Format data cleanup: dedup typos, backfill orphans, fix mis-typed LP/Singles (#370 Phase 2a)

[holden]

Context

Splitting the data half of #370 Phase 2 out from the UI half. Has standalone value (cleaner edit forms, accurate exports, better search) and unblocks future sub-facet UI work without forcing it.

Live audit of record_formats.name per record_type for user 1 (20,380 records):

45s (14,710 records):
  13,221  Singles: 7-inch
   1,147  Promotional Singles
     273  EPs: 7-inch
      44  Gold Standard Singles with '447' prefix
      15  Picture Disc Singles
       7  EPs: 7-inch 45 rpm           ← likely typo dup of EPs: 7-inch
       2  Gold Standard Singles with 'GB' prefix
       1  Plastic Soundsheets/Flexi-Discs ← misclassified (belongs under Picture Sleeves)

LPs (4,536 records):
   4,320  LPs: 10/12-inch
     135  Singles: 10/12-inch          ← these are 12\" Singles; arguably wrong type
      64  Singles: 12-inch             ← same
       6  EPs: 10-inch
       3  Promotional 12-inch Singles
       3  EPs: 12-inch
       2  Singles: 12-inch 33/45
       2  Promotional Singles: 12-inch
       1  Promotional LPs              ← redundant; merge into LPs: 10/12-inch?

Picture Sleeves (1,007 records):
   1,002  Picture Sleeves
       5  Gold Standard Picture Sleeves

78s (123 records):
     121  Singles: 78 rpm
       1  78 rpm Album                 ← typo dup
       1  Albums: 78 rpm               ← typo dup

ORPHANS (record_type_id = NULL):
       1  EPs: 7-inch 33/45            → assign to 45s
       1  LPs 10/12-inch               → typo dup of \"LPs: 10/12-inch\", reassign records and drop
       1  Singles: 14-inch             → assign to LPs (closest current bucket)
       1  Picture Sleeve               → typo dup of \"Picture Sleeves\", reassign records and drop

Scope

Idempotent Rake task: bin/rails db:cleanup:record_formats (or a one-shot data migration). Must:

A. Backfill 4 orphan formats with record_type_id

Format	Action
EPs: 7-inch 33/45	record_type_id = 45s id
LPs 10/12-inch	(handled in B — merge into canonical)
Singles: 14-inch	record_type_id = LPs id (least bad fit)
Picture Sleeve	(handled in B — merge into canonical)

B. Merge typo / casing dups

For each pair, repoint records.record_format_id to the canonical row, then delete the duplicate row from record_formats:

Duplicate	Canonical
LPs 10/12-inch (no colon)	LPs: 10/12-inch
Picture Sleeve (singular)	Picture Sleeves
78 rpm Album	Albums: 78 rpm (or vice versa — pick one)
EPs: 7-inch 45 rpm	EPs: 7-inch (likely intent — confirm)

C. Reclassify mis-typed LP/Singles records

The 199 records currently filed under LPs but tagged Singles: 10/12-inch / Singles: 12-inch / Singles: 12-inch 33/45 are 12" Singles — physically distinct from LPs. Two options:

1. Add a new record_type 'Singles: 12-inch' and reassign these formats to it
2. Leave them under LPs (status quo, but flag in the cleanup task report)

Recommendation: option 1, with the migration creating the new type. Need product confirmation before shipping.

D. Drop / merge 1-record stragglers

• Plastic Soundsheets/Flexi-Discs (1 record under 45s) → reassign record to Picture Sleeves and drop the format row, OR keep as-is if intentional
• Promotional LPs (1 record under LPs) → merge into LPs: 10/12-inch if user intent was the same, or keep
• Singles: 14-inch (1 record orphan, after backfill) → keep but flag

These are subjective — the cleanup task should report them and ask for sign-off, not auto-merge.

Implementation notes

• Wrap repointing + row-delete in a transaction per merge to keep referential integrity safe
• Print before/after counts for each action
• Add a --dry-run flag (default) so the task can be reviewed before mutating
• After the task runs, expected record_formats row count drops from ~57 → ~12 (counting only formats with at least one record assigned)

Out of scope

• Sub-format UI (tracked in companion issue, gated on this)
• Schema changes beyond optionally adding the "12" Single" record_type row
• Long-term taxonomy redesign (e.g. an orthogonal "intent" column for Promo / EP / Single)

References

• Original Phase 2 spec: Records index: replace granular Format filter with Record Type (with later format sub-facet) #370
• Companion: "Cascading sub-format filter for 45s only (Records index: replace granular Format filter with Record Type (with later format sub-facet) #370 Phase 2b)"