Execute spatial predicates with RowFn

[connortsui20]

Rationale for this change

Moves contains and intersects to prepared row execution. This lets each batch reuse work derived from constant geometries.

• Depends on: Execute spatial distance with RowFn #9349
• Progress towards: Epic: Row-oriented scalar functions #9128

What changes are included in this PR?

contains caches constant bounding boxes and creates a prepared geometry only when a relate path needs it. intersects hoists constant bounding boxes for conservative rejection. Tests cover operand order, constants, nulls, NaN coordinates, bounding boxes, and exact predicate agreement.

Rust 1.97.1 one-CGU fat-LTO results are mixed: most predicate cases remain within 8%, two constant-left contains cases regress by 16%, and sparse-null cases improve by 14–20%. The unchanged bounding-box controls remain near parity.

What APIs are changed? Are there any user-facing changes?

There are no public API changes. The predicates keep their existing OGC behavior and lazy scalar-function interfaces.

[codspeed-hq]

Merging this PR will regress 1 benchmark

⚠️ 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.

⚡ 14 improved benchmarks
❌ 1 regressed benchmark
✅ 1966 untouched benchmarks
🆕 1 new benchmark
⏩ 442 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	words_gather_scalar[65536]	8.3 µs	9.4 µs	-11.75%
⚡	Simulation	nullable_points_x_constant	1,402.2 µs	717.2 µs	+95.5%
⚡	Simulation	constant_x_nullable_points	1,402.9 µs	720.7 µs	+94.67%
⚡	Simulation	constant_x_nullable_polygons_disjoint	1,531.9 µs	793.7 µs	+93.02%
⚡	Simulation	points_x_constant	1,278.8 µs	664.4 µs	+92.48%
⚡	Simulation	polygons_overlapping_x_constant	1,300.3 µs	677.5 µs	+91.93%
⚡	Simulation	constant_x_points	1,278.9 µs	668.4 µs	+91.34%
⚡	Simulation	nullable_polygons_disjoint_x_constant	1,539 µs	809.8 µs	+90.05%
⚡	Simulation	polygons_disjoint_x_constant	1,291.7 µs	680.8 µs	+89.74%
⚡	Simulation	constant_x_polygons_disjoint	1,295.3 µs	683.4 µs	+89.55%
⚡	Simulation	nullable_polygons_x_nullable_points	597.3 µs	416.1 µs	+43.54%
⚡	Simulation	nullable_polygons_90pct_x_constant_point	528.6 µs	405.9 µs	+30.22%
⚡	Simulation	column_x_column_polygons	321.6 µs	269.7 µs	+19.25%
⚡	Simulation	column_x_column_polygons	324 µs	275.6 µs	+17.58%
⚡	Simulation	column_x_column_points	281.8 µs	251.3 µs	+12.12%
🆕	Simulation	constant_x_polygons_overlapping	N/A	841.8 µs	N/A

Tip

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

---

_{Comparing ct/row-fn-spatial-predicates (a37c59f) with ct/row-fn-spatial-distance (f77666b)}

Footnotes

1. 442 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. ↩

[connortsui20]

Here are the local spatial-predicate benchmark results from the final Rust 1.97.1 run.

The comparison used a develop baseline with benchmark-only backfills (1edd0fca) and the full RowFn stack (8ca9fc73). The machine was an AMD Ryzen 9 7950X running Linux. The build used rustc 1.97.1, LLVM 22.1.6, one CGU, fat LTO, and -C target-cpu=native.

Each target used two warm runs and seven alternating measured pairs. Each process used 100 samples with a 0.25–0.5 second measurement window on CPU 4. Negative changes are faster. The change is the median paired ratio, not the ratio of the two displayed medians.

Most predicate cases remain within 8%. Two constant-left contains cases regress by about 16%, while sparse-null cases improve by 14.04–20.45%. The bounding-box controls remain close to parity. The unchanged envelope control moved substantially for multipolygons, which indicates binary-layout sensitivity outside the predicate implementation.

Contains and intersects: 15 cases

Benchmark	Baseline	RowFn stack	Change
contains/constant_x_polygons_disjoint	68.210 µs	79.370 µs	+16.36%
contains/constant_x_points	70.630 µs	81.040 µs	+15.89%
intersects/polygons_overlapping_x_constant	69.820 µs	74.980 µs	+7.82%
intersects/polygons_disjoint_x_constant	68.160 µs	73.160 µs	+7.80%
intersects/points_x_constant	70.370 µs	75.390 µs	+6.86%
contains/polygons_x_constant_point	12.530 µs	12.930 µs	+3.94%
intersects/nullable_points_x_constant	74.250 µs	76.270 µs	+3.36%
contains/constant_x_nullable_points	74.560 µs	76.820 µs	+2.96%
contains/column_x_column_polygons	17.980 µs	18.250 µs	+1.84%
intersects/nullable_polygons_disjoint_x_constant	77.340 µs	78.060 µs	+1.36%
contains/constant_x_nullable_polygons_disjoint	77.420 µs	78.250 µs	+1.13%
contains/column_x_column_points	12.290 µs	12.440 µs	+0.90%
intersects/column_x_column_polygons	16.150 µs	16.200 µs	+0.62%
contains/nullable_polygons_90pct_x_constant_point	15.880 µs	13.640 µs	-14.04%
contains/nullable_polygons_x_nullable_points	20.220 µs	16.100 µs	-20.45%

Envelope controls: 7 cases

Benchmark	Baseline	RowFn stack	Change
multipolygon_random_nulls	21.220 µs	28.500 µs	+34.64%
multipolygon_mixed_validity	27.140 µs	35.780 µs	+32.22%
multipolygon_non_nullable	28.060 µs	37.000 µs	+31.95%
multipoint_non_nullable	19.540 µs	21.760 µs	+11.44%
point_random_nulls	1.239 µs	1.269 µs	+2.36%
point_non_nullable	1.209 µs	1.229 µs	+0.84%
point_mixed_validity	1.269 µs	1.269 µs	-0.79%

Bounding-box controls: 8 cases

Benchmark	Baseline	RowFn stack	Change
contains_bbox_disjoint	1.584 µs	1.589 µs	+2.20%
intersects_bbox_disjoint	1.584 µs	1.589 µs	+2.20%
intersects_bbox_candidate	48.990 µs	48.780 µs	+0.02%
contains_exact_disjoint	45.910 µs	45.910 µs	+0.00%
intersects_exact_disjoint	46.130 µs	46.130 µs	+0.00%
contains_bbox_candidate	48.750 µs	48.690 µs	-0.18%
intersects_exact_candidate	43.910 µs	47.700 µs	-0.62%
contains_exact_candidate	45.990 µs	46.650 µs	-0.76%