@the-mikedavis Please consider having this functionality as string:distance/3 with the first argument as DistanceType :: jaro with the potential for adding the atoms jaro_winkler | levenshtein | damerau_levenshtein in the future (as described here) and the return value as an integer for the number of edits or characters in common. A second function string:distance_normalized/3 could be added for the [0.0 .. 1.0] return value. Providing people with a choice of the algorithm can relate to the expected string contents and the expected algorithm latency.

@okeuday I don't like that, they have very different properties and return values, (I have implemented most of them).
Also the Levenstein algorithms are 10 times slower then jaro, so I don't really see the need of them.

For what would you use the Levenstein algorithm when jaro is 10 times faster?

@dgud Applications of the Levenshtein algorithm are described here and most usages are likely uncommon in Erlang, but a shell feature that could provide help based on module or function spelling mistakes could be an Erlang use-case. If the strings are relatively short the latency may be justifiable with getting a better result. Not attempting to advocate for a slower shell. It just seems best to care about more than 1 string distance algorithm.

> a shell feature that could provide help based on module or function spelling mistakes could be an Erlang use-case.

Which is one reason why we want to add jaro, and why we considered the other ones.
I'm still not convinced, that a Levenshtein variant is significantly better at that.

@the-mikedavis We have another variant of the code coming, we have concluded that the elixir variant calculates the transpositions "wrongly", at least it's different than the defacto standard, for example:
jaro_distance("Sunday", "Saturday") gives another answer than the algorithm on rosetta code.

The original paper also seems to agree more with rosetta algorithms, then elixirs, though the original code
does an integer division of 2 which would give the same result as elixirs of the example above.

But I'll keep this PR open for inspiration, i.e. test changes and docs.

Thanks, for the work.
Closed this in favor of #7879