countmap with comparisons other than ==

[ParadaCarleton]

Sometimes I'd like to build a countmap that compares using an operator other than == -- for instance, I might want to build a countmap using ≈ to decide whether two objects are (close enough) to equal, or maybe isseteq. It might be a good idea to add a third argument, defaulting to ==, that lets you do this.

[mschauer]

countmap is dictionary or Radix sort based, so there are no mutual comparisons with == either as far as I can see.

[nalimilan]

Grouping using ≈ is a very difficult and ill-defined problem as it's not a transitive operation. Grouping with == is also tricky due to things like NaN == NaN returning false, not to mention missing.

Anyway we would need a precise use case to evaluate whether something can be done about it.

[mschauer]

Maybe nice to point out, that grouping in approximately equal groups is a form of clustering and you can look at https://github.com/JuliaStats/Clustering.jl and perhaps try https://en.wikipedia.org/wiki/K-means_clustering