Class: CSV

CSV Data

CSV (comma-separated values) data is a text representation of a table:

• A row separator delimits table rows. A common row separator is the newline character "\n".

• A column separator delimits fields in a row. A common column separator is the comma character ",".

This CSV String, with row separator "\n" and column separator ",", has three rows and two columns:

"foo,0\nbar,1\nbaz,2\n"

Despite the name CSV, a CSV representation can use different separators.

For more about tables, see the Wikipedia article “) {CSV::Table (information)}”, especially its section “Simple table”

Class CSV

Class CSV provides methods for:

• Parsing CSV data from a String object, a File (via its file path), or an IO object.

• Generating CSV data to a String object.

To make CSV available:

require 'csv'

All examples here assume that this has been done.

Keeping It Simple

A CSV object has dozens of instance methods that offer fine-grained control of parsing and generating CSV data. For many needs, though, simpler approaches will do.

This section summarizes the singleton methods in CSV that allow you to parse and generate without explicitly creating CSV objects. For details, follow the links.

Simple Parsing

Parsing methods commonly return either of:

• An Array of Arrays of Strings:

  • The outer Array is the entire “table”.

  • Each inner Array is a row.

  • Each String is a field.

• A Table object. For details, see \ with Headers.

Parsing a String

The input to be parsed can be a string:

string = "foo,0\nbar,1\nbaz,2\n"

Method .parse returns the entire CSV data:

CSV.parse(string) # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

Method .parse_line returns only the first row:

CSV.parse_line(string) # => ["foo", "0"]

CSV extends class String with instance method String#parse_csv, which also returns only the first row:

string.parse_csv # => ["foo", "0"]

Parsing Via a File Path

The input to be parsed can be in a file:

string = "foo,0\nbar,1\nbaz,2\n"
path = 't.csv'
File.write(path, string)

Method .read returns the entire CSV data:

CSV.read(path) # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

Method .foreach iterates, passing each row to the given block:

CSV.foreach(path) do |row|
  p row
end

Output:

["foo", "0"]
["bar", "1"]
["baz", "2"]

Method .table returns the entire CSV data as a Table object:

CSV.table(path) # => #<CSV::Table mode:col_or_row row_count:3>

Parsing from an Open IO Stream

The input to be parsed can be in an open IO stream:

Method .read returns the entire CSV data:

File.open(path) do |file|
  CSV.read(file)
end # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

As does method .parse:

File.open(path) do |file|
  CSV.parse(file)
end # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

Method .parse_line returns only the first row:

File.open(path) do |file|
 CSV.parse_line(file)
end # => ["foo", "0"]

Method .foreach iterates, passing each row to the given block:

File.open(path) do |file|
  CSV.foreach(file) do |row|
    p row
  end
end

Output:

["foo", "0"]
["bar", "1"]
["baz", "2"]

Method .table returns the entire CSV data as a Table object:

File.open(path) do |file|
  CSV.table(file)
end # => #<CSV::Table mode:col_or_row row_count:3>

Simple Generating

Method .generate returns a String; this example uses method #<< to append the rows that are to be generated:

output_string = CSV.generate do |csv|
  csv << ['foo', 0]
  csv << ['bar', 1]
  csv << ['baz', 2]
end
output_string # => "foo,0\nbar,1\nbaz,2\n"

Method .generate_line returns a String containing the single row constructed from an Array:

CSV.generate_line(['foo', '0']) # => "foo,0\n"

CSV extends class Array with instance method Array#to_csv, which forms an Array into a String:

['foo', '0'].to_csv # => "foo,0\n"

“Filtering” CSV

Method .filter provides a Unix-style filter for CSV data. The input data is processed to form the output data:

in_string = "foo,0\nbar,1\nbaz,2\n"
out_string = ''
CSV.filter(in_string, out_string) do |row|
  row[0] = row[0].upcase
  row[1] *= 4
end
out_string # => "FOO,0000\nBAR,1111\nBAZ,2222\n"

CSV Objects

There are three ways to create a CSV object:

• Method .new returns a new CSV object.

• Method .instance returns a new or cached CSV object.

• Method CSV() also returns a new or cached CSV object.

Instance Methods

CSV has three groups of instance methods:

• Its own internally defined instance methods.

• Methods included by module Enumerable.

• Methods delegated to class IO. See below.

Delegated Methods

For convenience, a CSV object will delegate to many methods in class IO. (A few have wrapper “guard code” in CSV.) You may call:

• IO#binmode

• #binmode?

