NAME

SYNOPSIS

  use Alzabo::ChangeTracker;

  my $x = 0;
  my $y = 1;
  sub foo
  {
     my $tracker = Alzabo::ChangeTracker->new;
     $tracker->add( sub { $x = 0; } );

     $x = 1;

     bar();

     eval { something; };

     $tracker->backout if $@;
  }

  sub bar
  {
     my $tracker = Alzabo::ChangeTracker->new;
     $tracker->add( sub { $y = 1; } );

     $y = 2;
  }

DESCRIPTION

We only want to have one object of this type at any one time. In addition, only the stack frame that created it should be able to clear it (except through a backout). Why? Here's an example in pseudo-code to help explain it:

 sub foo
 {
   create a tracker;
   store some change info in the tracker;

   call sub bar;

   store some change info in the tracker;

   # point Y

   clear changes in tracker;
 }

 sub bar
 {
   create a tracker; # internally, we really just increment our stack count

   store some change info in the tracker;

   clear changes in tracker; # point X
 }

If at point X we were to really clear out the changes, even the changes just from sub bar, we'd have a problem. Because if at point Y, things go to hell and we want to back out the changes, we want to back out the changes from sub foo _AND_ sub bar. However, if bar is also an entry point we want to be able to track changes in bar and clear them from bar.

AUTHOR