[Haskell-cafe] Lambda-case / lambda-if

Donn Cave donn at avvanta.com
Tue Oct 5 17:25:44 EDT 2010


Quoth Ketil Malde <ketil at malde.org>,
...
> Just that they seem to be natural generalizations.  If it's just the
> single form of paramtrizing the condition, I think it's better served by
> a regular function, 'bool' or (??) or whatever.

Well, yes, there's some logic to that.  Like,

bool b c a = if a
        then b
        else c

  getArgs >>= bool (putStrLn "long") (putStrLn "short") . (> 0) . length

And I agree that's competitive with lambda-if as I understand it -
though possibly not for the same reasons.

For me, Haskell is not Lisp.  Haskell's syntax takes a different direction,
a mix of S-expression with stuff like if-then-else, and it works.  If the
lambda-if feature is actually useful in a way that takes advantage of
the strength of the if-then-else notation, then I'm all for it.

The problem is that due to the rarity of True/False as ... terminal
value of a computation (I just made that up!), neither of these
constructs is going to be worth much.  Forget about lambda-if, even
the regular function looks like hell -

     bool (putStrLn "long") (putStrLn "short") . (> 0) . length

Compared to

     \ t -> if length t > 0 then putStrLn "long" else putStrLn "short"
 
... and much more so, with less trivial examples.

In a brief survey of my own very small code base, I see only "hIsEOF"
as a place where I could really use lambda-if.  There, it would be
vastly better than a regular bool function, but that's a pretty minimal
use case.

	Donn


More information about the Haskell-Cafe mailing list