[robert3005]

Summary

Make stats validation cheap, skips it where values are valid by construction, and removes Scalar construction and copies from stats reads and aggregate partial merging.

Changes

• Scalar::try_new and Scalar::validate run an inline fast check. Error messages are built in a cold path.
• Constructors that are valid by construction use new_unchecked. VarBinView reads no longer check UTF-8 again.
• Stats reads and merges build fewer scalars. Cached bool and count stats are read without a Scalar.
• SumV2 parses partials from the tuple values directly.
• Accumulators made from one AggregateFnRef share their resolved dtypes, so their partial dtypes compare by pointer.
• Fix: merge_is_constant always overwrote its result with exact(false).
• New scalar_construction bench.

ns/op (M4 Max)	Before	After
try_new, primitive	8.6	1.3
execute_scalar, primitive / utf8 / struct8	15.9 / 26.2 / 196	7.5 / 11.8 / 114
statistics().get(Min)	18.0	9.7
StatsSet::merge_ordered	704	300
SumV2 combine_partials	99	31

API Changes

• AggregateFnVTable::partial_from_scalar, DynAccumulator::combine_partials and MinMaxResult::from_scalar take &Scalar.
• New StructFields::field_dtypes() returns &DType without cloning.
• After a stats merge, IsConstant can now be exact(true). When the result cannot be known, the stat is cleared.

[codspeed]

Merging this PR will improve performance by 31.08%

⚠️ 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.

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 21 improved benchmarks
✅ 2102 untouched benchmarks
🆕 10 new benchmarks
⏩ 518 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	sum_v2	91 µs	27.1 µs	×3.4
⚡	Simulation	density_sweep_dense_runs[0.001]	47.7 µs	30.3 µs	+57.3%
⚡	Simulation	filter_powerlaw_by_mostly_true[250000]	150 µs	99.5 µs	+50.79%
⚡	Simulation	is_valid_per_element[1024]	405.9 µs	280.5 µs	+44.71%
⚡	Simulation	is_valid_per_element[256]	104.5 µs	72.3 µs	+44.48%
⚡	Simulation	repeated[primitive]	238.8 µs	176 µs	+35.62%
⚡	Simulation	execute_scalar_struct_wide	444.9 µs	333.2 µs	+33.52%
⚡	Simulation	once[primitive]	272.9 µs	211.1 µs	+29.27%
⚡	Simulation	repeated[struct]	701.3 µs	557.7 µs	+25.75%
⚡	Simulation	once[dict]	497.6 µs	401.5 µs	+23.93%
⚡	Simulation	repeated[dict]	508.2 µs	410.7 µs	+23.72%
⚡	Simulation	once[struct]	785.5 µs	643.7 µs	+22.03%
⚡	Simulation	scalar_access[(1024, false, false)]	727.7 µs	599 µs	+21.48%
⚡	Simulation	execute_scalar_struct_simple	99.1 µs	83.9 µs	+18.06%
⚡	Simulation	scalar_access[(1024, true, false)]	787.8 µs	670.5 µs	+17.49%
⚡	Simulation	search_index_mixed_out_of_range_chunked	373.7 µs	330.5 µs	+13.09%
⚡	Simulation	search_index_below_min_chunked	375.4 µs	332.8 µs	+12.78%
⚡	Simulation	search_index_full_range_random_chunked	405.6 µs	363.1 µs	+11.68%
⚡	Simulation	probe_scalar_fn_valid_once[vortex.binary(bool?, len=100000)]	317.3 µs	288 µs	+10.18%
⚡	Simulation	probe_scalar_fn_once[vortex.binary(bool?, len=100000)]	262.1 µs	238.2 µs	+10.02%
...	...	...	...	...	...

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

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing rk/scalar-construction-perf (5087c16) with develop (58a3c2f)}

Footnotes

1. 518 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. ↩

[myrrc]

LGTM