Quick Navigator

Search Site

Unix VPS
A - Starter
B - Basic
C - Preferred
D - Commercial
MPS - Dedicated
Previous VPSs
* Sign Up! *

Contact Us
Online Help
Domain Status
Man Pages

Virtual Servers

Topology Map

Server Agreement
Year 2038

USA Flag



Man Pages

Manual Reference Pages  -  TEXT::CSV_XS (3)

.ds Aq ’


Text::CSV_XS - comma-separated values manipulation routines



 # Functional interface
 use Text::CSV_XS qw( csv );
 # Read whole file in memory as array of arrays
 my $aoa = csv (in => "data.csv");
 # Write array of arrays as csv file
 csv (in => $aoa, out => "file.csv", sep_char=> ";");

 # Object interface
 use Text::CSV_XS;

 my @rows;
 my $csv = Text::CSV_XS->new ({ binary => 1, auto_diag => 1 });
 open my $fh, "<:encoding(utf8)", "test.csv" or die "test.csv: $!";
 while (my $row = $csv->getline ($fh)) {
     $row->[2] =~ m/pattern/ or next; # 3rd field should match
     push @rows, $row;
 close $fh;

 $csv->eol ("\r\n");
 open $fh, ">:encoding(utf8)", "new.csv" or die "new.csv: $!";
 $csv->print ($fh, $_) for @rows;
 close $fh or die "new.csv: $!";


Text::CSV_XS provides facilities for the composition and decomposition of comma-separated values. An instance of the Text::CSV_XS class will combine fields into a CSV string and parse a CSV string into fields.

The module accepts either strings or files as input and support the use of user-specified characters for delimiters, separators, and escapes.

    Embedded newlines

<B>Important NoteB>: The default behavior is to accept only ASCII characters in the range from 0x20 (space) to 0x7E (tilde). This means that the fields can not contain newlines. If your data contains newlines embedded in fields, or characters above 0x7E (tilde), or binary data, you <B>B>must<B>B> set binary => 1 in the call to new. To cover the widest range of parsing options, you will always want to set binary.

But you still have the problem that you have to pass a correct line to the parse method, which is more complicated from the usual point of usage:

 my $csv = Text::CSV_XS->new ({ binary => 1, eol => $/ });
 while (<>) {           #  WRONG!
     $csv->parse ($_);
     my @fields = $csv->fields ();

this will break, as the while might read broken lines: it does not care about the quoting. If you need to support embedded newlines, the way to go is to <B>notB> pass eol in the parser (it accepts \n, \r, <B>andB> \r\n by default) and then

 my $csv = Text::CSV_XS->new ({ binary => 1 });
 open my $io, "<", $file or die "$file: $!";
 while (my $row = $csv->getline ($io)) {
     my @fields = @$row;

The old(er) way of using global file handles is still supported

 while (my $row = $csv->getline (*ARGV)) { ... }


Unicode is only tested to work with perl-5.8.2 and up.

On parsing (both for getline and parse), if the source is marked being UTF8, then all fields that are marked binary will also be marked UTF8.

For complete control over encoding, please use Text::CSV::Encoded:

 use Text::CSV::Encoded;
 my $csv = Text::CSV::Encoded->new ({
     encoding_in  => "iso-8859-1", # the encoding comes into   Perl
     encoding_out => "cp1252",     # the encoding comes out of Perl

 $csv = Text::CSV::Encoded->new ({ encoding  => "utf8" });
 # combine () and print () accept *literally* utf8 encoded data
 # parse () and getline () return *literally* utf8 encoded data

 $csv = Text::CSV::Encoded->new ({ encoding  => undef }); # default
 # combine () and print () accept UTF8 marked data
 # parse () and getline () return UTF8 marked data

On combining (print and combine): if any of the combining fields was marked UTF8, the resulting string will be marked as UTF8. Note however that all fields before the first field marked UTF8 and contained 8-bit characters that were not upgraded to UTF8, these will be bytes in the resulting string too, possibly causing unexpected errors. If you pass data of different encoding, or you don’t know if there is different encoding, force it to be upgraded before you pass them on:

 $csv->print ($fh, [ map { utf8::upgrade (my $x = $_); $x } @data ]);


While no formal specification for CSV exists, RFC 4180 1) describes the common format and establishes text/csv as the MIME type registered with the IANA. RFC 7111 2 adds fragments to CSV.

Many informal documents exist that describe the CSV format. How To: The Comma Separated Value (CSV) File Format 3) provides an overview of the CSV format in the most widely used applications and explains how it can best be used and supported.


The basic rules are as follows:

<B>CSVB> is a delimited data format that has fields/columns separated by the comma character and records/rows separated by newlines. Fields that contain a special character (comma, newline, or double quote), must be enclosed in double quotes. However, if a line contains a single entry that is the empty string, it may be enclosed in double quotes. If a field’s value contains a double quote character it is escaped by placing another double quote character next to it. The CSV file format does not require a specific character encoding, byte order, or line terminator format.
o Each record is a single line ended by a line feed (ASCII/LF=0x0A) or a carriage return and line feed pair (ASCII/CRLF=0x0D 0x0A), however, line-breaks may be embedded.
o Fields are separated by commas.
o Allowable characters within a CSV field include 0x09 (TAB) and the inclusive range of 0x20 (space) through 0x7E (tilde). In binary mode all characters are accepted, at least in quoted fields.
o A field within CSV must be surrounded by double-quotes to contain a separator character (comma).
Though this is the most clear and restrictive definition, Text::CSV_XS is way more liberal than this, and allows extension:
o Line termination by a single carriage return is accepted by default
o The separation-, escape-, and escape- characters can be any ASCII character in the range from 0x20 (space) to 0x7E (tilde). Characters outside this range may or may not work as expected. Multibyte characters, like UTF U+060C (ARABIC COMMA), U+FF0C (FULLWIDTH COMMA), U+241B (SYMBOL FOR ESCAPE), U+2424 (SYMBOL FOR NEWLINE), U+FF02 (FULLWIDTH QUOTATION MARK), and U+201C (LEFT DOUBLE QUOTATION MARK) (to give some examples of what might look promising) work for newer versions of perl for sep_char, and quote_char but not for escape_char.

If you use perl-5.8.2 or higher these three attributes are utf8-decoded, to increase the likelihood of success. This way U+00FE will be allowed as a quote character.

o A field in CSV must be surrounded by double-quotes to make an embedded double-quote, represented by a pair of consecutive double-quotes, valid. In binary mode you may additionally use the sequence "0 for representation of a NULL byte. Using 0x00 in binary mode is just as valid.
o Several violations of the above specification may be lifted by passing some options as attributes to the object constructor.



(Class method) Returns the current module version.


(Class method) Returns a new instance of class Text::CSV_XS. The attributes are described by the (optional) hash ref \%attr.

 my $csv = Text::CSV_XS->new ({ attributes ... });

The following attributes are available:

 my $csv = Text::CSV_XS->new ({ eol => $/ });
           $csv->eol (undef);
 my $eol = $csv->eol;

The end-of-line string to add to rows for print or the record separator for getline.

When not passed in a <B>parserB> instance, the default behavior is to accept \n, \r, and \r\n, so it is probably safer to not specify eol at all. Passing undef or the empty string behave the same.

When not passed in a <B>generatingB> instance, records are not terminated at all, so it is probably wise to pass something you expect. A safe choice for eol on output is either $/ or \r\n.

