to create a new
|
|
|
|
# \CSV::Row object which is then appended to the table:
|
|
|
|
# table << ['bam', 4]
|
|
|
|
# table[4] # => #<CSV::Row "Name":"bam" "Value":4>
|
2018-05-09 07:39:16 +03:00
|
|
|
def <<(row_or_array)
|
|
|
|
if row_or_array.is_a? Array # append Array
|
|
|
|
@table << Row.new(headers, row_or_array)
|
|
|
|
else # append Row
|
|
|
|
@table << row_or_array
|
|
|
|
end
|
|
|
|
|
|
|
|
self # for chaining
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
2020-08-12 03:48:00 +03:00
|
|
|
# :call-seq:
|
|
|
|
# table.push(*rows_or_arrays) -> self
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-12 03:48:00 +03:00
|
|
|
# A shortcut for appending multiple rows. Equivalent to:
|
|
|
|
# rows.each {|row| self << row }
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-12 03:48:00 +03:00
|
|
|
# Each argument may be either a \CSV::Row object or an \Array:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# rows = [
|
|
|
|
# CSV::Row.new(table.headers, ['bat', 3]),
|
|
|
|
# ['bam', 4]
|
|
|
|
# ]
|
|
|
|
# table.push(*rows)
|
|
|
|
# table[3..4] # => [#<CSV::Row "Name":"bat" "Value":3>, #<CSV::Row "Name":"bam" "Value":4>]
|
2018-05-09 07:39:16 +03:00
|
|
|
def push(*rows)
|
|
|
|
rows.each { |row| self << row }
|
|
|
|
|
|
|
|
self # for chaining
|
|
|
|
end
|
|
|
|
|
2020-08-17 23:55:27 +03:00
|
|
|
# :call-seq:
|
|
|
|
# table.delete(*indexes) -> deleted_values
|
|
|
|
# table.delete(*headers) -> deleted_values
|
|
|
|
#
|
|
|
|
# If the access mode is <tt>:row</tt> or <tt>:col_or_row</tt>,
|
|
|
|
# and each argument is either an \Integer or a \Range,
|
|
|
|
# returns deleted rows.
|
|
|
|
# Otherwise, returns deleted columns data.
|
|
|
|
#
|
|
|
|
# In either case, the returned values are in the order
|
|
|
|
# specified by the arguments. Arguments may be repeated.
|
|
|
|
#
|
|
|
|
# ---
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# Returns rows as an \Array of \CSV::Row objects.
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# One index:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# deleted_values = table.delete(0)
|
|
|
|
# deleted_values # => [#<CSV::Row "Name":"foo" "Value":"0">]
|
|
|
|
#
|
|
|
|
# Two indexes:
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# deleted_values = table.delete(2, 0)
|
|
|
|
# deleted_values # => [#<CSV::Row "Name":"baz" "Value":"2">, #<CSV::Row "Name":"foo" "Value":"0">]
|
|
|
|
#
|
|
|
|
# ---
|
|
|
|
#
|
|
|
|
# Returns columns data as column Arrays.
|
|
|
|
#
|
|
|
|
# One header:
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# deleted_values = table.delete('Name')
|
|
|
|
# deleted_values # => ["foo", "bar", "baz"]
|
|
|
|
#
|
|
|
|
# Two headers:
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# deleted_values = table.delete('Value', 'Name')
|
|
|
|
# deleted_values # => [["0", "1", "2"], ["foo", "bar", "baz"]]
|
2018-05-09 07:39:16 +03:00
|
|
|
def delete(*indexes_or_headers)
|
|
|
|
if indexes_or_headers.empty?
