Case insensitive match pattern

rushmorem · 2018-07-24T16:20:15.230Z

I’m wondering whether it’s feasible to add a special pattern that can be used with match in a case insensitive way without allocating. One horrible way to do it outside the compiler would be to generate all possible cases of a string at compile time but this generates so many possibilities making this infeasible.

I got this idea while working on optimising the psl crate’s performance. Assuming input is in lowercase already, List::domain can process 10 million domains in less than 3 seconds. However, if I lowercase the input first to do a case insensitive match, it takes about 13 seconds.

mbrubeck · 2018-07-24T16:28:46.717Z

Servo uses unicase for ASCII-case-insensitive comparison, and caseless for comparison with Unicode case folding and normalization. Both work without allocating new strings.

Update: I see that unicase now supports Unicode case folding too. (It used to support only ASCII case.)

rushmorem · 2018-07-24T16:41:00.392Z

Thanks. Can caseless be compiled with no_std? Currently psl is no_std.

MajorBreakfast · 2018-07-24T17:58:33.152Z

@rushmorem I just browsed the code: No it does not support no_std. Its dependency unicode-normalization uses a std::collections::VecDeque somewhere.

rushmorem · 2018-07-24T18:04:04.454Z

@MajorBreakfast Thanks. I was planning to check the source code later. You have saved me a bit of time

mbrubeck · 2018-07-24T18:21:25.960Z

I submitted a PR to allow caseless to work with no_std, minus the functions that require unicode_normalization:

github.com/unicode-rs/rust-caseless

Support no_std builds

by mbrubeck on 06:19PM - 24 Jul 18

2 commits changed 4 files with 31 additions and 3 deletions.

rushmorem · 2018-07-25T12:22:02.148Z

Cool, thanks!

system · 2019-03-25T08:30:35.790Z

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.