Common values for eol are "\012" (\n or Line Feed), "\015\012" (\r\n or Carriage Return, Line Feed), and "\015" (\r or Carriage Return). The eol attribute cannot exceed 7 (ASCII) characters.

If both $/ and eol equal "\015", parsing lines that end on only a Carriage Return without Line Feed, will be parsed correct.


 my $csv = Text::CSV_XS->new ({ sep_char => ";" });
         $csv->sep_char (";");
 my $c = $csv->sep_char;

The char used to separate fields, by default a comma. (,). Limited to a single-byte character, usually in the range from 0x20 (space) to 0x7E (tilde). When longer sequences are required, use sep.

The separation character can not be equal to the quote character or to the escape character.

See also CAVEATS


 my $csv = Text::CSV_XS->new ({ sep => "\N{FULLWIDTH COMMA}" });
           $csv->sep (";");
 my $sep = $csv->sep;

The chars used to separate fields, by default undefined. Limited to 8 bytes.

When set, overrules sep_char. If its length is one byte it acts as an alias to sep_char.

See also CAVEATS


 my $csv = Text::CSV_XS->new ({ quote_char => "" });
         $csv->quote_char (undef);
 my $c = $csv->quote_char;

The character to quote fields containing blanks or binary data, by default the double quote character ("). A value of undef suppresses quote chars (for simple cases only). Limited to a single-byte character, usually in the range from 0x20 (space) to 0x7E (tilde). When longer sequences are required, use quote.

quote_char can not be equal to sep_char.


 my $csv = Text::CSV_XS->new ({ quote => "\N{FULLWIDTH QUOTATION MARK}" });
             $csv->quote ("");
 my $quote = $csv->quote;

The chars used to quote fields, by default undefined. Limited to 8 bytes.

When set, overrules quote_char. If its length is one byte it acts as an alias to quote_char.

See also CAVEATS


 my $csv = Text::CSV_XS->new ({ escape_char => "\\" });
         $csv->escape_char (undef);
 my $c = $csv->escape_char;

The character to escape certain characters inside quoted fields. This is limited to a single-byte character, usually in the range from 0x20 (space) to 0x7E (tilde).

The escape_char defaults to being the double-quote mark ("). In other words the same as the default quote_char. This means that doubling the quote mark in a field escapes it:

 "foo","bar","Escape ""quote mark"" with two ""quote marks""","baz"

If you change the quote_char without changing the escape_char, the escape_char will still be the double-quote ("). If instead you want to escape the quote_char by doubling it you will need to also change the escape_char to be the same as what you have changed the quote_char to.

The escape character can not be equal to the separation character.


 my $csv = Text::CSV_XS->new ({ binary => 1 });
         $csv->binary (0);
 my $f = $csv->binary;

If this attribute is 1, you may use binary characters in quoted fields, including line feeds, carriage returns and NULL bytes. (The latter could be escaped as "0.) By default this feature is off.

If a string is marked UTF8, binary will be turned on automatically when binary characters other than CR and NL are encountered. Note that a simple string like "\x{00a0}" might still be binary, but not marked UTF8, so setting { binary => 1 } is still a wise option.


 my $csv = Text::CSV_XS->new ({ decode_utf8 => 1 });
         $csv->decode_utf8 (0);
 my $f = $csv->decode_utf8;

This attributes defaults to TRUE.

While parsing, fields that are valid UTF-8, are automatically set to be UTF-8, so that

  $csv->parse ("\xC4\xA8\n");

results in

  PV("\304\250"\0) [UTF8 "\x{128}"]

Sometimes it might not be a desired action. To prevent those upgrades, set this attribute to false, and the result will be



 my $csv = Text::CSV_XS->new ({ auto_diag => 1 });
         $csv->auto_diag (2);
 my $l = $csv->auto_diag;

Set this attribute to a number between 1 and 9 causes error_diag to be automatically called in void context upon errors.

In case of error 2012 - EOF, this call will be void.

If auto_diag is set to a numeric value greater than 1, it will die on errors instead of warn. If set to anything unrecognized, it will be silently ignored.

Future extensions to this feature will include more reliable auto-detection of autodie being active in the scope of which the error occurred which will increment the value of auto_diag with 1 the moment the error is detected.


 my $csv = Text::CSV_XS->new ({ diag_verbose => 1 });
         $csv->diag_verbose (2);
 my $l = $csv->diag_verbose;

Set the verbosity of the output triggered by auto_diag. Currently only adds the current input-record-number (if known) to the diagnostic output with an indication of the position of the error.


 my $csv = Text::CSV_XS->new ({ blank_is_undef => 1 });
         $csv->blank_is_undef (0);
 my $f = $csv->blank_is_undef;

Under normal circumstances, CSV data makes no distinction between quoted- and unquoted empty fields. These both end up in an empty string field once read, thus

 1,"",," ",2

is read as

 ("1", "", "", " ", "2")

When writing CSV files with always_quote set, the unquoted empty field is the result of an undefined value. To enable this distinction when reading CSV data, the blank_is_undef attribute will cause unquoted empty fields to be set to undef, causing the above to be parsed as

 ("1", "", undef, " ", "2")


 my $csv = Text::CSV_XS->new ({ empty_is_undef => 1 });
         $csv->empty_is_undef (0);
 my $f = $csv->empty_is_undef;

Going one step further than blank_is_undef, this attribute converts all empty fields to undef, so

 1,"",," ",2

is read as

 (1, undef, undef, " ", 2)

Note that this effects only fields that are originally empty, not fields that are empty after stripping allowed whitespace. YMMV.


 my $csv = Text::CSV_XS->new ({ allow_whitespace => 1 });
         $csv->allow_whitespace (0);
 my $f = $csv->allow_whitespace;

When this option is set to true, the whitespace (TAB’s and SPACE’s) surrounding the separation character is removed when parsing. If either TAB or SPACE is one of the three characters sep_char, quote_char, or escape_char it will not be considered whitespace.

Now lines like:

 1 , "foo" , bar , 3 , zapp

are parsed as valid CSV, even though it violates the CSV specs.

Note that <B>allB> whitespace is stripped from both start and end of each field. That would make it more than a feature to enable parsing bad CSV lines, as

 1,   2.0,  3,   ape  , monkey

will now be parsed as

 ("1", "2.0", "3", "ape", "monkey")

even if the original line was perfectly acceptable CSV.


 my $csv = Text::CSV_XS->new ({ allow_loose_quotes => 1 });
         $csv->allow_loose_quotes (0);
 my $f = $csv->allow_loose_quotes;

By default, parsing unquoted fields containing quote_char characters like

 1,foo "bar" baz,42

would result in parse error 2034. Though it is still bad practice to allow this format, we cannot help the fact that some vendors make their applications spit out lines styled this way.