• IO#close

• IO#close_read

• IO#close_write

• IO#closed?

• #eof

• #eof?

• IO#external_encoding

• IO#fcntl

• IO#fileno

• #flock

• IO#flush

• IO#fsync

• IO#internal_encoding

• #ioctl

• IO#isatty

• #path

• IO#pid

• IO#pos

• IO#pos=

• IO#reopen

• #rewind

• IO#seek

• #stat

• IO#string

• IO#sync

• IO#sync=

• IO#tell

• #to_i

• #to_io

• IO#truncate

• IO#tty?

Options

The default values for options are:

DEFAULT_OPTIONS = {
  # For both parsing and generating.
  col_sep:            ",",
  row_sep:            :auto,
  quote_char:         '"',
  # For parsing.
  field_size_limit:   nil,
  converters:         nil,
  unconverted_fields: nil,
  headers:            false,
  return_headers:     false,
  header_converters:  nil,
  skip_blanks:        false,
  skip_lines:         nil,
  liberal_parsing:    false,
  nil_value:          nil,
  empty_value:        "",
  strip:              false,
  # For generating.
  write_headers:      nil,
  quote_empty:        true,
  force_quotes:       false,
  write_converters:   nil,
  write_nil_value:    nil,
  write_empty_value:  "",
}

Options for Parsing

Options for parsing, described in detail below, include:

• #row_sep: Specifies the row separator; used to delimit rows.

• #col_sep: Specifies the column separator; used to delimit fields.

• #quote_char: Specifies the quote character; used to quote fields.

• #field_size_limit: Specifies the maximum field size + 1 allowed. Deprecated since 3.2.3. Use #max_field_size instead.

• #max_field_size: Specifies the maximum field size allowed.

• #converters: Specifies the field converters to be used.

• unconverted_fields: Specifies whether unconverted fields are to be available.

• #headers: Specifies whether data contains headers, or specifies the headers themselves.

• return_headers: Specifies whether headers are to be returned.

• #header_converters: Specifies the header converters to be used.

• skip_blanks: Specifies whether blanks lines are to be ignored.

• #skip_lines: Specifies how comments lines are to be recognized.

• strip: Specifies whether leading and trailing whitespace are to be stripped from fields. This must be compatible with col_sep; if it is not, then an ArgumentError exception will be raised.

• liberal_parsing: Specifies whether CSV should attempt to parse non-compliant data.

• nil_value: Specifies the object that is to be substituted for each null (no-text) field.

• empty_value: Specifies the object that is to be substituted for each empty field.

Options for Generating

Options for generating, described in detail below, include:

• #row_sep: Specifies the row separator; used to delimit rows.

• #col_sep: Specifies the column separator; used to delimit fields.

• #quote_char: Specifies the quote character; used to quote fields.

• write_headers: Specifies whether headers are to be written.

• force_quotes: Specifies whether each output field is to be quoted.

• quote_empty: Specifies whether each empty output field is to be quoted.

• write_converters: Specifies the field converters to be used in writing.

• write_nil_value: Specifies the object that is to be substituted for each nil-valued field.

• write_empty_value: Specifies the object that is to be substituted for each empty field.

CSV with Headers

CSV allows to specify column names of CSV file, whether they are in data, or provided separately. If headers are specified, reading methods return an instance of Table, consisting of Row.

# Headers are part of data
data = CSV.parse(<<~ROWS, headers: true)
  Name,Department,Salary
  Bob,Engineering,1000
  Jane,Sales,2000
  John,Management,5000
ROWS

data.class      #=> CSV::Table
data.first      #=> #<CSV::Row "Name":"Bob" "Department":"Engineering" "Salary":"1000">
data.first.to_h #=> {"Name"=>"Bob", "Department"=>"Engineering", "Salary"=>"1000"}

# Headers provided by developer
data = CSV.parse('Bob,Engineering,1000', headers: %i[name department salary])
data.first      #=> #<CSV::Row name:"Bob" department:"Engineering" salary:"1000">

Converters

By default, each value (field or header) parsed by CSV is formed into a String. You can use a field converter or header converter to intercept and modify the parsed values:

• See Field Converters.

• See Header Converters.

Also by default, each value to be written during generation is written ‘as-is’. You can use a write converter to modify values before writing.

• See Write Converters.

Specifying Converters

You can specify converters for parsing or generating in the options argument to various CSV methods:

• Option #converters for converting parsed field values.

• Option #header_converters for converting parsed header values.

• Option write_converters for converting values to be written (generated).

