-- Phase 1: remove existing duplicate rows.
-- "Duplicate" = same (vehicle_id, category_id, oem_code, name, position).
-- Backfill + reactive prefetch can re-drill the same category multiple times;
-- without a unique constraint that re-insert kept all rows, leaving 8–34%
-- duplicate per source (~108k extra rows of ~588k). Keep the oldest row per
-- group (preserves original created_at) and drop the rest.
DELETE FROM "parts" WHERE id IN (
  SELECT id FROM (
    SELECT id, row_number() OVER (
      PARTITION BY vehicle_id, category_id, oem_code, name, position
      ORDER BY created_at ASC, id ASC
    ) AS rn FROM "parts"
  ) t WHERE rn > 1
);
--> statement-breakpoint

-- Phase 2: prevent future dupes. NULLS NOT DISTINCT (Postgres 15+) so NULL
-- position / NULL vehicle_id collapse like equal values rather than each
-- counting as a separate "distinct" row.
CREATE UNIQUE INDEX IF NOT EXISTS "parts_dedup_idx" ON "parts"
  (vehicle_id, category_id, oem_code, name, position) NULLS NOT DISTINCT;