On Saturday 29 January 2011 16:36:09, Tim Baumgartner wrote:
Daniel,
thanks for your explanations! For some time, I thaught laziness might be some kind of holy grail.
There is no such thing. Laziness makes it possible to write many algorithms in a concise and elegant manner which can't be so expressed in a strict language. But on the other hand, laziness makes it harder to predict performance and memory usage (experience helps a lot with that).
But now I realize that it takes both a lot of time to understand properly and that it has more limitations than I expected. Another example for this: I wrote a function
recursiveContents :: FilePath -> IO [FilePath]
and later on I thaught it was possible to run some actions on every file from the list, in a lazy fashion. Then I had to find out that all files had to be found before I could process the first one.
That depends. Often you can use unsafeInterleaveIO (which is not nearly as unsafe as unsafePerformIO) to get the list lazily even in IO. Then, if your consumption pattern allows it, you can process the first file before the second is found (if, however, processing the files can influence the result of recursiveContents, you can see why there is an unsafe in unsafeInterleaveIO).
By now, I think I understand why, but it stops me from being too enthusiastic about laziness
Well, as with most things, there are advantages and disadvantages.
:-(
Regards Tim