If there is <B>reallyB> bad CSV data, like

 1,"foo "bar" baz",42


 1,""foo bar baz"",42

there is a way to get this data-line parsed and leave the quotes inside the quoted field as-is. This can be achieved by setting allow_loose_quotes <B>ANDB> making sure that the escape_char is not equal to quote_char.


 my $csv = Text::CSV_XS->new ({ allow_loose_escapes => 1 });
         $csv->allow_loose_escapes (0);
 my $f = $csv->allow_loose_escapes;

Parsing fields that have escape_char characters that escape characters that do not need to be escaped, like:

 my $csv = Text::CSV_XS->new ({ escape_char => "\\" });
 $csv->parse (qq{1,"my bar\s",baz,42});

would result in parse error 2025. Though it is bad practice to allow this format, this attribute enables you to treat all escape character sequences equal.


 my $csv = Text::CSV_XS->new ({ allow_unquoted_escape => 1 });
         $csv->allow_unquoted_escape (0);
 my $f = $csv->allow_unquoted_escape;

A backward compatibility issue where escape_char differs from quote_char prevents escape_char to be in the first position of a field. If quote_char is equal to the default " and escape_char is set to \, this would be illegal:


Setting this attribute to 1 might help to overcome issues with backward compatibility and allow this style.


 my $csv = Text::CSV_XS->new ({ always_quote => 1 });
         $csv->always_quote (0);
 my $f = $csv->always_quote;

By default the generated fields are quoted only if they need to be. For example, if they contain the separator character. If you set this attribute to 1 then all defined fields will be quoted. (undef fields are not quoted, see blank_is_undef). This makes it quite often easier to handle exported data in external applications. (Poor creatures who are better to use Text::CSV_XS. :)


 my $csv = Text::CSV_XS->new ({ quote_space => 1 });
         $csv->quote_space (0);
 my $f = $csv->quote_space;

By default, a space in a field would trigger quotation. As no rule exists this to be forced in CSV, nor any for the opposite, the default is true for safety. You can exclude the space from this trigger by setting this attribute to 0.


 my $csv = Text::CSV_XS->new ({ quote_empty => 1 });
         $csv->quote_empty (0);
 my $f = $csv->quote_empty;

By default the generated fields are quoted only if they need to be. An empty (defined) field does not need quotation. If you set this attribute to 1 then empty defined fields will be quoted. (undef fields are not quoted, see blank_is_undef). See also always_quote.

escape_null or quote_null (deprecated)

 my $csv = Text::CSV_XS->new ({ escape_null => 1 });
         $csv->escape_null (0);
 my $f = $csv->escape_null;

By default, a NULL byte in a field would be escaped. This option enables you to treat the NULL byte as a simple binary character in binary mode (the { binary => 1 } is set). The default is true. You can prevent NULL escapes by setting this attribute to 0.


 my $csv = Text::CSV_XS->new ({ quote_binary => 1 });
         $csv->quote_binary (0);
 my $f = $csv->quote_binary;

By default, all unsafe bytes inside a string cause the combined field to be quoted. By setting this attribute to 0, you can disable that trigger for bytes >= 0x7F.


 my $csv = Text::CSV_XS->new ({ keep_meta_info => 1 });
         $csv->keep_meta_info (0);
 my $f = $csv->keep_meta_info;

By default, the parsing of input records is as simple and fast as possible. However, some parsing information - like quotation of the original field - is lost in that process. Setting this flag to true enables retrieving that information after parsing with the methods meta_info, is_quoted, and is_binary described below. Default is false for performance.

If you set this attribute to a value greater than 9, than you can control output quotation style like it was used in the input of the the last parsed record (unless quotation was added because of other reasons).

 my $csv = Text::CSV_XS->new ({
    binary         => 1,
    keep_meta_info => 1,
    quote_space    => 0,

 my $row = $csv->parse (q{1,,"", ," ",f,"g","h""h",heelp,"he\k:'\h |\n:ulp"});

 $csv->print (*STDOUT, \@row);
 # 1,,, , ,f,g,"h""h",h?lp,h?lp
 $csv->keep_meta_info (11);
 $csv->print (*STDOUT, \@row);
 # 1,,"", ," ",f,"g","h""h",h?lp,"h?lp"


 my $csv = Text::CSV_XS->new ({ verbatim => 1 });
         $csv->verbatim (0);
 my $f = $csv->verbatim;

This is a quite controversial attribute to set, but makes some hard things possible.

The rationale behind this attribute is to tell the parser that the normally special characters newline (NL) and Carriage Return (CR) will not be special when this flag is set, and be dealt with as being ordinary binary characters. This will ease working with data with embedded newlines.

When verbatim is used with getline, getline auto-chomp’s every line.

Imagine a file format like

 M^^Hans^Janssen^Klas 2\n2A^Ja^11-06-2007#\r\n

where, the line ending is a very specific "#\r\n", and the sep_char is a ^ (caret). None of the fields is quoted, but embedded binary data is likely to be present. With the specific line ending, this should not be too hard to detect.

By default, Text::CSV_XS’ parse function is instructed to only know about "\n" and "\r" to be legal line endings, and so has to deal with the embedded newline as a real end-of-line, so it can scan the next line if binary is true, and the newline is inside a quoted field. With this option, we tell parse to parse the line as if "\n" is just nothing more than a binary character.

For parse this means that the parser has no more idea about line ending and getline chomps line endings on reading.

types A set of column types; the attribute is immediately passed to the types method.
callbacks See the Callbacks section below.
To sum it up,

 $csv = Text::CSV_XS->new ();

is equivalent to

 $csv = Text::CSV_XS->new ({
     eol                   => undef, # \r, \n, or \r\n
     sep_char              => ,,
     sep                   => undef,
     quote_char            => ",
     quote                 => undef,
     escape_char           => ",
     binary                => 0,
     decode_utf8           => 1,
     auto_diag             => 0,
     diag_verbose          => 0,
     blank_is_undef        => 0,
     empty_is_undef        => 0,
     allow_whitespace      => 0,
     allow_loose_quotes    => 0,
     allow_loose_escapes   => 0,
     allow_unquoted_escape => 0,
     always_quote          => 0,
     quote_empty           => 0,
     quote_space           => 1,
     escape_null           => 1,
     quote_binary          => 1,
     keep_meta_info        => 0,
     verbatim              => 0,
     types                 => undef,
     callbacks             => undef,

For all of the above mentioned flags, an accessor method is available where you can inquire the current value, or change the value

 my $quote = $csv->quote_char;
 $csv->binary (1);

It is not wise to change these settings halfway through writing CSV data to a stream. If however you want to create a new stream using the available CSV object, there is no harm in changing them.

If the new constructor call fails, it returns undef, and makes the fail reason available through the error_diag method.

 $csv = Text::CSV_XS->new ({ ecs_char => 1 }) or
     die "".Text::CSV_XS->error_diag ();

error_diag will return a string like

 "INI - Unknown attribute ecs_char"


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

Similar to combine + string + print, but much more efficient. It expects an array ref as input (not an array!) and the resulting string is not really created, but immediately written to the $io object, typically an IO handle or any other object that offers a print method.

For performance reasons print does not create a result string, so all string, status, fields, and error_input methods will return undefined information after executing this method.

If $colref is undef (explicit, not through a variable argument) and bind_columns was used to specify fields to be printed, it is possible to make performance improvements, as otherwise data would have to be copied as arguments to the method call:

 $csv->bind_columns (\($foo, $bar));
 $status = $csv->print ($fh, undef);

A short benchmark

 my @data = ("aa" .. "zz");
 $csv->bind_columns (\(@data));

 $csv->print ($io, [ @data ]);   # 11800 recs/sec
 $csv->print ($io,  \@data  );   # 57600 recs/sec
 $csv->print ($io,   undef  );   # 48500 recs/sec


 $status = $csv->say ($io, $colref);

Like print, but eol defaults to $\.


 $csv->print_hr ($io, $ref);

Provides an easy way to print a $ref (as fetched with getline_hr) provided the column names are set with column_names.

It is just a wrapper method with basic parameter checks over

 $csv->print ($io, [ map { $ref->{$_} } $csv->column_names ]);


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

This method constructs a CSV record from @fields, returning success or failure. Failure can result from lack of arguments or an argument that contains an invalid character. Upon success, string can be called to retrieve the resultant CSV string. Upon failure, the value returned by string is undefined and error_input could be called to retrieve the invalid argument.


