>>>>> "WR" == William Ricker <[EMAIL PROTECTED]> writes:
WR> This problem of transposing a matrix bigger than you want to WR> consider slurping feels familiar. The techniques used to sort WR> "big" files in the bad old days, disk or tape based sort-merge, WR> might be adaptble. The trick is to do it in 2 (or small * log N) WR> passes,by creating a simpler suitable intermediate form. interestingly, when doing tape merges with multiple tapes, the most efficient way is based on the fibonacci sequence. but in that case you can see the first records of each tape in a set to do the merge. in the large matrix you can't see all the needed elements so that may not be a viable direction to pursue. but here is another thought based on the old disk sort/merges and such. you can transpose sections of the input matrix and then join/merge those. so if you can only fit one part of the matrix into ram (including output buffers), you transpose it and output it to a file. do the same for the other sections and output them. then it is a fairly easy merge where you can read the first records of each of the transposed sections and merge into a single record and output it to a final file. see, i am not all about stem :) sorry i missed the meeting but i have a nasty cold i am fighting off. uri -- Uri Guttman ------ [EMAIL PROTECTED] -------- http://www.stemsystems.com --Perl Consulting, Stem Development, Systems Architecture, Design and Coding- Search or Offer Perl Jobs ---------------------------- http://jobs.perl.org _______________________________________________ Boston-pm mailing list [EMAIL PROTECTED] http://mail.pm.org/mailman/listinfo/boston-pm
