Miko O'Sullivan

NAME

JSON::Relaxed -- An extension of JSON that allows for better human-readability.

SYNOPSIS

 my ($rjson, $hash, $parser);
 
 # raw RJSON code
 $rjson = <<'(RAW)';
 /* Javascript-like comments are allowed */
 {
   // single or double quotes allowed
   a : 'Larry',
   b : "Curly",
   
   // nested structures allowed like in JSON
   c: [
      {a:1, b:2},
   ],
   
   // like Perl, trailing commas are allowed
   d: "more stuff",
 }
 (RAW)
 
 # subroutine parsing
 $hash = from_rjson($rjson);
 
 # object-oriented parsing
 $parser = JSON::Relaxed::Parser->new();
 $hash = $parser->parse($rjson);

INSTALLATION

JSON::Relaxed can be installed with the usual routine:

 perl Makefile.PL
 make
 make test
 make install

DESCRIPTION

JSON::Relaxed is a lightweight parser and serializer for an extension of JSON called Relaxed JSON (RJSON). The intent of RJSON is to provide a format that is more human-readable and human-editable than JSON. Most notably, RJSON allows the use of JavaScript-like comments. By doing so, configuration files and other human-edited files can include comments to indicate the intention of each configuration.

JSON::Relaxed is currently only a parser that reads in RJSON code and produces a data structure. JSON::Relaxed does not currently encode data structures into JSON/RJSON. That feature is planned.

Why Relaxed JSON?

There's been increasing support for the idea of expanding JSON to improve human-readability. "Relaxed" JSON is a term that has been used to describe a JSON-ish format that has some features that JSON doesn't. Although there isn't yet any kind of official specification, descriptions of Relaxed JSON generally include the following extensions to JSON:

  • comments

    RJSON supports JavaScript-like comments:

     /* inline comments */
     // line-based comments
  • trailing commas

    Like Perl, RJSON allows treats commas as separators. If nothing is before, after, or between commas, those commas are just ignored:

     [
        , // nothing before this comma
        "data",
        , // nothing after this comma
     ]
  • single quotes, double quotes, no quotes

    Strings can be quoted with either single or double quotes. Space-less strings are also parsed as strings. So, the following data items are equivalent:

     [
        "Starflower",
        'Starflower',
        Starflower
     ]

    Note that unquoted boolean values are still treated as boolean values, so the following are NOT the same:

     [
        "true",  // string
        true,    // boolean true
        
        "false", // string
        false,   // boolean false
        
        "null", // string
        null, // what Perl programmers call undef
     ]

    Because of this ambiguity, unquoted non-boolean strings should be considered sloppy and not something you do in polite company.

  • documents that are just a single string

    Early versions of JSON require that a JSON document contains either a single hash or a single array. Later versions also allow a single string. RJSON follows that later rule, so the following is a valid RJSON document:

     "Hello world"
  • hash keys without values

    A hash in JSON can have a key that is followed by a comma or a closing } without a specified value. In that case the hash element is simply assigned the undefined value. So, in the following example, a is assigned 1, b is assigned 2, and c is assigned undef:

     {
        a: 1,
        b: 2,
        c
     }

from_rjson()

from_rjson() is the simple way to quickly parse an RJSON string. Currently from_rjson() only takes a single parameter, the string itself. So in the following example, from_rjson() parses and returns the structure defined in $rjson.

 $structure = from_rjson($rjson);

Object-oriented parsing

To parse using an object, create a JSON::Relaxed::Parser object, like this:

 $parser = JSON::Relaxed::Parser->new();

Then call the parser's <code>parse</code> method, passing in the RJSON string:

 $structure = $parser->parse($rjson);

Methods

  • $parser->extra_tokens_ok()

    extra_tokens_ok() sets/gets the extra_tokens_ok property. By default, extra_tokens_ok is false. If by extra_tokens_ok is true then the multiple-structures isn't triggered and the parser returns the first structure it finds. So, for example, the following code would return undef and sets the multiple-structures error:

     $parser = JSON::Relaxed::Parser->new();
     $structure = $parser->parse('{"x":1} []');

    However, by setting multiple-structures to true, a hash structure is returned, the extra code after that first hash is ignored, and no error is set:

     $parser = JSON::Relaxed::Parser->new();
     $parser->extra_tokens_ok(1);
     $structure = $parser->parse('{"x":1} []');

Error codes

When JSON::Relaxed encounters a parsing error it returns undef and sets two global variables:

  • $JSON::Relaxed::err_id

    $err_id is a unique code for a specific error. Every code is set in only one place in JSON::Relaxed.

