The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


JSON::DWIW - JSON converter that Does What I Want


 use JSON::DWIW;
 my $json_obj = JSON::DWIW->new;
 my $data = $json_obj->from_json($json_str);
 my $str = $json_obj->to_json($data);
 my ($data, $error_string) = $json_obj->from_json($json_str);
 my $data = JSON::DWIW::deserialize($json_str);
 my $error_str = JSON::DWIW::get_error_string();
 use JSON::DWIW qw/deserialize_json from_json/
 my $data = deserialize_json($json_str);
 my $error_str = JSON::DWIW::get_error_string();
 my $error_string = $json_obj->get_error_string;
 my $error_data = $json_obj->get_error_data;
 my $stats = $json_obj->get_stats;
 my $data = $json_obj->from_json_file($file)
 my $ok = $json_obj->to_json_file($data, $file);
 my $data = JSON::DWIW->from_json($json_str);
 my $str = JSON::DWIW->to_json($data);
 my $data = JSON::DWIW->from_json($json_str, \%options);
 my $str = JSON::DWIW->to_json($data, \%options);
 my $true_value = JSON::DWIW->true;
 my $false_value = JSON::DWIW->false;
 my $data = { var1 => "stuff", var2 => $true_value,
              var3 => $false_value, };
 my $str = JSON::DWIW->to_json($data);

 my $data = JSON::DWIW::deserialize($str, { start_depth => 1,
                                            start_depth_handler => $handler });


Other JSON modules require setting several parameters before calling the conversion methods to do what I want. This module does things by default that I think should be done when working with JSON in Perl. This module also encodes and decodes faster than and JSON::Syck in my benchmarks.

This means that any piece of data in Perl (assuming it's valid unicode) will get converted to something in JSON instead of throwing an exception. It also means that output will be strict JSON, while accepted input will be flexible, without having to set any options.

For a list of changes in recent versions, see the documentation for JSON::DWIW::Changes.

This module can be downloaded from


Perl objects get encoded as their underlying data structure, with the exception of Math::BigInt and Math::BigFloat, which will be output as numbers, and JSON::DWIW::Boolean, which will get output as a true or false value (see the true() and false() methods). For example, a blessed hash ref will be represented as an object in JSON, a blessed array will be represented as an array. etc. A reference to a scalar is dereferenced and represented as the scalar itself. Globs, Code refs, etc., get stringified, and undef becomes null.

Scalars that have been used as both a string and a number will be output as a string. A reference to a reference is currently output as an empty string, but this may change.

You may notice there is a deserialize function, but not a serialize one. The deserialize function was written as a full rewrite (the parsing is in a separate, event-based library now) of from_json (now from_json calls deserialize). In the future, there will be a serialize function that is a rewrite of to_json.


Input is expected to utf-8. When decoding, null, true, and false become undef, 1, and 0, repectively. Numbers that appear to be too long to be supported natively are converted to Math::BigInt or Math::BigFloat objects, if you have them installed. Otherwise, long numbers are turned into strings to prevent data loss.

The parser is flexible in what it accepts and handles some things not in the JSON spec:


Both single and double quotes are allowed for quoting a string, e.g.,

    [ "string1", 'string2' ]
bare keys

Object/hash keys can be bare if they look like an identifier, e.g.,

    { var1: "myval1", var2: "myval2" }
extra commas

Extra commas in objects/hashes and arrays are ignored, e.g.,


becomes a 4 element array containing 1, 2, 3, and 4.

escape sequences

Latin1 hexadecimal escape sequences (\xHH) are accepted, as in Javascript. Also, the vertical tab escape \v is recognized (\u000b).


C, C++, and shell-style comments are accepted. That is

 /* this is a comment */
 // this is a comment

 # this is also a comment



Create a new JSON::DWIW object.

%options is an optional hash of parameters that will change the bahavior of this module when encoding to JSON. You may also pass these options as the second argument to to_json() and from_json(). The following options are supported:


If set to a true value, keys in hashes will not be quoted when converted to JSON if they look like identifiers. This is valid Javascript in current browsers, but not in JSON.


If set to a true value, errors found when converting to or from JSON will result in die() being called with the error message. The default is to not use exceptions.


