Donald Bruce Stewart wrote:
Ha! I don't think "pure lazy fp" would be the phrase I would choose to describe this code.
An example (from fannkuch):
t <- IO $ \s -> case readIntOffAddr# p1# 0# s of (# s, p1 #) -> case readIntOffAddr# p1# (n# -# 1#) s of (# s, pn #) -> (# s, not (p1 ==# 0# || pn ==# (n# -# 1#)) #)
Ok, I'll bite ;) This is a very atypical "example", as fannkuch is the only entry written this way (and it could be rewritten, with careful attention to the Core). There are also many lovely pure, lazy entries.
Not so atypical. More examples (I didn't look /that/ hard.. :)) ----------------------------------- * From reverse-complement reverseit iubc strand i l s = if i >=# l then (# s, (I# i, I# l) #) else case readWord8OffAddr# strand i s of { (# s, c #) -> case readWord8OffAddr# strand l s of { (# s, x #) -> case readWord8OffAddr# iubc (word2Int# x) s of { (# s, y#) [snip] * From k-nucleotide. eqmem i ptr1 ptr2 s = if i ==# 0# then (# s , True #) else case readInt8OffAddr# ptr1 0# s of { (# s, i8a #) -> case readInt8OffAddr# ptr2 0# s of { (# s, i8b #) -> if i8a ==# i8b then eqmem (i -# 1#) (plusAddr# ptr1 1#) (plusAddr# ptr2 1#) s else (# s, False #) } } * From n-body. kineticE i = let i' = (.|. (shiftL i 3)) in do m <- mass i vx <- unsafeRead b (i' vx) vy <- unsafeRead b (i' vy) vz <- unsafeRead b (i' vz) return $! 0.5 * m * (vx*vx + vy*vy + vz*vz) ---------------------- *I* am certainly not implying that Haskell is anything less than the most wonderous language in the entire world. I'm saying that there's a stark difference in style between the programs submitted to the shootout, and the ones I would show to people that I myself was trying to introduce to the wonders of purely lazy functional programming. :). I think there's a big-fat-lesson about the tension between abstraction and implementation in these entries. On one hand we've got "This is what I want", on the other it's "What do I have to do to implement it". Ben.