2 Commits (dev)

Author	SHA1	Message	Date
Broque Thomas	3f965f48cd	Personalized playlists: ruff B905 — explicit strict= on zip() ... CI ruff check failed on the seasonal_mix tuple-row coercion path where a `zip(columns, row)` call lacked an explicit `strict=`. Set `strict=False` to preserve the original intent (tolerant if the row shape ever drifts from the column tuple). The SELECT always returns 8 columns so the lengths match in practice; using strict=False just avoids a future raise if a generator drift changes that. Live happy path stays unchanged: rows from sqlite3.Row hit the `hasattr(r, 'keys')` branch above and never reach the zip line. The zip branch only runs for plain-tuple rows in tests.	2 weeks ago
Broque Thomas	53284ee7c8	Personalized playlists (2/N): all 8 generators wired through manager ... Adds the per-kind generator modules and registers them with the PlaylistKindRegistry so the manager's `refresh_playlist` can dispatch to any of them. Generators (each in its own module under `core/personalized/generators/`): Singletons (variant=''): - hidden_gems -> wraps service.get_hidden_gems - discovery_shuffle -> wraps service.get_discovery_shuffle - popular_picks -> wraps service.get_popular_picks Variant-bearing kinds: - time_machine -> variant = decade label ('1980s', '1990s', ...). Variant resolver returns 7 standard decades. Generator parses '1980s' -> 1980 + delegates to service.get_decade_playlist. - genre_playlist -> variant = URL-safe genre key ('electronic_dance', 'hip_hop_rap', ...). Resolver normalizes parent-genre keys from service.GENRE_MAPPING; free-form keywords pass through to service.get_genre_playlist. - daily_mix -> variant = top-genre rank ('1' / '2' / '3' / '4'). Generator looks up user's Nth-ranked library genre and returns discovery picks within it. Library half (was a stub returning []) is intentionally dropped: tracks table has no source IDs, so library rows can't sync. Fixed the stub to return [] cleanly without the misleading log warning. - fresh_tape -> Spotify Release Radar. Reads curated track IDs from discovery_curated_playlists (tries 'release_radar_<source>' first, falls back to 'release_radar') and hydrates against the discovery pool. - archives -> Spotify Discover Weekly. Same hydration path as fresh_tape but uses 'discovery_weekly'. - seasonal_mix -> variant = season key ('halloween' / 'christmas' / 'valentines' / 'summer' / 'spring' / 'autumn'). Reads curated IDs via SeasonalDiscoveryService then hydrates from seasonal_tracks (which carries full track_data_json). Each module: - Defines `generate(deps, variant, config) -> List[Track]`. - Defines `SPEC = PlaylistKindSpec(...)` and registers it on import (idempotent — re-import safe via `if registry.get(...) is None`). - For variant-bearing kinds, also defines `variant_resolver(deps)`. Shared helpers in `_common.py`: - `get_service(deps)` pulls the legacy `PersonalizedPlaylistsService` instance (deps.service or deps['service']). - `coerce_tracks(rows)` runs each dict through `Track.from_dict`, tolerates None / non-list inputs. Tests (50 new, total 85 across personalized subsystem): - Singletons: registration + display name + dispatch + limit forwarding + empty/None tolerance + missing-deps error + dict-form deps acceptance (16 tests). - Variants: variant_resolver listing + label parsing + invalid variant errors + parent-key normalization + free-form passthrough (13 tests). - Curated/hybrid: daily_mix rank-to-genre resolution + rank-out-of- range empty + invalid-variant error; fresh_tape & archives hydration order + missing-id skip + source-specific-then-fallback key dispatch + limit + missing-database-dep error; seasonal_mix curated-id hydration order + missing-id skip + JSON round-trip + empty-curated empty + limit + missing-service error (21 tests). 3304+ tests pass. No regression on existing 62 personalized tests.	2 weeks ago