 $line = $csv->string ();

This method returns the input to parse or the resultant CSV string of combine, whichever was called more recently.


 $colref = $csv->getline ($io);

This is the counterpart to print, as parse is the counterpart to combine: it parses a row from the $io handle using the getline method associated with $io and parses this row into an array ref. This array ref is returned by the function or undef for failure. When $io does not support getline, you are likely to hit errors.

When fields are bound with bind_columns the return value is a reference to an empty list.

The string, fields, and status methods are meaningless again.


 $arrayref = $csv->getline_all ($io);
 $arrayref = $csv->getline_all ($io, $offset);
 $arrayref = $csv->getline_all ($io, $offset, $length);

This will return a reference to a list of getline ($io) results. In this call, keep_meta_info is disabled. If $offset is negative, as with splice, only the last abs ($offset) records of $io are taken into consideration.

Given a CSV file with 10 lines:

 lines call
 ----- ---------------------------------------------------------
 0..9  $csv->getline_all ($io)         # all
 0..9  $csv->getline_all ($io,  0)     # all
 8..9  $csv->getline_all ($io,  8)     # start at 8
 -     $csv->getline_all ($io,  0,  0) # start at 0 first 0 rows
 0..4  $csv->getline_all ($io,  0,  5) # start at 0 first 5 rows
 4..5  $csv->getline_all ($io,  4,  2) # start at 4 first 2 rows
 8..9  $csv->getline_all ($io, -2)     # last 2 rows
 6..7  $csv->getline_all ($io, -4,  2) # first 2 of last  4 rows


The getline_hr and column_names methods work together to allow you to have rows returned as hashrefs. You must call column_names first to declare your column names.

 $csv->column_names (qw( code name price description ));
 $hr = $csv->getline_hr ($io);
 print "Price for $hr->{name} is $hr->{price} EUR\n";

getline_hr will croak if called before column_names.

Note that getline_hr creates a hashref for every row and will be much slower than the combined use of bind_columns and getline but still offering the same ease of use hashref inside the loop:

 my @cols = @{$csv->getline ($io)};
 $csv->column_names (@cols);
 while (my $row = $csv->getline_hr ($io)) {
     print $row->{price};

Could easily be rewritten to the much faster:

 my @cols = @{$csv->getline ($io)};
 my $row = {};
 $csv->bind_columns (\@{$row}{@cols});
 while ($csv->getline ($io)) {
     print $row->{price};

Your mileage may vary for the size of the data and the number of rows. With perl-5.14.2 the comparison for a 100_000 line file with 14 rows:

            Rate hashrefs getlines
 hashrefs 1.00/s       --     -76%
 getlines 4.15/s     313%       --


 $arrayref = $csv->getline_hr_all ($io);
 $arrayref = $csv->getline_hr_all ($io, $offset);
 $arrayref = $csv->getline_hr_all ($io, $offset, $length);

This will return a reference to a list of getline_hr ($io) results. In this call, keep_meta_info is disabled.


 $status = $csv->parse ($line);

This method decomposes a CSV string into fields, returning success or failure. Failure can result from a lack of argument or the given CSV string is improperly formatted. Upon success, fields can be called to retrieve the decomposed fields. Upon failure calling fields will return undefined data and error_input can be called to retrieve the invalid argument.

You may use the types method for setting column types. See types’ description below.


This function tries to implement RFC7111 (URI Fragment Identifiers for the text/csv Media Type) -

 my $AoA = $csv->fragment ($io, $spec);

In specifications, * is used to specify the last item, a dash (-) to indicate a range. All indices are 1-based: the first row or column has index 1. Selections can be combined with the semi-colon (;).

When using this method in combination with column_names, the returned reference will point to a list of hashes instead of a list of lists. A disjointed cell-based combined selection might return rows with different number of columns making the use of hashes unpredictable.

 $csv->column_names ("Name", "Age");
 my $AoH = $csv->fragment ($io, "col=3;8");

If the after_parse callback is active, it is also called on every line parsed and skipped before the fragment.




cell In cell-based selection, the comma (,) is used to pair row and column


The range operator (-) using cells can be used to define top-left and bottom-right cell location


The * is only allowed in the second part of a pair

 cell=3,2-*,2    # row 3 till end, only column 2
 cell=3,2-3,*    # column 2 till end, only row 3
 cell=3,2-*,*    # strip row 1 and 2, and column 1

Cells and cell ranges may be combined with ;, possibly resulting in rows with different number of columns


Disjointed selections will only return selected cells. The cells that are not specified will not be included in the returned set, not even as undef. As an example given a CSV like

 :            :

with cell=1,1-2,2;3,3-4,4;1,4;4,1 will return:


Overlapping cell-specs will return those cells only once, So cell=1,1-3,3;2,2-4,4;2,3;4,2 will return:


RFC7111 <> does not allow different types of specs to be combined (either row or col or cell). Passing an invalid fragment specification will croak and set error 2013.


Set the keys that will be used in the getline_hr calls. If no keys (column names) are passed, it will return the current setting as a list.

column_names accepts a list of scalars (the column names) or a single array_ref, so you can pass the return value from getline too:

 $csv->column_names ($csv->getline ($io));

column_names does <B>noB> checking on duplicates at all, which might lead to unexpected results. Undefined entries will be replaced with the string "\cAUNDEF\cA", so

 $csv->column_names (undef, "", "name", "name");
 $hr = $csv->getline_hr ($io);

Will set $hr->{"\cAUNDEF\cA"} to the 1st field, $hr->{""} to the 2nd field, and $hr->{name} to the 4th field, discarding the 3rd field.

column_names croaks on invalid arguments.


Takes a list of scalar references to be used for output with print or to store in the fields fetched by getline. When you do not pass enough references to store the fetched fields in, getline will fail with error 3006. If you pass more than there are fields to return, the content of the remaining references is left untouched.

 $csv->bind_columns (\$code, \$name, \$price, \$description);
 while ($csv->getline ($io)) {
     print "The price of a $name is \x{20ac} $price\n";

To reset or clear all column binding, call bind_columns with the single argument undef. This will also clear column names.

 $csv->bind_columns (undef);

If no arguments are passed at all, bind_columns will return the list of current bindings or undef if no binds are active.


 $eof = $csv->eof ();

If parse or getline was used with an IO stream, this method will return true (1) if the last call hit end of file, otherwise it will return false (’’). This is useful to see the difference between a failure and end of file.


 $csv->types (\@tref);

This method is used to force that (all) columns are of a given type. For example, if you have an integer column, two columns with doubles and a string column, then you might do a

 $csv->types ([Text::CSV_XS::IV (),
               Text::CSV_XS::NV (),
               Text::CSV_XS::NV (),
               Text::CSV_XS::PV ()]);

Column types are used only for decoding columns while parsing, in other words by the parse and getline methods.

You can unset column types by doing a

 $csv->types (undef);

or fetch the current type settings with

 $types = $csv->types ();

IV Set field type to integer.
NV Set field type to numeric/float.
PV Set field type to string.


 @columns = $csv->fields ();

This method returns the input to combine or the resultant decomposed fields of a successful parse, whichever was called more recently.

Note that the return value is undefined after using getline, which does not fill the data structures returned by parse.


 @flags = $csv->meta_info ();

This method returns the flags of the input to combine or the flags of the resultant decomposed fields of parse, whichever was called more recently.

