Provided by: libalgorithm-permute-perl_0.16-1_amd64 bug

NAME

       Algorithm::Permute - Handy and fast permutation with object oriented interface

SYNOPSIS

         use Algorithm::Permute;

         # default is to create n of n objects permutation generator
         my $p = Algorithm::Permute->new(['a'..'d']);

         # but also you can create r of n objects permutation generator, where r <= n
         my $p = Algorithm::Permute->new([1..4], 3);

         while (my @res = $p->next) {
           print join(", ", @res), "\n";
         }

         # and this one is the speed demon:
         my @array = (1..9);
         Algorithm::Permute::permute { print "@array\n" } @array;

DESCRIPTION

       This handy module makes performing permutation in Perl easy and fast, although perhaps its
       algorithm is not the fastest on the earth.  It supports permutation r of n objects where 0
       < r <= n.

METHODS

       new [@list]
           Returns a permutor object for the given items.

       next
           Returns a list of the items in the next permutation.  The order of the resulting
           permutation is the same as of the previous version of "Algorithm::Permute".

       peek
           Returns the list of items which will be returned by next(), but doesn't advance the
           sequence. Could be useful if you wished to skip over just a few unwanted permutations.

       reset
           Resets the iterator to the start. May be used at any time, whether the entire set has
           been produced or not. Has no useful return value.

CALLBACK STYLE INTERFACE

       Starting with version 0.03, there is a function - not exported by default - which supports
       a callback style interface:

       permute BLOCK ARRAY
           A block of code is passed, which will be executed for each permutation. The array will
           be changed in place, and then changed back again before "permute" returns. During the
           execution of the callback, the array is read-only and you'll get an error if you try
           to change its length. (You can change its elements, but the consequences are liable to
           confuse you and may change in future versions.)

           You have to pass an array, it can't just be a list. It does work with special arrays
           and tied arrays, though unless you're doing something particularly abstruse you'd be
           better off copying the elements into a normal array first. Example:

            my @array = (1..9);
            permute { print "@array\n" } @array;

           The code is run inside a pseudo block, rather than as a normal subroutine. That means
           you can't use "return", and you can't jump out of it using "goto" and so on. Also,
           "caller" won't tell you anything helpful from inside the callback. Such is the price
           of speed.

           The order in which the permutations are generated is not guaranteed, so don't rely on
           it.

           The low-level hack behind this function makes it currently the fastest way of doing
           permutation among others.

COMPARISON

       I've collected some Perl routines and modules which implement permutation, and do some
       simple benchmark. The whole result is the following.

       Permutation of eight scalars:

         Abigail's                     :  9 wallclock secs ( 8.07 usr +  0.30 sys =  8.37 CPU)
         Algorithm::Permute            :  5 wallclock secs ( 5.72 usr +  0.00 sys =  5.72 CPU)
         Algorithm::Permute qw(permute):  2 wallclock secs ( 1.65 usr +  0.00 sys =  1.65 CPU)
         List::Permutor                : 27 wallclock secs (26.73 usr +  0.01 sys = 26.74 CPU)
         Memoization                   : 32 wallclock secs (32.55 usr +  0.02 sys = 32.57 CPU)
         perlfaq4                      : 36 wallclock secs (35.27 usr +  0.02 sys = 35.29 CPU)

       Permutation of nine scalars (the Abigail's routine is commented out, because it stores all
       of the result in memory, swallows all of my machine's memory):

         Algorithm::Permute            :  43 wallclock secs ( 42.93 usr +  0.04 sys = 42.97 CPU)
         Algorithm::Permute qw(permute):  15 wallclock secs ( 14.82 usr +  0.00 sys = 14.82 CPU)
         List::Permutor                : 227 wallclock secs (226.46 usr +  0.22 sys = 226.68 CPU)
         Memoization                   : 307 wallclock secs (306.69 usr +  0.43 sys = 307.12 CPU)
         perlfaq4                      : 272 wallclock secs (271.93 usr +  0.33 sys = 272.26 CPU)

       The benchmark script is included in the bench directory. I understand that speed is not
       everything. So here is the list of URLs of the alternatives, in case you hate this module.

       •   Memoization is discussed in chapter 4 Perl Cookbook, so you can get it from O'Reilly:
           ftp://ftp.oreilly.com/published/oreilly/perl/cookbook

       •   Abigail's: http://www.foad.org/~abigail/Perl

       •   List::Permutor: http://www.cpan.org/modules/by-module/List

       •   The classic way, usually used by Lisp hackers: perldoc perlfaq4

ACKNOWLEDGEMENT

       In Edwin's words: Yustina Sri Suharini - my ex-fiance-now-wife, for providing the
       permutation problem to me.

SEE ALSO

Data Structures, Algorithms, and Program Style Using C - Korsh and Garrett

       • Algorithms from P to NP, Vol. I - Moret and Shapiro

AUTHOR

       Edwin Pratomo <edpratomo@cpan.org> was the original author.

       Stephan Loyd <sloyd@cpan.org> is co-maintainer after version 0.12.

       The object oriented interface is taken from Tom Phoenix's "List::Permutor".  Robin Houston
       <robin@kitsite.com> invented and contributed the callback style interface.

COPYRIGHT AND LICENSE

       This software is copyright (c) 1999 by Edwin Pratomo.

       This is free software; you can redistribute it and/or modify it under the same terms as
       the Perl 5 programming language system itself.