On Thu, Jan 10, 2008 at 10:10:21AM -0800, Don Stewart wrote:
rendel:
jerzy.karczmarczuk@info.unicaen.fr wrote:
Although it could be argued that laziness is the cause of some very obscure bugs... <g> Niko
Example, PLEASE.
Prelude> sum [1..1000000] *** Exception: stack overflow
Prelude> Data.List.foldl' (+) 0 [1..1000000] 500000500000
See, http://hackage.haskell.org/trac/ghc/ticket/1997
Strictness for for atomic numeric types is inconsitently applied across the base library. Fixing the inconsitencies would let fix a range of similar issues.
Having followed this discussion, I agree with your analysis, but also think that rendel has chosen a good example of a pretty obscure bug caused by laziness in code written by folks who are actually decent Haskell programmers. It's unfortunate that there's no way to include strictness behavior in function types (at least that I'm aware of), so one has to rely on possibly-undocumented (and certainly never checked by a compiler) strictness behavior of many functions in order to write truly correct code. I wish there were a nice way around this issue (but can't really even imagine one). -- David Roundy Department of Physics Oregon State University