add b->a instance for Alignable

[cscheid]

I'm keeping the failed attempt to get an instance via defaultBoundary as comments, mostly so you can see what I was trying to get at. I don't think something like that will work easily, but with this pull request, diagrams-lib emits warnings during compilation, which is bad.

Ideas?

[byorgey]

Thanks, I'll look at this more carefully soon. I have to page back in the way Alignable works first.

[byorgey]

Aha, I think I understand a bit better now. All the other "derived" Alignable instances (for [a], Set a, Map a, etc.) are for finite containers, and work by aligning to the combined boundary of all the contained Alignable things. This can't work for functions, however, since a function may be an "infinite" container (if its domain is infinite). Instead of aligning to the combined boundary, we just want to align each output individually. So there is no sensible defaultBoundary for functions. We can probably just implement it as \_ _ -> origin, with an explanatory note; since we don't use it for implementing alignBy it doesn't really matter (unless you try to do something like, say, align Sets of functions... which would simply have no effect). The given implementation of alignBy looks perfect.

[cscheid]

Ok, see if that's good enough - it at least compiles cleanly for me.

[byorgey]

Looks good -- thanks!