This options indicates what should be done if bad characters are found, e.g., bad utf-8 sequence. The default is to return an error and drop all the output.

The following values for bad_char_policy are supported:


default action, i.e., drop any output built up and return an error


Convert to a utf-8 char using the value of the byte as a code point. This is basically the same as assuming the bad character is in latin-1 and converting it to utf-8.


Ignore the error and pass through the raw bytes (invalid JSON)


If set to a true value, escape all multi-byte characters (e.g., \u00e9) when converting to JSON.


Synonym for escape_multi_byte


Add white space to the output when calling to_json() to make the output easier for humans to read.


When converting from JSON, return objects for booleans so that "true" and "false" can be maintained when encoding and decoding. If this flag is set, then "true" becomes a JSON::DWIW::Boolean object that evaluates to true in a boolean context, and "false" becomes an object that evaluates to false in a boolean context. These objects are recognized by the to_json() method, so they will be output as "true" or "false" instead of "1" or "0".


Don't escape solidus characters ("/") in strings. The output is still legal JSON with this option turned on.


Only do required escaping in strings (solidus and quote). Tabs, newlines, backspaces, etc., will not be escaped with this optioned turned on (but the output will still be valid JSON).


Set to a true value to sort hash keys (alphabetically) when converting to JSON.


A subroutine reference to call when parsing a number. The subroutine will be provided one string that is the number being parsed. The return value from the subroutine will be used to populate the return data instead of converting to a number.


  my $json = '{ "a": 6.3e-10 }';
  my $cb = sub { my ($val) = @_; return "I got the number '$val'"; };
  my $data = JSON::DWIW::deserialize($json, { parse_number => $cb });


A subroutine reference to call when parsing a constant (true, false, or null). The subroutine will be provided one string that is the constant being parsed. The return value from the subroutine will be used to populate the return data instead of converting to a boolean or undef. See the "parse_number" option.


Depth at which start_depth_handler should be called. See "start_depth_handler".


A reference to a subroutine to called when parsing and at level start_depth in the data structure. When specified along with start_depth, the parser does not return the entire data structure. Instead, it calls start_depth_handler for each element in the array when the parser is at level start_depth. This is useful for parsing a very large array without loading all the data into memory (especially when using deserialize_file).

E.g., with start_depth set to 1 and start_depth_handler set to $handler:

    my $str = '[ { "foo": "bar", "cat": 1 }, { "concat": 1, "lambda" : [ "one", 2, 3 ] } ]';
    my $foo = { foo => [ ] };
    my $handler = sub { push @{$foo->{foo}}, $_[0]; return 1; };
    my $data = JSON::DWIW::deserialize($str, { start_depth => 1,
                                               start_depth_handler => $handler });
    print STDERR Data::Dumper->Dump([ $foo ], [ 'foo' ]) . "\n";
    print STDERR Data::Dumper->Dump([ $data ], [ 'leftover_data' ]) . "\n";

    # Output
    $foo = {
             'foo' => [
                          'cat' => 1,
                          'foo' => 'bar'
                          'lambda' => [
                          'concat' => 1

    $leftover_data = [];


Returns the JSON representation of $data (arbitrary datastructure). See for details.

Called in list context, this method returns a list whose first element is the encoded JSON string and the second element is an error message, if any. If $error_msg is defined, there was a problem converting to JSON. You may also pass a second argument to to_json() that is a reference to a hash of options -- see new().

     my $json_str = JSON::DWIW->to_json($data);
     my ($json_str, $error_msg) = JSON::DWIW->to_json($data);
     my $json_str = JSON::DWIW->to_json($data, { use_exceptions => 1 });

Aliases: toJson, toJSON, objToJson

deserialize($json_str, \%options)

Returns the Perl data structure for the given JSON string. The value for true becomes 1, false becomes 0, and null gets converted to undef.

This function should not be called as a method (for performance reasons). Unlike from_json(), it returns a single value, the data structure resulting from the conversion. If the return value is undef, check the result of the get_error_string() function/method to see if an error is defined.

deserialize_file($file, \%options)

Same as deserialize, except that it takes a file as an argument. On Unix, this mmap's the file, so it does not load a big file into memory all at once, and does less buffer copying.


