Skip to content

layer2f.md

Latest commit

 

History

History
52 lines (44 loc) · 2.73 KB

layer2f.md

File metadata and controls

52 lines (44 loc) · 2.73 KB
Raw

Handoff: layer2f

Status: completed Timestamp: 2026-04-01T00:30:00Z

Completed

multidj/dedupe.py

  • Added table_exists, ensure_not_empty to imports
  • _keeper_sort_key: timesplayed → play_count
  • _find_groups (artist-title): removed JOIN, FROM tracks, deleted = 0, path AS filepath, play_count
  • _find_groups (filesize): removed JOIN, FROM tracks, deleted = 0, path AS filepath, play_count
  • Track dicts in both branches: timesplayed key → play_count
  • groups_output keeper/duplicate dicts: timesplayed → play_count
  • dedupe(): added wrong-DB guard + ensure_not_empty at entry
  • UPDATE library SET mixxx_deleted=1 → UPDATE tracks SET deleted=1

multidj/analyze.py

  • Added table_exists, ensure_not_empty to imports
  • analyze_key(): added wrong-DB guard + ensure_not_empty at entry
  • candidate_sql: FROM tracks, path AS filepath, deleted = 0, removed l. prefix, removed JOIN
  • count_sql: FROM tracks, deleted = 0, removed JOIN
  • where_clause: removed l. prefix from column references
  • UPDATE library SET key=? → UPDATE tracks SET key=?
  • detect_key and _write_tag untouched (pure audio, no schema refs)

tests/test_dedupe.py (6 tests, all pass)

  • test_dedupe_finds_duplicate_group — tracks 1 and 4 in same group
  • test_dedupe_keeper_selection — track 1 wins (play_count=12 vs 1)
  • test_dedupe_dry_run_no_write — no deleted=1 rows after dry-run
  • test_dedupe_apply_marks_duplicate — track 4 deleted=1, track 1 deleted=0
  • test_dedupe_no_false_positives — tracks 2 and 8 not in any group
  • test_dedupe_mode_field — dry_run / apply

tests/test_analyze.py (6 tests, all pass)

  • test_analyze_dry_run_returns_candidates — 5 tracks without key
  • test_analyze_dry_run_no_write — no keys written
  • test_analyze_force_includes_keyed — all 9 active tracks when force=True
  • test_analyze_apply_writes_key — mocked detect_key("11B") written to DB
  • test_analyze_limit — asserts result["processed"] == 2 (not total_candidates, which is unaffected by LIMIT)
  • test_analyze_mode_field — dry_run

Decisions Made

  • test_analyze_limit asserts result["processed"] not result["total_candidates"]: LIMIT only caps the fetched rows, total_candidates is always the full unfiltered count
  • librosa is lazily imported inside detect_key(), so from multidj.analyze import analyze_key is safe without librosa; the try/except import guard in test_analyze.py is a belt-and-suspenders precaution
  • Wrong-DB guard in dedupe() opens a separate guard connection before calling _find_groups() (which opens its own connection)

Test Results

  • New tests: 12/12 pass
  • Full suite: 73/73 pass (no regressions)

Remaining

Nothing — task complete.