  • $JSON::Relaxed::err_msg

    $err_msg is an English description of the code. It would be cool to migrate towards multi-language support for $err_msg.

Following is a list of all error codes in JSON::Relaxed:

  • missing-parameter

    The string to be parsed was not sent to $parser->parse(). For example:

     $parser->parse()
  • undefined-input

    The string to be parsed is undefined. For example:

     $parser->parse(undef)
  • zero-length-input

    The string to be parsed is zero-length. For example:

     $parser->parse('')
  • space-only-input

    The string to be parsed has no content beside space characters. For example:

     $parser->parse('   ')
  • no-content

    The string to be parsed has no content. This error is slightly different than space-only-input in that it is triggered when the input contains only comments, like this:

     $parser->parse('/* whatever */')
  • unclosed-inline-comment

    A comment was started with /* but was never closed. For example:

     $parser->parse('/*')
  • invalid-structure-opening-character

    The document opens with an invalid structural character like a comma or colon. The following examples would trigger this error.

     $parser->parse(':')
     $parser->parse(',')
     $parser->parse('}')
     $parser->parse(']')
  • multiple-structures

    The document has multiple structures. JSON and RJSON only allow a document to consist of a single hash, a single array, or a single string. The following examples would trigger this error.

     $parse->parse('{}[]')
     $parse->parse('{} "whatever"')
     $parse->parse('"abc" "def"')
  • unknown-token-after-key

    A hash key may only be followed by the closing hash brace or a colon. Anything else triggers unknown-token-after-key. So, the following examples would trigger this error.

     $parse->parse("{a [ }") }
     $parse->parse("{a b") }
  • unknown-token-for-hash-key

    The parser encountered something besides a string where a hash key should be. The following are examples of code that would trigger this error.

     $parse->parse('{{}}')
     $parse->parse('{[]}')
     $parse->parse('{]}')
     $parse->parse('{:}')
  • unclosed-hash-brace

    A hash has an opening brace but no closing brace. For example:

     $parse->parse('{x:1')
  • unclosed-array-brace

    An array has an opening brace but not a closing brace. For example:

     $parse->parse('["x", "y"')
  • unexpected-token-after-colon

    In a hash, a colon must be followed by a value. Anything else triggers this error. For example:

     $parse->parse('{"a":,}')
     $parse->parse('{"a":}')
  • missing-comma-between-array-elements

    In an array, a comma must be followed by a value, another comma, or the closing array brace. Anything else triggers this error. For example:

     $parse->parse('[ "x" "y" ]')
     $parse->parse('[ "x" : ]')
  • unknown-array-token

    This error exists just in case there's an invalid token in an array that somehow wasn't caught by missing-comma-between-array-elements. This error shouldn't ever be triggered. If it is please let me know.

  • unclosed-quote

    This error is triggered when a quote isn't closed. For example:

     $parse->parse("'whatever")
     $parse->parse('"whatever') }

INTERNALS

The following documentation is for if you want to edit the code of JSON::Relaxed itself.

JSON::Relaxed

JSON::Relaxed is the parent package. Not a lot actually happens in JSON::Relaxed, it mostly contains from_rjson() and definitions of various structures.

Special character and string definitions

The following hashes provide information about characters and strings that have special meaning in RJSON.

  • Escape characters

    The %esc hash defines the six escape characters in RJSON that are changed to single characters. %esc is defined as follows.

     our %esc = (
       'b'   => "\b",    #  Backspace
       'f'   => "\f",    #  Form feed
       'n'   => "\n",    #  New line
       'r'   => "\r",    #  Carriage return
       't'   => "\t",    #  Tab
       'v'   => chr(11), #  Vertical tab
     );
  • Structural characters

    The %structural hash defines the six characters in RJSON that define the structure of the data object. The structural characters are defined as follows.

     our %structural = (
       '[' => 1, # beginning of array
       ']' => 1, # end of array
       '{' => 1, # beginning of hash
       '}' => 1, # end of hash
       ':' => 1, # delimiter between name and value of hash element
       ',' => 1, # separator between elements in hashes and arrays
     );
  • Quotes

    The %quotes hash defines the two types of quotes recognized by RJSON: single and double quotes. JSON only allows the use of double quotes to define strings. Relaxed also allows single quotes. %quotes is defined as follows.

     our %quotes = (
       '"' => 1,
       "'" => 1,
     );
  • End of line characters