Similar to deserialize(), but expects to be called as a method.

Called in list context, this method returns a list whose first element is the data and the second element is the error message, if any. If $error_msg is defined, there was a problem parsing the JSON string, and $data will be undef. You may also pass a second argument to from_json() that is a reference to a hash of options -- see new().

     my $data = from_json($json_str)
     my ($data, $error_msg) = from_json($json_str)

Aliases: fromJson, fromJSON, jsonToObj


Similar to deserialize_file(), except that it expects to be called a a method, and it also returns the error, if any, when called in list context.

my ($data, $error_msg) = $json->from_json_file($file, \%options)


Converts $data to JSON and writes the result to the file $file. Currently, this is simply a convenience routine that converts the data to a JSON string and then writes it to the file.

 my ($ok, $error) = $json->to_json_file($data, $file, \%options);


Returns the error message from the last call, if there was one, e.g.,

 my $data = JSON::DWIW->from_json($json_str)
     or die "JSON error: " . JSON::DWIW->get_error_string;
 my $data = $json_obj->from_json($json_str)
     or die "JSON error: " . $json_obj->get_error_string;

Aliases: get_err_str(), errstr()


Returns the error details from the last call, in a hash ref, e.g.,

 $error_data = {
                'byte' => 23,
                'byte_col' => 23,
                'col' => 22,
                'char' => 22,
                'version' => '0.15a',
                'line' => 1

This is really only useful when decoding JSON.

Aliases: get_error(), error()


Returns statistics from the last method called to encode or decode. E.g., for an encoding (to_json() or to_json_file()),

    $stats = {
               'bytes' => 78,
               'nulls' => 1,
               'max_string_bytes' => 5,
               'max_depth' => 2,
               'arrays' => 1,
               'numbers' => 6,
               'lines' => 1,
               'max_string_chars' => 5,
               'strings' => 6,
               'bools' => 1,
               'chars' => 78,
               'hashes' => 1


Returns an object that will get output as a true value when encoding to JSON.


Returns an object that will get output as a false value when encoding to JSON.

json_to_xml($json, \%params)

This function (not a method) converts the given JSON to XML. Hash/object keys become tag names. Arrays that are hash values are output as multiple tags with the hash key as the tag name.

Any characters in hash keys not in [\w-] (i.e., letters, numbers, underscore, or dash), get converted to underscore ("_") when output as XML tags.

Valid parameters in \%params are the same as for passing to deserialize() or from_json(), plus the pretty option, which will add newlines and indentation to the XML to make it more human-readable.


Following are some methods I use for debugging and testing.


Returns true if the given string is flagged as utf-8.


Flags the given string as utf-8.


Clears the flag that tells Perl the string is utf-8.


Returns true if the given string is valid utf-8 (regardless of the flag).


Converts the string to utf-8, assuming it is latin1. This effects $str itself in place, but also returns $str.


Returns a utf8 string containing the byte sequence for the given code point.


Returns a string representing the byte sequence for $cp encoding in utf-8. E.g.,

 my $hex_bytes = JSON::DWIW->code_point_to_hex_bytes(0xe9);
 print "$hex_bytes\n"; # \xc3\xa9


Returns a reference to an array of code points from the given string, assuming the string is encoded in utf-8.


Dumps the internal structure of the given scalar.


Need new benchmarks here.


Perl 5.6 or later


If you find a bug, please file a tracker request at

When decoding a JSON string, it is a assumed to be utf-8 encoded. The module should detect whether the input is utf-8, utf-16, or utf-32.


Don Owens <>


Thanks to Asher Blum for help with testing.

Thanks to Nigel Bowden for helping with compilation on Windows.

Thanks to Robert Peters for discovering and tracking down the source of a number parsing bug.

Thanks to Mark Phillips for helping with a bug under Solaris on Sparc.

Thanks to Josh for helping debug [ #47344].


Copyright (c) 2007-2010 Don Owens <>. All rights reserved.

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

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.


The JSON home page:
The JSON spec:
The JSON-RPC spec:
JSON::Syck (included in YAML::Syck)