For each field, a meta_info field will hold flags that inform something about the field returned by the fields method or passed to the combine method. The flags are bit-wise-or’d like:
0x0001 The field was quoted.
0x0002 The field was binary.
See the is_*** methods below.


 my $quoted = $csv->is_quoted ($column_idx);

Where $column_idx is the (zero-based) index of the column in the last result of parse.

This returns a true value if the data in the indicated column was enclosed in quote_char quotes. This might be important for fields where content ,20070108, is to be treated as a numeric value, and where ,"20070108", is explicitly marked as character string data.

This method is only valid when keep_meta_info is set to a true value.


 my $binary = $csv->is_binary ($column_idx);

Where $column_idx is the (zero-based) index of the column in the last result of parse.

This returns a true value if the data in the indicated column contained any byte in the range [\x00-\x08,\x10-\x1F,\x7F-\xFF].

This method is only valid when keep_meta_info is set to a true value.


 my $missing = $csv->is_missing ($column_idx);

Where $column_idx is the (zero-based) index of the column in the last result of getline_hr.

 $csv->keep_meta_info (1);
 while (my $hr = $csv->getline_hr ($fh)) {
     $csv->is_missing (0) and next; # This was an empty line

When using getline_hr, it is impossible to tell if the parsed fields are undef because they where not filled in the CSV stream or because they were not read at all, as <B>allB> the fields defined by column_names are set in the hash-ref. If you still need to know if all fields in each row are provided, you should enable keep_meta_info so you can check the flags.


 $status = $csv->status ();

This method returns the status of the last invoked combine or parse call. Status is success (true: 1) or failure (false: undef or 0).


 $bad_argument = $csv->error_input ();

This method returns the erroneous argument (if it exists) of combine or parse, whichever was called more recently. If the last invocation was successful, error_input will return undef.


 Text::CSV_XS->error_diag ();
 $csv->error_diag ();
 $error_code               = 0  + $csv->error_diag ();
 $error_str                = "" . $csv->error_diag ();
 ($cde, $str, $pos, $rec, $fld) = $csv->error_diag ();

If (and only if) an error occurred, this function returns the diagnostics of that error.

If called in void context, this will print the internal error code and the associated error message to STDERR.

If called in list context, this will return the error code and the error message in that order. If the last error was from parsing, the rest of the values returned are a best guess at the location within the line that was being parsed. Their values are 1-based. The position currently is index of the byte at which the parsing failed in the current record. It might change to be the index of the current character in a later release. The records is the index of the record parsed by the csv instance. The field number is the index of the field the parser thinks it is currently trying to parse. See examples/csv-check for how this can be used.

If called in scalar context, it will return the diagnostics in a single scalar, a-la $!. It will contain the error code in numeric context, and the diagnostics message in string context.

When called as a class method or a direct function call, the diagnostics are that of the last new call.


 $recno = $csv->record_number ();

Returns the records parsed by this csv instance. This value should be more accurate than $. when embedded newlines come in play. Records written by this instance are not counted.


 $csv->SetDiag (0);

Use to reset the diagnostics if you are dealing with errors.



This function is not exported by default and should be explicitly requested:

 use Text::CSV_XS qw( csv );

This is the second draft. This function will stay, but the arguments might change based on user feedback.

This is an high-level function that aims at simple (user) interfaces. This can be used to read/parse a CSV file or stream (the default behavior) or to produce a file or write to a stream (define the out attribute). It returns an array- or hash-reference on parsing (or undef on fail) or the numeric value of error_diag on writing. When this function fails you can get to the error using the class call to error_diag

 my $aoa = csv (in => "test.csv") or
     die Text::CSV_XS->error_diag;

Alternative invocations:

 my $aoa = Text::CSV_XS::csv (in => "file.csv");

 my $csv = Text::CSV_XS->new ();
 my $aoa = $csv->csv (in => "file.csv"); # ignore object attributes

This function takes the arguments as key-value pairs. This can be passed as a list or as an anonymous hash:

 my $aoa = csv (  in => "test.csv", sep_char => ";");
 my $aoh = csv ({ in => $fh, headers => "auto" });

The arguments passed consist of two parts: the arguments to csv itself and the optional attributes to the CSV object used inside the function as enumerated and explained in new.

If not overridden, the default option used for CSV is

 auto_diag => 1

The option that is always set and cannot be altered is

 binary    => 1


Used to specify the source. in can be a file name (e.g. "file.csv"), which will be opened for reading and closed when finished, a file handle (e.g. $fh or FH), a reference to a glob (e.g. \*ARGV), the glob itself (e.g. *STDIN), or a reference to a scalar (e.g. \q{1,2,"csv"}).

When used with out, in should be a reference to a CSV structure (AoA or AoH) or a CODE-ref that returns an array-reference or a hash-reference. The code-ref will be invoked with no arguments.

 my $aoa = csv (in => "file.csv");

 open my $fh, "<", "file.csv";
 my $aoa = csv (in => $fh);

 my $csv = [ [qw( Foo Bar )], [ 1, 2 ], [ 2, 3 ]];
 my $err = csv (in => $csv, out => "file.csv");

If called in void context without the out attribute, the resulting ref will be used as input to a subsequent call to csv:

 csv (in => "file.csv", filter => { 2 => sub { length > 2 }})

will be a shortcut to

 csv (in => csv (in => "file.csv", filter => { 2 => sub { length > 2 }}))

where, in the absence of the out attribute, this is a shortcut to

 csv (in  => csv (in => "file.csv", filter => { 2 => sub { length > 2 }}),
      out => *STDOUT)


In output mode, the default CSV options when producing CSV are

 eol       => "\r\n"

The fragment attribute is ignored in output mode.

out can be a file name (e.g. "file.csv"), which will be opened for writing and closed when finished, a file handle (e.g. $fh or FH), a reference to a glob (e.g. \*STDOUT), or the glob itself (e.g. *STDOUT).

 csv (in => sub { $sth->fetch },            out => "dump.csv");
 csv (in => sub { $sth->fetchrow_hashref }, out => "dump.csv",
      headers => $sth->{NAME_lc});

When a code-ref is used for in, the output is generated per invocation, so no buffering is involved. This implies that there is no size restriction on the number of records. The csv function ends when the coderef returns a false value.


If passed, it should be an encoding accepted by the :encoding() option to open. There is no default value. This attribute does not work in perl 5.6.x. encoding can be abbreviated to enc for ease of use in command line invocations.


If this attribute is not given, the default behavior is to produce an array of arrays.

If headers is supplied, it should be either an anonymous list of column names, an anonymous hashref or a flag: auto or skip. When skip is used, the header will not be included in the output.

 my $aoa = csv (in => $fh, headers => "skip");

If auto is used, the first line of the CSV source will be read as the list of field headers and used to produce an array of hashes.

 my $aoh = csv (in => $fh, headers => "auto");

If headers is an anonymous list, the entries in the list will be used instead

 my $aoh = csv (in => $fh, headers => [qw( Foo Bar )]);
 csv (in => $aoa, out => $fh, headers => [qw( code description price }]);

If headers is an hash reference, this implies auto, but header fields for that exist as key in the hashref will be replaced by the value for that key. Given a CSV file like

 post-kode,city,name,id number,fubble


 csv (headers => { "post-kode" => "pc", "id number" => "ID" }, ...

will return an entry like

 { pc       => "1234AA",
   city     => "Duckstad",
   name     => "Donald",
   ID       => "13",
   fubble => "X313DF",


If passed, will default headers to "auto" and return a hashref instead of an array of hashes.

 my $ref = csv (in => "test.csv", key => "code");

with test.csv like


will return

  { 1   => {
        code    => 1,
        color   => gray,
        price   => 850,
        product => pc
    2   => {
        code    => 2,
        color   => white,
        price   => 12,
        product => keyboard
    3   => {
        code    => 3,
        color   => black,
        price   => 5,
        product => mouse


Only output the fragment as defined in the fragment method. This option is ignored when generating CSV. See out.

