Re: [Haskell-cafe] Patterns for processing large but finite streams

This sound exactly like what attribute grammars, like the system developed at Utrecht University [1], are useful for. Erik [1] http://www.cs.uu.nl/wiki/HUT/AttributeGrammarSystem On Fri, Jul 1, 2011 at 10:54, Eugene Kirpichov wrote:

Alexey, your definition of "mean" does not look like "liftS2 (/) sum length" - you have to manually "fuse" these computations.

I'm asking for a formalism that does this fusion automatically (and guaranteedly).

2011/7/1 Alexey Khudyakov :

...

On Fri, Jul 1, 2011 at 12:21 PM, Eugene Kirpichov wrote:

...

I meant the average of the whole list - given a sumS and lengthS ("S" for "Stream"), write meanS as something like liftS2 (/) sumS lengthS.

Or is that possible with lazy lists too?

Sure you can. Sum, length and mean could be calculated as left fold. If you need to calculate more that one statistic at time you can combine accumulators

...

sum = foldl (+) 0 length = foldl (\n _ -> n+1) 0 data Mean Double Int mean = foldl (\(Mean m n) x -> Mean (m + (x - m) / fromIntegral (n+1)) (n+1)) (Mean 0 0)

AFAIU iteratees basically use same technique.

-- Eugene Kirpichov Principal Engineer, Mirantis Inc. http://www.mirantis.com/ Editor, http://fprog.ru/

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe