soulsync/tests/discovery
Broque Thomas 27fa96fe97 PR5b: lift _run_youtube_discovery_worker to core/discovery/youtube.py
Second lift in the PR5 discovery-workers series. Pulls the 332-line
YouTube discovery worker out of `web_server.py` into its own focused
module under `core/discovery/`. Pure 1:1 lift — wrappers keep the
original entry-point name so the two callers
(`youtube_discovery_executor.submit(_run_youtube_discovery_worker, ...)`)
continue to work without changes.

What the YouTube discovery worker does:

1. Pause enrichment workers (release shared resources).
2. For each YouTube playlist track:
   - Cancellation check (phase != 'discovering' aborts).
   - Discovery cache lookup; cache hit short-circuits the search.
   - Strategy 1: matching_engine search queries with confidence scoring
     against Spotify (preferred) or iTunes (fallback).
   - Strategy 2: swapped artist/title query.
   - Strategy 3: raw (untokenized) query.
   - Strategy 4: extended search with limit=50.
   - On match → save to discovery cache.
   - On miss → build Wing It stub from raw source data.
3. After loop: phase='discovered', sort results by index, and for mirrored
   playlists write extra_data back to the DB.
4. Activity feed entry with match summary.
5. On error → state['status']='error', phase='fresh'.
6. Finally: resume enrichment workers.

Dependencies injected via `YoutubeDiscoveryDeps` (16 fields) —
youtube_playlist_states, spotify_client, matching_engine, plus 13
callable helpers (pause/resume enrichment, get_active_discovery_source,
get_metadata_fallback_client, discovery cache key/validate, extract
artist name, spotify_rate_limited, discovery_score_candidates,
get_metadata_cache, build_discovery_wing_it_stub, get_database,
add_activity_item).

Diff vs original after `deps.X` → global X normalization is **zero
differences** — 332 lines orig = 332 lines lifted, byte-identical body
(including all whitespace).

Pre-existing bug preserved (not fixed): if `youtube_playlist_states[url_hash]`
raises KeyError on entry, the outer except handler tries to mutate
`state` which is unbound → secondary UnboundLocalError. Same bug in
the original. Documented here for future cleanup but out of scope
for the lift.

Tests: 14 new under tests/discovery/test_discovery_youtube.py covering
cache hit short-circuit, Strategy 1 confidence match, Wing It fallback,
iTunes fallback path (Spotify unauthenticated and rate-limited),
cancellation (phase changed), skip_discovery flag, completion phase
update, activity feed entry, mirrored playlist DB writeback, non-mirrored
no-writeback, enrichment workers pause/resume, error-during-loop resume,
results sorted by index after retry.

Full suite: 1082 passing (was 1068). Ruff clean.
2026-04-28 11:58:10 -07:00
..
__init__.py PR5a: lift _run_sync_task to core/discovery/sync.py 2026-04-28 11:20:47 -07:00
test_discovery_sync.py PR5a: lift _run_sync_task to core/discovery/sync.py 2026-04-28 11:20:47 -07:00
test_discovery_youtube.py PR5b: lift _run_youtube_discovery_worker to core/discovery/youtube.py 2026-04-28 11:58:10 -07:00