1For the latest version of this document, please see [https://github.com/bos/aeson/blob/master/changelog.md](https://github.com/bos/aeson/blob/master/changelog.md).
5* Optimizations of generics, thanks to Rémy Oudompheng, here are some numbers for GHC 8.4:
6 * Compilation time: G/BigProduct.hs is 25% faster, G/BigRecord.hs is 2x faster.
7 * Runtime performance: BigRecord/toJSON/generic and BigProduct/encode/generic are more than 2x faster.
8* Added To/FromJSON instances for `Void` and Generics's `V1`, thanks to Will Yager
9* Added To/FromJSON instances for `primitive`'s `Array`, `SmallArray`, `PrimArray` and `UnliftedArray`, thanks to Andrew Thad.
10* Fixes handling of `UTCTime` wrt. leap seconds , thanks to Adam Schønemann
11* Warning and documentation fixes thanks to tom-bop, Gabor Greif, Ian Jeffries, and Mateusz Curyło.
15This release introduces bounds on the size of `Scientific` numbers when they are converted to other arbitrary precision types that do not represent them efficiently in memory.
17This means that trying to decode a number such as `1e1000000000` into an `Integer` will now fail instead of using a lot of memory. If you need to represent large numbers you can add a newtype (preferably over `Scientific`) and providing a parser using `withScientific`.
19The following instances are affected by this:
21* `FromJSONKey Natural`
23* `FromJSONKey Integer`
24* `FromJSON NominalDiffTime`
26For the same reasons the following instances & functions have been removed:
27* Remove `FromJSON Data.Attoparsec.Number` instance. Note that `Data.Attoparsec.Number` is deprecated.
28* Remove deprecated `withNumber`, use `withScientific` instead.
30Finally, encoding integral values with large exponents now uses scientific notation, this saves space for large numbers.
34* Catch 0 denominators when parsing Ratio
38* Fix bug in generically derived `FromJSON` instances that are using `unwrapUnaryRecords`, thanks to Xia Li-yao
39* Allow base-compat 0.10.*, thanks to Oleg Grenrus
44* `GKeyValue` has been renamed to `KeyValuePair`, thanks to Xia Li-yao
45* Removed unused `FromJSON` constraint in `withEmbeddedJson`, thanks to Tristan Seligmann
48* Optimizations of TH toEncoding, thanks to Xia Li-yao
49* Optimizations of hex decoding when using the default/pure unescape implementation, thanks to Xia Li-yao
50* Improved error message on `Day` parse failures, thanks to Gershom Bazerman
51* Add `encodeFile` as well as `decodeFile*` variants, thanks to Markus Hauck
52* Documentation fixes, thanks to Lennart Spitzner
53* CPP cleanup, thanks to Ryan Scott
57* Add `Ord` instance for `JSONPathElement`, thanks to Simon Hengel.
62* Added `withEmbeddedJSON` to help parse JSON embedded inside a JSON string, thanks to Jesse Kempf.
63* Memory usage improvements to the default (pure) parser, thanks to Jonathan Paugh. Also thanks to Neil Mitchell & Oleg Grenrus for contributing a benchmark.
64* `omitNothingFields` now works for the `Option` newtype, thanks to Xia Li-yao.
65* Some documentation fixes, thanks to Jonathan Paug & Philippe Crama.
69* Add `FromJSON` and `ToJSON` instances for
70 * `DiffTime`, thanks to Víctor López Juan.
71 * `CTime`, thanks to Daniel Díaz.
72* Fix handling of fractions when parsing Natural, thanks to Yuriy Syrovetskiy.
73* Change text in error messages for Integral types to make them follow the common pattern, thanks to Yuriy Syrovetskiy.
74* Add missing `INCOHERENT` pragma for `RecordToPair`, thanks to Xia Li-yao.
75* Everything related to `Options` is now exported from `Data.Aeson`, thanks to Xia Li-yao.
76* Optimizations to not escape text in clear cases, thanks to Oleg Grenrus.
77* Some documentation fixes, thanks to Phil de Joux & Xia Li-yao.
81* Add `parserThrowError` and `parserCatchError` combinators, thanks to Oleg Grenrus.
83* Add `Generic` instance for `Value`, thanks to Xia Li-yao.
85* Fix a mistake in the 1.2.0.0 changelog, the `cffi` flag is disabled by default! Thanks to dbaynard.
89* `tagSingleConstructors`, an option to encode single-constructor types as tagged sums was added to `Options`. It is disabled by default for backward compatibility.
91* The `cffi` flag is now turned off (`False`) by default, this means C FFI code is no longer used by default. You can flip the flag to get C implementation.
