epiphany/crates/epiphany-core
Levi Neuwirth cf81074ca0 Text Projection: the Chapter-5 value layer, from one field list
`project` and `parse` for every value an operation payload can embed.

The codec macros now emit a `TextValue` impl beside the `Codec` impl, from the
same invocation: 116 types whose field order cannot disagree between the binary
form and the text, at zero call-site churn. That is the companion's own rationale
applied to code -- a rule cannot drift from the listing it reads, and two listings
of one struct is the drift P13-I1 already cost us. `struct_codec!` rebuilds through
a struct literal and `cstyle_enum_codec!` matches exhaustively, so a field or
variant added later fails to compile rather than silently vanishing from the text.

The other 44 types have hand-written codecs and so need hand-written projections.
Their field order is verified by a mechanical diff of the identifier sequence in
each `fn enc` against the one in each `project`; all 44 agree. This matters because
a `project`/`parse` pair that agrees with itself on a *wrong* order round-trips
perfectly -- neither the compiler nor any round-trip test can see it. The
neighbouring blind spot, a mistyped constructor symbol, is closed by
`textvalue_names.rs`, which recovers each type's Rust name from its derived `Debug`
and compares it against the symbol actually emitted.

Strictness turned out to need only one of its two layers, and mutation testing is
what established that. Every per-site check is live: the set/map strictly-increasing
walk, `RationalTime`'s lowest-terms compare before construction, the catalog-id NFC
intern-and-compare, and `EventArena`'s ascending-`EventId` walk. Every whole-value
`ensure_canonical` guard was dead -- `Tempo::new`, `ReferencePitch::new`,
`SpellingPrecedence::new` and `EventOrderingDAG::try_new` reject rather than adjust,
so an accepted value re-projects to exactly its input and the guard could never
fire. A probe confirmed `try_new` returns its input map unchanged. Helper and all
four call sites removed: a check that cannot fail invites weakening the real one.

Also moves `catalog_name` out of the grammar *test* and into
`operation_kind_tag_vocabulary!`, where the discriminant and decoder already live.
It was a hand-maintained list parallel to an enum -- the exact shape that has cost
this project four bugs.

Method note recorded in DECISIONS: the work list came from compiler errors, but the
compiler reports only the frontier. `AnchorOffset`, `VoiceSelector`, `PowerOfTwo`,
`OctaveOffset` and `NonZeroU16` were each hidden behind a type that had not compiled
yet, so the list must be iterated to a fixpoint.

Gate green -- clippy 0, 1109 tests, doc 0, conformance 8/8, no golden churn.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-21 12:17:49 -04:00
..
src Text Projection: the Chapter-5 value layer, from one field list 2026-07-21 12:17:49 -04:00
tests Text Projection: the Chapter-5 value layer, from one field list 2026-07-21 12:17:49 -04:00
Cargo.toml A B C D F 2026-06-19 12:42:31 -04:00
DECISIONS.md Text Projection: the Chapter-5 value layer, from one field list 2026-07-21 12:17:49 -04:00
README.md Phase 2: spelling + decomposition pre-passes (Agent H) with F merge gate 2026-06-23 16:33:39 -04:00

README.md

epiphany-core

The Epiphany score graph: the in-memory representation of all musical content in a score, implementing the normative requirements of Chapters 25 of the core specification (spec/core_spec.pdf). This is Agent B's crate per spec/QUICKSTART.md — the largest scope — building only on Agent A's epiphany-determinism.

The graph is the canonical truth about the music; layout, serialization, and editing operations are downstream projections and consumers of it. — Chapter 5, Design Principles

