fix(worker): resolve MB to the original edition + loosen duration gate

Root cause of albums re-downloading forever despite completing (Linkin Park "Hybrid
Theory"): the MB resolver read the tracklist from an arbitrary releases[0], often a
later reissue/deluxe. HT resolved to a 2023 13-track reissue (~52 min); a complete
2000 12-track download (~38 min) then failed _download_problem's duration gate
(2267s < 2877s×0.85) and fell through peer after peer, never importing.

- _pick_release: choose the ORIGINAL standard edition (Official, earliest date)
  instead of releases[0], so completeness/duration are judged against the edition
  sources actually deliver.
- _DURATION_MIN_RATIO 0.85 -> 0.65: a source legitimately delivering a shorter
  edition than MB's release shouldn't be rejected; only genuinely truncated/preview
  downloads (a fraction of expected runtime) should be.

Together with the est-time peer ranking, this stops the download→reject→re-download
loop. 334 tests pass.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Jonathan
2026-07-20 22:27:44 +02:00
parent d2a86932e0
commit 5f59283bf0
3 changed files with 49 additions and 5 deletions
+20 -2
View File
@@ -75,6 +75,23 @@ def _best_release_group(artist: str, album: str, groups: list) -> dict | None:
return best
def _pick_release(releases: list) -> dict | None:
"""Choose which release of a release-group to read the tracklist from. MB returns them in an
arbitrary order and ``releases[0]`` is often a later reissue/deluxe/anniversary edition with
more tracks and a longer runtime than the standard album a source actually delivers — which
then makes a *complete* download fail the pipeline's track-count/duration completeness checks
and re-download forever (Linkin Park "Hybrid Theory": MB's arbitrary pick was a 2023 13-track
reissue at ~52 min vs the 2000 original's 12 tracks / ~38 min). Prefer the ORIGINAL standard
edition: an Official release, then the earliest date, then a dated one. Pure — unit-tested."""
if not releases:
return None
def _key(r: dict) -> tuple:
official = 0 if (r.get("status") == "Official") else 1 # official editions first
date = r.get("date") or ""
return (official, 0 if date else 1, date) # then dated, then earliest date
return min(releases, key=_key)
class MusicBrainzResolver:
"""Real MbResolver using the MusicBrainz webservice via musicbrainzngs.
@@ -119,9 +136,10 @@ class MusicBrainzResolver:
if rgid:
rgfull = _with_retry(lambda: musicbrainzngs.get_release_group_by_id(rgid, includes=["releases"]))
releases = rgfull.get("release-group", {}).get("release-list", [])
if releases:
chosen = _pick_release(releases)
if chosen:
rel = _with_retry(
lambda: musicbrainzngs.get_release_by_id(releases[0]["id"], includes=["recordings"])
lambda: musicbrainzngs.get_release_by_id(chosen["id"], includes=["recordings"])
).get("release", {})
tracks = []
for medium in rel.get("medium-list", []):
+4 -2
View File
@@ -22,8 +22,10 @@ _AUDIO_EXT = {".flac", ".mp3", ".m4a", ".opus", ".ogg", ".aac", ".wav"}
# Reject a download whose total playtime falls below this fraction of MusicBrainz's expected
# total — catches truncated files / preview clips substituted for real tracks that keep the
# track COUNT right. Generous, so edition/encoding differences don't false-positive; bonus
# tracks (over-long) are always fine.
_DURATION_MIN_RATIO = 0.85
# tracks (over-long) are always fine. Kept well below 1.0 because a source legitimately delivers a
# different edition than MB's resolved release (shorter radio edits, no bonus/live tracks), and only
# a genuinely truncated/preview-filled download (playtime a fraction of expected) should be caught.
_DURATION_MIN_RATIO = 0.65
# Cap how many ranked candidates a single job will actually download+verify before giving up.
# Without this, the incomplete-download fall-through can grind through dozens of sources (a popular
# album returns ~200 Soulseek candidates), each a slow attempt that also leaves an abandoned slskd