Combining all of them could give something like

 use Text::CSV_XS qw( csv );
 my $aoh = csv (
     in       => "test.txt",
     encoding => "utf-8",
     headers  => "auto",
     sep_char => "|",
     fragment => "row=3;6-9;15-*",
 say $aoh->[15]{Foo};


Callbacks enable actions triggered from the inside of Text::CSV_XS.

While most of what this enables can easily be done in an unrolled loop as described in the SYNOPSIS callbacks can be used to meet special demands or enhance the csv function.

 $csv->callbacks (error => sub { $csv->SetDiag (0) });

the error callback is invoked when an error occurs, but only when auto_diag is set to a true value. A callback is invoked with the values returned by error_diag:

 my ($c, $s);

 sub ignore3006
     my ($err, $msg, $pos, $recno, $fldno) = @_;
     if ($err == 3006) {
         # ignore this error
         ($c, $s) = (undef, undef);
         Text::CSV_XS->SetDiag (0);
     # Any other error
     } # ignore3006

 $csv->callbacks (error => \&ignore3006);
 $csv->bind_columns (\$c, \$s);
 while ($csv->getline ($fh)) {
     # Error 3006 will not stop the loop


 $csv->callbacks (after_parse => sub { push @{$_[1]}, "NEW" });
 while (my $row = $csv->getline ($fh)) {
     $row->[-1] eq "NEW";

This callback is invoked after parsing with getline only if no error occurred. The callback is invoked with two arguments: the current CSV parser object and an array reference to the fields parsed.

The return code of the callback is ignored unless it is a reference to the string skip, in which case the record will be skipped in getline_all.

 sub add_from_db
     my ($csv, $row) = @_;
     $sth->execute ($row->[4]);
     push @$row, $sth->fetchrow_array;
     } # add_from_db

 my $aoa = csv (in => "file.csv", callbacks => {
     after_parse => \&add_from_db });

This hook can be used for validation:
FAIL Die if any of the records does not validate a rule:

 after_parse => sub {
     $_[1][4] =~ m/^[0-9]{4}\s?[A-Z]{2}$/ or
         die "5th field does not have a valid Dutch zipcode";

DEFAULT Replace invalid fields with a default value:

 after_parse => sub { $_[1][2] =~ m/^\d+$/ or $_[1][2] = 0 }

SKIP Skip records that have invalid fields (only applies to getline_all):

 after_parse => sub { $_[1][0] =~ m/^\d+$/ or return \"skip"; }


 my $idx = 1;
 $csv->callbacks (before_print => sub { $_[1][0] = $idx++ });
 $csv->print (*STDOUT, [ 0, $_ ]) for @members;

This callback is invoked before printing with print only if no error occurred. The callback is invoked with two arguments: the current CSV parser object and an array reference to the fields passed.

The return code of the callback is ignored.

 sub max_4_fields
     my ($csv, $row) = @_;
     @$row > 4 and splice @$row, 4;
     } # max_4_fields

 csv (in => csv (in => "file.csv"), out => *STDOUT,
     callbacks => { before print => \&max_4_fields });

This callback is not active for combine.

Callbacks for csv ()

The csv allows for some callbacks that do not integrate in XS internals but only feature the csv function.

  csv (in        => "file.csv",
       callbacks => {
           filter       => { 6 => sub { $_ > 15 } },    # first
           after_parse  => sub { say "AFTER PARSE";  }, # first
           after_in     => sub { say "AFTER IN";     }, # second
           on_in        => sub { say "ON IN";        }, # third

  csv (in        => $aoh,
       out       => "file.csv",
       callbacks => {
           on_in        => sub { say "ON IN";        }, # first
           before_out   => sub { say "BEFORE OUT";   }, # second
           before_print => sub { say "BEFORE PRINT"; }, # third

filter This callback can be used to filter records. It is called just after a new record has been scanned. The callback accepts a hashref where the keys are the index to the row (the field number, 1-based) and the values are subs to return a true or false value.

 csv (in => "file.csv", filter => {
            3 => sub { m/a/ },       # third field should contain an "a"
            5 => sub { length > 4 }, # length of the 5th field minimal 5

If the keys to the filter contain any character that in not a digit it will also implicitly set headers to auto unless headers was already passed as argument. When headers are active, returning an array of hashes, the filter is not applicable to the header itself.

 csv (in => "file.csv", filter => { foo => sub { $_ > 4 }});

All sub results should match, as in AND.

The context of the callback sets $_ localized to the field indicated by the filter. The two arguments are as with all other callbacks, so the other fields in the current row can be seen:

 filter => { 3 => sub { $_ > 100 ? $_[1][1] =~ m/A/ : $_[1][6] =~ m/B/ }}

If the context is set to return a list of hashes (headers is defined), the current record will also be available in the localized %_:

 filter => { 3 => sub { $_ > 100 && $_{foo} =~ m/A/ && $_{bar} < 1000  }}

If the filter is used to alter the content by changing $_, make sure that the sub returns true in order not to have that record skipped:

 filter => { 2 => sub { $_ = uc }}

will upper-case the second field, and then skip it if the resulting content evaluates to false. To always accept, end with truth:

 filter => { 2 => sub { $_ = uc; 1 }}

after_in This callback is invoked for each record after all records have been parsed but before returning the reference to the caller. The hook is invoked with two arguments: the current CSV parser object and a reference to the record. The reference can be a reference to a HASH or a reference to an ARRAY as determined by the arguments.

This callback can also be passed as an attribute without the callbacks wrapper.

before_out This callback is invoked for each record before the record is printed. The hook is invoked with two arguments: the current CSV parser object and a reference to the record. The reference can be a reference to a HASH or a reference to an ARRAY as determined by the arguments.

This callback can also be passed as an attribute without the callbacks wrapper.

on_in This callback acts exactly as the after_in or the before_out hooks.

This callback can also be passed as an attribute without the callbacks wrapper.

csv The function csv can also be called as a method or with an existing Text::CSV_XS object. This could help if the function is to be invoked a lot of times and the overhead of creating the object internally over and over again would be prevented by passing an existing instance.

 my $csv = Text::CSV_XS->new ({ binary => 1, auto_diag => 1 });

 my $aoa = $csv->csv (in => $fh);
 my $aoa = csv (in => $fh, csv => $csv);

both act the same. Running this 20000 times on a 20 lines CSV file, showed a 53% speedup.


Combine (...)
Parse (...)
The arguments to these internal functions are deliberately not described or documented in order to enable the module authors make changes it when they feel the need for it. Using them is highly discouraged as the API may change in future releases.