    The %newlines hash defines the three ways a line can end in a RJSON document. Lines in Windows text files end with carriage-return newline ("\r\n"). Lines in Unixish text files end with newline ("\n"). Lines in some operating systems end with just carriage returns ("\n"). %newlines is defined as follows.

     our %newlines = (
       "\r\n" => 1,
       "\r" => 1,
       "\n" => 1,
     );
  • Boolean

    The %boolean hash defines strings that are boolean values: true, false, and null. (OK, 'null' isn't just a boolean value, but I couldn't think of what else to call this hash.) %boolean is defined as follows.

     our %boolean = (
       'null' => 1,
       'true' => 1,
       'false' => 1,
     );

JSON::Relaxed::Parser

A JSON::Relaxed::Parser object parses the raw RJSON string. You don't need to instantiate a parser if you just want to use the default settings. In that case just use from_rjson().

You would create a JSON::Relaxed::Parser object if you want to customize how the string is parsed. I say "would" because there isn't actually any customization in these early releases. When there is you'll use a parser object.

To parse in an object oriented manner, create the parser, then parse.

 $parser = JSON::Relaxed::Parser->new();
 $structure = $parser->parse($string);
new

JSON::Relaxed::Parser-new()> creates a parser object. Its simplest and most common use is without any parameters.

 my $parser = JSON::Relaxed::Parser->new();
option: unknown

The unknown option sets the character which creates the unknown object. The unknown object exists only for testing JSON::Relaxed. It has no purpose in production use.

 my $parser = JSON::Relaxed::Parser->new(unknown=>'~');
Parser "is" methods

The following methods indicate if a token has some specific property, such as being a string object or a structural character.

  • is_string()

    Returns true if the token is a string object, i.e. in the class JSON::Relaxed::Parser::Token::String.

  • is_struct_char()

    Returns true if the token is one of the structural characters of JSON, i.e. one of the following:

     { } [ ] : ,
  • is_unknown_char()

    Returns true if the token is the unknown character.

  • is_list_opener()

    Returns true if the token is the opening character for a hash or an array, i.e. it is one of the following two characters:

     { [
  • is_comment_opener()

    Returns true if the token is the opening character for a comment, i.e. it is one of the following two couplets:

     /*
     //
parse()

parse() is the method that does the work of parsing the RJSON string. It returns the data structure that is defined in the RJSON string. A typical usage would be as follows.

 my $parser = JSON::Relaxed::Parser->new();
 my $structure = $parser->parse('["hello world"]');

parse() does not take any options.

parse_chars()

parse_chars() parses the RJSON string into either individual characters or two-character couplets. This method returns an array. The only input is the raw RJSON string. So, for example, the following string:

 $raw = qq|/*x*/["y"]|;
 @chars = $parser->parse_chars($raw);

would be parsed into the following array:

 ( "/*", "x", "*/", "[", "\"", "y", "\""", "]" )

Most of the elements in the array are single characters. However, comment delimiters, escaped characters, and Windows-style newlines are parsed as two-character couplets:

  • \ followed by any character

  • \r\n

  • //

  • /*

  • */

parse_chars() should not produce any fatal errors.

tokenize()

tokenize() organizes the characters from parse_chars() into tokens. Those tokens can then be organized into a data structure with structure().

Each token represents an item that is recognized by JSON. Those items include structural characters such as { or }, or strings such as "hello world". Comments and insignificant whitespace are filtered out by tokenize().

For example, this code:

 $parser = JSON::Relaxed::Parser->new();
 $raw = qq|/*x*/ ["y"]|;
 @chars = $parser->parse_chars($raw);
 @tokens = $parser->tokenize(\@chars);

would produce an array like this:

 (
   '[',
   JSON::Relaxed::Parser::Token::String::Quoted=HASH(0x20bf0e8),
   ']'
 )

Strings are tokenized into string objects. When the parsing is complete they are returned as scalar variables, not objects.

tokenize() should not produce any fatal errors.

structure()

$parser-structure()> organizes the tokens from tokenize() into a data structure. $parser-structure()> returns a single string, single array reference, a single hash reference, or (if there are errors) undef.

JSON::Relaxed::Parser::Structure::Hash

This package parses Relaxed into hash structures. It is a static package, i.e. it is not instantiated.

build()

This static method accepts the array of tokens and works through them building the hash reference that they represent. When build() reaches the closing curly brace (}) it returns the hash reference.

get_value

This static method gets the value of a hash element. This method is called after a hash key is followed by a colon. A colon must be followed by a value. It may not be followed by the end of the tokens, a comma, or a closing brace.

