There's a couple of tricks I want to try.


One, is generic behaviours that dispatch to specific implementations based on the atom tag of a data type. Should work nicely for monoids and other simple stuff.


Another, is taking the idea a step further by creating several alternate arithmetic operator implementations. I'm thinking rationals, complex, and maybe polynomials for laughs. The first one will be useful, at least...

The second idea is driven by LFE letting you override operators really easy. You can import them from a specific module, or do a local override (with or without a macro).

I'm curious how far this idea can be taken. Can you write a histogram that works on any type that's a field - and then make that histogram a field itself, and construct a histogram of histograms?



If the code works well, it'll play a significant part in an upcoming metrics library.