93* The `Options` constructor is no longer exposed to prevent new options from being breaking changes, use `defaultOptions` instead.
95* The contents of `GToJSON` and `GToEncoding` are no longer exposed.
97* Some INLINE pragmas were removed to avoid GHC running out of simplifier ticks.
101* Fix an accidental change in the format of `deriveJSON`. Thanks to Xia Li-yao!
103* Documentation improvements regarding `ToJSON`, `FromJSON`, and `SumEncoding`. Thanks to Xia Li-yao and Lennart Spitzner!
107* Added a pure implementation of the C FFI code, the C FFI code. If you wish to use the pure haskell version set the `cffi` flag to `False`. This should make aeson compile when C isn't available, such as for GHCJS. Thanks to James Parker & Marcin Tolysz!
109* Using the `fast` flag can no longer cause a test case to fail. As far as we know this didn't affect any users of the library itself. Thanks to Xia Li-yao!
113* Added instances for `UUID`.
115* The operators for parsing fields now have named aliases:
116 - `.:` => `parseField`
117 - `.:?` => `parseFieldMaybe`
118 - `.:!` => `parseFieldMaybe'`
119 - These functions now also have variants with explicit parser functions: `explicitParseField`, `explicitParseFieldMaybe`, "explicitParseFieldMaybe'`
120Thanks to Oleg Grenrus.
122* `ToJSONKey (Identity a)` and `FromJSONKey (Identity a)` no longer require the unnecessary `FromJSON a` constraint. Thanks to Oleg Grenrus.
124* Added `Data.Aeson.Encoding.pair'` which is a more general version of `Data.Aeson.Encoding.pair`. Thanks to Andrew Martin.
126* `Day`s BCE are properly encoded and `+` is now a valid prefix for `Day`s CE. Thanks to Matt Parsons.
128* Some commonly used ToJSON instances are now specialized in order to improve compile time. Thanks to Bartosz Nitka.
131[JSONTestSuite](https://github.com/nst/JSONTestSuite) cleanups, all
132motivated by tighter RFC 7159 compliance:
134* The parser now rejects numbers for which
135 [the integer portion contains a leading zero](https://github.com/bos/aeson/commit/3fb7c155f2255482b1b9566ec5c1eaf9895d630e).
136* The parser now rejects numbers for which
137 [a decimal point is not followed by at least one digit](https://github.com/bos/aeson/commit/ecfca35a45286dbe2bbaf5f62354be393bc59b66),
138* The parser now rejects documents that contain [whitespace outside the
139 set {space, newline, carriage return, tab}](https://github.com/bos/aeson/commit/8ef622c2ad8d4a109884e17c2792238a2a320e44).
141Over 90% of JSONTestSuite tests currently pass. The remainder can be
142categorised as follows:
144* The string parser is strict with Unicode compliance where the RFC
145 leaves room for implementation-defined behaviour (tests prefixed
146 with "`i_string_`". (This is necessary because the `text` library
147 cannot accommodate invalid Unicode.)
149* The parser does not (and will not) support UTF-16, UTF-32, or byte
151* The parser accepts unescaped control characters, even though the RFC
152 states that control characters must be escaped. (This may change at
153 some point, but doesn't seem important.)
157* Fixes a regression where a bunch of valid characters caused an
158 "Invalid UTF8-Stream" error when decoding. Thanks to Vladimir
159 Shabanov who investigated and fixed this.
163* Fixes a regression where it was no longer possible to derive
164 instances for types such as `data T a = T { f1 :: a, f2 :: Maybe a
167Thanks to Sean Leather for fixing this, and to Ryan Scott for helping out.
171* Decoding performance has been significantly improved (see
172 https://github.com/bos/aeson/pull/452). Thanks to @winterland1989.
174* Add `ToJSON`/`FromJSON` instances for newtypes from
175 `Data.Semigroup`: `Min`, `Max`, `First`, `Last`, `WrappedMonoid`,
176 `Option`. Thanks to Lennart Spitzner.
178* Make the documentation for `.:!` more accurate. Thanks to Ian Jeffries.
184* Introduced new `FromJSONKey` and `ToJSONKey` type classes that are
185 used to encode maps without going through HashMap. This also allows arbitrary
186 serialization of keys where a string-like key will encode into an object and
187 other keys will encode into an array of key-value tuples.
189* Added higher rank classes: `ToJSON1`, `ToJSON2`, `FromJSON1`, and
192* Added `Data.Aeson.Encoding` with functions to safely write `ToJSON`
193 instances using `toEncoding`.
