On Thu, Jun 05, 2003 at 10:03:55AM +0100, Graham Klyne wrote:
At 19:50 04/06/03 -0400, Derek Elkins wrote:
You (Graham) also have some parentheses issues; e.g. in foo ++ (combinations 5 l) the parentheses are superfluous. I'm tempted to argue that being superfluous doesn't mean they shouldn't be there.
Keith already made one argument for less parentheses, here's my take on the subject: Given that: (++) :: [a] -> [a] -> [a] -- contatenates lists foo, l :: [a] -- are lists bar :: Int -> [a] -> [a] -- produces a list Which of the following make sense?
(foo ++) bar 5 l -- No, because bar isn't a list, (foo ++ bar) 5 l -- and 5 and l would be applied to a list, (foo ++ bar 5) l -- (as opposed to a function,) which make no sense
foo (++ bar 5 l) -- Can't apply a function to a value, -- though (++ bar 5 l) foo has the same effect as what we intended
foo ++ (bar 5 l) -- which is just foo ++ bar 5 l
Because of the static type checking that takes place, you can't easily (not unless you were _trying_ ;) produce an ambiguous expression such that the removal of brackets keeps it well-typed, yet is not equivalent to the original. So as opposed to the C code where you (and I) would put in extra brackets `just to be sure', I wouldn't bother with them unless I know the expression's going to be ambiguous. (or if the compiler tells me so. ;-) (I suppose you could argue it's not necessarily obvious that foo is a list and bar is a 2-ary function of an Int and a list. My response would be to rename foo and bar so that this is the case. ;-)
I'd just about figured the ShowS idea, but I've yet to get a handle on this idea of [a] 'monoid'.
Might http://www.engr.mun.ca/~theo/Misc/haskell_and_monads.htm be of any help? later, /Liyang -- who managed to sneak into Category Theory lectures, but still has no idea what a monad is. ^_^; -- .--| Liyang HU |--| http://nerv.cx/ |--| Caius@Cam |--| ICQ: 39391385 |--. | :::::::::::::::::::::: This is not a signature. :::::::::::::::::::::: |