in reply to The Ovidian Transform

There are lots of ways to get about a 50% speed increase over an ST. Creating all of those tiny anonymous arrays isn't a speedy task. It does lead to rather simple and non-error-prone code, however.

One alternative that is as general purpose is to build a separate array of keys to sort on and then sort a list of indices from that: 

my @sorted= do {
    my @sorton= map { get_date($_) } @data;
    my @indices= sort { $sorton[$a] cmp $sorton[$b] } 0..@data;
    @data[@indices];
};

[download]
Another alternative that can be slightly faster but is a bit harder to customize is to combine the key to sort on with the original data into a single string in a reversible manner and use: 
my @sorted= map { restore_orig_data($_) }
    sort map { prepend_sort_key($_) } @data;

[download]
See also http://www.sysarch.com/perl/sort_paper.html or even http://www.perl.com/doc/FMTEYEWTK/sort.html.

        - tye (but my friends call me "Tye")

Replies are listed 'Best First'.
Re(2): (tye): The Ovidian Transform
by dmmiller2k (Chaplain) on Dec 31, 2001 at 23:21 UTC

    Good one, tye. I've had great success with your second alternative, prepending the sort key to the original data, making a single string and sorting that.

    It has been at *least* 50% faster than the "standard" ST technique, in the instances where I've used it (your mileage may vary, of course :). Mostly, those instances were cases where the data was complex enough that sorting was slow to begin with...

    Update: While this approach probably cannot compete with the raw speed of C, the cost-benefit ratio is far more favorable (to me), given the size of the learning curve I would require in order to be effective using XS (which is essentially what using C in this context boils down to).

    dmm

    You can give a man a fish and feed him for a day ...
    Or, you can     teach him to fish and feed him for a lifetime
[reply]

In Section Meditations