|
|
|
|
raise ArgumentError, "wrong number of arguments (given 0, expected 1+)"
|
|
|
|
end
|
|
|
|
deleted_values = indexes_or_headers.map do |index_or_header|
|
|
|
|
if @mode == :row or # by index
|
|
|
|
(@mode == :col_or_row and index_or_header.is_a? Integer)
|
|
|
|
@table.delete_at(index_or_header)
|
|
|
|
else # by header
|
2018-12-23 10:00:35 +03:00
|
|
|
if index_or_header.is_a? Integer
|
|
|
|
@headers.delete_at(index_or_header)
|
|
|
|
else
|
|
|
|
@headers.delete(index_or_header)
|
|
|
|
end
|
2018-05-09 07:39:16 +03:00
|
|
|
@table.map { |row| row.delete(index_or_header).last }
|
|
|
|
end
|
|
|
|
end
|
|
|
|
if indexes_or_headers.size == 1
|
|
|
|
deleted_values[0]
|
|
|
|
else
|
|
|
|
deleted_values
|
|
|
|
end
|
|
|
|
end
|
|
|
|
|
2020-08-17 23:55:27 +03:00
|
|
|
# Removes rows or columns for which the block returns a truthy value;
|
|
|
|
# returns +self+.
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# Removes rows when the access mode is <tt>:row</tt> or <tt>:col_or_row</tt>;
|
|
|
|
# calls the block with each \CSV::Row object:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table.by_row! # => #<CSV::Table mode:row row_count:4>
|
|
|
|
# table.size # => 3
|
|
|
|
# table.delete_if {|row| row['Name'].start_with?('b') }
|
|
|
|
# table.size # => 1
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# Removes columns when the access mode is <tt>:col</tt>;
|
|
|
|
# calls the block with each column as a 2-element array
|
|
|
|
# containing the header and an \Array of column fields:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table.by_col! # => #<CSV::Table mode:col row_count:4>
|
|
|
|
# table.headers.size # => 2
|
|
|
|
# table.delete_if {|column_data| column_data[1].include?('2') }
|
|
|
|
# table.headers.size # => 1
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# Returns a new \Enumerator if no block is given:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table.delete_if # => #<Enumerator: #<CSV::Table mode:col_or_row row_count:4>:delete_if>
|
2018-05-09 07:39:16 +03:00
|
|
|
def delete_if(&block)
|
|
|
|
return enum_for(__method__) { @mode == :row or @mode == :col_or_row ? size : headers.size } unless block_given?
|
|
|
|
|
|
|
|
if @mode == :row or @mode == :col_or_row # by index
|
|
|
|
@table.delete_if(&block)
|
|
|
|
else # by header
|
|
|
|
deleted = []
|
|
|
|
headers.each do |header|
|
|
|
|
deleted << delete(header) if yield([header, self[header]])
|
|
|
|
end
|
|
|
|
end
|
|
|
|
|
|
|
|
self # for chaining
|
|
|
|
end
|
|
|
|
|
|
|
|
include Enumerable
|
|
|
|
|
2020-08-17 23:55:27 +03:00
|
|
|
# Calls the block with each row or column; returns +self+.
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# When the access mode is <tt>:row</tt> or <tt>:col_or_row</tt>,
|
|
|
|
# calls the block with each \CSV::Row object:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# table.by_row! # => #<CSV::Table mode:row row_count:4>
|
|
|
|
# table.each {|row| p row }
|
|
|
|
# Output:
|
|
|
|
# #<CSV::Row "Name":"foo" "Value":"0">
|
|
|
|
# #<CSV::Row "Name":"bar" "Value":"1">
|
|
|
|
# #<CSV::Row "Name":"baz" "Value":"2">
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# When the access mode is <tt>:col</tt>,
|
|
|
|
# calls the block with each column as a 2-element array
|
|
|
|
# containing the header and an \Array of column fields:
|
|
|
|
# table.by_col! # => #<CSV::Table mode:col row_count:4>
|
|
|
|
# table.each {|column_data| p column_data }
|
|
|
|
# Output:
|
|
|
|
# ["Name", ["foo", "bar", "baz"]]
|
|
|
|
# ["Value", ["0", "1", "2"]]
|
2018-05-09 07:39:16 +03:00
|
|
|
#
|
2020-08-17 23:55:27 +03:00
|
|
|
# Returns a new \Enumerator if no block is given:
|
|
|
|
# table.each # => #<Enumerator: #<CSV::Table mode:col row_count:4>:each>
|
2018-05-09 07:39:16 +03:00
|
|
|
def each(&block)
|
|
|
|
return enum_for(__method__) { @mode == :col ? headers.size : size } unless block_given?
