On Mon, Feb 21, 2005 at 10:32:15PM -0800, Jonathan Lang wrote: : ...then you've got the notion of Fuzzy Logic Sets, where the key would be : the prospective element and the value would be the degree of membership. : For fuzzy sets, hashes seem to be a better fit than junctions, which have : no obvious means of supplying the degree-of-membership info. You're still : stuck with the "keys must be strings" problem, so it's not an ideal : solution; but it's better than nothing.
Hash keys are only strings by default. Just as any array can have a shape, so can any hash, and the shape of the hash is determined by the subtypes of its dimensional indices. (See S9 for more on this.) But using values for degree of membership is an interesting idea. On the other hand, if we ever have numeric datatypes with built-in error bars, those could generalize to keys with an associated certainty of some sort, and then you can leave the values out of it. In which case a junction might have varying degrees of likelihood varying between noneness and allness. It's all too powerful for a Pooh like me, though... :-) Larry