197* A Cabal `fast` flag was added to disable building with optimizations. This drastically speeds up compiling both aeson ***and*** libraries using aeson so it is recommended to enable it during development. With cabal-install you can `cabal install aeson -ffast` and with stack you can add a flag section to your stack.yaml:
204* Added list specific members to `ToJSON` and `FromJSON` classes. In
205 the same way `Read` and `Show` handle lists specifically. This
206 removes need for overlapping instances to handle `String`.
208* Added a new `sumEncoding` option `UntaggedValue` which prevents
209 objects from being tagged with the constructor name.
211* JSONPaths are now tracked in instances derived with template-haskell
214* Get rid of redundancy of JSONPath error messages in nested records.
216 `eitherDecode "{\"x\":{\"a\": [1,2,true]}}" :: Either String Y`
218 `Error in $.x.a[2]: failed to parse field" x: failed to parse field a: expected Int, encountered Boolean`
219 and now yields `Error in $.x.a[2]: expected Int, encountered Boolean"`.
221 Some users might prefer to insert `modifyFailure` themselves to
222 customize error messages, which previously prevented the use of
225* Backwards compatibility with `bytestring-0.9` using the
226 `bytestring-builder` compatibility package.
228* Export `decodeWith`, `decodeStrictWith`, `eitherDecodeWith`, and
229 `eitherDecodeStrictWith` from `Data.Aeson.Parser`. This allows
230 decoding using explicit parsers instead of using `FromJSON`
233* Un-orphan internal instances to render them in haddocks.
237* Integral `FromJSON` instances now only accept integral
238 values. E.g. parsing `3.14` to `Int` fails instead of succeeding
241* Over/underflows are now caught for bounded numeric types.
243* Remove the `contents` field encoding with `allNullaryToStringTag = False`,
244 giving us `{ "tag" : "c1" }` instead of `{ "tag" : "c1", contents : [] }`.
245 The contents field is optional when parsing so this is only a breaking
246 change for ToJSON instances.
248* Fix a bug where `genericToEncoding` with `unwrapUnaryRecords = True`
249 would produce an invalid encoding: `"unwrap\":""`.
251* `ToJSON` instances using `genericToEncoding` and `omitNothingFields`
252 no longer produce invalid JSON.
254* Added instances for `DList`, `Compose`, `Product`, `Sum`.
258* Enable `PolyKinds` to generalize `Proxy`, `Tagged`, and `Const` instances.
259* Add `unsafeToEncoding` in `Data.Aeson.Types`, use with care!
263* Fix build with `base >= 4.8` and `unordered-containers < 0.2.6`.
267* Fix build on TH-less GHCs
271* Fix build with `base < 4.8` and `unordered-containers < 0.2.6`.
272* Add missing field in docs for `defaultOptions`.
276* Fixes a bug where the hashes of equal values could differ.
280The only changes are added instances.
283* `ToJSON a => ToJSON (NonEmpty a)`
284* `FromJSON a => FromJSON (NonEmpty a)`
286* `FromJSON (Proxy a)`
287* `ToJSON b => ToJSON (Tagged a b)`
288* `FromJSON b => FromJSON (Tagged a b)`
289* `ToJSON a => ToJSON (Const a b)`
290* `FromJSON a => FromJSON (Const a b)`
292These are now available for older GHCs:
298This release should be close to backwards compatible with aeson 0.9.
300If you are upgrading from aeson 0.10 it might be easier to go back in
301history to the point you were still using 0.9.
305* Revert `.:?` to behave like it did in 0.9. If you want the 0.10
306 behavior use `.:!` instead.
308* Revert JSON format of `Either` to 0.9, `Left` and `Right` are now
309 serialized with an initial uppercase letter. If you want the names
310 in lowercase you can add a newtype with an instance.
312* All `ToJSON` and `FromJSON` instances except for `[a]` are no longer
313 `OVERLAPPABLE`. Mark your instance as `OVERLAPPING` if it overlaps
314 any of the other aeson instances.
316* All `ToJSON` and `FromJSON` instances except for `[Char]` are no
317 longer incoherent, this means you may need to replace your
318 incoherent instances with a newtyped instance.
322* Introduce `.:!` that behaves like `.:?` did in 0.10.
324* Allow `HH:MM` format for `ZonedTime` and `UTCTime`.
325 This is one of the formats allowed by
326 [ISO 8601](https://en.wikipedia.org/wiki/ISO_8601#Times).
328* Added `ToJSON` and `FromJSON` instances for the
329 `Version`, `Ordering`, and `Natural` types.
333* JSONPath identifiers are now escaped if they contain invalid characters.
