we could add some tests as well, just in case?

Happy to, do you mean for _cached_regex, or tests for TextWrapper, or?

A

I think tests with accessing / resetting / deleting the attributes (they are publicly named even though they are not exposed, just in case someone is subclassing them [again they shouldn't be part of the public API but well.. you never know; and it would be good to just check that the patterns are the expected ones.

But yes, also a small test for the descriptor with a fake pattern and check that its implementation is correct (well it should be AFAIK).

This change is an optimization, you should provide a benchmark to prove that the change is worth it.

@vstinner sorry for not including them. Benchmarks show a consistent large decrease in import time, around 10ms on a standard release build in Windows. See below for detailed numbers:

Using -X importtime -Sc 'import textwrap'

Old: import time: 907 | 10796 | textwrap_current
New: import time: 311 | 311 | textwrap_new

The same command, on the default PCbuild\build.bat build:

Old: import time: 5272 | 29988 | textwrap_current
New: import time: 1076 | 1076 | textwrap_new

With hyperfine, Python 3.13.2:

PS>  hyperfine -N --warmup 5 "python -c ''" "python -c 'import textwrap_current'" "python -c 'import textwrap_new'"
Benchmark 1: python -c ''
  Time (mean ± σ):      24.1 ms ±   0.9 ms    [User: 5.4 ms, System: 3.4 ms]
  Range (min … max):    22.3 ms …  26.4 ms    124 runs

Benchmark 2: python -c 'import textwrap_current'
  Time (mean ± σ):      34.6 ms ±   1.5 ms    [User: 8.3 ms, System: 4.0 ms]
  Range (min … max):    29.5 ms …  38.1 ms    85 runs

Benchmark 3: python -c 'import textwrap_new'
  Time (mean ± σ):      24.2 ms ±   0.9 ms    [User: 8.3 ms, System: 3.7 ms]
  Range (min … max):    21.8 ms …  27.0 ms    124 runs

A

All commit authors signed the Contributor License Agreement.

This PR is stale because it has been open for 30 days with no activity.

Can this change be made simpler by using lazy import re?

Can this change be made simpler by using lazy import re?

I don't think so as the class body of the textwrap class will be executed at import time, so re will be reified directly.