More optimisations to fuzzy matching: happy to contribute but need community advice about fuzzywuzzy

[josegonzalez]

Make the PR. If it works and passes tests, why wouldn't it be merged?

[dchaplinsky]

Mostly because of two reasons:

1. Memoization is not always a desirable behavior. If you have different data to compare each time, it would cost you memory to memoize results you won't be using anyway.
2. I applied my ideas only to one method and did that barbarically using microoptimisations here and there (for example, I eliminated call to ratio and pasted appropriate chunk of code directly to the method and that gave me like another 1.5x speedup). If you have an idea how to generalize that to other methods without doing copy-paste spaghetti style coding — I'm happy to.

Maybe I'll come up with some ideas myself, just need some time to think it over and fresh brain.

[josegonzalez]

Make memoization optional, off by default. Or provide alternative functions that are memoized.

If you can get a few of them working, even by copy-paste, we can try working through them to get it in decent shape :)

Closing for now since it's not actionable, but please feel free to open a pull request.