sub do_msalist {    
   my (%country, %state, %msaname, @msakeys);

   sub by_americentric {    
      $country{$b} cmp $country{$a}  or
      $state{$a}   cmp $state{$b}    or    
      $msaname{$a} cmp $msaname{$b}  or    
      $a <=> $b
   }
 
   foreach (sort by_americentric @msakeys) {
      ...
   }
}
[download]

You need to have the variables accessible by by_americentric(). There are a few ways to do this. One way is to put both of the subroutines inside an enclosing block, like this:

{
    my (%country, %state, %msaname);

sub do_msalist {

    my @msakeys = @_;

    foreach(sort by_americentric @msakeys) { 
        #...
    }

}

sub by_americentric {
    $country{$b} cmp $country{$a}
    or
        $state{$a} cmp $state{$b}
    or
        $msaname{$a} cmp $msaname{$b}
    or
        $a <=> $b;
}

}
[download]

Or, you can define the associative arrays as package variables with 'use vars'. Or you can just put by_americentric in as a blockin do_msalist, and call it that way:

sub do_msalist {

    my @msakeys = @_;

    foreach(sort {
    $country{$b} cmp $country{$a}
    or
        $state{$a} cmp $state{$b}
    or
        $msaname{$a} cmp $msaname{$b}
    or
        $a <=> $b;
} @msakeys) { 
        #...
    }

}
[download]

stephen

{
    my (%country, %state, %msaname, @msakeys);
    sub do_msalist {
        ...
        foreach(sort by_americentric @msakeys) { 
            ...
        }
    } #do_msalist

    sub by_americentric {
        $country{$b} cmp $country{$a}
        or
        $state{$a} cmp $state{$b}
        or
        $msaname{$a} cmp $msaname{$b}
        or
        $a <=> $b
    } #by_americentric
}
[download]

Use an in-line sort block instead (like @output = sort { BLOCK } @input) and all will be well.

-- Randal L. Schwartz, Perl hacker

sub do_msalist {
    my (%country, %state, %msaname, @msakeys);
     ...
    my @sorted = map  { $$_[0] }
                 sort { $$a[1] cmp $$b[1] }
                 map  { [$_, americentric_key(
                             $country{$_},
                             $state{$_},
                             $msaname{$_},
                             $_
                 )]} @msakeys;
    for (@sorted) {
     ...
    }
}

sub americentric_key {
 # This part may need adjusting depending on your data
 sprintf("%10s%10s%10s%10d", @_;
}
[download]

I knew about sort {BLOCK} already -- but what I was hoping to achieve was something recyclable -- where if I add other do_blah subs besides do_msalist, they can call the same sort routine. Hence, I like Stephen's suggestion of giving the hashes package scope with use vars.

This sort of triggers a new question -- I know package scope is a bad, bad thing under mod_perl in most cases, but can I work around it by explicitly undefing the hashes before the script returns?

Spud Zeppelin * spud@spudzeppelin.com

Perhaps someone else can answer more authoritatively, but my understanding of global variables is that
• a) they're slower to access than local counterparts; and
• b) in mod_perl they are persistent and take up space while your script isn't executing, and leave the environment "dirty" for the next time the CGI is called
Emptying the hashes will fix problem (b) but they can't help you with (a). Unless you're doing a lot of data crunching, though, this might be an acceptable loss, though in my opinion it's still slightly poor coding practice, but you could do a lot worse.