Add generic deriving

[carymrobbins]

No description provided.

[paf31]

👍

[garyb]

@paf31 Would you say this is a breaking change?

I mean, it isn't, but the new version won't work in older version of psc.

[paf31]

I think it should be a conscious decision to upgrade, and a major bump is the only way we have to ensure that, so yeah.

[garyb]

In that case:

• purescript-quickcheck
• purescript-maps
• purescript-catenable-lists
• purescript-free
• purescript-sets
• purescript-graphs
• purescript-machines
• purescript-strongcheck
• purescript-argonaut-core
• purescript-argonaut-codecs
• purescript-argonaut-traversals
• purescript-argonaut

I'm not sure how comprehensive this tool is that I use to generate these, there must be some kind of bug in it as when I was doing the strings update it didn't list everything. I really should rewrite it in PS and get it doing all that other stuff we talked about...

[garyb]

I'd suggest we add more derived generic instances and release them all as a batch if we're going down this route though, to prevent having to bump a bunch more times.

[paf31]

Agreed, but what else needs a generic instance? I don't know if one for Map is appropriate for example. Anything with a Generic instance necessarily leaks its internal representation.

[garyb]

Good point, perhaps there isn't anything else that would need it :) Let's just be sure before we commit!

[paf31]

We have various newtypes which might benefit from it. In purescipt-maybe for example.

[paf31]

Although of course that module can't depend on generics 😄

[garyb]

Good point, monoid has a bunch more too.

[carymrobbins]

So, in regards to Map, we could just implement a Generic instance that doesn't expose its internal representation. It could defer to fromList, see this commit - carymrobbins/purescript-maps@5f8c41e

[paf31]

But then, e.g. gShow would be show on lists, which might be a bit odd.

[carymrobbins]

That's how it basically is now, except for the qualified names and the extra, yet proper, parens -

> singleton "foo" 1
fromList Cons (Tuple ("foo") (1)) (Nil)

-- With Generic (Map k v)
> data W = W (Map String Int)
> derive instance genericW :: Generic W
> instance showW :: Show W where show = gShow
> W (singleton "foo" 1)
W (Data.Map.Map.fromList (Cons (Data.Tuple.Tuple "foo" 1) Nil))

[paf31]

@garyb Do we want to include this in 1.0? It seems harmless enough to me.

[garyb]

Yeah, I think so!