perf(array): reuse probe state in primitive and compressed arrays

[joseph-isaacs]

Uses the scalar probe API from #9843 to retain preparation across repeated lookups, including recursive child access.

• Primitive: unit local state; reads the value after shared validity checking.
• FastLanes RLE: retains the slice base offset and probes for indices, offsets, and values.
• RunEnd: retains ends and values child probes across binary-search comparisons and lookups.
• PCO: retains validity, prefix ranks, page boundaries, and the most recently decoded page.

let mut probe = array.probe(ProbeUsage::Repeated);
probe.execute_scalar(10, &mut ctx)?;
probe.execute_scalar(11, &mut ctx)?; // Reuses preparation at each visited slot.

Includes nested RunEnd/PCO cache-reuse and drop tests, nullable/sliced access, lazy-validity regressions, an example, and scalar-probe benchmarks. The benchmark note labels older measurements as historical rather than results for the current implementation.

Validation: 3,925 crate tests passed (one skipped), array doctests, cargo +nightly fmt --all, Python Rust formatting, and cargo clippy --all-targets --all-features. The scalar-probe benchmark suite completed with short samples; this is not a controlled before/after performance comparison.

[codspeed-hq]

Merging this PR will regress 5 benchmarks

⚠️ 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.

⚡ 3 improved benchmarks
❌ 5 regressed benchmarks
✅ 2189 untouched benchmarks
🆕 64 new benchmarks
⏩ 218 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	filtered_owned_i64_avx2[OneNullInEight]	21.9 µs	25.8 µs	-14.81%
❌	Simulation	is_valid_per_element[256]	103.5 µs	121.3 µs	-14.66%
❌	Simulation	is_valid_per_element[1024]	401.5 µs	469.9 µs	-14.54%
❌	Simulation	execute_scalar_struct_simple	106.4 µs	121.3 µs	-12.24%
❌	Simulation	execute_scalar_struct_wide	502.9 µs	564.9 µs	-10.97%
⚡	WallTime	arrow_checked_add_u32_neon[16384]	20.3 µs	12.3 µs	+65.25%
⚡	WallTime	words_gather_scalar_avx2[65536]	9.4 µs	8.3 µs	+13.3%
⚡	WallTime	filtered_owned_i64_avx512[OneNullInEight]	26.6 µs	24 µs	+10.52%
🆕	Simulation	pco_execute_scalar[(1, false, false)]	N/A	211.2 µs	N/A
🆕	Simulation	pco_execute_scalar[(1, true, false)]	N/A	210.7 µs	N/A
🆕	Simulation	pco_execute_scalar[(1024, false, false)]	N/A	132.8 ms	N/A
🆕	Simulation	pco_execute_scalar[(1024, false, true)]	N/A	132.8 ms	N/A
🆕	Simulation	pco_execute_scalar[(1024, true, false)]	N/A	101.6 ms	N/A
🆕	Simulation	pco_execute_scalar[(1024, true, true)]	N/A	98.8 ms	N/A
🆕	Simulation	pco_execute_scalar[(64, false, false)]	N/A	8.4 ms	N/A
🆕	Simulation	pco_execute_scalar[(64, false, true)]	N/A	8.4 ms	N/A
🆕	Simulation	pco_execute_scalar[(64, true, false)]	N/A	6.8 ms	N/A
🆕	Simulation	pco_execute_scalar[(64, true, true)]	N/A	6.2 ms	N/A
🆕	Simulation	pco_probe[(1, false, false)]	N/A	213.8 µs	N/A
🆕	Simulation	pco_probe[(1, true, false)]	N/A	210.8 µs	N/A
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

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

---

_{Comparing ji/array-probe-encodings (7c04bd8) with ji/array-probe-api (4eae754)}

Footnotes

1. 218 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