There are three forms for specifying converters:

• A converter proc: executable code to be used for conversion.

• A converter name: the name of a stored converter.

• A converter list: an array of converter procs, converter names, and converter lists.

Converter Procs

This converter proc, strip_converter, accepts a value field and returns field.strip:

strip_converter = proc {|field| field.strip }

In this call to .parse, the keyword argument converters: string_converter specifies that:

• Proc string_converter is to be called for each parsed field.

• The converter’s return value is to replace the field value.

Example:

string = " foo , 0 \n bar , 1 \n baz , 2 \n"
array = CSV.parse(string, converters: strip_converter)
array # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

A converter proc can receive a second argument, field_info, that contains details about the field. This modified strip_converter displays its arguments:

strip_converter = proc do |field, field_info|
  p [field, field_info]
  field.strip
end
string = " foo , 0 \n bar , 1 \n baz , 2 \n"
array = CSV.parse(string, converters: strip_converter)
array # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

Output:

[" foo ", #<struct CSV::FieldInfo index=0, line=1, header=nil>]
[" 0 ", #<struct CSV::FieldInfo index=1, line=1, header=nil>]
[" bar ", #<struct CSV::FieldInfo index=0, line=2, header=nil>]
[" 1 ", #<struct CSV::FieldInfo index=1, line=2, header=nil>]
[" baz ", #<struct CSV::FieldInfo index=0, line=3, header=nil>]
[" 2 ", #<struct CSV::FieldInfo index=1, line=3, header=nil>]

Each FieldInfo object shows:

• The 0-based field index.

• The 1-based line index.

• The field header, if any.

Stored Converters

A converter may be given a name and stored in a structure where the parsing methods can find it by name.

The storage structure for field converters is the Hash Converters. It has several built-in converter procs:

• :integer: converts each String-embedded integer into a true Integer.

• :float: converts each String-embedded float into a true Float.

• :date: converts each String-embedded date into a true Date.

• :date_time: converts each String-embedded date-time into a true DateTime

. This example creates a converter proc, then stores it:

strip_converter = proc {|field| field.strip }
CSV::Converters[:strip] = strip_converter

Then the parsing method call can refer to the converter by its name, :strip:

string = " foo , 0 \n bar , 1 \n baz , 2 \n"
array = CSV.parse(string, converters: :strip)
array # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

The storage structure for header converters is the Hash HeaderConverters, which works in the same way. It also has built-in converter procs:

• :downcase: Downcases each header.

• :symbol: Converts each header to a Symbol.

There is no such storage structure for write headers.

In order for the parsing methods to access stored converters in non-main-Ractors, the storage structure must be made shareable first. Therefore, Ractor.make_shareable(CSV::Converters) and Ractor.make_shareable(CSV::HeaderConverters) must be called before the creation of Ractors that use the converters stored in these structures. (Since making the storage structures shareable involves freezing them, any custom converters that are to be used must be added first.)

Converter Lists

A converter list is an Array that may include any assortment of:

• Converter procs.

• Names of stored converters.

• Nested converter lists.

Examples:

numeric_converters = [:integer, :float]
date_converters = [:date, :date_time]
[numeric_converters, strip_converter]
[strip_converter, date_converters, :float]

Like a converter proc, a converter list may be named and stored in either CSV::Converters or HeaderConverters:

CSV::Converters[:custom] = [strip_converter, date_converters, :float]
CSV::HeaderConverters[:custom] = [:downcase, :symbol]

There are two built-in converter lists:

CSV::Converters[:numeric] # => [:integer, :float]
CSV::Converters[:all] # => [:date_time, :numeric]

Field Converters

With no conversion, all parsed fields in all rows become Strings:

string = "foo,0\nbar,1\nbaz,2\n"
ary = CSV.parse(string)
ary # => # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

When you specify a field converter, each parsed field is passed to the converter; its return value becomes the stored value for the field. A converter might, for example, convert an integer embedded in a String into a true Integer. (In fact, that’s what built-in field converter :integer does.)

There are three ways to use field converters.

