fix[file]: read the one-row pruning result in can_prune instead of re…

[tomsanbear]

Summary

Closes: #8366

VortexFile::can_prune stopped pruning for every and/or and eq predicate after #7575 removed bottom-up constant-folding. It evaluates a predicate's stat-falsification expression over the one-row file-stats array, but only read a folded Columnar::Constant result and mapped Columnar::Canonical(_) to false, discarding the computed boolean. Post-#7575, composite falsifications (boolean and/or trees, and the or(min > lit, lit > max) that eq expands to) execute to a one-row Canonical instead of folding, so pruning silently stopped for them. Bare gt/lt comparisons still fold through the compare kernel's one-level constant fast path, so only composite and eq predicates regressed.

It went unnoticed upstream because can_prune has no in-repo callers, the scan path prunes through FileStatsLayoutReader::evaluate_file_stats, which already reads the one-row result out correctly. This change makes can_prune mirror that read-out (execute to Canonical, take the row-0 scalar) and adds an end-to-end regression test covering bare, and/or, and eq predicates with non-falsifiable controls.

API Changes

No signature changes. This restores the runtime behavior of the public VortexFile::can_prune: it now returns true (prunable) for and/or/eq predicates that file statistics prove cannot match, where it previously returned false. No engine integration is affected. The scan path already pruned correctly via evaluate_file_stats; only direct callers of can_prune saw the regression. No docs changes needed.

Testing

• New test_can_prune_composite_predicates (vortex-file/src/tests.rs): writes a file with stat-bounded age/price columns and asserts pruning for bare, and, or, and eq falsifiable predicates, plus non-falsifiable controls that must not prune.
• Verified non-vacuous: the test fails against the pre-fix Columnar::Canonical(_) => false arm and passes after.
• cargo nextest run -p vortex-file — 62/62 pass (existing suite + new test).
• cargo +nightly fmt -p vortex-file and cargo clippy -p vortex-file --all-targets clean.

AI assistance

Developed with assistance from Claude Code (agentic AI) for root-cause analysis, the fix, and the regression test. All changes were reviewed and verified by me against the Vortex source and a downstream test suite.

[robert3005]

you can skip this into_bool

[joseph-isaacs]

Great find thanks!

Can we keep the execute::<Columnar> and just check if the canonical first value is a boolean

[codspeed-hq]

Merging this PR will degrade performance by 12.89%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚡ 4 improved benchmarks
❌ 11 regressed benchmarks
✅ 1517 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	chunked_bool_canonical_into[(1000, 10)]	20.4 µs	35.5 µs	-42.59%
❌	Simulation	take_10k_random	196.7 µs	254.5 µs	-22.73%
❌	Simulation	patched_take_10k_contiguous_patches	230.3 µs	289.2 µs	-20.36%
❌	Simulation	patched_take_10k_random	242.3 µs	301.2 µs	-19.55%
❌	Simulation	decompress_rd[f64, (10000, 0.0)]	111.3 µs	138.3 µs	-19.51%
❌	Simulation	decompress_rd[f64, (10000, 0.1)]	111.3 µs	138.1 µs	-19.4%
❌	Simulation	decompress_rd[f64, (10000, 0.01)]	111 µs	137.7 µs	-19.4%
❌	Simulation	take_10k_contiguous	252.3 µs	309.9 µs	-18.58%
❌	Simulation	chunked_varbinview_canonical_into[(1000, 10)]	161.6 µs	197.9 µs	-18.33%
❌	Simulation	chunked_varbinview_into_canonical[(1000, 10)]	176.9 µs	213.1 µs	-16.99%
❌	Simulation	decompress_rd[f32, (10000, 0.1)]	80.7 µs	89.8 µs	-10.14%
⚡	WallTime	cuda/bitpacked_u8/unpack/3bw[100M]	351.3 µs	298.9 µs	+17.53%
⚡	Simulation	decompress_rd[f64, (100000, 0.0)]	980.4 µs	845.5 µs	+15.96%
⚡	Simulation	bitwise_not_vortex_buffer_mut[128]	244.4 ns	215.3 ns	+13.55%
⚡	Simulation	bitwise_not_vortex_buffer_mut[1024]	304.7 ns	275.6 ns	+10.58%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing tomsanbear:fix/can-prune-one-row-result (4662318) with develop (0dd63f0)}

[tomsanbear]

Great find thanks!

Can we keep the execute::<Columnar> and just check if the canonical first value is a boolean

for sure, let me make a quick patch to address this and robert's comment

[tomsanbear]

@joseph-isaacs @robert3005 updated with the change to maintain the previous match arm