NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/lang/perl/7606

Text File | 1992-12-24 | 2.6 KB | 75 lines

Newsgroups: comp.lang.perl
Path: sparky!uunet!psinntp!balltown!rodney
From: [e-mail] (Rodney Peck)
Subject: Re: Looking for a "numerical/alpha" sort routine.
Message-ID: <[e-mail]>
Organization: New York State Institute for Sebastian Cabot Studies
References: <[e-mail]>
Date: Thu, 24 Dec 1992 03:32:59 GMT
Lines: 64

[e-mail] (Jesse Buckley) writes:
: Does anyone have a sort/compare routine that will work for numerical
: alpha data?  For example, if I have 1, 2, 3a, 3b, 3c, I want them sorted
: that way.  Right now the <=> compare ignored the letters after the
: numbers.  Thanks.

Well, I played with this question for about an hour and came up with this:

  $a cmp $b unless $a <=> $b;

It's efficient because the cmp is only called if they are numerically
equal.  Otherwise, the value of the numeric compare is returned.

Orginally, I spent a lot of time trying to get just the characters and
then compare them with cmp, but since the numbers are equal, the
beginning of the strings will be the same and cmp will properly rank
them even though the numbers are still on the front of the string.

However, since I spent all that time working on this, and figured out
a few nifty things that other people might want to study or whatever.

Here's the general idea I came up with orginally before trying to
remove the local variables:
sub newcmp {
# works like <=>, but also deals with trailing letters
# use the <=> first, but it lies about things being
# equal so check with a cmp on the trailing letters

local ($ans) = ($a <=> $b);
  if ($ans == 0)
  {
    local($aa, $bb);
    ($aa) = ($a =~ /\d*(\D*)/);
    ($bb) = ($b =~ /\d*(\D*)/);
    $ans = ($aa cmp $bb);
  }
  $ans;
}

Here are four different one liners trying to get the letters away
from the numbers before I realized that the numbers would be the same
if the numbers compared equally.  It might be worth someone's while
to read them and figure out the different intentions.  The big thing
was to avoid local variables, and to reduce size.

   "$a:$b" =~ /\d*(\D*):\d*(\D*)/, $1 cmp $2 unless $a <=> $b;

   @_ = "$a:$b:" =~ /\d(\D):/g, shift cmp shift unless $a <=> $b;

   split(/:\d+/,":$a:$b", shift, shift cmp shift unless $a <=> $b;

   ($junk,@_) = split(/:\d+/,":$a:$b"), shift cmp shift unless $a <=> $b;

And finally, the obJAPH using the new sort routine:

#----
#!/bin/perl
$_= '2bn3ap1aj3cr3dl2h 2eh3e 4cc1bu2gr3be4ah';
$_.='4ba4dk2aa4ee2fe1dt2co1cs2dt1e 4fr9a\9bn';
while(($a,$b,$_)=/^(..)(.)(.*)$/) {$a{$a}=$b;}
grep($f.=$a{$_},sort{$a cmp $b unless $a<=>$b}
keys%a);                 eval("print \"$f\"");
#----
Rodney