Add an [exec_partial_detailed] function to re.

[mben-romdhane]

This is similar to exec_partial, but it allows users to see the groups that matched if a match exists, and if not, then it tells you an index before which a match cannot start.

Motivation

We had some code that was searching for a relatively small snippet within a large collection of text using regexes. This collection of text arrived in chunks, and the snippet could lie within the boundary of the chunks. We would only stop consuming chunks if we found the snippet. The most generic solution was to keep appending the chunks to a buffer, and redo the regex search from the beginning. However, that lead to $O(n^2)$ performance around larger inputs, and was resulting in significant slowdowns. Using this method, we could start the search at a later index, resulting in effectively $O(n)$ improvement.

Alternative solutions

We could export the current state of the match, but that seemed very complicated to fit into the existing code. The solution proposed here seemed to work well in practice.

Testing

Existing tests pass. Added additional tests for the new function.

[rgrinberg]

We don't call this function in every branch, but we allocate the closure eagerly. Can we hoist it up outside of this function?

[mben-romdhane]

Sorry for the delay, somehow I missed the notification for this, and I went back to check on it, and found that there was a response. I pushed a new version that hoists this function.