0.10.0.0

Performance improvements

• Direct encoding via the new toEncoding method is over 2x faster than toJSON. (You must write or code-gen a toEncoding implementation to unlock this speedup. See below for details.)

• Improved string decoding gives a 12% speed win in parsing string-heavy JSON payloads (very common).

• Encoding and decoding of time-related types are 10x faster (!!) as a result of bypassing Data.Time.Format and the arbitrary-precision Integer type.

• When using toEncoding, [Char] can be encoded without a conversion to Text. This is fast and efficient.

• Parsing into an Object is now 5% faster and more allocation-efficient.

SUBTLE API CHANGES, READ CAREFULLY

With the exception of long-deprecated code, the API changes below should be upwards compatible from older versions of aeson. If you run into upgrade problems, please file an issue with details.

• The ToJSON class has a new method, toEncoding, that allows direct encoding from a Haskell value to a lazy bytestring without construction of an intermediate Value.

  The performance benefits of direct encoding are significant: more than 2x faster than before, with less than 1/3 the memory usage.

  To preserve API compatibility across upgrades from older versions of this library, the default implementation of toEncoding uses toJSON. You will not see any performance improvement unless you write an implementation of toEncoding, which can be very simple:

  instance ToJSON Coord where
    toEncoding = genericToEncoding defaultOptions
  

  (Behind the scenes, the encode function uses toEncoding now, so if you implement toEncoding for your types, you should see a speedup immediately.)

  If you use Template Haskell or GHC Generics to auto-generate your ToJSON instances, you'll benefit from fast toEncoding implementations for free!

• When converting from a Value to a target Haskell type, FromJSON instances now provide much better error messages, including a complete JSON path from the root of the object to the offending element. This greatly eases debugging.

• It is now possible to use Template Haskell to generate FromJSON and ToJSON instances for types in data families.

• If you use Template Haskell or generics, and used to use the camelTo function to rename fields, the new camelTo2 function is smarter. For example, camelTo will rename CamelAPICase to camelapi_case (ugh!), while camelTo2 will map it to camel_api_case (yay!).

• New ToJSON and FromJSON instances for the following time-related types: Day, LocalTime.

• The Result type is now an instance of Foldable and Traversable.

• The Data.Aeson.Generic module has been removed. It was deprecated in late 2013.

• GHC 7.2 and older are no longer supported.

• The instance of Monad for the Result type lacked an implementation of fail (oops). This has been corrected.

0.9.0.1

• A stray export of encodeToBuilder got away!

0.9.0.0

• The json and json' parsers are now synonyms for value and value', in conformance with the looser semantics of RFC 7159.

• Renamed encodeToByteStringBuilder to the more compact encodeToBuilder.

0.8.1.1

• The dependency on the unordered-containers package was too lax, and has been corrected.

0.8.1.0

• Encoding a Scientific value with a huge exponent is now handled efficiently. (This would previously allocate a huge arbitrary-precision integer, potentially leading to a denial of service.)

• Handling of strings that contain backslash escape sequences is greatly improved. For a pathological string containing almost a megabyte of consecutive backslashes, the new implementation is 27x faster and uses 42x less memory.

• The ToJSON instance for UTCTime is rendered with higher (picosecond) resolution.

• The value parser now correctly handles leading whitespace.

• New instances of ToJSON and FromJSON for Data.Sequence and Data.Functor.Identity. The Value type now has a Read instance.

• ZonedTime parser ordering now favours the standard JSON format, increasing efficiency in the common case.

• Encoding to a Text.Builder now escapes '<' and '>' characters, to reduce XSS risk.

0.8.0.2

• Fix ToJSON instance for 15-tuples (see #223).

0.8.0.1

• Support time-1.5.

0.8.0.0

• Add ToJSON and FromJSON instances for tuples of up to 15 elements.

0.7.1.0

• Major compiler and library compatibility changes: we have dropped support for GHC older than 7.4, text older than 1.1, and bytestring older than 0.10.4.0. Supporting the older versions had become increasingly difficult, to the point where it was no longer worth it.

0.7.0.0

• The performance of encoding to and decoding of bytestrings have both improved by up to 2x, while also using less memory.

• New dependency: the scientific package lets us parse floating point numbers more quickly and accurately.

• eitherDecode, decodeStrictWith: fixed bugs.

• Added FromJSON and ToJSON instances for Tree and Scientific.

• Fixed the ToJSON instances for UTCTime and ZonedTime.

0.6 series

• Much improved documentation.

• Angle brackets are now escaped in JSON strings, to help avoid XSS attacks.

• Fixed up handling of nullary constructors when using generic encoding.

• Added ToJSON/FromJSON instances for:

  • The Fixed class
  • ISO-8601 dates: UTCTime, ZonedTime, and TimeZone

• Added accessor functions for inspecting Values.

• Added eitherDecode function that returns an error message if decoding fails.

0.5 to 0.6

• This release introduces a slightly obscure, but backwards-incompatible, change.

  In the generic APIs of versions 0.4 and 0.5, fields whose names began with a "_" character would have this character removed. This no longer occurs, as it was both buggy and surprising (https://github.com/bos/aeson/issues/53).

• Fixed a bug in generic decoding of nullary constructors (https://github.com/bos/aeson/issues/62).

0.4 to 0.5

• When used with the UTF-8 encoding performance improvements introduced in version 0.11.1.12 of the text package, this release improves aeson's JSON encoding performance by 33% relative to aeson 0.4.

  As part of achieving this improvement, an API change was necessary. The fromValue function in the Data.Aeson.Encode module now uses the text package's Builder type instead of the blaze-builder package's Builder type.

0.3 to 0.4

• The new decode function complements the longstanding encode function, and makes the API simpler.

• New examples make it easier to learn to use the package (https://github.com/bos/aeson/tree/master/examples).

• Generics support

  aeson's support for data-type generic programming makes it possible to use JSON encodings of most data types without writing any boilerplate instances.

  Thanks to Bas Van Dijk, aeson now supports the two major schemes for doing datatype-generic programming:

  • the modern mechanism, built into GHC itself (http://www.haskell.org/ghc/docs/latest/html/users_guide/generic-programming.html)

  • the older mechanism, based on SYB (aka "scrap your boilerplate")

  The modern GHC-based generic mechanism is fast and terse: in fact, its performance is generally comparable in performance to hand-written and TH-derived ToJSON and FromJSON instances. To see how to use GHC generics, refer to examples/Generic.hs.

  The SYB-based generics support lives in Data.Aeson.Generic and is provided mainly for users of GHC older than 7.2. SYB is far slower (by about 10x) than the more modern generic mechanism. To see how to use SYB generics, refer to examples/GenericSYB.hs.

• We switched the intermediate representation of JSON objects from Data.Map to Data.HashMap which has improved type conversion performance.

• Instances of ToJSON and FromJSON for tuples are between 45% and 70% faster than in 0.3.

• Evaluation control

  This version of aeson makes explicit the decoupling between identifying an element of a JSON document and converting it to Haskell. See the Data.Aeson.Parser documentation for details.

  The normal aeson decode function performs identification strictly, but defers conversion until needed. This can result in improved performance (e.g. if the results of some conversions are never needed), but at a cost in increased memory consumption.

  The new decode' function performs identification and conversion immediately. This incurs an up-front cost in CPU cycles, but reduces reduce memory consumption.