base64-2.1: CHANGELOG.md

# Revision history for base64

## 2.1

This release completes the typed parser API for URL-safe Base64. It removes
`isBase64Url` and `isValidBase64Url`; see [MIGRATION-2.1.md](MIGRATION-2.1.md)
for replacement examples.

* Adds `parseBase64Padded` and `parseBase64Unpadded` to all six URL-safe
  modules: strict, lazy, and short ByteString and Text. URL Parsers validate the URL alphabet, length, padding policy, and canonical padding bits before returning the original encoded value. Lazy parsers validate
  the whole input and preserve its chunk boundaries; Text parsers do not
  require the decoded payload to be valid UTF-8.
* Removed both URL `Bool` validators in line with the `2.0` work. No one was using them. 

## 2.0

This release removes two public functions. See [MIGRATION-2.0.md](MIGRATION-2.0.md)
for replacement examples and the stricter validation behavior.

* Replace `isBase64` and `isValidBase64` with `parseBase64` in all six
  standard-alphabet modules. The parser returns
  `Either Text (Base64 'StdPadded a)`, preserving the original encoded input
  after checking its alphabet, length, padding, and canonicity.
* For a Boolean check, migrate to `isRight . parseBase64`. This is stricter
  than the old shape-only `isValidBase64` check. URL-specific validators
  are unchanged.
* Remove the predicates without a deprecation period. A source scan of 45
  published direct dependents found no occurrences of either function;
  private projects and other historical versions were not covered.
* Add an optional, manual `simd` flag (off by default) for bundled ARM64 NEON
  and runtime-detected x86-64 AVX2/SSSE3 encoding and strict decoding.
* The SIMD path retains scalar fallbacks, encoding and decoding behavior,
  padding/canonicity checks, and decoding error offsets.
* Add flag-on/off test and benchmark variants, SIMD boundary regressions,
  and native memory checks in CI.

## 1.0.1

* Support GHC 9.6 through 9.14, including `base-4.22`.
* Require GHC 9.6 or later (`base >= 4.18`) and `bytestring >= 0.11`.
* Refresh CI for current stable GHC releases and GitHub Actions; retire the i386 CI job.
* Update maintainer metadata and copyright notices through 2026.
* No changes to the public API.

## 1.0

New epoch time! lots of changes to the API. Please see [MIGRATION-1.0.md](MIGRATION-1.0.md) for more info:

* Add support for GHC 9.x
* Bumped bounds to more modern dependency set
* Introduced `Base64` newtype, located in `Data.Base64.Types`, as well as ADT for associated alphabets.
* Signature for `encodeBase64` has changed so that it produces a `Base64` wrapped value.
* `decodeBase64` and its URLsafe variants now consume a value of type `Base64` and produces a decoded value of the underlying type.
* The *old* `decodeBase64` and its URLsafe variants is now renamed to `decodeBase64Untyped*`.
* Fix for 32-bit systems in which memory barriers were not respected in the encoe loop (see: #56)
* Optimize encode and decode loops to achieve *significant speedup* across the board.

Thanks to everyone who contributed (especially @sofia-m-a!).

## 0.4.2.4

* Support for GHC 9.2.x
* Remove dependency on `ghc-byteorder`
* Bump base lower bound to GHC 8.10.x to make sure endianness is properly support
  (see: https://gitlab.haskell.org/ghc/ghc/-/issues/20338)
* Bump upper bound for `deepseq`

## 0.4.2.3

* Tighter length calculations in unpadded base64url [#35](https://github.com/emilypi/base64/pull/35)

## 0.4.2.2

* Add `NFData`, `Exception`, and `Generic` instances for `Base64Error` + `@since` annotations for new instances. ([#28](https://github.com/emilypi/base64/pull/28))
* Doc improvements and add `-XTrustworty` and `-XSafe` annotations where needed. ([#27](https://github.com/emilypi/base64/pull/27))
* Improve URL canonicity validation and correctness checking (now supports correct checking for unpadded Base64url) ([#26](https://github.com/emilypi/base64/pull/26))
* Fixed perf regressions in decode
* Test coverage is at 98%

## 0.4.2.1

* Security fix: reject non-canonical base64 encoded values - ([#25](https://github.com/emilypi/base64/pull/25))

* Perf improvements

## 0.4.2

* Added support for `Data.ByteString.Short`, `Data.ByteString.Lazy`, `Data.Text.Short`, and `Data.Text.Lazy`. ([#17](https://github.com/emilypi/base64/pull/17))
* Optimize decode algorithm (now beats `base64-bytestring` in every category!) - ([#13](https://github.com/emilypi/base64/pull/13))
* Use `decodeLatin1` when decoding to text, so that functions are total - ([#13](https://github.com/emilypi/base64/pull/13))
* Added `decodeWith*` variants and a `Base64Error` type to handle decoding errors when decoding base64 values - ([#13](https://github.com/emilypi/base64/pull/13))
* Improved error reporting: all offsets are now precisely accurate. - ([#13](https://github.com/emilypi/base64/pull/13))
* Validations added to head, rejecting invalid corner cases (such as bytestrings of length `l == 1 mod 4`, which are never correct) - ([#16](https://github.com/emilypi/base64/pull/16))
* Added `decodeBase64Padded` for symmetry - ([#13](https://github.com/emilypi/base64/pull/13))


## 0.4.1 -- 2020-02-04

* Optimize loops for 32-bit and 64-bit architectures
* Restructure project to be more amenable to swapping head/tail/loops
* Fix module header alignment

## 0.4.0 -- 2020-01-26

* With this major version release, we remove the redundant `encodeBase64Unpadded` and `decodeBase64Unpadded` functions from `Base64.hs`. This is for two reasons:
  1. There is no reason for them to exist, since all std base64 is expected to be padded (in contrast to base64url)
  2. it was literally redundant with `decodeBase64`.

* Use a specialized `Bool` type to give better visual cues regarding which functions add padding

## 0.3.1.1 -- 2020-01-15

* Make sure benchmark code builds

## 0.3.1.0 -- 2020-01-08

* Bug fix for `isBase64` and `isBase64Url` - wrong alphabet was used
* Added `isValidBase64` and `isValidBase64Url` for alphabet conformity. The `isBase64*` functions now tell if it's *correct* base64 now in the sense that it's decodable and valid.
* Dropped Cabal version to 2.0 for backcompat with Stack
* Better documentation

## 0.3.0.0 -- 2020-01-07

* After a discussion with lexilambda, we're making 'encodeBase64' be `ByteString -> Text` by default, offering `ByteString -> ByteString` as
  a secondary format.
* Add `decodeBase64Lenient` to the API for phadej
* Fix unpadded decoding bug where garbage was appended to the end of garbage inputs. A cleaner way to do this is to simply encode as Base64 with
  padding and then strip padding chars until I come up with a workflow specific to unpadded inputs (I used to have this, so I'll have to dig it up)
* Added `isBase64` and `isBase64Url` to the API
* Performance is stable

## 0.2.0.0 -- 2020-01-05

* After a discussion with phadej, we're doing away with the flags, and splitting the optics out into their own separate library
* Removed unnecessary inline pragmas

## 0.1.0.0 -- 2020-01-05

* Do away with the typeclasses, and just provide prisms + synonyms
* Continued performance improvements to decoding
* Corrected benchmarks

## 0.0.1.0 -- 2020-01-03

* First version. Released on an unsuspecting world.
* Preliminary release