What's here

Area Items Spec
Identifiers ReplicaId (+ SYSTEM_DERIVED), OperationId, the typed 128-bit family (EventId, PitchId, VoiceId, …), TypedObjectId, IdentityContext, derive_system_id Ch. 5 §"Identifiers"; Ch. 6 §"Operation Identity"
Time RationalTime (inline-or-promoted), MusicalPosition/MusicalDuration (typed algebra), WallClockTime/WallClockDuration, TimeAnchor/AnchorOffset, EventPosition/EventDuration/ConcreteDuration, TimeSignature/BeatGroup, NotatedComponent/NoteValue Ch. 3
Tempo TempoMap, Tempo, TempoSegment, TempoShape with closed-form musical_to_wallclock/wallclock_to_musical over constant/linear/exponential segments (curve deferred → TempoError) Ch. 3 §"Tempo Map"
Pitch Pitch, ScalePosition, IdentifiedPitch, PitchSpelling, the spelling-attachment subsystem, ReferencePitch, spell (single-pitch simplest spelling), all three equivalences (scale_position_equivalent, enharmonic_equivalent, sounding_equivalent) Ch. 2; Ch. 4 registry ids
Pre-passes derive_annotations: the real spelling pre-pass (Temperley line-of-fifths) + notational-decomposition pre-pass (metric greedy-aligned splitting) as canonical derived annotations (not stored), the eligibility TaxonomyReport, and resolve_spelling (authored-override precedence) Ch. 2 §"Spelling Pre-Pass"; Ch. 3 §"Notational Decomposition"
Events the Event taxonomy (7 variants) and the slotmap-backed EventArena Ch. 5 §"The Event Arena"
Graph Canvas, Region, Staff vs StaffInstance, Voice/VoiceOrigin, Measure, BarlineAlignmentGroup, aleatoric EventOrderingDAG (acyclic by construction), the full cross-cutting registry, the full top-level Score Ch. 5
Indexes ScoreIndexes: the four mandatory indexes (event-time, cross-cutting-reference, measure, spelling-attachment) Ch. 5 §"Indexes"
Invariants check_invariants over all 19 enumerated graph invariants, with a typed InvariantViolation witness per check Ch. 5 §"Graph Invariants"
Generators generators::valid_score/valid_score_rich (positive), violating_score (negative, per invariant), shrink (witness minimizer) QUICKSTART, Agent B hand-off

The identity discipline this crate enforces

  1. Replica + counter, big-endian canonical bytes. Every typed identifier is (replica << 64) | counter; its canonical 16-byte form is to_be_bytes() (8-byte replica, 8-byte counter) and the numeric Ord is the Appendix-D lexicographic byte order. Identity is exact, never tolerant.
  2. A reserved system namespace. ReplicaId::SYSTEM_DERIVED is rejected by ReplicaId::generate/from_entropy; system-derived ids (derive_system_id, derive_promoted_voice_id) live only in that namespace, with counters trunc64(BLAKE3(domain_tag || canonical_inputs)) via epiphany-determinism.
  3. Cross-kind confusion is a compile error. Each object kind has its own newtype; TypedObjectId tags them apart with a discriminant that is part of canonical content.

Graph invariants as property tests

The Chapter 5 invariants are property tests in CI, not runtime assertions in release builds (QUICKSTART). check_invariants returns every violation with a small witness. For each invariant generators provides:

  • a positive generator (valid_score / arbitrary_graph_corpus) whose output passes every check, and
  • a negative generator (violating_score) plus a shrinker (shrink) that minimizes a violating graph to a small witness while retaining only the structure the violation needs.

Generation is deterministic (a vendored SplitMix64), so a failing case reproduces from its seed — no platform entropy enters generation (Appendix D §"Randomness").

Implementation decisions

Per QUICKSTART "Decisions you'll need to make" (full rationale in DECISIONS.md):

  • Replica entropy: getrandom (decision 1). ReplicaId::generate re-draws until the value is not the reserved namespace.
  • Event-arena storage: slotmap (decision 2) plus a hash index for the required O(1) EventId lookup and generation-checked stale handles.
  • Sync only (decision 4): no async anywhere.
  • Current stable Rust (decision 5); MSRV pinned at the workspace's 1.77.
  • RationalTime's promoted arm uses num-rational's BigRational, the spec's reference design (Ch. 3 §"Recommended Implementation").
  • unsafe is forbidden crate-wide (#![forbid(unsafe_code)]).

Hand-off criteria (QUICKSTART, Agent B)

  • Every invariant has both a generator and a shrinker (generators::{valid_score, violating_score, shrink}; one per invariant, property-tested), plus targeted tests for the cross-cutting/anchor/tie sub-rules.
  • The arbitrary-graph corpus runs clean (generators::tests::positive_corpus_runs_clean, 500 graphs), and a breadth corpus (valid_score_rich: concurrent metric/proportional/ aleatoric regions, measures, triplet, tie, spanner, marker, chord symbol, decomposition, tombstones) runs clean over 200 seeds.
  • cargo test -p epiphany-core clean (69 unit + 5 integration).
  • cargo clippy --all-targets -- -D warnings clean; cargo doc clean under RUSTDOCFLAGS="-D warnings".

