Improve responsiveness when invoked via Python

[ofek]

This saves a handful of milliseconds on Windows and even more on other platforms when running python -m ruff. On non-Windows systems the process is replaced directly (impossible on Windows unfortunately).

❯ docker run --rm python:3.11 bash -c "for i in {1..15}; do python -m timeit -n 1 -r 1 'from pathlib import Path'; done"
1 loop, best of 1: 25.7 msec per loop
1 loop, best of 1: 3.07 msec per loop
1 loop, best of 1: 3.16 msec per loop
1 loop, best of 1: 3.06 msec per loop
1 loop, best of 1: 3.32 msec per loop
1 loop, best of 1: 3.93 msec per loop
1 loop, best of 1: 3.26 msec per loop
1 loop, best of 1: 3.73 msec per loop
1 loop, best of 1: 3.1 msec per loop
1 loop, best of 1: 3.29 msec per loop
1 loop, best of 1: 3.12 msec per loop
1 loop, best of 1: 3.05 msec per loop
1 loop, best of 1: 3.04 msec per loop
1 loop, best of 1: 3.19 msec per loop
1 loop, best of 1: 3.04 msec per loop

❯ docker run --rm python:3.11 bash -c "for i in {1..15}; do python -m timeit -n 1 -r 1 'import subprocess'; done"
1 loop, best of 1: 31.2 msec per loop
1 loop, best of 1: 3.75 msec per loop
1 loop, best of 1: 4.71 msec per loop
1 loop, best of 1: 3.88 msec per loop
1 loop, best of 1: 4.08 msec per loop
1 loop, best of 1: 4.35 msec per loop
1 loop, best of 1: 3.94 msec per loop
1 loop, best of 1: 4.06 msec per loop
1 loop, best of 1: 3.88 msec per loop
1 loop, best of 1: 3.85 msec per loop
1 loop, best of 1: 3.84 msec per loop
1 loop, best of 1: 4.01 msec per loop
1 loop, best of 1: 4.21 msec per loop
1 loop, best of 1: 4.07 msec per loop
1 loop, best of 1: 4.11 msec per loop

❯ python -m timeit -n 1 -r 1 "from pathlib import Path"
1 loop, best of 1: 5.25 msec per loop

❯ python -m timeit -n 1 -r 1 "import subprocess"
1 loop, best of 1: 7.61 msec per loop

[zanieb]

Hey :)

Can you show the performance changes for invoking Ruff itself rather than microbenchmarks of the import? It's hard to tell if these changes are worthwhile in our use-case. I generally prefer the readability of undeferred imports and use of Path when it's not critical.

[ofek]

❯ hyperfine -m 10 --warmup 1 "python -m ruff --help"
Benchmark 1: python -m ruff --help
  Time (mean ± σ):      78.3 ms ±   1.8 ms    [User: 0.0 ms, System: 0.0 ms]
  Range (min … max):    75.7 ms …  83.3 ms    34 runs

❯ hyperfine -m 10 --warmup 1 "python -m ruff --help"
Benchmark 1: python -m ruff --help
  Time (mean ± σ):      73.4 ms ±   1.0 ms    [User: 0.0 ms, System: 1.2 ms]
  Range (min … max):    71.7 ms …  75.7 ms    36 runs

I can't easily test right now in Docker, these are on my Windows machine.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

Formatter (stable)

✅ ecosystem check detected no format changes.

Formatter (preview)

✅ ecosystem check detected no format changes.

[charliermarsh]

This seems reasonable to me.