# Migration Notes

## doc_id formula change — 2026-03-13

**Changed:** `doc_id` is now computed as `UUID5(NamespaceURL, url)` — a deterministic UUID derived from the canonical URL, stable across crawls and reruns.

**Before (shards below):** `doc_id` was the UUID extracted from the WARC-Record-ID header:
```
doc_id = TrimPrefix(warc_record_id, "<urn:uuid:")   # e.g. <urn:uuid:abc-123> → abc-123
```

### Affected shards — CC-MAIN-2026-08 (232 shards, need re-export)

| Range | Count |
|---|---|
| 00000–00047 | 48 |
| 00058 | 1 |
| 00101–00148 | 48 |
| 00251–00291 | 41 |
| 00401–00424 | 24 |
| 00551–00562 | 12 |
| 00701–00720 | 20 |
| 00851–00888 | 38 |
| **Total** | **232** |

These shards will be re-exported automatically as the pipeline processes them again with `--republish`.

### Client-side workaround (no re-export needed)

The `url` column is correct and stable in **all** shards. Recompute `doc_id` client-side:

```python
import uuid
doc_id = str(uuid.uuid5(uuid.NAMESPACE_URL, row["url"]))
```

```go
import "github.com/google/uuid"
docID := uuid.NewSHA1(uuid.NameSpaceURL, []byte(row.URL)).String()
```