Re: [ghc-steering-committee] TypeApplications for Overloaded Literals (#129), Recommendation: reject

Hi, it seems this proposal does not get much support. I kinda like the high-level goal, but not enough to defend it vehemently. I am changing my recommendation to reject, to facilitate consensus. I should give those time to speak up who were in silent agreement with my initial recommendation, and if nothing happens, will reject this proposal in a week's time. Cheers, Joachim Am 8. Mai 2018 06:40:30 GMT-04:00 schrieb Simon Peyton Jones :

I’m on the fence:

* Benefit: slight * Cost: slight

But the spec is distressingly ad-hoc. Magical behaviour if

* RebindableSyntax is no * The ‘fromInteger’ in scope is not ‘Prelude.fromInteger’

So I lean to rejection. Unless some stronger motivations emerge.

Simon

From: ghc-steering-committee On Behalf Of Richard Eisenberg Sent: 06 May 2018 20:27 To: Iavor Diatchki Cc: ghc-steering-committee@haskell.org; Joachim Breitner Subject: Re: [ghc-steering-committee] TypeApplications for Overloaded Literals (#129), Recommendation: accept

I'm also unconvinced about this proposal. It's hard to specify, requiring that Prelude.fromInteger be magical. Note that if someone defined a separate Num class (under -XNoImplicitPrelude -XRebindableSyntax) that's fully equivalent to Prelude.Num, it wouldn't work. Also, Joachim's point about `:t 5` is a poor motivation: `:t` does not give you helpful information w.r.t. type applications. You need `:t +v`. With -fprint-explicit-foralls,

λ> :t +v 5 5 :: forall {p}. Num p => p

The appearance of the braces in the output means that 5 is not available for visible type application (the variable p is *inferred*). So if there's any unexpected behavior in the current implementation, that's due to ill-informed expectations, not misbehavior.

In the end, I would prefer to reject this proposal in favor of Joachim's casual proposal here: https://mail.haskell.org/pipermail/ghc-steering-committee/2018-May/000541.ht... (and expanded upon by my subsequent post).

Richard

On May 6, 2018, at 12:27 PM, Iavor Diatchki mailto:iavor.diatchki@gmail.com> wrote:

I am not convinced, if that's the sole motivation. By that reasoning you might expect that `(-5) @Int` or `(5+1) @Int` would also work, but they wouldn't.

On Sun, May 6, 2018 at 8:47 AM Joachim Breitner mailto:mail@joachim-breitner.de> wrote: Hi,

Am Sonntag, den 06.05.2018, 15:42 +0000 schrieb Iavor Diatchki:

...

Is there a reason why we should write `5 @Int` as opposed to `5 :: Int`?

Because we can. Or rather, we should can, because

Prelude> :t 5 5 :: Num t => t Prelude> :t 5 @Double <interactive>:1:1: error: • Cannot apply expression of type ‘t0’ to a visible type argument ‘Double’ • In the expression: 5 @Double

is confusing and inconsistent with the user’s expectation after learning about when they can use ExplicitTypeApplications (unless one knows about the syntactic sugar involved.)

...

Also, there appears to be a typo in the spec, the part which specifies translations for integers (1 turned into 0?)

thanks, fixed.

(BTW, all of you are owners of the repository and should have the necessary permissions to edit pull requests directly.)

Cheers, Joachim

-- Joachim Breitner mail@joachim-breitner.demailto:mail@joachim-breitner.de http://www.joachim-breitner.de/ _______________________________________________ ghc-steering-committee mailing list ghc-steering-committee@haskell.orgmailto:ghc-steering-committee@haskell.org https://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-steering-committee _______________________________________________ ghc-steering-committee mailing list ghc-steering-committee@haskell.orgmailto:ghc-steering-committee@haskell.org https://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-steering-committee