Re: [ghc-steering-committee] Proposal #638: Prefix form for MkSolo# (Recommend Accept)

I also agree that we should accept. We need some name for the unit unboxed tuple data constructor, and MkSolo# seems to fit with what we currently have. Simon's suggestion that we rethink the naming of the tuple type constructors seems to be a separate question. I think it warrants a new proposal/amendment if anyone feels strongly enough, rather than blocking this proposal, especially given that the original proposal's type names are already implemented. Adam On 14/03/2024 10:33, Matthías Páll Gissurarson wrote:

I agree with the sentiment here, having Type0 and Type1 as the canonical names would have been preferable in the original proposal. However, this amendment doesn't touch on that: it only changes the constructor.

We'd still want MkSolo# even if Solo was the synonym, due to the ambiguity described in the amendment. Renaming the canonical types would be a further, separate amendment to the original proposal.

I believe we should accept the amendment, and consider a separate amendment later.

On Tue, 12 Mar 2024 at 09:49, Simon Peyton Jones mailto:simon.peytonjones@gmail.com> wrote:

Unless I'm misreading, the proposal is only about the constructors' name. Which you don't propose to change, do you?

Yes. I was questioning the proposal itself rather than the amendment.

S

On Tue, 12 Mar 2024 at 09:43, Arnaud Spiwack mailto:arnaud.spiwack@tweag.io> wrote:

Unless I'm misreading, the proposal is only about the constructors' name. Which you don't propose to change, do you?

(that being said, I think I agree with your comment that the name of the type ought to have been `Tuple1`, it'd make more sense)

On Tue, 12 Mar 2024 at 10:38, Simon Peyton Jones mailto:simon.peytonjones@gmail.com> wrote:

Well this proposal deepens the commitment to an exception for Solo and Solo#.   But I'm not really objecting, just asking.

Simon

On Tue, 12 Mar 2024 at 09:34, Arnaud Spiwack mailto:arnaud.spiwack@tweag.io> wrote:

In favour.

Simon: I don't think your objection pertains to this particular proposal amendment, does it? Rather it's a further change to the original proposal that you'd like to see.

On Mon, 11 Mar 2024 at 11:48, Simon Peyton Jones mailto:simon.peytonjones@gmail.com> wrote:

Thanks Matthias

I'm generally supportive, but please see my comment exploring a minor alternative https://github.com/ghc-proposals/ghc-proposals/pull/638#issuecomment-1988147....

Simon

On Sat, 9 Mar 2024 at 00:12, Matthías Páll Gissurarson mailto:mpg@mpg.is> wrote:

Greetings committee!

In [proposal #638](https://github.com/ghc-proposals/ghc-proposals/pull/638 https://github.com/ghc-proposals/ghc-proposals/pull/638), @int-index proposes that we introduce a prefix form of MkSolo#, and apparent oversight in proposal #475 [Non-punning list and tuple syntax](https://github.com/ghc-proposals/ghc-proposals/pull/475 https://github.com/ghc-proposals/ghc-proposals/pull/475).

Previously, you would write `(# a #)` to construct a `Solo# a`. But the question is: what would be the prefix form of this constructor? It can't be `(# #)`, because this is already defined as a constructor of `Unit#`!

This amendment proposes the `MkSolo#` constructor, having us write `MkSolo# a` for the prefix form. The discussion seems unanimous, after care was taken to clarify that a fully applied `MkSolo# a` would still be pretty printed as `(# a #)`, avoiding programmer confusion.

It seems quite straightforward to me, so:

I recommend accepting this amendment to #475.

-- -- Matthías Páll Gissurarson http://mpg.is/

-- Adam Gundry, Haskell Consultant Well-Typed LLP, https://www.well-typed.com/ Registered in England & Wales, OC335890 27 Old Gloucester Street, London WC1N 3AX, England