Speed up _sre matching of runs of literal characters #161335

Feature or enhancement

Proposal

In SRE(match) (Modules/_sre/sre_lib.h), SRE_OP_LITERAL and SRE_OP_LITERAL_IGNORE match one character and then return to the opcode dispatch. A literal string in a pattern compiles to one LITERAL opcode per character, so "Content-Length: " costs 16 dispatches.

This proposes looping inside the LITERAL / LITERAL_IGNORE handler while the next opcode is the same literal opcode, so a run of N literal characters costs one dispatch instead of N. Matching behavior is unchanged. The extra read of the next opcode stays in bounds, because every compiled pattern ends with SUCCESS or FAILURE.

This is an internal speed-up only: no API or behavior change.

Benchmarks

pyperformance regex benchmarks, before vs after, compared with pyperf compare_to:

8 of 9 results are significant gains. The one slowdown (Intel regex_v8, -1.2%) is small.

Testing

• test_re passes with the change.
• Match results are identical before and after on 87,222 regex cases.

Implementation: #161332

Has this already been discussed elsewhere?

No response given

Links to previous discussion of this feature:

No response

Linked PRs

• gh-161335: Collapse runs of adjacent LITERAL opcodes into one _sre dispatch #161332