j-jzk / klang-re Goto Github PK

The README is outdated. It should contain a quick reference for which syntax klang-re currently supports.

In Klang, NFAs are compared quite often, so it might be good to optimize it.

An approach that comes to my mind is to use something like the StateFactory used for FA states - a singleton that assigns the objects an incrementing ID. This might not be thread-safe, however (what if two NFAs are created at a same time?).

Another approach would be to simply remove the data attribute from the class, which would make it fall back to simple pointer comparison (I think).

It would probably be good to see if this doesn't affect existing code (tests, mostly), and maybe write some benchmarks.

Given multiple regexes, the library should be able to tell which ones match the start of a given string, as fast as possible.