fix(core): GPIF imports use string 1 = highest and read staff tunings

[PhysShell]

What

GPIF imports (GP6 .gpx, GP7/8 .gp) now follow griff's string convention — string 1 = highest (ADR-0018) — and GP7 imports get the right pitch.

GPIF lists a tuning lowest string first and numbers a note's string from 0 = lowest. Two defects followed:

	before	after
GP6, tuning in track properties (177 guitar tracks in a 410-file corpus)	pitches right, tuning low-first, every position mirrored (string 1 = lowest)	tuning highest first, positions renumbered
GP7, tuning only in <Staves><Staff> (314 tracks)	guitarpro 0.4.2 never reads the staff tuning, falls back to a high-first Standard E while notes stay numbered from the low string — 0.2% of notes had the pitch their GPIF Midi property states	tuning read from the staff; (string, fret, pitch) = tuning + fret for 100% of 329,501 non-tie notes, = Midi for 98.1%

The remaining 1.9% Midi mismatches are notes where GPIF's Midi is not tuning + fret at all (octave-transposed and percussion parts) — out of scope here.

How

import_gp_score reads the GPIF document itself (guitarpro::io::gpx::read_gpx / read_gp), runs the crate's read_gpif, then renumbers each track once at the boundary:

• tuning from the GPIF track's own properties, else the first staff carrying one;
• track.strings highest first; a note's raw string r of n becomes n − r; out-of-range → 0, refused like any invalid GP string;
• a track without a GPIF tuning (drums, keys) keeps the crate's fallback, read in GPIF order.

Every source now shares the GP3/4/5 convention, so the per-format zero_indexed plumbing is gone. The unit tests that pinned the mirrored numbering (GP6 raw string 0 → griff string 1) are replaced by tests of the renumbering.

Tests

• Red first (1af6df3), then green (8d1dcb2).
• core/tests/gp_gpif_strings.rs: authored GPIF text (no copyrighted tab) packed with the crate's public BCFS/BCFZ writer — track-level tuning, staff-only tuning, staff tuning beside a tuning-less track property block, seven strings. Each asserts tuning order, positions, pitches, and pitch_at(position) == pitch.
• Unit tests for mirror_gpif_string and normalise_track_strings (GPIF tuning, crate fallback, out-of-range strings).
• Corpus check against the GPIF Midi property as above (local, corpus not committed).
• cargo test --workspace: green locally (1,692 tests). missing_file_golden was skipped locally only: its golden holds the English OS error text and the test host reports it in Russian — unrelated to this change. cargo clippy -p griff-core --all-targets -- -D warnings and cargo fmt --check clean.

Notes

• Found by the Constraint Lab fingering audit (lab: optimization phase — fingering optimality gap (CP-SAT oracle vs production DP and human tabs) #197). Its GP7 human-tab lines were affected; the audit's numbers are being re-measured on this fix, and the solver-gap result does not depend on it.
• Upstream: the staff-tuning fallback is a guitarpro crate defect; this adapter no longer depends on it. No upstream issue filed yet.

🤖 Generated with Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 576bb175-2db2-4d0c-9d0f-e385a5af213b

---

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.}