`csv.reader` calls the state machine for every character needlessly

Bug report

Bug description:

The state machine:

https://github.com/python/cpython/blob/bbcb75c986c47887e6c0757e63d59cd7af544f39/Modules/_csv.c#L726

is called for every character processed by csv.reader:

https://github.com/python/cpython/blob/bbcb75c986c47887e6c0757e63d59cd7af544f39/Modules/_csv.c#L969-L974

Even putting aside sophisticated SIMD or branching optimizations, it could be more efficient.

Most time is likely to be spent in a field (IN_FIELD, IN_QUOTED_FIELD). It's more efficient to find interesting characters (ie: escapes, quotes) and just copy the whole slice in between.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-138214

从 Modules/_csv.c 中第 726 行附近的状态机以及第 969-974 行附近的 csv.reader 处理开始。开始前先查看关联的 PR gh-138214，然后验证任何更改都能保持 csv.reader 的行为，同时避免不必要的逐字符状态机调用。