v0.0.6

Other

• TKEY initial-key typed accessor (spec §4.2.1 / §4.2.3)
• ID3v2.2 write support — frame-id demotion + 6-byte header + PIC layout (spec id3v2-00 §3.2/§4.15)
• typed TFLT file-type accessor file_type() + FileType enum
• typed language accessor + Language enum for COMM/USLT/USER/SYLT (spec v2.3/v2.4 structure §3)
• typed TMED media-type accessor + MediaType enum (spec v2.3 §4.6.3 / v2.4 §4.2.3)
• complete ID3v2.2.0 §4 frame-table read support + §3.1 compression-bit fix
• add the compression dimension to the write-options matrix
• frame-level zlib compression both directions + v2.3 format-flag additions (spec v2.3 §3.3 / v2.4 §4.1.2)
• typed TCON content-type accessor + ContentType enum (v2.3 §4.2.1 / v2.4 §4.2.3)
• typed SYTC tempo accessor + SytcTempo enum (spec v2.4 §4.7)
• drop release-plz.toml — use release-plz defaults across the workspace
• typed ETCO event-type accessor (spec v2.3 §4.6 / v2.4 §4.5)
• Add typed Equ2Interpolation accessor for EQU2 interpolation method
• typed RVA2 channel_type accessor (spec v2.4 §4.11)
• typed SYLT content-type + COMR received-as accessors (spec §4.9 / §4.24)
• surface v2.4 §4.2 text frames in to_key_value_pairs mapping
• typed TIPL/TMCL/IPLS pair-list accessors (spec §4.2.2)
• Round 209: Criterion bench harness for parse/write hot paths
• surface v2.4 extended-header is_update + restrictions sub-fields
• structural IPLS involved-people-list frame (spec v2.3 §4.4)
• structural EQUA equalisation frame (spec v2.3 §4.13)
• structural RVAD relative volume adjustment frame (spec v2.3 §4.12)
• Add structured RVRB reverb frame (spec v2.3 §4.13 / v2.4 §4.13)
• structural MLLT MPEG location lookup table frame
• ID3v2.4 footer emission + parse validation (spec §3.4)
• cover extended-header CRC × unsync combinations
• extended-header CRC-32 verification + emission (spec §3.2)
• drop enumerated reader names from fuzz.yml comment
• daily CI workflow + curated ID3 seed corpus for parse target
• writer-side unsynchronisation (WholeTag + v2.4 PerFrame)
• cargo-fuzz target driving panic-freedom over the ID3 parser
• typed TimestampUnit accessor for ETCO/SYTC/SYLT/POSS
• add ASPI audio seek point index frame (v2.4 §4.30)
• add ENCR encryption method registration frame
• add GRID group identification registration frame
• structural MCDI/ETCO/SYLT/POSS/RBUF/SEEK/SIGN/AENC/LINK frames
• structural USER/OWNE/COMR/SYTC/RVA2/EQU2 frames
• structural POPM / PCNT / PRIV / GEOB / UFID frames
• replace never-match regex with semver_check = false
• drop enable_miri input (miri now manual-only via workflow_dispatch)
• grant release-plz shim contents+pull-requests write
• migrate to OxideAV/.github reusable workflows
• pin release-plz to patch-only bumps

