Infix operators in Hamlet: yes or no?

[Michael Snoyman]

We discussed this a bit last week, but let's get it out in the open:
do we want infix operators in Hamlet?

I'm strongly leaning towards taking them all out (sans $, as Max mentioned).

Relevant Github issue: https://github.com/yesodweb/yesod/issues/278

[Luite Stegeman]

I personally like to have them, but only if they actually work in all
situations (or at least when there's a very clear distinction between
what's supported an what's not). So if the current implementation is
giving problems, I'd say remove support until there's something that
actually does work.

luite

[Vagif Verdi]

If it is an implementation problem, then it is understandable.
Otherwise i do not see the reason to treat infix functions differently.

[Michael Snoyman]

That's a funny comment: the whole feature is to tree infix operators
differently! That's at the crux of the matter: is this an extra
difference that should be reflected in Hamlet? If we're following the
dumb language theory, it might be overkill.

[Vagif Verdi]

Perhaps i missed the relevant discussion. Could you please point me to it so i can see what are the reasons to treat infix operations differently.

On Wednesday, March 28, 2012 9:34:39 AM UTC-7, Michael Snoyman wrote:

That's a funny comment: the whole feature is to tree infix operators
differently! That's at the crux of the matter: is this an extra
difference that should be reflected in Hamlet? If we're following the
dumb language theory, it might be overkill.

[Michael Snoyman]

Sure: https://groups.google.com/forum/#!topic/yesodweb/LstCHKG84_Q/discussion

[Felipe Almeida Lessa]

On Wed, Mar 28, 2012 at 6:37 AM, Michael Snoyman <mic...@snoyman.com> wrote:
> I'm strongly leaning towards taking them all out (sans $, as Max mentioned).

+1 for taking them out.

--
Felipe.

[Max Cantor]

There seems to be people who think that Hamlet should stick with the x . y . z $ v style convention. Maybe we include (.) and ($) and just treat them interchangeably in hamlet. its kind of a hack but might appease them

max

[mifrai]

I agree with Luite. They're nice to have, but if they're causing problems - I'd rather see them go than cause more confusion.

But, just to be clear, if I wanted to use an infix operator - would I still be able to do the following?

  $if (==) a b

    etc

On Wednesday, March 28, 2012 2:37:16 AM UTC-7, Michael Snoyman wrote:

On Wednesday, March 28, 2012 2:37:16 AM UTC-7, Michael Snoyman wrote:

On Wednesday, March 28, 2012 2:37:16 AM UTC-7, Michael Snoyman wrote:

[Max Cantor]

This is an important point, but should be easy to support. We just need to make sure we parse (..) as a normal function..

Max

[mifrai]

I'm not sure I like this. While it would work in that specific case - it can lead to a lot of confusion if anybody tries to use it for anything else. For example, in lenses or for actual function composition:

$for x <- map (a . b) $ getL (la . lb) c

  #{x}

While it's arguable whether you should be doing something like that in the templates, I would say that treating (.) as ($) makes hamlet context switching more difficult than I would feel comfortable with.

[Patrick Hurst]

I'm in favor of keeping them; it would be absurd IMO if you had to introduce a name for the function (* 5) in the surrounding context just so that you could multiply things by five.

Could you use haskell-src-exts for doing the parsing and then just evaluate the AST?

[Greg Weber]

we don't want a haskell-src-extras dependency. A new version was just
released that should work well with 7.4. Historically it has lagged
behind GHC releases. If someone wanted to create a haskell-src-extras
version of Haskell we are very open to that possibility, but we don't
think we can make it the default.

[Michael Snoyman]

I've removed infix operators from master:
https://github.com/yesodweb/shakespeare/commit/40b6496ace781aabbefaa470f10a7ddab283b5f6

Michael

[Patrick Hurst]

Is $ support still a planned feature?

On Wednesday, March 28, 2012 at 11:31 PM, Michael Snoyman wrote:

[Michael Snoyman]

Yes, it's already there. $ was never actually including as part of
infix operators, but as syntactic sugar of Shakespeare. You can see
here[1] that $ was explicitly excluded from the infix parsing.

[1] https://github.com/yesodweb/shakespeare/blob/40b6496ace781aabbefaa470f10a7ddab283b5f6/shakespeare/Text/Shakespeare/Base.hs#L107