base32-1.0: CHANGELOG.md

# Revision history for base32

## 1.0

This release introduces a typed encoding API with breaking signature changes.
See [MIGRATION-1.0.md](MIGRATION-1.0.md) for the changes from 0.4.

* Introduce alphabet- and padding-tagged `Base32` values in `Data.Base32.Types`.
  Encoders return wrapped values; typed decoders consume them. Checked decoders
  now use the `Untyped` suffix for consuming raw Base32 encoded bytes.

  Typed decoding uses unchecked scalar and SIMD loops, with streaming lazy decoding.
  `StdAlphabet` and `HexAlphabet` admit both padding forms in general typed decoders;
  padding-specific typed decoders require exact tags.

* Replace `isBase32`, `isValidBase32`, `isBase32Hex`, and `isValidBase32Hex` with
  `parseBase32Padded` and `parseBase32Unpadded` in all twelve APIs.
  Parsers preserve input, accept the decoder's existing case-insensitive alphabets,
  and check unused pad bits. Noncanonical pad bits are now rejected by decoders too. These were removed because they had no measurable downstream usage, and were of questionable utility.  
* Added an optional, default-off `simd` flag with NEON, SSSE3, and AVX2 prefixes
  and a portable scalar fallback, for both standard and extended-hex alphabets.
* Remove `Data.ByteString.Base32.Internal`; use `Internal.Head` for codec entry
  points and `Internal.Utils` for `validateLastNPads`. They were relics of older development. 
* Support GHC 9.6–9.14; update copyright through 2026, CI, bounds, and benchmarks.

## 0.4

* Support for GHC 9.8
* Drop support for GHC <9.x
* Bump `bytestring` lower bound to 0.11 (thanks @vlix!)

## 0.3.1.0

* Small bug fix
* Doc updates

## 0.3

* Bump lower bounds for `bytestring` and `text` to latest significant major/major version
* Update CI for 8.10.x thru 9.6 coverage
* Correct the update the latin1 decode default for text modules to `utf8` (which is now pure for `text-2.0`).

## 0.2.2.0

* Add support for GHC 9.2.x
* Drop support for GHC <8.10
* Drop dependency on `ghc-byteorder`
* Bump bounds for `deepseq`

## 0.2.1.0

* Expose `Data.ByteString.Base32.Internal` API
* Use closed-form encoding/decoding length calculations, reducing branching and improving performance
* Update tests:
  * Migrate to using lower heap-footprint concrete dict passing
  * Support for 8.10.5 testing


## 0.2.0.0

* Bugfix: fix unpadded base32 encoding ([#4](https://github.com/emilypi/Base32/pull/4))
* Bugfix: Use `decodeLatin1` when decoding, defer to `decodeWith*` primitives for finer-grained
          `Text` to `ByteString` conversions. ([#5](https://github.com/emilypi/Base32/pull/5))
* Add short and lazy variants for `Text` and `ByteString` ([#6](https://github.com/emilypi/Base32/pull/6))
* Expanded test coverage to 96% ([#7](https://github.com/emilypi/Base32/pull/7))
* Improved performance and better error reporting in final quanta ([#9](https://github.com/emilypi/Base32/pull/9))
* Improved docs + trustworthy/safe annotations ([#10](https://github.com/emilypi/Base32/pull/10))

## 0.1.1.1 -- 2020-02-16

* Documentation now references correct RFC section

## 0.1.1 -- 2020-02-16

* Textual interface now uses correct unpadded version

## 0.1.0.0 -- 2020-02-16

* First version. Released on an unsuspecting world.