perf: avoid unnecessary flat

[juanrgm]

Summary

splitProps calls Array.prototype.flat unnecessarily and this causes some delays when there are many keys.

How did you test this change?

var values = Array.from({ length: 100 }).map((v, i) => `value${i}`);
var tests = 1000;

// 24.216064453125 ms
function flat() {
  console.time("flat");
  for (let x = 1; x <= tests; ++x) new Set(values.flat());
  console.timeEnd("flat");
}

// 3.248046875 ms
function noflat() {
  console.time("noflat");
  for (let x = 1; x <= tests; ++x) new Set(values);
  console.timeEnd("noflat");
}

function benchmark() {
  flat();
  noflat();
}

benchmark();

[changeset-bot]

🦋 Changeset detected

Latest commit: 1b5ea07

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
solid-js	Patch
test-integration	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[ankit-gautam23]

LGTM

[ryansolid]

To be fair I'd be shocked if the keys ever got over 12. I'm sure it is faster to access the index in any case and size add is minimal.

[juanrgm]

To be fair I'd be shocked if the keys ever got over 12. I'm sure it is faster to access the index in any case and size add is minimal.

In SUID there are 100 style properties, not counting those of the component itself, so almost all components are suffering a noticeable performance loss.

I'm rewriting the splitProps and mergeProps functions and the results are promising:

✓ test/component.bench.ts (10) 7674ms
   ✓ splitProps(0, 15) (2) 2189ms
     name                            hz     min      max    mean     p75     p99    p995    p999     rme  samples
   · oldSplitProps(0, 15)    646,064.99  0.0011  19.4825  0.0015  0.0013  0.0045  0.0051  0.0100  ±7.74%   323033
   · newSplitProps(0, 15)  3,573,490.18  0.0001   3.0986  0.0003  0.0003  0.0006  0.0007  0.0014  ±1.90%  1786746   fastest
   ✓ splitProps(0, 100) (2) 2423ms
     name                             hz     min     max    mean     p75     p99    p995    p999     rme  samples
   · oldSplitProps(0, 100)    140,244.40  0.0061  0.8396  0.0071  0.0067  0.0128  0.0154  0.0569  ±0.99%    70123
   · newSplitProps(0, 100)  3,655,819.15  0.0002  2.4475  0.0003  0.0003  0.0006  0.0008  0.0016  ±1.45%  1827910   fastest
   ✓ splitProps(25, 100) (2) 3357ms
     name                           hz     min     max    mean     p75     p99    p995    p999     rme  samples
   · oldSplitProps(25, 100)  32,577.70  0.0265  0.5104  0.0307  0.0277  0.0671  0.0942  0.2191  ±0.74%    16289
   · newSplitProps(25, 100)  97,826.43  0.0094  0.5183  0.0102  0.0097  0.0194  0.0228  0.0567  ±0.51%    48914   fastest  
   ✓ splitProps(50, 100) (2) 4270ms
     name                           hz     min     max    mean     p75     p99    p995    p999     rme  samples
   · oldSplitProps(50, 100)  16,848.36  0.0489  2.0058  0.0594  0.0585  0.1376  0.1900  0.3668  ±1.17%     8430
   · newSplitProps(50, 100)  37,215.99  0.0212  0.5338  0.0269  0.0225  0.0810  0.0947  0.1797  ±0.81%    18608   fastest  
   ✓ splitProps(100, 25) (2) 5179ms
     name                           hz     min     max    mean     p75     p99    p995    p999     rme  samples
   · oldSplitProps(100, 25)   9,644.99  0.0952  0.5339  0.1037  0.1000  0.2092  0.2367  0.4495  ±0.72%     4823
   · newSplitProps(100, 25)  21,117.39  0.0425  1.0419  0.0474  0.0441  0.1081  0.1488  0.3723  ±0.88%    10560   fastest  
   · SplitProps bench (0)


 BENCH  Summary

  newSplitProps(0, 15) - test/component.bench.ts > splitProps(0, 15)
    5.53x faster than oldSplitProps(0, 15)

  newSplitProps(0, 100) - test/component.bench.ts > splitProps(0, 100)
    26.07x faster than oldSplitProps(0, 100)

  newSplitProps(25, 100) - test/component.bench.ts > splitProps(25, 100)
    3.00x faster than oldSplitProps(25, 100)

  newSplitProps(50, 100) - test/component.bench.ts > splitProps(50, 100)
    2.21x faster than oldSplitProps(50, 100)

  newSplitProps(100, 25) - test/component.bench.ts > splitProps(100, 25)
    2.19x faster than oldSplitProps(100, 25)

[coveralls]

Pull Request Test Coverage Report for Build 4824514257

Warning: This coverage report may be inaccurate.

This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

• For more information on this, see Tracking coverage changes with pull request builds.
• To avoid this issue with future PRs, see these Recommended CI Configurations.
• For a quick fix, rebase this PR at GitHub. Your next report should be accurate.

Details

• 1 of 1 (100.0%) changed or added relevant line in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.02%) to 94.011%

---

Totals
Change from base Build 4640851655:	-0.02%
Covered Lines:	4077
Relevant Lines:	4267

---

💛 - Coveralls