David Owens II

@owensd It's rare, contextual, & never done outside of a protocol for me. [1]

@allonsykraken I don't find it rare at all. I often deal with collections of types and that requires type erasure a lot.

@allonsykraken so I really just end up coding against the protocol itself and the specific type is not really even needed to be known.

@owensd Right! I haven't needed to deal with a "collection of types" scenario yet but personally I find type erasure very useful sometimes.

@owensd Maybe I'm missing something here too. Sometimes I see convos indicating bad things about type erasure. I'm for it TBH

@allonsykraken I find it easier to forgo the additional type information early on so I don't have to deal with the boilerplate later.

@owensd True story. It's why I do it. I don't think it indicates anything when you have to keep using it. I could be wrong, but I believe...

@owensd it suggests that we badly need generalized existentials

@anandabits and once you have that, all you know again is that you have a Power from calling attack().

@owensd the amount of type info will depend on the constraints you provide. Right solution depends on your problem and your design of course

@anandabits @owensd Being only as precise as you need to be lets you stay flexible and avoid unnecessary refactoring.

@jckarter @owensd agree. And learning how to make the right tradeoffs takes time and experience.

@anandabits @owensd Absent other constraints, I think it's better to err on the side of more general types.