`csv.Sniffer._guess_delimiter()` iterates over all ASCII on each line

[maurycy]

Bug report

Bug description:

I know that there were discussions about rewriting CSV Sniffer, and they seem to be on hold, to avoid breaking any code relying on its behavior.

I believe it's possible, though, to still make it significantly faster.

The current code iterates over 127 ASCII characters and counts their occurences on each line, even if they're not present:

cpython/Lib/csv.py

Line 382 in b36d23f

for char in ascii:

It's highly inefficient. We can count only present characters, and backfill zeros.

CPython versions tested on:

3.15

Operating systems tested on:

macOS

Linked PRs

• gh-137627: Make csv.Sniffer.sniff() 2x faster #137628