    Reading a CSV file line by line:

 my $csv = Text::CSV_XS->new ({ binary => 1, auto_diag => 1 });
 open my $fh, "<", "file.csv" or die "file.csv: $!";
 while (my $row = $csv->getline ($fh)) {
     # do something with @$row
 close $fh or die "file.csv: $!";

Reading only a single column

 my $csv = Text::CSV_XS->new ({ binary => 1, auto_diag => 1 });
 open my $fh, "<", "file.csv" or die "file.csv: $!";
 # get only the 4th column
 my @column = map { $_->[3] } @{$csv->getline_all ($fh)};
 close $fh or die "file.csv: $!";

with csv, you could do

 my @column = map { $_->[0] }
     @{csv (in => "file.csv", fragment => "col=4")};

    Parsing CSV strings:

 my $csv = Text::CSV_XS->new ({ keep_meta_info => 1, binary => 1 });

 my $sample_input_string =
     qq{"I said, ""Hi!""",Yes,"",2.34,,"1.09","\x{20ac}",};
 if ($csv->parse ($sample_input_string)) {
     my @field = $csv->fields;
     foreach my $col (0 .. $#field) {
         my $quo = $csv->is_quoted ($col) ? $csv->{quote_char} : "";
         printf "%2d: %s%s%s\n", $col, $quo, $field[$col], $quo;
 else {
     print STDERR "parse () failed on argument: ",
         $csv->error_input, "\n";
     $csv->error_diag ();

    Printing CSV data

The fast way: using print

An example for creating CSV files using the print method:

 my $csv = Text::CSV_XS->new ({ binary => 1, eol => $/ });
 open my $fh, ">", "foo.csv" or die "foo.csv: $!";
 for (1 .. 10) {
     $csv->print ($fh, [ $_, "$_" ]) or $csv->error_diag;
 close $fh or die "$tbl.csv: $!";

The slow way: using combine and string

or using the slower combine and string methods:

 my $csv = Text::CSV_XS->new;

 open my $csv_fh, ">", "hello.csv" or die "hello.csv: $!";

 my @sample_input_fields = (
     You said, "Hello!",   5.67,
     "Surely",   ,   3.14159);
 if ($csv->combine (@sample_input_fields)) {
     print $csv_fh $csv->string, "\n";
 else {
     print "combine () failed on argument: ",
         $csv->error_input, "\n";
 close $csv_fh or die "hello.csv: $!";

    Rewriting CSV

Rewrite CSV files with ; as separator character to well-formed CSV:

 use Text::CSV_XS qw( csv );
 csv (in => csv (in => "bad.csv", sep_char => ";"), out => *STDOUT);

As STDOUT is now default in csv, a one-liner converting a UTF-16 CSV file with BOM and TAB-separation to valid UTF-8 CSV could be:

 $ perl -C3 -MText::CSV_XS=csv -we\
    csv(in=>"utf16tab.csv",encoding=>"utf16",sep=>"\t") >utf8.csv

    Dumping database tables to CSV

Dumping a database table can be simple as this (TIMTOWTDI):

 my $dbh = DBI->connect (...);
 my $sql = "select * from foo";

 # using your own loop
 open my $fh, ">", "foo.csv" or die "foo.csv: $!\n";
 my $csv = Text::CSV_XS->new ({ binary => 1, eol => "\r\n" });
 my $sth = $dbh->prepare ($sql); $sth->execute;
 $csv->print ($fh, $sth->{NAME_lc});
 while (my $row = $sth->fetch) {
     $csv->print ($fh, $row);

 # using the csv function, all in memory
 csv (out => "foo.csv", in => $dbh->selectall_arrayref ($sql));

 # using the csv function, streaming with callbacks
 my $sth = $dbh->prepare ($sql); $sth->execute;
 csv (out => "foo.csv", in => sub { $sth->fetch            });
 csv (out => "foo.csv", in => sub { $sth->fetchrow_hashref });

Note that this does not discriminate between empty values and NULL-values from the database, as both will be the same empty field in CSV. To enable distinction between the two, use quote_empty.

 csv (out => "foo.csv", in => sub { $sth->fetch }, quote_empty => 1);

If the database import utility supports special sequences to insert NULL values into the database, like MySQL/MariaDB supports \N, use a filter or a map

 csv (out => "foo.csv", in => sub { $sth->fetch },
                     on_in => sub { $_ //= "\\N" for @$_[1] });

 while (my $row = $sth->fetch) {
     $csv->print ($fh, [ map { $_ // "\\N" } @$row ]);

these special sequences are not recognized by Text::CSV_XS on parsing the CSV generated like this, but map and filter are your friends again

 while (my $row = $csv->getline ($io)) {
     $sth->execute (map { $_ eq "\\N" ? undef : $_ } @$row);

 csv (in => "foo.csv", filter => { 1 => sub {
     $sth->execute (map { $_ eq "\\N" ? undef : $_ } @{$_[1]}); 0; }});

    The examples folder

For more extended examples, see the examples/ 1) sub-directory in the original distribution or the git repository 2).


The following files can be found there: This can be used as a boilerplate to parse invalid CSV and parse beyond (expected) errors alternative to using the error callback.

 $ perl examples/ bad.csv >good.csv

csv-check This is a command-line tool that uses techniques to check the CSV file and report on its content.

 $ csv-check files/utf8.csv
 Checked with examples/csv-check 1.5 using Text::CSV_XS 0.81
 OK: rows: 1, columns: 2
     sep = <,>, quo = <">, bin = <1>

csv2xls A script to convert CSV to Microsoft Excel. This requires Date::Calc and Spreadsheet::WriteExcel. The converter accepts various options and can produce UTF-8 Excel files.
csvdiff A script that provides colorized diff on sorted CSV files, assuming first line is header and first field is the key. Output options include colorized ANSI escape codes or HTML.

 $ csvdiff --html --output=diff.html file1.csv file2.csv


Text::CSV_XS is not designed to detect the characters used to quote and separate fields. The parsing is done using predefined (default) settings. In the examples sub-directory, you can find scripts that demonstrate how you could try to detect these characters yourself.

    Microsoft Excel

The import/export from Microsoft Excel is a risky task, according to the documentation in Text::CSV::Separator. Microsoft uses the system’s list separator defined in the regional settings, which happens to be a semicolon for Dutch, German and Spanish (and probably some others as well). For the English locale, the default is a comma. In Windows however, the user is free to choose a predefined locale, and then change every individual setting in it, so checking the locale is no solution.

As of version 1.17, a lone first line with just


will be recognized and honored when parsing with getline.


More Errors & Warnings New extensions ought to be clear and concise in reporting what error has occurred where and why, and maybe also offer a remedy to the problem.

error_diag is a (very) good start, but there is more work to be done in this area.

Basic calls should croak or warn on illegal parameters. Errors should be documented.

setting meta info Future extensions might include extending the meta_info, is_quoted, and is_binary to accept setting these flags for fields, so you can specify which fields are quoted in the combine/string combination.

 $csv->meta_info (0, 1, 1, 3, 0, 0);
 $csv->is_quoted (3, 1);

Metadata Vocabulary for Tabular Data <> (a W3C editor’s draft) could be an example for supporting more metadata.

Parse the whole file at once Implement new methods or functions that enable parsing of a complete file at once, returning a list of hashes. Possible extension to this could be to enable a column selection on the call:

 my @AoH = $csv->parse_file ($filename, { cols => [ 1, 4..8, 12 ]});

Returning something like

 [ { fields => [ 1, 2, "foo", 4.5, undef, "", 8 ],
     flags  => [ ... ],
   { fields => [ ... ],

Note that the csv function already supports most of this, but does not return flags. getline_all returns all rows for an open stream, but this will not return flags either. fragment can reduce the required rows or columns, but cannot combine them.

Cookbook Write a document that has recipes for most known non-standard (and maybe some standard) CSV formats, including formats that use TAB, ;, |, or other non-comma separators.

