Hey! I never added `DateTime` support for a variety of reasons. They can be summed up to a) no need to pull in `std/times` b) the current code is a bit annoying to extend for more types and c) I never needed it. Unix timestamps as integers & formatted as strings (to e.g. group by a bunch of dates) works perfectly fine in my use cases.
In any case, for very specific types of operations, I would then always recommend to use custom code and not rely on a DF library. Nim code is fast after all, unlike Python where a DF library provides not only convenience, but also massive performance gains. Regarding point b) from above: I've tried multiple approaches at making it easier to extend Datamancer for new types. One runs into the "problem of runtime time ⇔ compile time interfacing" in one way or another, given that a DF library is usually inherently designed to not require manually annotating types for every single operation. The last real approach kind of worked, but ended up as being too difficult to work with in terms of development and has some ugly warts when using it. Your post made me consider the problem again. I'm toying around with something at the moment, but no guarantees it will lead somewhere.
