Improve performance of http.cookiejar.join_header_words

[picnixz]

Feature or enhancement

Proposal:

Currently, we use if not re.search(r"^\w+$", v):. This can be replaced by if not v.isalnum() and '_' not in v. Benchmarks are as follows:

$ ./python -m pyperf timeit -s 'x = [[("text/plain", None), ("charset", "iso-8859-1\"")]]; from http.cookiejar import join_header_words as f' 'f(x)'

Mean +- std dev: [ref] 795 ns +- 8 ns -> [new] 604 ns +- 20 ns: 1.32x faster

It doesn't help having a compiled regex by the way.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

No response

Linked PRs

• gh-128982: Substitute regular expression in http.cookiejar.join_header_words for an efficient alternative #128983
• gh-128982: Revert "#128982: Substitute regular expression in http.cookiejar.join_header_words for an efficient alternative (GH-128983)" and add tests #130584

[encukou]

Thank you!

[serhiy-storchaka]

This is incorrect change. For example, it produces incorrect results for "foo;_" and "foo_bar".

[encukou]

Oof, thanks for noticing.
I'll revert and add a test.

[serhiy-storchaka]

\w here looks arbitrary. And it is. There are other red flags. See #130631.

[picnixz]

I'll close this one and we'll move to 130631