CHANGES.md

v3.1.0

• Dialyzer and api spec improvements

v3.0

• drop support for OTP versions before 17.0
• remove definition options for disabling maps globally, {return_maps, false} is still an accepted option to decode/2

v2.8.2

• enable debug_info for rebar3

v2.8.1

• enable debug_info when used via mix
• accept erlang:timestamp as input to the parser

v2.8.0

• add JSX_FORCE_MAPS env var for forcing decoding to maps rather than attempting to autodetect

v2.7.2

• fix an issue where tuples were assumed to be jsx ast and not checked
• mask a function_clause error in encoder with a badarg error for api unity

v2.7.1

• support for milliseconds in datetimes

v2.7.0

• return_tail option
• fixes for edoc generation

v2.6.2

• ensure maps are always enabled when compiling via mix

v2.6.1

• hex.pm maintenance release

v2.6.0

• equivalent to v2.5.3 but created for semver reasons

v2.5.3

• add a mix.exs to be buildable by both mix and rebar
• minor README updates

v2.5.2

• fix regression parsing <<"-0e...">> (thanks @c-bik)

v2.5.1

• assume all datetimes are UTC time and add Z designator to indicate
• fix parsing issue with datetimes in arrays

v2.5.0

• consult/2 function for reading a file directly to a json term
• maps_always build flag for always returning maps on platforms that support them
• dialyzer fixes

v2.4.0

• enough performance improvements to justify a new version. 2-3x speedup depending on mode of operation

v2.3.1

• fixes an issue where astral plane json escape sequences were inadvertently being converted to the unicode replacement character

v2.3

• switched to a faster implementation of string parsing in both the decoder and encoder
• expand uescape option to the decoder
• allow control codes in json passed to decoder (contrary to the spec, yes)

v2.2

• return_maps option
• uescape option for 7-bit clean output
• add Makefile for slightly better erlang.mk compatibility
• add maps_support/0 call to determine whether jsx was compiled with support for maps or not

v2.1.1

• faster generation of json via iolists
• repeat_keys option

v2.1

• force the end of streams with end_json in addition to end_stream
• support for encoding erlang datetime tuples to iso8601 format
• allow a single trailing comma in objects and arrays

v2.0.4

• more typespec adjustments

v2.0.3

• update some typespecs to make them more comprehensive

v2.0.2

• fixes travis-ci spec

v2.0.1

• fix regression in output of empty objects/arrays

v2.0

• jsx is much more pragmatic by default; common json errors are silently ignored (and fixed). stricter parsing must be enabled with options
• add support for encoding otp 17.0's new maps data type
• removed pre_encode and post_decode options in favour of making jsx functions easier to wrap and customize
• streaming behavior is now disabled by default and must be requested explicitly
• removed deprecated function names (to_json, to_term, term_to_json, etc)
• expanded test coverage

v1.4.5

• various fixes to typespecs uncovered by dialyzer
• allow integer keys during encoding
• convert atoms (other than true, false and null) to strings during encoding

v1.4.4

• typespec for json_term/0 fixed
• incorrect boolean shortcircuiting fixed in multibyte escape processing

v1.4.3

• add empty rebar.config for mix build tool
• add attempt_atom option for decoding json objects
• fix a bug related to multibyte codepoints and streaming input
• add a missing error state in the encoder

v1.4.2

• build apparatus cleaned up and streamlined
• new {raw, <<"json goes here">>} intermediate form to support direct generation of json
• bugfixes involving inappropriate exceptions from jsx functions

v1.4.1

• fixes a bug with interaction between dirty_strings and even numbers of escape characters
• performance enhancements

v1.4

• radically refactored decoder
• dirty_strings now behaves intuitively in decoding. bad codepoints, bad utf8, illegal characters and escapes (except " and ' if single_quoted_strings is enabled) are ignored completely
• incomplete_handler & error_handler are now available for use, see documentation in README

v1.3.3

• pre_encode now orders input in the order you'd expect

v1.3.2

• pre_encode is now able to handle tuples correctly

v1.3.1

• pre_encode is now able to handle tuples

v1.3

• introduces prettify/1 and minify/1, shortcuts for format/2
• introduce encode/1,2 and decode/1,2 as primary interface to built in tokenizers. to_json/1,2 and to_term/1,2 remain accessible but not advertised
• new parser/3 function exposes syntactic analysis stage for use with user defined tokenizers
• improved documentation

v1.2.1

• fixes incorrect handling of escaped forward slashes, thanks bob ippolito

v1.2

• rewritten handling of string escaping to improve performance
• pre_encode and post_decode hooks, see README
• relax option

v1.1.2

• add dirty_strings option
• more fixes for invalid unicode in strings

v1.1.1

• fixes bug regarding handling of invalid unicode in R14Bxx

v1.1

• improvements to string escaping and json generation performance

v1.0.2

• fixes to function specs
• rewritten README
• comments option

v1.0.1

• rebar fix