(Please don't top-post... and please don't quote all of the previous message unless you really need to It breaks the flow of conversation.) On 09/07/2015 01:43 AM, Tikhon Jelvis wrote:
+1 from me. At a high level, one of the things that attracts me to Haskell over other languages is that it's willing to change and improve at a faster rate (although, in absolute terms, it's still pretty slow and concerned about backwards compatibility). Assuming a tweak makes sense, rolling it out as an extension and then folding it into the language (maybe as part of the mythical Haskell 2020 standard :)) is perfect. DoAndIfThenElse is a reasonable point of comparison and seems to be a good deal.
Haskell keeps evolving at a brisk pace, see the depenently typed stuff that's coming up, etc. I question the value of constantly changing the *syntax*, which is mostly a triviality. The other syntactic extensions that we brought up saved a lot more than one character, btw!
And I *do* think this tweak makes sense. The new behavior is more inline with my expectations. I feel that do and lambdas are self-contained expressions and inherently group their contents together; needing an extra set of parentheses or a $ does not make sense. A good way to think about it is that I see do as having braces semantically even when they're syntactically optional. I think most people would agree that requiring parentheses aroud
foo (do { x; y; z })
is redundant and not useful.
So what would you write if foo were changed to take two arguments? foo (do { x; y; z }) (do { a; b; c }) See... no surprises! Regards,