perf(tables): seek keyset row pages instead of scanning from the first key

[waleedlatif1]

Summary

• The table row drain and the export page reader advanced with order_key IS NULL OR (order_key, id) > (anchor). The OR keeps the unkeyed tail reachable but stops the planner from seeking (table_id, order_key, id), so every page scanned from the table's first key and filtered out every row before the anchor — page cost grew with page depth.
• Split the seek into its two disjoint halves (keyed rows past the anchor, unkeyed tail), each a limited index range, and UNION ALL them under the same ORDER BY/LIMIT/OFFSET. Inner limits cover ask + offset, so a compound cursor resuming inside the unkeyed tail lands on the same row.
• No schema or index change.

Type of Change

• Performance improvement

Testing

• Old vs new query executed side by side on Postgres 17 (tied keys, unkeyed tail, pages entirely inside the tail, page sizes 1–1001, offsets 0–40): 2,904 page comparisons, identical rows and order
• EXPLAIN ANALYZE on a deep page: ~1.6 s → ~9 ms; plan becomes a Merge Append over two index range scans with nothing filtered out
• New integration case pages keyed + unkeyed rows through both readers at several page sizes (fails if either half or the inner limit is dropped); replaced a mock test that asserted the old SQL text
• service-filter-threading.test.ts passes locally; biome on changed files

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing (new tests pass the test-audit authoring gate)
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Oct 10, 2026 10:58pm UTC

[cubic-dev-ai]

No issues found across 5 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

<sub>Heads up: you’ve reached your flex budget. Increase your flex budget or wait for usage to reset.

Turn on auto-fix | Re-trigger cubic</sub>

[greptile-apps]

Confidence Score: 5/5

[High impact] The PR appears safe to merge; no actionable issues were found.

Summary

The PR replaces one broad seek condition with two limited queries joined by UNION ALL.

• Table row pages and exports seek keyed rows and the NULL-key tail separately.

No actionable issues were found. Tests were inspected but not run during this review.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Read next page] --> B{Anchor kind}
  B -->|No anchor| C[Read ordered rows]
  B -->|Keyed anchor| D[Read keyed rows after anchor]
  B -->|Keyed anchor| E[Read unkeyed tail]
  D --> F[UNION ALL and order by key and id]
  E --> F
  F --> G[Apply page limit and row-reader offset]
  B -->|Unkeyed export anchor| H[Read unkeyed rows after anchor id]

Loading

_{Reviews (1) · Last reviewed commit: "perf(tables): seek keyset row pages inst..." · Reviewed by Greptile}