buffer: re-enable Fast API for Buffer.write

[ronag]

Re-enables fast Fast API for Buffer.write after fixing UTF8 handling.

13:05:25 buffers/buffer-write-string-short.js n=1000000 len=1 encoding='ascii'                            ***    266.52 %       ±7.27%  ±9.75% ±12.84%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=1 encoding='latin1'                           ***    201.16 %       ±4.74%  ±6.33%  ±8.29%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=1 encoding='utf8'                             ***    242.07 %       ±6.29%  ±8.44% ±11.14%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=16 encoding='ascii'                           ***    264.81 %       ±5.88%  ±7.86% ±10.29%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=16 encoding='latin1'                          ***    197.28 %       ±6.96%  ±9.32% ±12.26%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=16 encoding='utf8'                            ***    349.02 %       ±8.80% ±11.84% ±15.68%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=32 encoding='ascii'                           ***    271.75 %       ±8.70% ±11.69% ±15.43%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=32 encoding='latin1'                          ***    214.58 %       ±5.85%  ±7.84% ±10.32%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=32 encoding='utf8'                            ***    351.17 %       ±7.64% ±10.28% ±13.60%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=8 encoding='ascii'                            ***    273.30 %       ±8.65% ±11.63% ±15.38%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=8 encoding='latin1'                           ***    191.62 %       ±5.62%  ±7.52%  ±9.88%
13:05:25 buffers/buffer-write-string-short.js n=1000000 len=8 encoding='utf8'                             ***    224.55 %       ±4.72%  ±6.33%  ±8.33%

[lemire]

Having a non-deterministic test is unpleasant.

[anonrig]

You can trigger fast api call using the method @targos implemented in his previous pull requests.

[lemire]

This looks to me. I wish we could use a deterministic test. Having to loop a million times to trigger and issue is both slow and error prone.

[codecov]

Codecov Report

Attention: Patch coverage is 88.88889% with 8 lines in your changes missing coverage. Please review.

Project coverage is 87.60%. Comparing base (981c701) to head (0ae2ae1).
Report is 252 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_buffer.cc	88.88%	3 Missing and 5 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #54526      +/-   ##
==========================================
- Coverage   87.61%   87.60%   -0.01%     
==========================================
  Files         650      650              
  Lines      182835   182889      +54     
  Branches    35382    35389       +7     
==========================================
+ Hits       160185   160227      +42     
- Misses      15925    15927       +2     
- Partials     6725     6735      +10     

Files with missing lines	Coverage Δ
src/node_buffer.cc	71.41% <88.88%> (+1.46%)	⬆️

... and 32 files with indirect coverage changes

[jasnell]

Some lint issues but otherwise LGTM

[blexrob]

With the current commits, the following test will fail:

let i = 0;
const testStr = "\xc2\x80"; // when stored in 1-byte chars, this is a valid UTF-8 encoding
const expected = Buffer.from(testStr).toString("hex");
for(; i < 1_000_000; i++) {
  const buf = Buffer.from(testStr);
  const ashex = buf.toString("hex");
  if (ashex !== expected) {
    console.log(`Decoding changed in iteration ${i} when changing to FastWriteStringUTF8, got ${ashex}, expected ${expected}`);
    break;
  }
}

if(i<1_000_000) {
  console.error("FAILED after %d iterations",i);
} else
  console.log("PASSED after %d iterations",i);

The source string is stored as UTF16 (compressed to 1 byte), but still needs to be converted to UTF-8. The fast path can only be implemented as memcpy when only characters from the range 0-127 are present.

[ronag]

The source string is stored as UTF16 (compressed to 1 byte), but still needs to be converted to UTF-8. The fast path can only be implemented as memcpy when only characters from the range 0-127 are present.

I thought FastOneByteString means that it's only one byte chars?

[ronag]

< when only characters from the range 0-127 are present

@lemira does simdutf have this? validate_ascii?

[RedYetiDev]

Given the repeated activity since the failed commit, I assume it is known that the commit-queue originally failed, so I removed the label. Feel free to re-add.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61792/

[RedYetiDev]

Correct me if I'm wrong, but shouldn't the "PR-URL" meta be added by the bot?

[targos]

Correct me if I'm wrong, but shouldn't the "PR-URL" meta be added by the bot?

You're right. @ronag please remove the PR-URL from the commit.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61821/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61838/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61847/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61848/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61849/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61860/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61894/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61898/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61904/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/61926/

[nodejs-github-bot]

Landed in dc74f17