335* Fixed JSONPath messages for Seq to include indices.
337* Fixed JSONPath messages for Either to include `left`/`right`.
339* Fix missing quotes surrounding time encodings.
341* Fix #293: Type error in TH when using `omitNothingFields = True`.
345* Various updates to support GHC 8.
350## Performance improvements
352* Direct encoding via the new `toEncoding` method is over 2x faster
353 than `toJSON`. (You must write or code-gen a `toEncoding`
354 implementation to unlock this speedup. See below for details.)
356* Improved string decoding gives a 12% speed win in parsing
357 string-heavy JSON payloads (very common).
359* Encoding and decoding of time-related types are 10x faster (!!) as a
360 result of bypassing `Data.Time.Format` and the arbitrary-precision
363* When using `toEncoding`, `[Char]` can be encoded without a conversion to
364 `Text`. This is fast and efficient.
366* Parsing into an `Object` is now 5% faster and more
367 allocation-efficient.
369## SUBTLE API CHANGES, READ CAREFULLY
371With the exception of long-deprecated code, the API changes below
372**should be upwards compatible** from older versions of `aeson`. If you run
373into upgrade problems, please file an issue with details.
375* The `ToJSON` class has a new method, `toEncoding`, that allows
376 direct encoding from a Haskell value to a lazy bytestring without
377 construction of an intermediate `Value`.
379 The performance benefits of direct encoding are significant: more
380 than 2x faster than before, with less than 1/3 the memory usage.
382 To preserve API compatibility across upgrades from older versions of
383 this library, the default implementation of `toEncoding` uses
384 `toJSON`. You will *not* see any performance improvement unless you
385 write an implementation of `toEncoding`, which can be very simple:
388 instance ToJSON Coord where
389 toEncoding = genericToEncoding defaultOptions
392 (Behind the scenes, the `encode` function uses `toEncoding` now, so
393 if you implement `toEncoding` for your types, you should see a
394 speedup immediately.)
396 If you use Template Haskell or GHC Generics to auto-generate your
397 `ToJSON` instances, you'll benefit from fast toEncoding
398 implementations for free!
400* When converting from a `Value` to a target Haskell type, `FromJSON`
401 instances now provide much better error messages, including a
402 complete JSON path from the root of the object to the offending
403 element. This greatly eases debugging.
405* It is now possible to use Template Haskell to generate `FromJSON`
406 and `ToJSON` instances for types in data families.
408* If you use Template Haskell or generics, and used to use the
409 `camelTo` function to rename fields, the new `camelTo2` function is
410 smarter. For example, `camelTo` will rename `CamelAPICase` to
411 `camelapi_case` (ugh!), while `camelTo2` will map it to
412 `camel_api_case` (yay!).
414* New `ToJSON` and `FromJSON` instances for the following time-related
415 types: `Day`, `LocalTime`.
417* `FromJSON` `UTCTime` parser accepts the same values as for `ZonedTime`,
418 but converts any time zone offset into a UTC time.
420* The `Result` type is now an instance of `Foldable` and `Traversable`.
422* The `Data.Aeson.Generic` module has been removed. It was deprecated in
425* GHC 7.2 and older are no longer supported.
427* The instance of `Monad` for the `Result` type lacked an implementation
428 of `fail` (oops). This has been corrected.
430* Semantics of `(.:?)` operator are changed. It's doesn't anymore accept
431 present `Null` value.
433* Added `(Foldable t, ToJSON a) => ToJSON (t a)` overlappable instance.
434 You might see `No instance for (Foldable YourPolymorphicType) arising from a
435 use of ‘.=’` -errors due this change.
439* A stray export of `encodeToBuilder` got away!
443* The `json` and `json'` parsers are now synonyms for `value` and
444 `value'`, in conformance with the looser semantics of RFC 7159.
446* Renamed `encodeToByteStringBuilder` to the more compact
451* The dependency on the `unordered-containers` package was too lax,
452 and has been corrected.
456* Encoding a `Scientific` value with a huge exponent is now handled
457 efficiently. (This would previously allocate a huge
458 arbitrary-precision integer, potentially leading to a denial of
461* Handling of strings that contain backslash escape sequences is
462 greatly improved. For a pathological string containing almost a
463 megabyte of consecutive backslashes, the new implementation is 27x
464 faster and uses 42x less memory.
466* The `ToJSON` instance for `UTCTime` is rendered with higher
467 (picosecond) resolution.
469* The `value` parser now correctly handles leading whitespace.
471* New instances of `ToJSON` and `FromJSON` for `Data.Sequence` and
472 `Data.Functor.Identity`. The `Value` type now has a `Read` instance.