JSON::Relaxed::Parser::Structure::Array

This package parses Relaxed into array structures. It is a static package, i.e. it is not instantiated.

build()

This static method accepts the array of tokens and works through them building the array reference that they represent. When build() reaches the closing square brace (]) it returns the array reference.

missing_comma()

This static method build the missing-comma-between-array-elements error message.

invalid_array_token)

This static method build the unknown-array-token error message.

JSON::Relaxed::Parser::Token::String

Base class . Nothing actually happens in this package, it's just a base class for JSON::Relaxed::Parser::Token::String::Quoted and JSON::Relaxed::Parser::Token::String::Unquoted.

JSON::Relaxed::Parser::Token::String::Quoted

A JSON::Relaxed::Parser::Token::String::Quoted object represents a string in the document that is delimited with single or double quotes. In the following example, Larry and Curly would be represented by Quoted objects by Moe would not.

 [
    "Larry",
    'Curly',
    Moe
 ]

Quoted objects are created by $parser->tokenize() when it works through the array of characters in the document.

  • new()

    new() instantiates a JSON::Relaxed::Parser::Token::String::Quoted object and slurps in all the characters in the characters array until it gets to the closing quote. Then it returns the new Quoted object.

    A Quoted object has the following two properties:

    raw: the string that is inside the quotes. If the string contained any escape characters then the escapes are processed and the unescaped characters are in raw. So, for example, \n would become an actual newline.

    quote: the delimiting quote, i.e. either a single quote or a double quote.

  • as_perl()

    as_perl() returns the string that was in quotes (without the quotes).

JSON::Relaxed::Parser::Token::String::Unquoted

A JSON::Relaxed::Parser::Token::String::Unquoted object represents a string in the document that was not delimited quotes. In the following example, Moe would be represented by an Unquoted object, but Larry and Curly would not.

 [
    "Larry",
    'Curly',
    Moe
 ]

Unquoted objects are created by $parser->tokenize() when it works through the array of characters in the document.

An Unquoted object has one property, raw, which is the string. Escaped characters are resolved in raw.

  • new()

    new() instantiates a JSON::Relaxed::Parser::Token::String::Unquoted object and slurps in all the characters in the characters array until it gets to a space character, a comment, or one of the structural characters such as { or :.

  • as_perl()

    as_perl() returns the unquoted string or a boolean value, depending on how it is called.

    If the string is a boolean value, i.e. true, false, then the as_perl return 1 (for true), 0 (for false) or undef (for null), unless the always_string option is sent, in which case the string itself is returned. If the string does not represent a boolean value then it is returned as-is.

    $parser->structure() sends the always_string when the token is a key in a hash. The following example should clarify how always_string is used:

     {
        // key: the literal string "larry"
        // value: 1
        larry : true,
        
        // key: the literal string "true"
        // value: 'x'
        true : 'x',
        
        // key: the literal string "null"
        // value: 'y'
        null : 'y',
        
        // key: the literal string "z"
        // value: undef
        z : null,
     }

JSON::Relaxed::Parser::Token::Unknown

This class is just used for development of JSON::Relaxed. It has no use in production. This class allows testing for when a token is an unknown object.

To implement this class, add the 'unknown' option to JSON::Relaxed->new(). The value of the option should be the character that creates an unknown object. For example, the following option sets the tilde (~) as an unknown object.

 my $parser = JSON::Relaxed::Parser->new(unknown=>'~');

The "unknown" character must not be inside quotes or inside an unquoted string.

TERMS AND CONDITIONS

Copyright (c) 2014 by Miko O'Sullivan. All rights reserved. This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself. This software comes with NO WARRANTY of any kind.

AUTHOR

Miko O'Sullivan miko@idocs.com

VERSION

Version: 0.04

HISTORY

Version 0.01 Nov 30, 2014

Initial version.

Version 0.02 Dec 3, 2014

Fixed test.t so that it can load lib.pm when it runs.

Added $parser->extra_tokens_ok(). Removed error code invalid-structure-opening-string and allowed that error to fall through to multiple-structures.

Cleaned up documentation.

Version 0.03 Dec 6, 2014

Modified test for parse_chars to normalize newlines. Apparently the way Perl on Windows handles newline is different than what I expected, but as long as it's recognizing newlines and|or carriage returns then the test should pass.

Version 0.04 Apr 28, 2016

Fixed bug in which end of line did not terminate some line comments.

Minor cleanups of documentation.

Cleaned up test.pl.

Version 0.05 Apr 30, 2016

Fixed bug: Test::Most was not added to the prerequisite list. No changes to the functionality of the module itself.