feat(library): fold a missing track into its on-disk replacement (M485 #5286 #5287)
release / web (push) Successful in 2m21s
release / go (push) Successful in 2m34s
release / govulncheck (push) Successful in 40s
release / integration (push) Successful in 6m17s
release / android (push) Successful in 6m36s
release / Build signed APK (releases and dev) (push) Successful in 6m6s
release / Attach APK to the Release (tag releases only) (push) Skipped
release / Build + push container image (push) Successful in 1m19s
release / Verify release artifacts (tag releases only) (push) Skipped
release / web (push) Successful in 2m21s
release / go (push) Successful in 2m34s
release / govulncheck (push) Successful in 40s
release / integration (push) Successful in 6m17s
release / android (push) Successful in 6m36s
release / Build signed APK (releases and dev) (push) Successful in 6m6s
release / Attach APK to the Release (tag releases only) (push) Skipped
release / Build + push container image (push) Successful in 1m19s
release / Verify release artifacts (tag releases only) (push) Skipped
A track marked missing whose replacement is already on disk under another row — on the operator's library 355 of 451 missing tracks, nearly all Lidarr mp3 -> flac upgrades — is folded into the replacement: likes, plays, playlist entries and tags move across and the missing row goes. Move adoption could not catch these: the replacements were re-encodes (no shared audio hash) with no recording MBID at import. Pairs (ListMissingTrackPairs): the same recording MBID within the album group, or the same album row and title ignoring case. Each side must have exactly one candidate; conflicting MBIDs refuse a pair. No duration or track position gate: on the 142 pairs known to be one recording, 18% differed by over 2s and the poorly tagged set is where numbering is broken (spike #5274). Each pair folds in its own transaction after locking both rows and checking the pair still holds. Runs automatically (operator, 2026-10-07) after a full scan, after a watcher batch that added or updated tracks, and after an AcoustID pass that matched any track. The first scan after deploy repairs the existing rows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
@@ -0,0 +1,75 @@
|
||||
-- Missing-pair pass (M485). A track marked missing whose replacement is already
|
||||
-- on disk under another row — most often a Lidarr quality upgrade, mp3 → flac at
|
||||
-- a new path — is folded into that replacement, so its history moves across and
|
||||
-- re-acquisition stops chasing music the library already has.
|
||||
--
|
||||
-- Move adoption (#2528) cannot catch these: it links a new file to a missing row
|
||||
-- at the moment the file is first inserted, and the replacement arrives as a
|
||||
-- re-encode with no recording MBID, so it has nothing to match on. AcoustID may
|
||||
-- give it an MBID later; by then both rows exist and only a fold can join them.
|
||||
|
||||
-- name: ListMissingTrackPairs :many
|
||||
-- Every candidate (missing, present) pair, with how many candidates each side
|
||||
-- has. Two rules, measured on the operator's library on 2026-10-07 (#5274):
|
||||
--
|
||||
-- mbid the same recording MBID, within the album group (the album row, or
|
||||
-- albums sharing a release group). 142 of 359 pairs.
|
||||
-- title the same album row and the same title, ignoring case. 213 more. The
|
||||
-- replacements are often not the original file — another edition,
|
||||
-- re-encoded, untagged — so the metadata is what is left to go on.
|
||||
--
|
||||
-- Deliberately NOT gated on duration or track position. On the 142 pairs known
|
||||
-- to be one recording, 18% differed by over 2s (old mp3 durations are poor
|
||||
-- estimates), and the poorly tagged set is exactly where numbering is broken.
|
||||
--
|
||||
-- A pair whose MBIDs are both set and differ is not a title match: the tags say
|
||||
-- they are different recordings. The caller folds a pair only when each side has
|
||||
-- exactly one candidate; anything else is ambiguous and left alone.
|
||||
WITH grp AS (
|
||||
SELECT id AS album_id, COALESCE(release_group_mbid, id::text) AS g
|
||||
FROM albums
|
||||
),
|
||||
cand AS (
|
||||
SELECT m.id AS missing_id, p.id AS present_id, 'mbid'::text AS rule
|
||||
FROM tracks m
|
||||
JOIN grp gm ON gm.album_id = m.album_id
|
||||
JOIN grp gp ON gp.g = gm.g
|
||||
JOIN tracks p ON p.album_id = gp.album_id
|
||||
AND p.missing_since IS NULL
|
||||
AND p.mbid = m.mbid
|
||||
WHERE m.missing_since IS NOT NULL
|
||||
AND m.mbid IS NOT NULL
|
||||
UNION ALL
|
||||
SELECT m.id, p.id, 'title'::text
|
||||
FROM tracks m
|
||||
JOIN tracks p ON p.album_id = m.album_id
|
||||
AND p.missing_since IS NULL
|
||||
AND lower(p.title) = lower(m.title)
|
||||
WHERE m.missing_since IS NOT NULL
|
||||
AND (m.mbid IS NULL OR p.mbid IS NULL OR m.mbid = p.mbid)
|
||||
),
|
||||
pairs AS (
|
||||
-- A pair both rules found is one candidate, credited to the stronger rule.
|
||||
SELECT missing_id, present_id, min(rule)::text AS rule
|
||||
FROM cand
|
||||
GROUP BY missing_id, present_id
|
||||
)
|
||||
SELECT pr.missing_id, pr.present_id, pr.rule,
|
||||
m.file_path AS missing_path, p.file_path AS present_path,
|
||||
(count(*) OVER (PARTITION BY pr.missing_id))::int AS missing_candidates,
|
||||
(count(*) OVER (PARTITION BY pr.present_id))::int AS present_candidates
|
||||
FROM pairs pr
|
||||
JOIN tracks m ON m.id = pr.missing_id
|
||||
JOIN tracks p ON p.id = pr.present_id
|
||||
ORDER BY m.file_path, p.file_path;
|
||||
|
||||
-- name: LockTracksForPairing :many
|
||||
-- Locks both rows of a pair for the fold's transaction, in id order so two
|
||||
-- passes cannot deadlock, and returns what the fold re-checks: the missing row
|
||||
-- must still be missing and the present one still present. A scan may have
|
||||
-- restored or adopted the first, or marked the second, since the pairs were listed.
|
||||
SELECT id, missing_since
|
||||
FROM tracks
|
||||
WHERE id = ANY(sqlc.arg(ids)::uuid[])
|
||||
ORDER BY id
|
||||
FOR UPDATE;
|
||||
Reference in New Issue
Block a user