474* `ZonedTime` parser ordering now favours the standard `JSON` format,
475 increasing efficiency in the common case.
477* Encoding to a `Text.Builder` now escapes `'<'` and `'>'` characters,
482* Fix `ToJSON` instance for 15-tuples (see #223).
490* Add `ToJSON` and `FromJSON` instances for tuples of up to 15
495* Major compiler and library compatibility changes: we have dropped
496 support for GHC older than 7.4, `text` older than 1.1, and
497 `bytestring` older than 0.10.4.0. Supporting the older versions had
498 become increasingly difficult, to the point where it was no longer
503* The performance of encoding to and decoding of bytestrings have both
504 improved by up to 2x, while also using less memory.
506* New dependency: the `scientific` package lets us parse floating point
507 numbers more quickly and accurately.
509* `eitherDecode`, `decodeStrictWith`: fixed bugs.
511* Added `FromJSON` and `ToJSON` instances for `Tree` and `Scientific`.
513* Fixed the `ToJSON` instances for `UTCTime` and `ZonedTime`.
517* Much improved documentation.
519* Angle brackets are now escaped in JSON strings, to help avoid XSS
522* Fixed up handling of nullary constructors when using generic
525* Added `ToJSON`/`FromJSON` instances for:
528 * ISO-8601 dates: `UTCTime`, `ZonedTime`, and `TimeZone`
530* Added accessor functions for inspecting `Value`s.
532* Added `eitherDecode` function that returns an error message if
537* This release introduces a slightly obscure, but
538 backwards-incompatible, change.
540 In the generic APIs of versions 0.4 and 0.5, fields whose names
541 began with a `"_"` character would have this character removed. This
542 no longer occurs, as it was both buggy and surprising
543 (https://github.com/bos/aeson/issues/53).
545* Fixed a bug in generic decoding of nullary constructors
546 (https://github.com/bos/aeson/issues/62).
550* When used with the UTF-8 encoding performance improvements
551 introduced in version 0.11.1.12 of the `text` package, this release
552 improves `aeson`'s JSON encoding performance by 33% relative to
555 As part of achieving this improvement, an API change was necessary.
556 The `fromValue` function in the `Data.Aeson.Encode` module now uses
557 the `text` package's `Builder` type instead of the `blaze-builder`
558 package's `Builder` type.
562* The new `decode` function complements the longstanding `encode`
563 function, and makes the API simpler.
565* New examples make it easier to learn to use the package
566 (https://github.com/bos/aeson/tree/master/examples).
570 `aeson`'s support for data-type generic programming makes it
571 possible to use JSON encodings of most data types without writing
572 any boilerplate instances.
574 Thanks to Bas Van Dijk, `aeson` now supports the two major schemes
575 for doing datatype-generic programming:
577 * the modern mechanism, built into GHC itself
578 (http://www.haskell.org/ghc/docs/latest/html/users_guide/generic-programming.html)
580 * the older mechanism, based on SYB (aka "scrap your
583 The modern GHC-based generic mechanism is fast and terse: in fact,
584 its performance is generally comparable in performance to
585 hand-written and TH-derived `ToJSON` and `FromJSON` instances. To
586 see how to use GHC generics, refer to `examples/Generic.hs`.
588 The SYB-based generics support lives in `Data.Aeson.Generic` and is
589 provided mainly for users of GHC older than 7.2. SYB is far slower
590 (by about 10x) than the more modern generic mechanism. To see how
591 to use SYB generics, refer to `examples/GenericSYB.hs`.
593* We switched the intermediate representation of JSON objects from
594 `Data.Map` to `Data.HashMap` which has improved type conversion
597* Instances of `ToJSON` and `FromJSON` for tuples are between 45% and 70%
602 This version of aeson makes explicit the decoupling between
603 *identifying* an element of a JSON document and *converting* it to
604 Haskell. See the `Data.Aeson.Parser` documentation for details.
606 The normal `aeson` `decode` function performs identification
607 strictly, but defers conversion until needed. This can result in
608 improved performance (e.g. if the results of some conversions are
609 never needed), but at a cost in increased memory consumption.
611 The new `decode'` function performs identification and conversion
612 immediately. This incurs an up-front cost in CPU cycles, but
613 reduces reduce memory consumption.
keysesc to close
- j k
- move down, up
- enter
- open
- esc
- back, or up a level
- 1 2 3 4
- readme, files, commits, refs
- /
- filter
- y
- copy the clone command
- c
- compare the chosen commit with another (or shift-click both)
- ?
- show or hide this list