docs: update query benchmarks (3.9.4)

[github-actions]

Automated query benchmark update for 3.9.4 from workflow run #737.

[greptile-apps]

Greptile Summary

This automated PR records query benchmark results for version 3.9.4, updating the summary table and latest-results section in generated/benchmarks/QUERY-BENCHMARKS.md. The WASM engine shows notably elevated fnImpact depth-3/5 latencies (~41–42% up) and diffImpact latency (~35% up) vs. 3.9.3.

• The mid benchmark target changed from enumNode (3.9.3) to ctx (3.9.4), which means all percentage deltas in the summary table compare measurements taken against different functions. Readers may misinterpret the regressions/improvements as purely engine-performance signals when some of the delta may reflect workload differences between those two symbols.

Confidence Score: 5/5

Safe to merge — documentation-only, generated by automation, no runtime impact.

All findings are P2 (readability/comparability concern about the mid target switch). The data itself is internally consistent and correctly formatted. No code paths are affected.

generated/benchmarks/QUERY-BENCHMARKS.md — the percentage deltas in the 3.9.4 rows are compared against a different mid benchmark target (enumNode → ctx), which may mislead readers interpreting the WASM fnImpact regressions.

Important Files Changed

Filename	Overview
generated/benchmarks/QUERY-BENCHMARKS.md	Adds 3.9.4 benchmark data; mid target changed from enumNode to ctx, making percentage deltas vs. 3.9.3 not strictly apples-to-apples; WASM fnImpact shows 41–42% increase.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Release Workflow Trigger] --> B[Run codegraph on itself]
    B --> C{For each engine}
    C --> D[Native Rust Engine]
    C --> E[WASM Engine]
    D --> F[Query: fnDeps depth 1/3/5\nhub=buildGraph mid=ctx leaf=docs]
    E --> F
    F --> G[Query: fnImpact depth 1/3/5]
    G --> H[Query: diffImpact]
    H --> I[Median over 5 runs]
    I --> J[Compute % delta vs prev version]
    J --> K[Append to QUERY-BENCHMARKS.md]
    K --> L[Open automated PR]

Loading

_{Reviews (1): Last reviewed commit: "docs: update query benchmarks (3.9.4)" | Re-trigger Greptile}

[greptile-apps]

Mid-target change makes percentage deltas not comparable

The mid benchmark target switched from enumNode (3.9.3) to ctx (3.9.4). All percentage arrows in the 3.9.4 rows are computed by dividing the new latency against 3.9.3 latency from a different function. A heavier or lighter ctx graph neighbourhood can explain part (or all) of the observed regression, so readers cannot tell from the table alone how much is engine performance vs. workload change. Consider adding a footnote like *mid target changed from enumNode; deltas not directly comparable* on the 3.9.4 rows, or re-running 3.9.3 with mid=ctx as a baseline.