Parse::CSV - Highly flexible CSV parser for large files


  # Simple headerless comma-seperated column parser
  my $simple = Parse::CSV->new(
      file => 'file.csv',
  while ( my $array_ref = $simple->fetch ) {
     # Do something...

... or a more complex example...

  # Parse a colon-seperated variables file  from a handle as a hash
  # based on headers from the first line.
  # Then filter, so we emit objects rather than the plain hash.
  my $objects = Parse::CSV->new(
      handle => $io_handle,
      sep_char   => ';',
      names      => 1,
      filter     => sub { My::Object->new( $_ ) },
  while ( my $object = $objects->fetch ) {


Surely the CPAN doesn't need yet another CSV parsing module.

Text::CSV_XS is the standard parser for CSV files. It is fast as hell, but unfortunately it can be a bit verbose to use.

A number of other modules have attempted to put usability wrappers around this venerable module, but they have all focused on parsing the entire file into memory at once.

This method is fine unless your CSV files start to get large. Once that happens, the only existing option is to fall back on the relatively slow and heavyweight XML::SAXDriver::CSV module.

Parse::CSV fills this functionality gap. It provides a flexible and light-weight streaming parser for large, extremely large, or arbitrarily large CSV files.

Main Features

Stream-Based Parser - All parsing a line at a time.

Array Mode - Parsing can be done in simple array mode, returning a reference to an array if the columns are not named.

Hash Mode - Parsing can be done in hash mode, putting the data into a hash and return a reference to it.

Filter Capability - All items returned can be passed through a custom filter. This filter can either modify the data on the fly, or drop records you don't need.

Writing Filters

A Parse::CSV filter is a subroutine reference that is passed the raw record as $_, and should return the alternative or modified record to return to the user.

The null filter (does not modify or drop any records) looks like the following.

  sub { $_ };

A filter which reversed the order of the columns (assuming they are passed as an array) might look like the following.

  sub { return [ reverse @$_ ] };

To drop the record, you should return undef from the filter. The parser will then keep pulling and parsing new records until one passes the filter.

  # Only keep records where foo is true
  sub { $_->{foo} ? $_ : undef }

To signal an error, throw an exception

  sub {
      $_->{foo} =~ /bar/ or die "Assumption failed";
      return $_;



The new constructor creates and initialise a new CSV parser.

It takes a number of params.

To specify the CSV data source, you should provide either the file param, which should be the name of the file to read, or the handle param, which should be a file handle to read instead.

The actual parsing is done using Text::CSV_XS. Any of its constructor/parsing params can also be provided to this new method, and they will be passed on.

Alternatively, they can be passed as a single HASH reference as the csv_attr param. For example:

  $parser = Parse::CSV->new(
      file     => 'file.csv',
      csv_attr => {
          sep_char   => ';',
          quote_char => "'",

An optional names param can be provided, which should be an array reference containing the names of the columns in the CSV file.

  $parser = Parse::CSV->new(
      file  => 'file.csv',
      names => [ 'col1', 'col2', 'col3' ],

If the names param is provided, the parser will map the columns to a hash where the keys are the field names provided, and the values are the values found in the CSV file.

If the names param is not provided, the parser will return simple array references of the columns.

If the names param is true and not a reference, the names will be automatically determined by reading the first line of the CSV file and using those values as the field names.

The optional filter param will be used to filter the records if provided. It should be a CODE reference or any otherwise callable scalar, and each value parsed (either array reference or hash reference) will be passed to the filter to be changed or converted into an object, or whatever you wish.

Returns a new Parse::CSV object, or throws an exception (dies) on error.


Once a Parse::CSV object has been created, the fetch method is used to parse and return the next value from the CSV file.

Returns an ARRAY, HASH or the output of the filter, based on the configuration of the object, or undef in a variety of situations.

Returning undef means either some part of the parsing and filtering process has resulted in an error, or that the end of file has been reached.

On receiving undef, you should the errstr method. If it is a null string you have reached the end of file. Otherwise the error message will be returned. Thus, the basic usage of Parse::CSV will look like the following.

  my $parser = Parse::CSV->new(
      file => 'file.csv',
  while ( my $value = $parser->fetch ) {
      # Do something...
  if ( $parser->errstr ) {
      # Handle errors...


The row method returns the current row of the CSV file.

This is a one-based count, so when you first create the parser, the value of row will be zero (unless you are using names on automatic in which case it will be 1).


  $status = $csv->combine(@columns);

The combine method is provided as a convenience, and is passed through to the underlying Text::CSV_XS object.


  $line = $csv->string;

The string method is provided as a convenience, and is passed through to the underlying Text::CSV_XS object.


  $status = $csv->print($io, $columns);

The print method is provided as a convenience, and is passed through to the underlying Text::CSV_XS object.


  @fields = $csv->fields;

The fields method is provided as a convenience, and is passed through to the underlying Text::CSV_XS object. It shows the actual row as an array.


  # Get the current column names in use
  my @names = $csv->names;
  # Change the column names on the fly mid stream
  $csv->names( 'fn1', 'fn2' );

The names method gets or sets the column name mapping for the parser.


On error, the errstr method returns the error that occured.

If the last action was NOT an error, returns the null string ''.


Bugs should always be reported via the CPAN bug tracker at

For other issues, or commercial enhancement or support, contact the author.


Adam Kennedy <>


Uwe Sarnowski <>




Copyright 2006 - 2012 Adam Kennedy.

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

The full text of the license can be found in the LICENSE file included with this module.