Hi bench mark watchers! Statz sez: > If we are doing a benchmark, that's fine. I'll go > with you on that. If we are trying to produce a GUI > app, that's a different issue with literally > thousands of other items thrown on the table. It's > like trying to hoe your vegetable garden with a > forklift and a bowl of Chile con carne. Bold words and true! I buy all that. HOWEVER we still need a self-contained application --- one which is essentially all muscle and no fat. A great benchmark, a Statzmark let us say. The application is called "WordCensus". When one drops a TEXT file called "myfile" on the smart "WordCensus" icon (FB variant), a TEXT file "myfile.count" is produced and an alert pops up saying: << See word census in "myfile.count" alongside of "myfile". Census done in 243 miliseconds. Use option key to bypass this alert. Mouse now to exit.>> That is self-explanatory. Right? If the option key is down while dropping, then no alert. Some necessary details. Assign to "myfile.count" the same creator as "myfile" has. Except if the creator is in a set of bad creators like "????", " ", to be completed by Bill, in which case one assigns the creator of a certain freeware or shareware editor like TexEditPlus having no file-size limit; again let Bill say which. No 32K limits. No TE window. No scrolling. But a really useful little utility that minds its own business. One can find the KJ bible and the works of Shakespeare in the Gutenberg public domain libraries. Cheers, Laurent S PS. Go for speed, or MetroWerks will wipe us out. PPS. Delighted I am to see Munger demos behaving; I've been bitten by that critter before. I'll bet though that Munger is not part of the fastest algorithm. PPPS. Bill, you write: > It is partially a sort benchmark, in that the output must be presented in > sorted order, but I'd expect most of the measurable time to be spent > searching the text. Dangerous. For 8000 words are you going to search the text 8000 times with Munger? I'ld suggest following up tedd's hints on binary trees. In a single pass build an ordered binary tree whose leaves are the distinct words each with its population count. But any great sort algorithhm seems a promising starting point.