Pull Request Explorer

Summary

Fix seed-dependent crash in checkpoint fuzz test

This change eliminates a seed-dependent crash in the checkpoint fuzz test, ensuring reliable test execution and preventing false negatives in CI pipelines. By guarding against divide‑by‑zero errors, the test suite remains stable, improving confidence in database checkpoint functionality.

Health Assessment

• • The PR had a long review period with minimal changes, indicating low complexity but delayed feedback.

Summary

Add numeric comparison with scalar sub-query

Enables expressive SQL predicates that compare numeric columns to scalar sub-queries, improving query flexibility and performance for analytical workloads.

Health Assessment

• • The PR required a 144‑hour cycle time with a 139‑hour first review, indicating a slow review process. Despite no post‑review commits, the large code churn (≈20k lines) and 20 files changed suggest high complexity and potential risk.

Summary

Reduce Ubuntu package download failures in CI

Configures CI jobs to use Hetzner mirrors, reducing timeouts and ensuring builds succeed.

Health Assessment

• • Fast cycle time and minimal review indicate smooth integration; large scope but low risk due to CI nature.

Summary

Bump gosu to 1.19-go1.25.12

Update Dockerfile to use newer Go patch for CVE fixes, improving security.

Health Assessment

• • Quick dependency bump with minimal changes, merged within hours.

Summary

Add week unit support to SAMPLE BY

Enables users to use 'SAMPLE BY w' without a leading digit, simplifying queries and reducing errors.

Health Assessment

• • PR resolved a parsing bug with minimal changes and quick review, indicating efficient issue handling.

Summary

Fix fuzz test failures due to non-ASCII logging

This change prevents intermittent test failures caused by logging non-ASCII values, ensuring reliable test runs and accurate CI results.

Health Assessment

• • Quick fix with minimal changes, no production impact, resolved in a single commit.

Summary

Fix dot_product result for transposed arrays

Corrects dot_product calculation when operands have non-vanilla layout, ensuring accurate results for transposed or sliced arrays without affecting performance.

Health Assessment

• • The PR was resolved quickly with a single commit and minimal code changes, indicating a straightforward bug fix with low risk to the system.

Summary

Fix query activity test hang on constrained CI

Prevents flaky test failures and thread leaks in CI by adding timeouts and yielding, improving stability and test reliability.

Health Assessment

• • Single test file change resolved flaky CI failures with minimal rework; quick review and merge indicates low risk.

Summary

Fix materialized view error message for missing sampling interval

Improves error handling for materialized view creation by providing clearer messages when a timestamp column is present but no sampling interval is specified, reducing confusion and aiding developers.

Health Assessment

• • Quick fix with minimal changes, approved after a single review, indicating low complexity and risk.

Summary

Fix covering index over parquet after column type change

Corrects a bug that caused WAL suspension when adding a POSTING covering index to parquet partitions with pending column type changes, ensuring reliable index creation and preventing data write stalls.

Health Assessment

• • The PR resolved a critical WAL suspension issue with a single commit and one approval after a 37‑hour review, indicating a focused change but moderate review latency.

Summary

Fix POSTING index covering scan thread migration bug

Resolves crashes and resource leaks during multi-worker HTTP parquet exports and pgwire queries, ensuring reliable data export and preventing client connection drops.

Health Assessment

• • The PR required extensive rework and multiple review cycles, indicating complex concurrency bug fixes and significant code churn.

Summary

Resolve Parquet column file mismatch after ALTER

Fixes storage-policy TO PARQUET conversion failures caused by column type changes, ensuring reliable data export and preventing pipeline stalls.

Health Assessment

• • The PR required multiple review cycles and several commits to address complex metadata handling after column type changes, indicating significant rework but ultimately resolved a critical storage-policy failure.

Summary

Fix socket leaks during shutdown

Ensures proper cleanup of socket descriptors when queries are parked or retried, preventing resource leaks in long‑running JVMs.

Health Assessment

• • PR addressed a subtle resource leak with minimal rework and quick approval, indicating a focused bug fix with low complexity.

Summary

Fix PIVOT and expression column names with double quotes

Corrects SQL compiler to strip protective quotes from column aliases, ensuring accurate column names in result sets and preventing leaks into metadata.

Health Assessment

• • Long review cycle with many commits before approval indicates significant rework; single review after extensive changes suggests high complexity and potential risk.

Summary

Fix catalogue queries after concurrent cache clear

Ensures accurate table listings by preventing incomplete data exposure during concurrent cache clears, improving reliability for users querying metadata.

Health Assessment

• • The PR resolved a rare concurrency bug with a single review round and moderate review time, indicating a focused fix with minimal risk.

Summary

Prevent stale egress plans from leaking to clients

Fixes a bug where QWP egress exposed stale query plans to clients, improving reliability and consistency with other protocols.

Health Assessment

• • The PR had a long cycle time (117h) and slow first review (94h), with a large code change affecting core backend logic and CI configuration, indicating significant effort and potential risk.

Summary

Fail fast on deterministic ingest rejections

Reduces wasted retries and improves client error visibility by treating permanent data errors as terminal, speeding up failure detection and reducing resource consumption.

Health Assessment

• • PR had a single review after a long review time, but no rework; moderate scope and normal cycle time.

Summary

Speed up sender coercion-error tests

Reduces CI runtime by eliminating unnecessary wait times in error-path tests, improving developer productivity and pipeline reliability.

Health Assessment

• • The change is a small, single-file test optimization that was reviewed and merged quickly, delivering a significant performance win with minimal risk.

Summary

Fix LATEST ON row selection bug

Corrects a bug that caused incorrect row selection in LATEST ON queries, ensuring accurate time-series data retrieval for users.

Health Assessment

• • The PR spanned nearly 12 days with a single review after many commits, indicating high complexity and potential risk. The bug fix is critical for accurate query results in time-series data.

Summary

Bump gosu version in Dockerfile

Updates the gosu utility to the latest patch, ensuring compatibility with Go 1.25.11 and maintaining container security.

Health Assessment

• • Fast cycle time and minimal changes indicate low risk and straightforward maintenance.

Summary

Fix crash on empty VARCHAR joins

Prevents assertion errors during hash joins, GROUP BY, or DISTINCT operations on empty VARCHAR keys, ensuring reliable query execution for customers.

Health Assessment

• • Long review cycle indicates complexity, but single review round and no rework suggest the fix was straightforward and well-contained.

Summary

Add connect timeout and tolerant client startup

Improves client reliability by adding cross‑platform connect timeout and tolerant startup, reducing connection hangs and enabling smoother failover.

Health Assessment

• • The PR required a series of iterative commits after a single review, indicating moderate complexity and refinement. The 7‑day cycle time and large line changes reflect a substantial backend and CI impact.

Summary

Prevent index corruption on .pk stat failure

Fixes a critical bug that could corrupt indexed data during column renames, ensuring data integrity and preventing query failures.

Health Assessment

• • Quick fix with minimal changes, resolved in a single commit after review, indicating low complexity and high confidence in stability.

Summary

Make DataID publication thread-safe

Ensures reliable replication and backup by preventing race conditions that could corrupt data identifiers, improving system stability and data integrity.

Health Assessment

• • The PR was reviewed and merged within a day, with a single commit and minimal back‑and‑forth, indicating a focused fix. The large line count reflects a comprehensive thread‑safety overhaul, but the quick turnaround suggests low complexity and high confidence in the change.