leven

Without this feature, we'll have to calculate all distances to find the closest. If we know all the options, we can skip unnecessary calculations. To support this, we'll need to implement #14 first.

This is the most performant Levenshtein implementation for NodeJS of which I know, and I have a though on how to make it faster for many applications. When testing many string pairs by similarity (for example, when sorting, or filtering strings above or below a threshold), often we are happy to short-circuit on any pair with a distance greater than . At scale, given that the least similar strings are also very expensive to compute, there is a big potential savings. (For one, just comparing length is enough to discard many candidates.) (For pairs with a distance above , the distance returned can be , or - I don't have a strong opinion on which is best, but probably is the most useful, as it would require not extra handling for sorting or filtering.)