SJSON is a small library to read/write simplified JSON, as described originally on the Bitsquid blog.

SJSON is licensed under the two-clause BSD license. See LICENSE.txt for details.

SJSON is very similar to normal JSON (in fact, since release 1.2.0, the SJSON library will also load plain JSON). It mostly reduces the required markup a bit. The main differences are:

• Every file starts with an implicit object. That is, an empty SJSON file is equivalent to a JSON file containing {}.
• Commas after a key-value pair are optional.
• Keys don't have to be quoted as long as they are valid identifiers. An identifier consists of letters, digits, and _.
• = is allowed in addition to : for key-value separation. The canonical separator is =.
• C and C++ style comments are allowed.

In addition, this library provides support for raw string literals.

JSON:

{
    "foo" : 23,
    "bar" : [1, 2, 3],
    "baz" : {
        "key" : "value"
    }
}

SJSON:

foo = 23
bar = [1, 2, 3]
baz = {
    // SJSON also allows for comments
    key = "value"
}

As an extension, SJSON allows for raw string literals, in both Lua and Python flavors:

foo = [=[This is a raw literal with embedded " and stuff]=]

foo = """This is a raw literal with embedded " and stuff"""

The library provides four methods, similar to the Python JSON module. These are:

• dump: Encode an object as SJSON and write to a stream.
• dumps: Encode an object as SJSON and return a string.
• load: Decode a SJSON encoded object from a stream.
• loads: Decode a SJSON encoded object from a string.

• Bump minimum Python to 3.12.
• Add type information.
• Various internal cleanups & refactorings.

• Packaging changes only. Raised supported Python version to Python 3.12

• Add support for Python-style raw strings, delimited by """.
• Improve handling of unknown string escapes. Previously, those would raise an exception, now they get passed through.

• Re-release of 2.0.2.

• Packaging changes only. This release contains packaging changes only and has not been released to the public, use 2.0.3 instead.

• Add dump in addition to dumps for consistency with the Python JSON module.
• Additional PEP8 conformance tweaks.

• The library is now PEP8 compliant. This should not affect most users of this library, the only user-visible change is that ParseException.GetLocation has been renamed to get_location. The core functions have not been renamed and are API compatible.

• Keys did not get quoted properly during encoding if they contained special characters.
• List elements were incorrectly indented.
• List indentation now accepts either a string or a number (similar to the Python JSON module.)
• Both : and = are now supported as key-value separators, allowing the SJSON library to parse plain JSON files.

• Add support for C/C++ style comments.
• Line/column numbers start at 1 now (previously, the first character was in line 0, column 0).

• Parsing performance has been significantly improved.
• It is possible to parse a file-like stream or string now.

• Track position during parsing. This will likely reduce the performance a bit, but allows for much better error messages.
• Input is byte-oriented now.

• Add support for raw string literals. These are delimited by [=[ ]=] and don't require escaping inside the string.

• Strings with whitespace are now properly escaped.

• Various fixes to string encoding/decoding bugs.
• Encoding now uses collections.abc to identify sequences and mappings instead of testing directly against list and dict.

Initial PyPI release.