search

goldfirere / units

The home of the units Haskell package
94 stars 19 forks source link

Swap `#` with `##`? (and same for `%`) #16

Closed goldfirere closed 10 years ago

goldfirere commented 10 years ago

Right now, ## is monomorphic while # is polymorphic. This goes against the general design of making the monomorphic things more accessible than polymorphic things, as non-experts will want the monomorphic ones.

(Suggested by João Cristóvão)

nushio3 commented 10 years ago

I agree that we make monomorphic things more accessible to the users, so that units will be more available to non-experts. Moreover, isn't that even better if polymorphic ones are equally as accessible as the monomorphic ones?

For example, the type of 1/3 is

:t 1/3
1/3 :: Fractional a => a

while print $ 1/3 appears to choose a specific instance for a, namely Double.

I'm wondering how it is achieved and if the mechanism can be applied to our current problem, so that # and % are polymorphic and they choose DefaultLCSU when case of ambiguity. Is there something like default type instance declaration mechanism to Haskell?

goldfirere commented 10 years ago

Five minutes of thinking has yielded no way to do this. The Haskell Report actually defines the defaulting behavior that you reference, and it's quite a limited facility. See here.

The closest I can come to emulating this behavior is with -XIncoherentIInstances somewhere, but I think that this extension, when used to get defaulting, brings a real danger of misbehavior.

I'd welcome a good answer here, but I think we're stuck with having two different term-level operators.

One possible way forward is to have Data.Metrology.Mono and Data.Metrology.Poly and then the operators could have the same name. Is that better than the current design? Perhaps. Perhaps not. If we went with this idea, I would want Data.Metrology to re-export the Mono bindings, like how Data.Metrology.SI re-exports Data.Metrology.SI.Mono.

nushio3 commented 10 years ago

Ok, then I'm agreed to swap # with ## and so on.

(I've made this decision much earlier, but I forgot to write here because the Shonan meeting is so exciting.)

nushio3 commented 10 years ago

Well Richard, do you want to do the implementation since I'll be in the meeting for some few days? I'll be back in June.

goldfirere commented 10 years ago

I'm happy to do this, but I'm favoring the Data.Metrology.Mono vs. Data.Metrology.Poly approach.

nushio3 commented 10 years ago

Ok! Then let's do the Data.Metrology.Mono & Data.Metrology.Poly approach.