[connortsui20]

Tracking Issue: #10701

Top of the decimal Widen stack, on #11005. No release gate.

Summary

A same-scale decimal cast that grows the canonical width copied every value into the wider type, even when the caller never reads them all. It now returns a decimal Widen over the source, so the copy waits for execution, and compute after the cast runs on the narrow values. This is the first production code that builds a decimal Widen, so it lands after the serializer lowering (#11002) and the kernels (#11003, #11004).

Changes

The child keeps the statistics of the source, such as exact extrema, because it holds the same values. A cast of a decimal Widen passes to its child. Widen execution copies decimal values instead of yielding a cast, which also keeps it from yielding a cast that returns a Widen again. A test checks that the default file writer round-trips a decimal Widen.

Known cost: filter at precision 38 is slower on a Widen than on the canonical array (see #10946), and a cast result now takes that path.

Local benchmark results

cargo bench -p vortex-array --bench widen_decimal, Apple M4, medians. The numbers were measured at the top of this stack before the split, so they cover the whole stack, not this layer alone. The baseline is the canonical input, and the candidate is the same values as a Widen (child precision 5 under precision 10, 9 under 18, and 18 under 38). CodSpeed on this PR measures the layer itself.

bench	p10 canonical	p10 Widen	p18 canonical	p18 Widen	p38 canonical	p38 Widen
cast_widen	12.08 µs	6.958 µs	12.33 µs	6.958 µs	16.54 µs	14.58 µs

API Changes

Casting a DecimalArray to a higher precision at the same scale returns a Widen when the canonical width grows.

[codspeed]

Merging this PR will regress 1 benchmark

2 warnings

⚠️ Walltime results may be unreliable

Some benchmarks used the Walltime instrument outside CodSpeed Macro Runners, which can make the results inconsistent.

Learn more about Macro Runners

See affected benchmarks

⚠️ 1 benchmark is dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

⚡ 12 improved benchmarks
❌ 1 regressed benchmark
✅ 2347 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	CPU Simulation	filter_powerlaw_by_random[10000]	700.5 ns	787.9 ns	-11.09%
⚡	CPU Simulation	cast_widen[p38_childp18_widen]	2.9 ms	1.5 ms	+95.98%
⚡	CPU Simulation	widen_nullable[16384]	358.4 µs	225.4 µs	+59.03%
⚡	Walltime	filter_neon[64]	2.9 µs	1.9 µs	+57.4%
⚡	CPU Simulation	cast_widen[p10_childp5_widen]	1,210.9 µs	789.8 µs	+53.32%
⚡	CPU Simulation	cast_widen[p18_childp9_widen]	1,211 µs	790.2 µs	+53.26%
⚡	Walltime	threshold_neon[64]	3.7 µs	2.6 µs	+41.85%
⚡	Walltime	filter_neon[80]	3.7 µs	2.7 µs	+40.45%
⚡	CPU Simulation	cast_widen[p38_childp18_canonical]	2.4 ms	1.8 ms	+31.49%
⚡	Walltime	threshold_neon[80]	4.3 µs	3.5 µs	+22.42%
⚡	CPU Simulation	take_filter_primitive_small_uncached_random_mask_random_indices[4096, 10]	24.3 µs	21.6 µs	+12.21%
⚡	CPU Simulation	cast_widen[p18_childp9_canonical]	1,007.7 µs	913.2 µs	+10.35%
⚡	CPU Simulation	cast_widen[p10_childp5_canonical]	1,008.2 µs	913.7 µs	+10.35%

Tip

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

---

_{Comparing ct/decimal-widen-lazy-cast (c0ec9ab) with ct/decimal-widen-stored (eb6bfc5)}