semigroup rant again
Dec. 17th, 2011 10:48 amI got an impression that the traditional, distributed, map/reduce, while having its background in FP and, at the first glance, easily modeled by map and foldl, actually does not seem to have a mathematical foundation... or does it? I could not find any.
A simple example, frequency count. What exactly do we produce? Or, in general, having a collection of (key,value), where value is an element of a monoid M, what exactly do we produce by folding? Not a map K -> M?
A simple example, frequency count. What exactly do we produce? Or, in general, having a collection of (key,value), where value is an element of a monoid M, what exactly do we produce by folding? Not a map K -> M?