On Fri, Apr 3, 2009 at 9:30 PM, Heinrich Apfelmus

wrote:

Zachary Turner wrote:

...

I thought I would try to see if it were possible to write, in point-free form using no lambda functions and no intermediate functions, a function that takes 2 lists of Booleans, computes the pairwise logical AND of the two lists, and returns a list containing the 0 based indices of the elements where the logical and of the two was true. I know that at some point it becomes overkill and for the sake of readability one should know when to draw the line. So I want to see if someone with more experience than me can comment on whether or not this is over the line :P

trueIndices = curry $ map fst . filter snd . zip [0..] . map (uncurry (&&)) .. (uncurry zip)

So do all the uncurries and curries make it too hard to understand or is it pretty easy to read this? For me it takes me a while to figure out by looking at it because it's hard to trace all the currying and uncurrying. And is there a more elegant solution?

Looks very readable to me, though I'd write it as

trueIndices = (map fst . filter snd . zip [0..] .) . zipWith (&&)

or even simply as

trueIndices xs ys = map fst . filter snd . zip [0..] $ zipWith (&&) xs ys

because composing functions with more than one argument tends to be a bit messy.

With Conal's semantic editor combinators

http://conal.net/blog/posts/semantic-editor-combinators/

it would be written as

trueIndices = (result . result) (map fst . filter snd . zip [0..]) (zipWith (&&))

That was a pretty interesting blog post, and easily understandable which is always nice. Thanks for the link. I also had never even used the zipWith function, so thanks for pointing out that equivalence.

Chaddaï Fouché wrote:

On Sat, Apr 4, 2009 at 5:20 AM, Zachary Turner wrote:

...

...

With Conal's semantic editor combinators

http://conal.net/blog/posts/semantic-editor-combinators/

it would be written as

trueIndices = (result . result) (map fst . filter snd . zip [0..]) (zipWith (&&))

That was a pretty interesting blog post, and easily understandable which is always nice. Thanks for the link. I also had never even used the zipWith function, so thanks for pointing out that equivalence.

I'm not sure you really want to write this pointfree... Generally I tend to avoid it when there's two arguments of identical standing. Also the (map fst . filter snd...) is a Data.List function (findIndices) :

trueIndices xs ys = findindices id $ zipWith (&&) xs ys

I love the pointfree style and it might be a good exercise to try and transcribe any function to pointfree, but sometimes pointful is just clearer.

Or to put it differently, sometimes "pointfree is pointless" ;-) /M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe Haskell is an even 'redder' pill than Lisp or Scheme. -- PaulPotts