http://www.perlmonks.org?node_id=281412


in reply to Re: Perl Idioms Explained - keys %{{map{$_=>1}@list}}
in thread Perl Idioms Explained - keys %{{map{$_=>1}@list}}

This actually seems to be slightly quicker than the grep based solution:

Benchmark: timing 100000 iterations of broquaint, grep_seen, aristotle +, juerd... broquaint: 25 wallclock secs (21.62 usr + 0.01 sys = 21.63 CPU) @ 46 +22.99/s (n=100000) grep_seen: 12 wallclock secs (11.55 usr + 0.00 sys = 11.55 CPU) @ 86 +60.26/s (n=100000) aristotle: 10 wallclock secs ( 9.05 usr + 0.00 sys = 9.05 CPU) @ 11 +044.84/s (n=100000) juerd: 9 wallclock secs ( 8.66 usr + 0.00 sys = 8.66 CPU) @ 11 +543.35/s (n=100000)
on
my @list = qw/a b c d d a e b a b d e f a erti wen udfgn wei sdej usdjdh iu t k a b b b b a a a a c d f e f s r s gkl h u s y t d s w s d log u log ds/;

Jenda
Always code as if the guy who ends up maintaining your code will be a violent psychopath who knows where you live.
   -- Rick Osborne

Edit by castaway: Closed small tag in signature

Replies are listed 'Best First'.
Re^3: Perl Idioms Explained (explained)
by Aristotle (Chancellor) on Aug 06, 2003 at 14:47 UTC

    Yep. As I said, it does not iterate on the Perl side. The pertinent nodes in optree are only ever seen once, and the loop is implicit in the execution of that single op. With all other approaches, the pertinent ops have to be dispatched and executed as many times as there are items in the list.

    In fact, the keys %{{map{$_=>1}@list}} approach needs to first iterate over all elements of @list for the map, building a list as it goes, then build another list for keys.

    In contrast, the @uniq = grep ! $seen{$_}++, @list; approach only builds a list once, while iterating over the elements of @list.

    And of course as already explained, undef @seen{@list}; @uniq = keys %seen; just builds a single list - without iterating at all.

    Makeshifts last the longest.