whoops (#1272)

scripts/backfill_match_summary.py

@@ -96,58 +96,54 @@ async def main(dry_run: bool) -> None:
     gz_files = sorted(REPLAYS_DIR.glob("*/replay_data.json.gz"))
     log.info("Found %d replay files in %s", len(gz_files), REPLAYS_DIR)
 
-    pending: list[dict] = []
-    skipped = 0
-    unreadable = 0
-
-    for gz_path in gz_files:
-        session_id = gz_path.parent.name
-
-        if session_id in already:
-            skipped += 1
-            continue
-
-        replay = _load_replay(gz_path)
-        if replay is None:
-            unreadable += 1
-            continue
-
-        game = _game_dict_from_replay(replay, gz_path)
-        if game is None:
-            log.warning("Could not extract game dict from %s", gz_path)
-            unreadable += 1
-            continue
-
-        pending.append(game)
+    # Session ID is the directory name — no file reads needed to find missing ones
+    missing_paths = [p for p in gz_files if p.parent.name not in already]
+    skipped = len(gz_files) - len(missing_paths)
 
     log.info(
-        "Summary: %d to backfill | %d already in DB | %d unreadable",
-        len(pending), skipped, unreadable,
+        "Summary: %d to backfill | %d already in DB",
+        len(missing_paths), skipped,
     )
 
-    if not pending:
+    if not missing_paths:
         log.info("Nothing to do.")
         return
 
     if dry_run:
-        log.info("[DRY RUN] Would backfill %d matches — not writing.", len(pending))
-        for g in pending[:10]:
-            log.info("  %s  endTime=%s  map=%r", g["sessionIdHex"], g["endTime"], g["missionName"])
-        if len(pending) > 10:
-            log.info("  ... and %d more", len(pending) - 10)
+        log.info("[DRY RUN] Would backfill %d matches — not writing.", len(missing_paths))
+        for p in missing_paths[:10]:
+            log.info("  %s", p.parent.name)
+        if len(missing_paths) > 10:
+            log.info("  ... and %d more", len(missing_paths) - 10)
         return
 
-    # Process in batches of 50 to keep memory flat and give progress feedback
+    # Read files and write in batches of 50
     batch_size = 50
-    total = len(pending)
+    total = len(missing_paths)
     done = 0
+    unreadable = 0
+
     for i in range(0, total, batch_size):
-        batch = pending[i : i + batch_size]
+        batch_paths = missing_paths[i : i + batch_size]
+        batch: list[dict] = []
+
+        for gz_path in batch_paths:
+            replay = _load_replay(gz_path)
+            if replay is None:
+                unreadable += 1
+                continue
+            game = _game_dict_from_replay(replay, gz_path)
+            if game is None:
+                unreadable += 1
+                continue
+            batch.append(game)
+
+        if batch:
             await process_match_summaries(batch)
             done += len(batch)
             log.info("Backfilled %d / %d", done, total)
 
-    log.info("Done. %d match_summary rows written.", total)
+    log.info("Done. %d written, %d unreadable.", done, unreadable)
 
 
 if __name__ == "__main__":