|
|
|
|
|
|
|
|
if @mode == :col
|
|
|
|
headers.each { |header| yield([header, self[header]]) }
|
|
|
|
else
|
|
|
|
@table.each(&block)
|
|
|
|
end
|
|
|
|
|
|
|
|
self # for chaining
|
|
|
|
end
|
|
|
|
|
2020-08-17 23:55:27 +03:00
|
|
|
# Returns +true+ if all each row of +self+ <tt>==</tt>
|
|
|
|
# the corresponding row of +other_table+, otherwise, +false+.
|
|
|
|
#
|
|
|
|
# The access mode does no affect the result.
|
|
|
|
#
|
|
|
|
# Equal tables:
|
|
|
|
# source = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
|
|
|
|
# table = CSV.parse(source, headers: true)
|
|
|
|
# other_table = CSV.parse(source, headers: true)
|
|
|
|
# table == other_table # => true
|
|
|
|
#
|
|
|
|
# Different row count:
|
|
|
|
# other_table.delete(2)
|
|
|
|
# table == other_table # => false
|
|
|
|
#
|
|
|
|
# Different last row:
|
|
|
|
# other_table << ['bat', 3]
|
|
|
|
# table == other_table # => false
|
2018-05-09 07:39:16 +03:00
|
|
|
def ==(other)
|
|
|
|
return @table == other.table if other.is_a? CSV::Table
|
|
|
|
@table == other
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Returns the table as an Array of Arrays. Headers will be the first row,
|
2018-05-09 07:39:16 +03:00
|
|
|
# then all of the field rows will follow.
|
|
|
|
#
|
|
|
|
def to_a
|
|
|
|
array = [headers]
|
|
|
|
@table.each do |row|
|
|
|
|
array.push(row.fields) unless row.header_row?
|
|
|
|
end
|
|
|
|
|
|
|
|
array
|
|
|
|
end
|
|
|
|
|
|
|
|
#
|
2019-10-12 08:03:21 +03:00
|
|
|
# Returns the table as a complete CSV String. Headers will be listed first,
|
2018-05-09 07:39:16 +03:00
|
|
|
# then all of the field rows.
|
|
|
|
#
|
|
|
|
# This method assumes you want the Table.headers(), unless you explicitly
|
|
|
|
# pass <tt>:write_headers => false</tt>.
|
|
|
|
#
|
|
|
|
def to_csv(write_headers: true, **options)
|
2019-04-08 02:44:49 +03:00
|
|
|
array = write_headers ? [headers.to_csv(**options)] : []
|
2018-05-09 07:39:16 +03:00
|
|
|
@table.each do |row|
|
2019-04-08 02:44:49 +03:00
|
|
|
array.push(row.fields.to_csv(**options)) unless row.header_row?
|
2018-05-09 07:39:16 +03:00
|
|
|
end
|
|
|
|
|
|
|
|
array.join("")
|
|
|
|
end
|
|
|
|
alias_method :to_s, :to_csv
|
|
|
|
|
|
|
|
#
|
|
|
|
# Extracts the nested value specified by the sequence of +index+ or +header+ objects by calling dig at each step,
|
|
|
|
# returning nil if any intermediate step is nil.
|
|
|
|
#
|
|
|
|
def dig(index_or_header, *index_or_headers)
|
|
|
|
value = self[index_or_header]
|
|
|
|
if value.nil?
|
|
|
|
nil
|
|
|
|
elsif index_or_headers.empty?
|
|
|
|
value
|
|
|
|
else
|
|
|
|
unless value.respond_to?(:dig)
|
|
|
|
raise TypeError, "#{value.class} does not have \#dig method"
|
|
|
|
end
|
|
|
|
value.dig(*index_or_headers)
|
|
|
|
end
|
|
|
|
end
|
|
|
|
|
|
|
|
# Shows the mode and size of this table in a US-ASCII String.
|
|
|
|
def inspect
|
|
|
|
"#<#{self.class} mode:#{@mode} row_count:#{to_a.size}>".encode("US-ASCII")
|
|
|
|
end
|
|
|
|
end
|
2018-09-05 16:33:22 +03:00
|
|
|
end
|