v0.2.1

Changed

• Replace all custom list_reverse implementations with list.reverse from stdlib (14 modules)
• bignum.encode_int now uses list accumulator instead of O(n^2) string concatenation
• bignum.count_leading_char renamed to count_leading_zeros_str and uses char_value for zero alias handling (e.g. Crockford O->0)
• Shared find_index extracted into bignum module; Base58 Bitcoin/Flickr deduplicated
• z85.list_to_string uses string.join instead of recursive concatenation
• crockford.string_char_at and nopadding.find_char_pos use let assert for unreachable branches
• verify-readme.sh uses trap cleanup and indentation-safe grep patterns
• Example comments clarified to note they do not produce real Bitcoin addresses

Fixed

• Base91 EOF decode now flushes one final byte (matching C reference implementation)
• urlsafe_nopadding.decode now reports the earliest invalid character, not just =
• Crockford case-insensitive test was a tautology (compared same input to itself)
• README 8-to-5 bit compound modifier hyphenation corrected to 8-to-5-bit

Added

• base58check middle-position InvalidCharacter test

v0.2.0

Added

• Base2 (binary string), Base8 (octal), Base10 (decimal) encodings
• Base58 Flickr variant with multibase Z prefix
• Crockford Base32 check symbol support via CrockfordCheck variant and encode_check / decode_check
• Multibase prefixes 0 (base2), 7 (base8), 9 (base10), Z (base58flickr)
• Ascii85 btoa y abbreviation for 4 consecutive spaces (0x20202020)
• Adobe Ascii85 form-feed whitespace handling (PostScript compliance)
• NoPadding Base64 decoders now reject = with InvalidCharacter
• Shared internal/bignum module for radix-based encodings
• NIST test vectors for SHA-256 (448-bit, 896-bit, padding boundary cases)
• Base91 fixed vectors, error cases, and 48/64-byte roundtrip tests
• CI: README code snippet compilation check, examples compile check, OTP 26/27/28 matrix
• Multibase coverage matrix in README
• Examples for JWT, QR, Bitcoin, and multibase use cases
• Leading-zero preservation documented for big-integer encodings

Changed

• BREAKING: Base58 is now Base58(Bitcoin) in the Encoding ADT
• BREAKING: Ascii85, AdobeAscii85, Rfc1924Base85, Z85 are now Base85(Btoa), Base85(Adobe), Base85(Rfc1924), Base85(Z85)
• BREAKING: yabase.decode_as renamed to yabase.decode
• BREAKING: yabase.encode_with_prefix / yabase.decode (multibase) renamed to yabase.encode_multibase / yabase.decode_multibase
• BREAKING: bech32.encode(hrp, data) / bech32.encode_m(hrp, data) merged into bech32.encode(variant, hrp, data)
• BREAKING: CrockfordCheck added to Base32Variant for unified API access
• Crockford Base32 now uses number encoding per spec (was byte-stream chunking)
• Bech32 module documentation clarified as byte-payload convenience API
• yabase/core/dispatcher marked as internal module
• All encoder functions refactored from O(n^2) string concat to O(n) list accumulator

Fixed

• SHA-256 now uses full 64-bit length in padding (was 32-bit)
• Crockford Base32 encoding now matches the published specification
• Base91 data corruption on inputs longer than ~30 bytes (bigint bit-buffer overflow)
• Adobe Ascii85 now accepts form-feed as whitespace per PostScript spec

v0.1.0

Added

• Unified, type-safe Encoding type covering 19 binary-to-text encodings
• Three API layers: low-level modules, unified dispatch (yabase.encode / yabase.decode_as), and facade shortcuts
• Multibase prefix encoding and auto-detection (yabase.encode_with_prefix / yabase.decode)
• Base16 (hex) encoding and decoding
• Base32 variants: RFC4648, Hex, Crockford, Clockwork, z-base-32
• Base64 variants: Standard, URL-safe, NoPadding, DQ (Dragon Quest hiragana)
• Base36, Base45 (RFC 9285), Base58 (Bitcoin), Base62
• Base91, Ascii85, Adobe Ascii85 (PDF/PostScript), Z85, RFC 1924 Base85
• Base58Check (version byte + SHA-256 double-hash checksum, BIP reference)
• Bech32 (BIP 173) and Bech32m (BIP 350) with separate API (HRP + checksum)
• Pure Gleam SHA-256 implementation for Base58Check checksum computation
• CodecError type with specific error variants: InvalidCharacter, InvalidLength, Overflow, UnsupportedPrefix, UnsupportedMultibaseEncoding, InvalidChecksum, InvalidHrp
• Decoded type for multibase auto-detection results
• Bech32Decoded type with HRP, data, and variant auto-detection
• Base58CheckDecoded type with version byte and payload
• Strict input validation: padding position checks (Base32/Base64), trailing data rejection, pure-padding rejection, length constraints (Z85)
• 302 tests covering fixed vectors, roundtrips, error cases, and edge cases