NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/lang/perl/5324

Internet Message Format | 1992-08-14 | 1.8 KB

Path: sparky!uunet!gatech!rpi!bu.edu!bu-pub.bu.edu!jdh
From: [e-mail] (Jason Heirtzler)
Newsgroups: comp.lang.perl
Subject: Re: perl sort bug on Irix?
Message-ID: <[e-mail]>
Date: 14 Aug 92 19:32:16 GMT
References: <[e-mail]>
Sender: [e-mail]
Organization: Boston University, Information Technology
Lines: 44

In article <[e-mail]>, I wrote:
>Perl's sort seems to behave oddly on Irix.  What I would like
>to do is an inverse sort based on the length.  Is this program
>correct perl (ie. not some implementation dependant thing?)
>
>  #!/usr/local/bin/perl --    # -*-Perl-*-
>
>  push( @dirs, "usr" );
>  push( @dirs, "usr/bin" );
>  push( @dirs, "usr/bin/X11" );
>  push( @dirs, "usr/lib" );
>
>  @crud =  sort {length($a) < length($b);} @dirs;
>
>  foreach $p ( @crud ) {
>      print "$p\n";
>  }

Thanks to all the people who responded, and said to use:

  @crud =  sort {length($b) <=> length($a);} @dirs;

One interesting response came from [e-mail]
(Hal R. Pomeranz) who suggested:

  By the way, if @dirs or whatever you're sorting is going to be a long
  list and if you were using some more CPU intensive function than
  length(), you could speed up your code:

  #!/usr/local/bin/perl

  @dirs = ("usr", "usr/bin", "usr/bin/X11", "usr/lib", ... );    # whatever
  grep(($lengths{$_} = length($_)), @dirs);
  @crud =  sort {$lengths{$a} <=> $lengths{$b};} @dirs;
  print join("\n", @crud), "\n";

  The grep() statement is a little mysterious here, but it's really just
  a tight loop through the @dirs list which sets the values of an
  associative array to the lengths of the strings in your list.  Then
  you can compare the lengths by doing table lookups instead of
  recomputing the function in the sort clause each time.  This way you
  do O(n) function calls instead of O(n*log(n)).

Thanks, folks..