Added

• Typed TKEY initial-key accessor Id3Frame::initial_key() + the
  MusicalKey / KeyAccidental enums (spec v2.3 §4.2.1 / v2.4 §4.2.3).
  The frame "contains the musical key in which the sound starts",
  "represented as a string with a maximum length of three characters";
  the ground keys are A..G, the halfkeys b / #, minor m, and
  "off key is represented with an o only" (e.g. the spec example
  Dbm). The accessor decodes that grammar to MusicalKey::Key { tonic, accidental, minor }, MusicalKey::OffKey, or MusicalKey::Custom for
  any value outside the grammar (tonic not in A..G, unknown / out-of-
  order trailing character, or a value past the three-character maximum)
  so a forward-compatible or non-conforming source surfaces structurally
  rather than being dropped, matching the posture of file_type() /
  media_type(). The grammar paragraph is identical across v2.2 (TKE),
  v2.3, and v2.4 so the accessor is version-independent; the raw
  Id3Frame::Text::values is unchanged and round-trips losslessly
  through write_tag. Three new lib tests (spec-example grammar coverage
  for natural / minor / flat / sharp keys and the o off-key; non-
  conforming inputs collapsing to Custom; accessor None on non-TKEY
  frames) plus one integration round-trip test (write → parse under both
  v2.3 and v2.4 envelopes preserving both the typed view and the raw
  value) cover the matrix.
• ID3v2.2 write support (write_tag(&tag, Id3Version::V2_2) / the
  write_tag_with_options path), completing the v2.2 round-trip — the
  format was previously parse-only. Each four-char id is demoted to its
  three-character v2.2 id (the inverse of the existing v2.2→v2.3
  promotion table), the six-byte v2.2 frame header (3-char id + 3-byte
  big-endian size, no flags; spec id3v2-00 §3.2) is emitted, and the
  v2.2 PIC layout is reconstructed (fixed three-character image-format
  code per §4.15 in place of v2.3 APIC's NUL-terminated MIME). The §4
  frame set is closed, so a frame with no v2.2 equivalent (a v2.4-only
  addition or an unrecognised Unknown id) is skipped rather than
  emitted under an id a conformant v2.2 reader could not interpret. v2.2
  predates the extended header, footer, and per-frame flags byte, so the
  writer rejects with_crc / with_footer / with_update /
  with_restrictions / with_compression under a v2.2 target;
  UnsyncMode::WholeTag is supported via the header bit-7 unsync flag
  and PerFrame collapses to whole-tag as it does for v2.3. Six new
  round-trip tests cover text/comment/lyrics/URL/picture frames, the PIC
  three-char image format, whole-tag unsync over a false-sync payload,
  v2.4-only-frame skipping, the post-v2.2 option rejections, and a
  structured POPM frame; the fuzz target now also drives the v2.2
  write+reparse path for panic-freedom.
• Typed TFLT file-type accessor Id3Frame::file_type() + FileType
  enum (spec v2.3 §4.2.1 / v2.4 §4.2.3). The frame "indicates which type
  of audio this tag defines" via a predefined code optionally followed by
  /-separated refinements, "in a similar way to the predefined types in
  the TMED frame, but without parentheses". Because the wire form is
  identical in both versions (MPG/3 → code="MPG", refinements=["3"])
  and carries no parentheses or v2.3 free-text refinement, a single bare
  grammar covers both dialects; the only version difference is the
  v2.4-added MIME top-level code, which the predefined table resolves
  under either envelope. The four predefined codes (MIME/MPG/VQF/PCM)
  resolve to their spec descriptions; an out-of-table code surfaces as
  FileType::Predefined { name: None, .. } so a forward-compatible
  reference is preserved structurally rather than dropped, and a value
  with an empty top-level segment surfaces as FileType::Custom. The raw
  Id3Frame::Text::values is unchanged and round-trips losslessly,
  mirroring the TMED media_type() accessor.
• Typed TMED media-type accessor Id3Frame::media_type() + MediaType
  enum (spec v2.3 §4.6.3 / v2.4 §4.2.3). The frame "describes from which
  media the sound originated — either a text string or a reference to the
  predefined media types found in the list below." The accessor
  normalises both version dialects onto one vocabulary, mirroring the
  TCON content_types() accessor: v2.3 wraps a reference in (...)
  optionally followed by a free-text refinement ((MC) with four channels → media="MC", text=" with four channels"; (VID/PAL/VHS)
  → media="VID", refinements=["PAL","VHS"]) with a (( escape for a
  literal-( free-text name, while v2.4 drops the parentheses (the spec
  example VID/PAL/VHS parses to the same reference). The 15 predefined
  top-level codes (DIG/ANA/CD/LD/TT/MD/DAT/DCC/DVD/TV/VID/RAD/TEL/MC/REE)
  resolve to their spec descriptions; an out-of-table code surfaces as
  MediaType::Predefined { name: None, .. } so a forward-compatible
  reference is preserved rather than dropped. The raw text value is
  unchanged and round-trips losslessly through write_tag.
• Complete ID3v2.2.0 §4 frame-table read support. The v2.2 walker
  (3-char ids, 3-byte sizes, no frame flags) previously dispatched
  only text / URL / COM / ULT / PIC / REV / EQU; it now maps
  every declared v2.2 frame onto the crate's typed surface: UFI §4.1
  → Ufid, IPL §4.4 → Ipls, MCI §4.5 → MusicCdId, ETC §4.6
  → EventTimingCodes, MLL §4.7 → MpegLocationLookup, STC §4.8
  → SyncedTempo, SLT §4.10 → SyncedLyrics, GEO §4.16 → Geob,
  CNT §4.17 → PlayCounter, POP §4.18 → Popularimeter, BUF
  §4.19 → RecommendedBuffer, CRA §4.21 → AudioEncryption (all
  byte-identical to their v2.3 descendants' payload layouts), plus two
  v2.2-specific walkers — RVA §4.12 (right/left volume-change fields
  are unconditional; presence is not keyed on the inc/dec sign bits
  the way v2.3 RVAD gates its front block, so a both-decrement frame
  keeps its data) and LNK §4.22 (the linked frame identifier is
  always exactly 3 bytes, so a URL whose first byte is an uppercase
  id-class character can never be folded into the identifier the way
  the v2.3/v2.4 3-vs-4-byte heuristic would). CRM §4.20 (encrypted
  meta frame) has no v2.3/v2.4 descendant and is preserved verbatim
  via Id3Frame::Unknown. The v22_promote conversion table now
  covers the full §4 id list so pass-through writes promote correctly.
• Frame-level zlib compression, both directions (spec v2.3 §3.3
  format flag i / v2.4 §4.1.2 format flag k). parse_tag now
  inflates compressed frames transparently in both dialects — v2.3's
  4-byte big-endian decompressed-size header addition and v2.4's
  mandatory data-length indicator (a compressed v2.4 frame without
  the DLI bit is rejected per the spec's "requires the 'Data Length
  Indicator' bit to be set as well") — and dispatches the recovered
  payload structurally. The announced decompressed size is
  authoritative: an inflate-length mismatch drops the frame (earlier
  frames survive, matching the corrupted-frame posture), and the
  announce doubles as the allocation cap under a 64 MiB per-frame
  ceiling so a zlib bomb can't force a...

v0.0.5

Other

• drop Cargo.lock — this crate is a library
• bump oxideav-core / oxideav-codec dep examples to "0.1"
• bump to oxideav-core 0.1.1 + codec 0.1.1
• bump oxideav-core + oxideav-codec deps to "0.1"