Re: lists of pairs

*To*: mathgroup at smc.vnet.net*Subject*: [mg8545] Re: [mg8511] lists of pairs*From*: Russell Towle <rustybel at foothill.net>*Date*: Sat, 6 Sep 1997 23:16:13 -0400*Sender*: owner-wri-mathgroup at wolfram.com

Hi all, With reference to Jens' comparative Timings for two methods of discarding "reversals" from a list of pairs of integers, I have done some more tests, using larger lists of pairs. As mentioned before, this problem has to do with obtaining a list of the edges of a polyhedron (as indices into a list of vertices), with duplicates dropped. In my test, instead of the edges of a pentagonal dodecahedron, where the initial list contained 60 pairs, and the desired list has but 30 pairs, I used the edges of an n=12 polar zonohedron, where the initial list has 528 pairs of integers, and the desired list, 264 pairs. In practice, I shall be working at times with much more complicated polyhedra. Let the initial list of pairs be called "edges." I will list the timings for (1), my bad method, (2) Jens' method, and (3) the method proposed by C. Woll and others, which differs from the first two in that, rather than returning indices into "edges," it returns the desired subset of "edges." 1. 573 seconds (my method). 2. 298 seconds (Jens' method). 3. .0833 seconds (Woll's method). The drastic improvement of (3) over (1) and (2) may be ascribed to not having to use Position[] at all, I guess. Russell Towle Giant Gap Press: books on California history, digital topographic maps P.O. Box 141 Dutch Flat, California 95714 ------------------------------ Voice: (916) 389-2872 e-mail: rustybel at foothill.net ------------------------------