Re: while(<>) { ... } considered harmful

[BrowserUk]

'scuse my ignorance of nether regions below Perl's skirts, but would it impose a huge performance penalty for map and grep to localise $_ on entry?

I visualise them as subs with their own local scope. Would a local $_; at the top cost so very much?

---

Well It's better than the Abottoire, but Yorkshire!

[Ovid]

Unless I totally misunderstood the above comments -- which is possible as I just briefly skimmed them -- the problem here is not that map and grep don't localize $_, but that they may execute code that may alter the value of $_, which in this case is aliased to a constant. Consider this code:

  $_ = 'Ovid';
  %hash = map { $_ => 1 } qw/ foo bar baz /;
  print;

That code will print 'Ovid' because $_ has been localized within the map statement. Now, if you try to alter the $_, your code fails because you're trying to alter a constant:

  $_ = 'Ovid';
  %hash = map {
      $_ .= 'asdf';
      $_ => 1
  } qw/ foo bar baz /;
  print;

To get around that, you can do this:

  $_ = 'Ovid';
  %hash = map {
      local $_ = $_;
      $_ .= 'asdf';
      $_ => 1
  } qw/ foo bar baz /;
  print;

That allows you to assign the aliased constant to a real variable, thus avoiding the problems outlined in this thread.

Cheers,
Ovid

Join the Perlmonks Setiathome Group or just click on the the link and check out our stats.