Depth of the invariant checks

The checks are not surface-level. In particular: invariant 3 computes per-clock event intervals and detects both disorder and overlap; invariant 7 resolves region extents to absolute wall-clock coordinates (wall-clock leaves, plus event/region/measure-start anchors — an event anchor is its region origin plus its region-relative position) and only skips pairs that can't be placed without the deferred tempo map; invariant 9 sweeps every reachable anchor (region extents, meter changes, measure starts, clef/key changes, user breaks, spanners, spelling ranges); invariant 10 resolves all graph references — cross-cutting anchor targets, annotation layers, tuplet parents, graphic objects, and event-internal references (indeterminate alternatives, trajectory event-pitches, graphic-event objects, cue sources); invariant 11 covers every id kind, plus tombstone/live collisions, SYSTEM_DERIVED misuse (including the score's own identity context), and arena index/well-formedness integrity (catching post-get_mut corruption); invariant 17 validates explicit and implicit (pitch-id-ascending) tie pairings, per-class adjacency, and the cross-voice position rule; invariant 18 recomputes the deterministic promoted-voice derivation. Enharmonic equivalence is a sounding notion (octave matters: C4 ≠ C5). Empty pitched events are rejected at the arena boundary and re-checked; IdentityContext::try_new rejects the reserved replica and counters use checked_add so a counter is never silently reused.

Known bounded limitations (deferred dependencies)

  • Tempo conversion integrates the piecewise map in closed form for Constant/Linear/Exponential segments (Chapter 3 §"Conversion"); only TempoShape::Curve is deferred to the open numerical algorithm (TempoError::CurveIntegrationUnsupported). Segment boundaries that cannot be placed without the score graph, and malformed segment sequences, return a TempoError, never a wrong answer. The inverse round-trips ordinary rhythms via a documented continued-fraction approximation (DECISIONS P11-7).
  • Region time-overlap (invariant 7) resolves extents to absolute wall-clock, now including musical event positions placed through the region's effective tempo map (its local_tempo_map, else the score map). Extents that still cannot be placed (no tempo defined, or a deferred curve) are skipped rather than rejected. Sound (no false positives), incomplete (DECISIONS P11-4).
  • System-promoted voice derivation (invariant 18) retains the winning and losing operation ids on VoiceOrigin::SystemPromoted; the checker recomputes the exact four-input derivation used by epiphany-ops.
  • The Chapter 4 tuning catalogPitchSpace/TuningSystem/ AccidentalRegistry definitions, the built-in catalog, the hierarchical resolver, and the position→frequency resolution function — is not an Agent B deliverable (the QUICKSTART lists those as referenced-by-id). This crate models the identifiers and the score-level ScoreTuningContext; Pitch::sounding_equivalent takes a caller-supplied frequency resolver. See DECISIONS P11-5.

Scope boundaries

The full Chapter 5 top-level Score shape is modeled (metadata, instruments, staff groups, parts, tuning context, tempo map, analysis layers, views) along with the complete CrossCuttingRegistry (slurs, ties, beams, tuplets, spanners, markers, repeats, analytical annotations, comments, graphic gestures, lyrics, chord symbols). The reference- and identity-bearing fields are modeled in depth; deeper bodies (tuning resolution, tempo-curve integration, part layout, view recipes, glyph/engraving detail) are Chapters 3/4/7 and later companions.

Engraving-display detail (StemConfiguration, ClefChange, KeySignatureChange, articulations, dynamics, line styles, spanner/marker visual kinds) is introduced informally here and fully defined in Chapter 7 — it belongs to Agent E (epiphany-layout-ir), and this crate carries minimal, clearly-marked placeholders for it. Operation envelopes, stamps (HLC), causal contexts, the canonical reduction, tombstone tracking, and conflict records are Chapter 6 / Agent C (epiphany-ops); epiphany-core defines only the OperationId they hang off and the tombstone-aware invariants.

Ambiguities discovered while building are not resolved in code — they are batched as Pass 11 candidates in DECISIONS.md (QUICKSTART, Process notes).