Difference between revisions of "Worker wrapper"

From HaskellWiki
Jump to: navigation, search
(Adding link to the ww transformation page)
(Adding like to workerwrapper)
Line 53: Line 53:
==See also==
==See also==
* [[Avoiding parameter passing]]
* [[Avoiding parameter passing]]
* [[Worker Wrapper Transformation]]
* [http://www.workerwrapper.com The Worker Wrapper Transformation]

Revision as of 04:32, 8 May 2008

It is sometimes easier or more efficient to write functions which have particular "start arguments" or that pass state. When this is the case write wrappers rather than trying to code within the original signature.

Accumulator examples

e.g. the function reverse

reverse :: [a] -> [a]

could be written

reverse [] = []
reverse (x:xs) =  reverse xs ++ [x]

however this will be more efficient if it were written

reverse xs = revWorker [] xs

revWorker s [] = s
revWorker s (x:xs) = revWorker (x:s) xs

note often a worker will also return some state along with a result which can be stripped away.

revWorker could be considered a generally useful function. revWorker x y is equivalent to, but probably faster than, reverse y ++ x.

Other examples

Wrappers are often used to play the role of loop initialisation in imperative languages. For example:

fib n = fibWorker n 0 1

fibWorker n f1 f2
 | n == 0    = f1
 | otherwise = fibWorker (n-1) f2 (f1+f2)

Hiding the worker

Also, often one hides the worker(s) with a where (or let). E.g. :

fib = fibWorker 0 1
  fibWorker f0 f1 n
    | n == 0  = f0
    | True    = fibWorker f1 (f0 + f1) (n-1)

Of course, one then has to de-hide the worker if one want to test it with different initial arguments.

In some cases, though, the worker can be a generally useful function on its own merits, in that case one obviously shouldn't hide it.

One common case is where you want to subject the worker function to Unit testing. In such a situation, the test suite has to be able to get to the worker function. Another is where the worker might be a candidate for further abstraction. (See Higher order function for examples.)

Moving a worker to a higher level (which may require some Lambda lifting) is known as Let floating.

See also