perf(pycodestyle): Remove regex captures

[MichaReiser]

From the regex documentation:

Advice: Prefer in this order: is_match, find, captures.

source

This PR removes our usages of capture_iter with find_iter and and implements the whitespace-testing manually.

Performance

• no-logical: pycode style rules disabled
• pr3715: The base version with logical lines enabled
• pr3735: This PR with logical lines enabled.

group                                      no-logical                             pr3735                                  pr3715
-----                                      ----------                             -----                                  ------
linter/all-rules/large/dataset.py          1.00      8.5±0.01ms     4.8 MB/sec    1.10      9.4±0.05ms     4.3 MB/sec    1.10      9.4±0.15ms     4.3 MB/sec
linter/all-rules/numpy/ctypeslib.py        1.00      2.1±0.01ms     7.8 MB/sec    1.09      2.3±0.00ms     7.1 MB/sec    1.15      2.5±0.00ms     6.8 MB/sec
linter/all-rules/numpy/globals.py          1.00    247.9±3.25µs    11.9 MB/sec    1.10    272.6±1.25µs    10.8 MB/sec    1.16    286.9±3.46µs    10.3 MB/sec
linter/all-rules/pydantic/types.py         1.00      3.7±0.04ms     7.0 MB/sec    1.11      4.1±0.07ms     6.3 MB/sec    1.19      4.4±0.02ms     5.9 MB/sec
linter/default-rules/large/dataset.py      1.00      4.7±0.01ms     8.7 MB/sec    1.19      5.6±0.11ms     7.3 MB/sec    1.24      5.8±0.08ms     7.0 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.00    999.5±5.57µs    16.7 MB/sec    1.22   1216.5±2.35µs    13.7 MB/sec    1.31  1304.4±16.73µs    12.8 MB/sec
linter/default-rules/numpy/globals.py      1.00    101.0±0.41µs    29.2 MB/sec    1.36    137.5±1.36µs    21.5 MB/sec    1.44    145.7±1.31µs    20.2 MB/sec
linter/default-rules/pydantic/types.py     1.00      2.1±0.01ms    11.9 MB/sec    1.19      2.5±0.06ms    10.0 MB/sec    1.26      2.7±0.03ms     9.5 MB/sec

We're getting closer, but a 36% regression for some cases is still heavy....

[MichaReiser]

Current dependencies on/for this PR:

• main
  • PR Use unicode-width to determine line-length instead of character count #3714
    • PR perf(pycodestyle): Reduce allocations when computing logical lines #3715
      • PR perf(pycodestyle): Remove regex captures #3735 👈
        • PR perf(pycodestyle): Refactor checks to iterate over tokens insteadof text #3736
          • PR perf(pycodestyle): Introduce TokenKind #3745
            • PR perf(pycodestyle): Initialize Stylist from tokens #3757
          • PR Token Kind, remove text #3737

This comment was auto-generated by Graphite.

[github-actions]

PR Check Results

Ecosystem

✅ ecosystem check detected no changes.

Benchmark

Linux

group                                      main                                   pr
-----                                      ----                                   --
linter/all-rules/large/dataset.py          1.00     17.5±0.10ms     2.3 MB/sec    1.00     17.5±0.07ms     2.3 MB/sec
linter/all-rules/numpy/ctypeslib.py        1.02      4.6±0.04ms     3.6 MB/sec    1.00      4.5±0.02ms     3.7 MB/sec
linter/all-rules/numpy/globals.py          1.00    620.0±1.39µs     4.8 MB/sec    1.00    621.6±1.77µs     4.7 MB/sec
linter/all-rules/pydantic/types.py         1.01      7.8±0.07ms     3.3 MB/sec    1.00      7.7±0.05ms     3.3 MB/sec
linter/default-rules/large/dataset.py      1.01      9.2±0.05ms     4.4 MB/sec    1.00      9.2±0.06ms     4.4 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.00      2.1±0.01ms     8.1 MB/sec    1.00      2.1±0.01ms     8.1 MB/sec
linter/default-rules/numpy/globals.py      1.00    229.8±1.34µs    12.8 MB/sec    1.00    228.8±0.64µs    12.9 MB/sec
linter/default-rules/pydantic/types.py     1.00      4.3±0.02ms     5.9 MB/sec    1.00      4.3±0.04ms     5.9 MB/sec

Windows

group                                      main                                   pr
-----                                      ----                                   --
linter/all-rules/large/dataset.py          1.01     15.0±0.03ms     2.7 MB/sec    1.00     14.9±0.03ms     2.7 MB/sec
linter/all-rules/numpy/ctypeslib.py        1.01      4.1±0.02ms     4.1 MB/sec    1.00      4.0±0.02ms     4.1 MB/sec
linter/all-rules/numpy/globals.py          1.03    453.6±3.19µs     6.5 MB/sec    1.00    439.7±3.14µs     6.7 MB/sec
linter/all-rules/pydantic/types.py         1.01      6.7±0.02ms     3.8 MB/sec    1.00      6.6±0.02ms     3.9 MB/sec
linter/default-rules/large/dataset.py      1.04      8.3±0.03ms     4.9 MB/sec    1.00      8.0±0.03ms     5.1 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.02   1752.4±5.15µs     9.5 MB/sec    1.00   1722.7±8.36µs     9.7 MB/sec
linter/default-rules/numpy/globals.py      1.06    185.1±1.48µs    15.9 MB/sec    1.00    174.4±3.12µs    16.9 MB/sec
linter/default-rules/pydantic/types.py     1.03      3.8±0.02ms     6.7 MB/sec    1.00      3.7±0.02ms     6.9 MB/sec

[charliermarsh]

I trust you to resolve any necessary issues based on discussion around the deviations, but just tag me if you want eyes on anything as you follow up.

[MichaReiser]

I checked these numbers and they match pycodestyle

crates/ruff/resources/test/fixtures/pycodestyle/E27.py:28:2: E274 tab before keyword
crates/ruff/resources/test/fixtures/pycodestyle/E27.py:30:5: E274 tab before keyword

The other tests are fixed (very hacky solution but the next PR replaces it anyway)