excellent, glad to help... do you see everything that is going on there?
                - Ant

Not sure if the original poster understood, but I'm still a little lost, could you go into more detail?

Kickstart

I thought I was in the twilight zone earlier when I came to perlmonks to put down a question about sorting multidimensional hashes and the top question here was just that!

here's mine...
I have a 4 dimensional hash something like this: $logdata->{$log}->{$line}->{$category} = $value and I need to sort at the $log level by the value of a particular $category.
I already know what categories I need to sort by but don't want to write different subs for each.
My attempt went something like...

    for my $log ( sort {
                           if (!$query->param('merge')) { $a cmp $b }
                           else { # it's merged so we have to sort at 
+log level rather than line level
                               if ($query->param('sortby') eq 'time_as
+c') {
                                   for my $line ( keys %{$logdata->{$l
+og}} ) {
                                       my $timeA = $logdata->{$a}->{$l
+ine}->{'time'};
                                       my $timeB = $logdata->{$b}->{$l
+ine}->{'time'};
                                       $timeA cmp $timeB;
                                   }
                               }
                               elsif ($query->param('sortby') eq 'time
+_des') {
                                   for my $line ( keys %{$logdata->{$l
+og}} ) {
                                       my $timeA = $logdata->{$a}->{$l
+ine}->{'time'};
                                       my $timeB = $logdata->{$b}->{$l
+ine}->{'time'};
                                       $timeB cmp $timeA;
                                   }
                               }
                           }
                       } keys %$logdata ) {
        my $loglines = $logdata->{$log};
        for my $line ( sort keys %$loglines ) {
[download]

but since I am all "use strict;"-ed up, Perl moans that $log needs an explicit package name for $log in the for my $line ( keys %{$logdata->{$log}} ) { lines. This is more complicated than anything I have had to do to date and this is where I got stuck.

Against my better judgement I tried a foreach my $log (keys %$logdata) { above the two lines mentioned in the last paragraph but the output was not quite what I was looking for (a bunch of empty lines inamongst the UNsorted output).

I have a solution - make the first key in %$logdata "$log|$line" and remove a level from the hash but I figure there has to be a way to do this properly. Any help would be much appreciated.

larryk

Ok, first of all... you should put a sort that large in a subroutine, just for simplicity's sake...

for(sort sort_sub @foo) ...

sub sort_sub { 
  $a <=> $b;
}
[download]

Similar to that... make your code MUCH more readable :)

Perl moans about $log not being there because it isn't... $log isn't filled until after the sort is done... your $log data is in $a and $b.

So, let me get this straight... you want to sort at the first level of your hash, by a key in the 3rd level of your hash? The problem I see in this is... I assume you have many lines (i.e. $line data)? If that is true, then how do you find the category you need under many lines? Or is there just one line that has a 'time' category, and that is the one you need?

Can you give a small example of what your hash looks like?
                - Ant

I agree it _would_ make it more readable and I tried to put it in a sub but the sort option (8 to be precise) comes from $query-param('sortby') (CGI.pm) which, as far as I can tell by failed attempts, I cannot pass thusly: (while using sic strict)
foreach my $x (sort sortsub($query) keys %y) {...

Every line has a time on it and I have to sort a number of logs by like times of day. Data is as follows:


log1.log<br>
--------<br>
Time=00:00:00.001|Request=det|Category=btyjar<br>
Time=00:00:00.002|Request=sdf|Category=345<br>
Time=00:00:00.003|Request=fdgh|Category=cvn<br>
Time=00:00:00.004|Request=cv|Category=ryui<br>
log2.log<br>
--------<br>
Time=00:00:00.001|Request=h5|Category=56yjh<br>
Time=00:00:00.002|Request=hjk|Category=dr6<br>
Time=00:00:00.003|Request=qw|Category=345<br>
Time=00:00:00.004|Request=thgj|Category=234<br>
[download]

ok, that's example data. then to collect the data i use:


    for my $log (@logs) {
        if ($log =~ /\.gz$/) {
            eval "`gunzip $log`";
            next if $@;
            $log = substr $log, 0, -3;
        }
        open(LOG,$log) || warn "log not opened\n";
        while (<LOG>) {
            chomp;
            my @data = split /\|/;
            for (@data) {
                my($cat,$value) = split /=/;
                $logdata{$log}{$.}{lc($cat)} = $value;
            }
        }
        close LOG;
    }
[download]

output is in one of two forms: log by log (so i do a little "1 of 10", "previous", "next" to switch between) or Merged - all the files in one big table. both output formats need to be sortable. the log by log is easy because I am sorting at the line level but the merged table is where I got stuck.

I do have many lines and so I need to find a way to get the sort to iterate through those lines, pulling out the time and comparing it, while still at the log level to find out in which order to sort all of the lines of all of the logs.

cheers for taking the time to look at this.