• Using option converters with a parsing method:

  ary = CSV.parse(string, converters: :integer)
  ary # => [0, 1, 2] # => [["foo", 0], ["bar", 1], ["baz", 2]]

• Using option converters with a new CSV instance:

  csv = CSV.new(string, converters: :integer)
  # Field converters in effect:
  csv.converters # => [:integer]
  csv.read # => [["foo", 0], ["bar", 1], ["baz", 2]]

• Using method #convert to add a field converter to a CSV instance:

  csv = CSV.new(string)
  # Add a converter.
  csv.convert(:integer)
  csv.converters # => [:integer]
  csv.read # => [["foo", 0], ["bar", 1], ["baz", 2]]

Installing a field converter does not affect already-read rows:

csv = CSV.new(string)
csv.shift # => ["foo", "0"]
# Add a converter.
csv.convert(:integer)
csv.converters # => [:integer]
csv.read # => [["bar", 1], ["baz", 2]]

There are additional built-in converters, and custom converters are also supported.

Built-In Field Converters

The built-in field converters are in Hash Converters:

• Each key is a field converter name.

• Each value is one of:

  • A Proc field converter.

  • An Array of field converter names.

Display:

CSV::Converters.each_pair do |name, value|
  if value.kind_of?(Proc)
    p [name, value.class]
  else
    p [name, value]
  end
end

Output:

[:integer, Proc]
[:float, Proc]
[:numeric, [:integer, :float]]
[:date, Proc]
[:date_time, Proc]
[:all, [:date_time, :numeric]]

Each of these converters transcodes values to UTF-8 before attempting conversion. If a value cannot be transcoded to UTF-8 the conversion will fail and the value will remain unconverted.

Converter :integer converts each field that Integer() accepts:

data = '0,1,2,x'
# Without the converter
csv = CSV.parse_line(data)
csv # => ["0", "1", "2", "x"]
# With the converter
csv = CSV.parse_line(data, converters: :integer)
csv # => [0, 1, 2, "x"]

Converter :float converts each field that Float() accepts:

data = '1.0,3.14159,x'
# Without the converter
csv = CSV.parse_line(data)
csv # => ["1.0", "3.14159", "x"]
# With the converter
csv = CSV.parse_line(data, converters: :float)
csv # => [1.0, 3.14159, "x"]

Converter :numeric converts with both :integer and :float..

Converter :date converts each field that Date.parse accepts:

data = '2001-02-03,x'
# Without the converter
csv = CSV.parse_line(data)
csv # => ["2001-02-03", "x"]
# With the converter
csv = CSV.parse_line(data, converters: :date)
csv # => [#<Date: 2001-02-03 ((2451944j,0s,0n),+0s,2299161j)>, "x"]

Converter :date_time converts each field that DateTime.parse accepts:

data = '2020-05-07T14:59:00-05:00,x'
# Without the converter
csv = CSV.parse_line(data)
csv # => ["2020-05-07T14:59:00-05:00", "x"]
# With the converter
csv = CSV.parse_line(data, converters: :date_time)
csv # => [#<DateTime: 2020-05-07T14:59:00-05:00 ((2458977j,71940s,0n),-18000s,2299161j)>, "x"]

Converter :numeric converts with both :date_time and :numeric..

As seen above, method #convert adds converters to a CSV instance, and method #converters returns an Array of the converters in effect:

csv = CSV.new('0,1,2')
csv.converters # => []
csv.convert(:integer)
csv.converters # => [:integer]
csv.convert(:date)
csv.converters # => [:integer, :date]

Custom Field Converters

You can define a custom field converter:

strip_converter = proc {|field| field.strip }
string = " foo , 0 \n bar , 1 \n baz , 2 \n"
array = CSV.parse(string, converters: strip_converter)
array # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

You can register the converter in Converters Hash, which allows you to refer to it by name:

CSV::Converters[:strip] = strip_converter
string = " foo , 0 \n bar , 1 \n baz , 2 \n"
array = CSV.parse(string, converters: :strip)
array # => [["foo", "0"], ["bar", "1"], ["baz", "2"]]

Header Converters

Header converters operate only on headers (and not on other rows).

There are three ways to use header converters; these examples use built-in header converter :downcase, which downcases each parsed header.

• Option #header_converters with a singleton parsing method:

  string = "Name,Count\nFoo,0\n,Bar,1\nBaz,2"
  tbl = CSV.parse(string, headers: true, header_converters: :downcase)
  tbl.class # => CSV::Table
  tbl.headers # => ["name", "count"]

• Option #header_converters with a new CSV instance:

  csv = CSV.new(string, header_converters: :downcase)
  # Header converters in effect:
  csv.header_converters # => [:downcase]
  tbl = CSV.parse(string, headers: true)
  tbl.headers # => ["Name", "Count"]

