feat: add base16.decode_strict for canonical-form hex verification

[nao1215]

Summary

Adds `base16.decode_strict/1` to close the API symmetry gap with
`base32/rfc4648` and `base64/standard`, both of which already
expose `decode_strict`.

Changes

• New `pub fn decode_strict(input: String) -> Result(BitArray, CodecError)`
  in `src/yabase/base16.gleam`.
• Imports `NonCanonical` from `yabase/core/error`.
• Adds 6 regression tests in `test/base16_test.gleam`.

Design decisions

• Canonical form is RFC 4648 §8 uppercase (`encode/1`'s output).
  `encode_lowercase/1` is the documented opt-in non-canonical form,
  so `decode_strict` rejects its output even though the lenient
  `decode/1` accepts it. This matches the contract the package
  already advertises in `encode/1`'s docstring.
• Same shape as `base64/standard.decode_strict` and (post-base32/rfc4648.decode_strict accepts lowercase input — canonical-form contract violated (RFC 4648 §3.5/§6) #86)
  `base32/rfc4648.decode_strict`: the canonicality check is
  byte-equal (`encode(bytes) == input`). No `string.uppercase` or
  other normalization step — that is exactly the bug base32/rfc4648.decode_strict accepts lowercase input — canonical-form contract violated (RFC 4648 §3.5/§6) #86 closed,
  and reintroducing it here would defeat the purpose.
• `decode/1` itself is unchanged. The lenient read path still
  accepts both cases per the existing "the decoder remains
  case-insensitive" contract noted in the module-level docstring.

Verification

• `just check` passes (799 tests, README snippets verified).

Closes #87

Summary by CodeRabbit

• New Features

  • Added strict Base16 decoder to enforce canonical uppercase hexadecimal format validation.

• Tests

  • Added comprehensive test coverage for canonical form validation and error handling.

• Documentation

  • Updated changelog documenting strict decoding behavior.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 1c74cb93-161d-45d2-86f1-99a5624b97f5

📥 Commits

Reviewing files that changed from the base of the PR and between 14f98ab and bc1f6fe.

📒 Files selected for processing (3)

• CHANGELOG.md
• src/yabase/base16.gleam
• test/base16_test.gleam

---

📝 Walkthrough

Walkthrough

Adds a strict decoder base16.decode_strict/1 that enforces RFC 4648 canonical uppercase form by re-encoding decoded bytes and comparing against the original input, returning Error(NonCanonical) for non-canonical hex. Includes comprehensive tests and changelog documentation. Closes #87.

Changes

Base16 Strict Decoder Implementation

Layer / File(s)	Summary
Error Type Imports src/yabase/base16.gleam, test/base16_test.gleam	Import NonCanonical error constructor alongside existing codec error variants.
Strict Decoder Logic src/yabase/base16.gleam	New decode_strict/1 function decodes leniently via decode/1, re-encodes bytes with uppercase encode/1, and returns Error(NonCanonical) if the re-encoded string differs from the original input. Other errors propagate unchanged.
Test Coverage test/base16_test.gleam	Six unit tests verify canonical uppercase acceptance, rejection of lowercase and mixed-case with NonCanonical, propagation of InvalidCharacter and InvalidLength errors, and acceptance of empty input.
Changelog CHANGELOG.md	Document yabase/base16.decode_strict/1 with canonical-form semantics and closure of #87.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related issues

• base16 lacks decode_strict — canonical-form check unavailable for hex digests #87: Direct implementation of the requested base16.decode_strict/1 feature with exact specification matching the proposed API (compare encode(bytes) == input without pre-uppercasing).
• base32/rfc4648.decode_strict accepts lowercase input — canonical-form contract violated (RFC 4648 §3.5/§6) #86: Related prior-art issue documenting a similar strict-decoder bug in base32.decode_strict where incorrect input normalization allowed non-canonical lowercase; this PR implements the correct pattern for base16.

Poem

🐰 A hex decoder hopped with cheer,
Now strict and canonical, crystal clear—
Uppercase only, no mixed case,
Canonical checks keep crooks from the base,
DEADBEEF stays, deadbeef must go!

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: adding a new base16.decode_strict function for canonical-form verification.
Linked Issues check	✅ Passed	The implementation fully satisfies issue #87: decode_strict returns Error(NonCanonical) for non-canonical input, preserves decode/1's leniency, uses direct byte-equality check (encode(bytes) == input), and includes comprehensive test coverage.
Out of Scope Changes check	✅ Passed	All changes are directly scoped to implementing decode_strict for base16 as specified in #87; no unrelated modifications are present.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/issue-87-base16-decode-strict

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}