Update: In the days and weeks after the debate that followed this comment of mine, through reading and CB conversations I have convinced myself that INIT blocks are not "up to spec", so I'm avoiding them. See in particular TimToady's node in this thread.

In the TIMTOWTDI vein, I'm partial to INIT blocks myself:

use strict;
use warnings;

test() for (1..3);


{
  my $memory;
  INIT { $memory = 'a' }

  sub test {
    print "Value of static var is ", $memory++, "\n";
  }
}
__END__
Value of static var is a
Value of static var is b
Value of static var is c
[download]

Of course, if one doesn't like INIT or BEGIN blocks, one can always explicitly check in test that $memory has been initialized and remedy the situation if necessary, but that's ugly, IMO.

I'm partial to INIT blocks myself

Noooooooooooo! Please please please don't do that! It will break if you put the code using the INIT block in a module and that modules is directly or indirectly required (i.e., loaded after program compile-time). INITs are not executed right after the file in which it was found is done compiling. It's executed after the program file is done compiling. INIT is used when you need to do something delay execution of something to before program run-time.

Personally I go with adding a BEGIN before the bare block, thus avoiding to retype the variable name.

BEGIN {
  my $memory = 'a';

  sub test {
    print "Value of static var is ", $memory++, "\n";
  }
}
[download]

Update: added clarification.

ihb

See perltoc if you don't know which perldoc to read!

Yes, but I want this initialization to happen just before run time. There's absolutely no reason, in general, for the initialization of the typical static variable to happen at compile time.

the lowliest monk