• Method #header_convert adds a header converter to a CSV instance:

  csv = CSV.new(string)
  # Add a header converter.
  csv.header_convert(:downcase)
  csv.header_converters # => [:downcase]
  tbl = CSV.parse(string, headers: true)
  tbl.headers # => ["Name", "Count"]

Built-In Header Converters

The built-in header converters are in Hash HeaderConverters. The keys there are the names of the converters:

CSV::HeaderConverters.keys # => [:downcase, :symbol]

Converter :downcase converts each header by downcasing it:

string = "Name,Count\nFoo,0\n,Bar,1\nBaz,2"
tbl = CSV.parse(string, headers: true, header_converters: :downcase)
tbl.class # => CSV::Table
tbl.headers # => ["name", "count"]

Converter :symbol converts each header by making it into a Symbol:

string = "Name,Count\nFoo,0\n,Bar,1\nBaz,2"
tbl = CSV.parse(string, headers: true, header_converters: :symbol)
tbl.headers # => [:name, :count]

Details:

• Strips leading and trailing whitespace.

• Downcases the header.

• Replaces embedded spaces with underscores.

• Removes non-word characters.

• Makes the string into a Symbol.

Custom Header Converters

You can define a custom header converter:

upcase_converter = proc {|header| header.upcase }
string = "Name,Value\nfoo,0\nbar,1\nbaz,2\n"
table = CSV.parse(string, headers: true, header_converters: upcase_converter)
table # => #<CSV::Table mode:col_or_row row_count:4>
table.headers # => ["NAME", "VALUE"]

You can register the converter in HeaderConverters Hash, which allows you to refer to it by name:

CSV::HeaderConverters[:upcase] = upcase_converter
table = CSV.parse(string, headers: true, header_converters: :upcase)
table # => #<CSV::Table mode:col_or_row row_count:4>
table.headers # => ["NAME", "VALUE"]

Write Converters

When you specify a write converter for generating CSV, each field to be written is passed to the converter; its return value becomes the new value for the field. A converter might, for example, strip whitespace from a field.

Using no write converter (all fields unmodified):

output_string = CSV.generate do |csv|
  csv << [' foo ', 0]
  csv << [' bar ', 1]
  csv << [' baz ', 2]
end
output_string # => " foo ,0\n bar ,1\n baz ,2\n"

Using option write_converters with two custom write converters:

strip_converter = proc {|field| field.respond_to?(:strip) ? field.strip : field }
upcase_converter = proc {|field| field.respond_to?(:upcase) ? field.upcase : field }
write_converters = [strip_converter, upcase_converter]
output_string = CSV.generate(write_converters: write_converters) do |csv|
  csv << [' foo ', 0]
  csv << [' bar ', 1]
  csv << [' baz ', 2]
end
output_string # => "FOO,0\nBAR,1\nBAZ,2\n"

Character Encodings (M17n or Multilingualization)

This new CSV parser is m17n savvy. The parser works in the Encoding of the IO or ::String object being read from or written to. Your data is never transcoded (unless you ask Ruby to transcode it for you) and will literally be parsed in the Encoding it is in. Thus CSV will return Arrays or Rows of Strings in the Encoding of your data. This is accomplished by transcoding the parser itself into your Encoding.

Some transcoding must take place, of course, to accomplish this multiencoding support. For example, :col_sep, :row_sep, and :quote_char must be transcoded to match your data. Hopefully this makes the entire process feel transparent, since CSV’s defaults should just magically work for your data. However, you can set these values manually in the target Encoding to avoid the translation.

It’s also important to note that while all of CSV’s core parser is now Encoding agnostic, some features are not. For example, the built-in converters will try to transcode data to UTF-8 before making conversions. Again, you can provide custom converters that are aware of your Encodings to avoid this translation. It’s just too hard for me to support native conversions in all of Ruby’s Encodings.

Anyway, the practical side of this is simple: make sure IO and ::String objects passed into CSV have the proper Encoding set and everything should just work. CSV methods that allow you to open IO objects (CSV::foreach(), .open(), .read(), and .readlines()) do allow you to specify the Encoding.

One minor exception comes when generating CSV into a ::String with an Encoding that is not ASCII compatible. There’s no existing data for CSV to use to prepare itself and thus you will probably need to manually specify the desired Encoding for most of those cases. It will try to guess using the fields in a row of output though, when using .generate_line() or Array#to_csv().

I try to point out any other Encoding issues in the documentation of methods as they come up.

This has been tested to the best of my ability with all non-“dummy” Encodings Ruby ships with. However, it is brave new code and may have some bugs. Please feel free to report any issues you find with it.