Examples could be taken from W3C’s CSV on the Web: Use Cases and Requirements <>

Steal Steal good new ideas and features from PapaParse <> or csvkit <>.
Perl6 support I’m already working on perl6 support here <>. No promises yet on when it is finished (or fast). Trying to keep the API alike as much as possible.


combined methods Requests for adding means (methods) that combine combine and string in a single call will <B>notB> be honored (use print instead). Likewise for parse and fields (use getline instead), given the problems with embedded newlines.

    Release plan

No guarantees, but this is what I had in mind some time ago:
o DIAGNOSTICS section in pod to *describe* the errors (see below)


The current hard-coding of characters and character ranges makes this code unusable on EBCDIC systems. Recent work in perl-5.20 might change that.

Opening EBCDIC encoded files on ASCII+ systems is likely to succeed using Encode’s cp37, cp1047, or posix-bc:

 open my $fh, "<:encoding(cp1047)", "ebcdic_file.csv" or die "...";


Still under construction ...

If an error occurs, $csv->error_diag can be used to get information on the cause of the failure. Note that for speed reasons the internal value is never cleared on success, so using the value returned by error_diag in normal cases - when no error occurred - may cause unexpected results.

If the constructor failed, the cause can be found using error_diag as a class method, like Text::CSV_XS->error_diag.

The $csv->error_diag method is automatically invoked upon error when the contractor was called with auto_diag set to 1 or 2, or when autodie is in effect. When set to 1, this will cause a warn with the error message, when set to 2, it will die. 2012 - EOF is excluded from auto_diag reports.

Errors can be (individually) caught using the error callback.

The errors as described below are available. I have tried to make the error itself explanatory enough, but more descriptions will be added. For most of these errors, the first three capitals describe the error category:

Initialization error or option conflict.


Carriage-Return related parse error.


End-Of-File related parse error.


Parse error inside quotation.


Parse error inside field.


Combine error.


HashRef parse related error.

And below should be the complete list of error codes that can be returned:
o 1001 INI - sep_char is equal to quote_char or escape_char

The separation character cannot be equal to the quotation character or to the escape character, as this would invalidate all parsing rules.

o 1002 INI - allow_whitespace with escape_char or quote_char SP or TAB

Using the allow_whitespace attribute when either quote_char or escape_char is equal to SPACE or TAB is too ambiguous to allow.

o 1003 INI - \r or \n in main attr not allowed

Using default eol characters in either sep_char, quote_char, or escape_char is not allowed.

o 1004 INI - callbacks should be undef or a hashref

The callbacks attribute only allows one to be undef or a hash reference.

o 1005 INI - EOL too long

The value passed for EOL is exceeding its maximum length (16).

o 1006 INI - SEP too long

The value passed for SEP is exceeding its maximum length (16).

o 1007 INI - QUOTE too long

The value passed for QUOTE is exceeding its maximum length (16).

o 2010 ECR - QUO char inside quotes followed by CR not part of EOL

When eol has been set to anything but the default, like "\r\t\n", and the "\r" is following the <B>secondB> (closing) quote_char, where the characters following the "\r" do not make up the eol sequence, this is an error.

o 2011 ECR - Characters after end of quoted field

Sequences like 1,foo,"bar"baz,22,1 are not allowed. "bar" is a quoted field and after the closing double-quote, there should be either a new-line sequence or a separation character.

o 2012 EOF - End of data in parsing input stream

Self-explaining. End-of-file while inside parsing a stream. Can happen only when reading from streams with getline, as using parse is done on strings that are not required to have a trailing eol.

o 2013 INI - Specification error for fragments RFC7111

Invalid specification for URI fragment specification.

o 2021 EIQ - NL char inside quotes, binary off

Sequences like 1,"foo\nbar",22,1 are allowed only when the binary option has been selected with the constructor.

o 2022 EIQ - CR char inside quotes, binary off

Sequences like 1,"foo\rbar",22,1 are allowed only when the binary option has been selected with the constructor.

o 2023 EIQ - QUO character not allowed

Sequences like "foo "bar" baz",qu and 2023,",2008-04-05,"Foo, Bar",\n will cause this error.

o 2024 EIQ - EOF cannot be escaped, not even inside quotes

The escape character is not allowed as last character in an input stream.

o 2025 EIQ - Loose unescaped escape

An escape character should escape only characters that need escaping.

Allowing the escape for other characters is possible with the attribute allow_loose_escape.

o 2026 EIQ - Binary character inside quoted field, binary off

Binary characters are not allowed by default. Exceptions are fields that contain valid UTF-8, that will automatically be upgraded if the content is valid UTF-8. Set binary to 1 to accept binary data.

o 2027 EIQ - Quoted field not terminated

When parsing a field that started with a quotation character, the field is expected to be closed with a quotation character. When the parsed line is exhausted before the quote is found, that field is not terminated.

o 2030 EIF - NL char inside unquoted verbatim, binary off
o 2031 EIF - CR char is first char of field, not part of EOL
o 2032 EIF - CR char inside unquoted, not part of EOL
o 2034 EIF - Loose unescaped quote
o 2035 EIF - Escaped EOF in unquoted field
o 2036 EIF - ESC error
o 2037 EIF - Binary character in unquoted field, binary off
o 2110 ECB - Binary character in Combine, binary off
o 2200 EIO - print to IO failed. See errno
o 3001 EHR - Unsupported syntax for column_names ()
o 3002 EHR - getline_hr () called before column_names ()
o 3003 EHR - bind_columns () and column_names () fields count mismatch
o 3004 EHR - bind_columns () only accepts refs to scalars
o 3006 EHR - bind_columns () did not pass enough refs for parsed fields
o 3007 EHR - bind_columns needs refs to writable scalars
o 3008 EHR - unexpected error in bound fields
o 3009 EHR - print_hr () called before column_names ()
o 3010 EHR - print_hr () called with invalid arguments


IO::File, IO::Handle, IO::Wrap, Text::CSV, Text::CSV_PP, Text::CSV::Encoded, Text::CSV::Separator, Text::CSV::Slurp, Spreadsheet::CSV and Spreadsheet::Read, and of course perl.

If you are using perl6, you can have a look at Text::CSV in the perl6 ecosystem, offering the same features.


A CSV parser in JavaScript, also used by W3C <>, is the multi-threaded in-browser PapaParse <>.

csvkit <> is a python CSV parsing toolkit.


Alan Citterman <> wrote the original Perl module. Please don’t send mail concerning Text::CSV_XS to Alan, who is not involved in the C/XS part that is now the main part of the module.

Jochen Wiedmann <> rewrote the en- and decoding in C by implementing a simple finite-state machine. He added variable quote, escape and separator characters, the binary mode and the print and getline methods. See ChangeLog releases 0.10 through 0.23.

H.Merijn Brand <> cleaned up the code, added the field flags methods, wrote the major part of the test suite, completed the documentation, fixed most RT bugs, added all the allow flags and the csv function. See ChangeLog releases 0.25 and on.


 Copyright (C) 2007-2015 H.Merijn Brand.  All rights reserved.
 Copyright (C) 1998-2001 Jochen Wiedmann. All rights reserved.
 Copyright (C) 1997      Alan Citterman.  All rights reserved.

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

Search for    or go to Top of page |  Section 3 |  Main Index

perl v5.20.3 CSV_XS (3) 2015-05-21

Powered by GSP Visit the GSP FreeBSD Man Page Interface.
Output converted with manServer 1.07.