r? @ibraheemdev

rustbot has assigned @ibraheemdev.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

I think it is possible to share TwoWaySearcher, I suppose it would make most sense to have str use the slice code in this case?

Just a drive-by comment for now, but I believe there is and ought to be only one behavior difference, and it only applies when the needle is the empty string. In the str case, the empty string should only match at UTF-8 boundaries. But in the [u8] case, the empty string should match at every byte offset.

Apologies if you already knew this, but it's subtle and wanted to flag it. And also say that there shouldn't be any other behavior differences.

This is a lot of code that I'm not very familiar with so I'm going to pass this one on (feel free to take this one @BurntSushi if you are up for it). r? libs

Yeah makes sense, If there is a decent way of making this PR smaller we can look at that too.

Just a drive-by comment for now, but I believe there is and ought to be only one behavior difference, and it only applies when the needle is the empty string. In the str case, the empty string should only match at UTF-8 boundaries. But in the [u8] case, the empty string should match at every byte offset.

Yes, I hope I've found all the places where that matters. From what I can see, the TwoWaySearcher (a big part of the complexity of the pattern searching) does not actually handle UTF-8 boundaries. That is done in the StrSearcher.

So maybe a better first PR is to move TwoWaySearcher over to slice::byte_pattern, and use it from str::pattern?

I'll close this in favor of #135931 and other followups that break this change up into more manageable pieces