in reply to Re^13: How to completely destroy class attributes with Test::Most?
in thread How to completely destroy class attributes with Test::Most?

"The key motivation for independent iterator objects is that they allow the program to have multiple iterators even on the same category by removing the need for the File::Collector object to track the iterators."

Yes, but there is a big convenience to having File::Collector track the iterators, isn't there? I can call $s->selected_file from anywhere and refer to the same file from any of the various class methods. I don't have to worry about passing the iterator objects around. Or am I missing something?

$PM = "Perl Monk's";
$MCF = "Most Clueless Friar Abbot Bishop Pontiff Deacon Curate Priest Vicar";
$nysus = $PM . ' ' . $MCF;
Click here if you love Perl Monks

Replies are listed 'Best First'.
Re^15: How to completely destroy class attributes with Test::Most?
by jcb (Parson) on Aug 29, 2019 at 00:01 UTC

    You seem to be missing that you can put all of the "action" methods that your subclasses add onto File::Collector in the subclasses of File::Collector::Iterator instead.

    This gives a better-separated design, where subclasses of File::Collector add categories and subclasses of File::Collector::Iterator add actions. This way, File::Collector is more of a passive data store, while the iterator subclasses perform actions, using a Model/Controller pattern.

      Well, I played around with this some more. I got it working but it makes the code a lot uglier. For example, here's what I have to generate nested iterators now:

      my $it1 = $fp->next_parseable_file; while (my $file1 = $it1->next) { print $fp->short_name($file1) . "\n"; my $it2 = $fp->next_nonparseable_file; while (my $file2 = $it2->next) { print $fp->short_name($file2) . "\n"; } }

      Whereas before, I just had:

      while ($fp->next_parseable_file) { print $fp->short_name . "\n"; while ($fp->next_nonparseable_file) { print $fp->short_name . "\n"; } }

      The iterators now have to be created separately from the while loop. And I'm stuck passing around the $file to the short_name method.

      I get that having the separation in logic the way you describe is probably more ideal from a logical perspective. But maybe this is a case where some rules should be broken? I don't have enough experience to say. Or maybe there is some trickery I could invoke to get the best of both worlds?

      $PM = "Perl Monk's";
      $MCF = "Most Clueless Friar Abbot Bishop Pontiff Deacon Curate Priest Vicar";
      $nysus = $PM . ' ' . $MCF;
      Click here if you love Perl Monks

      Ok, I think I might be on to something good...or it could be colossally stupid, I'm not sure which. If I could indulge your attention for a little bit more, I'd appreciate it.

      So check this out:

      # prints out short name of all "txt-files" $da->bundle_txt_files->do->print_short_names;

      And here is the iterator built using your suggestions (I changed the "all" method name to "do," which seemed to make more sense):

      package File::Collector::Iterator ; use strict; use warnings; use Carp; use Log::Log4perl::Shortcuts qw(:all); my $collector; # private package variable for storing the Fi +le::Collector sub print_short_names { my $s = shift; print $collector->{files}{$s->selected_file}{short_path} . "\n"; # + uses the File::Collector object to look up short path } sub new { my $class = shift; $collector = shift; # new gets passed the $collector from File: +:Collector AUTOLOAD bless [@_], $class; } sub next { my $self = shift; shift @$self; return $self->[0]; } sub selected_file { (shift)->[0] } # do method of executing methods on child classes sub do { my $self = shift; bless \$self, 'File::Collector::Iterator::All'; } { package File::Collector::Iterator::All; use Log::Log4perl::Shortcuts qw(:all); sub AUTOLOAD { our $AUTOLOAD; my $self = shift; my @method = split /::/, $AUTOLOAD; my $method = pop @method; $$self->$method(@_) while ($$self->next); } }

      What do you think? Is this a good solution?

      $PM = "Perl Monk's";
      $MCF = "Most Clueless Friar Abbot Bishop Pontiff Deacon Curate Priest Vicar";
      $nysus = $PM . ' ' . $MCF;
      Click here if you love Perl Monks

        Ideally, the iterator should need no connection to its parent other than shared data, and I see no reason that File::Collector should be a singleton object, which means that it needs to be an instance variable if it is used at all.

        We can avoid that by transferring more data into the newly-constructed iterator. I now see that your application attaches some additional attributes to each file, where I had previously supposed that you care only about the file name. The solution is simple: add a name key to each item in $collector->{files} and change selected_file to:

        sub selected_file { (shift)->[0]->{name} }

        (also change the return value of next)

        When constructing an iterator, pass a list of the relevant entries in ->{files} (which are hashrefs, so the per-category lists should be aliases into the master file set) to new. This even allows another bit of AUTOLOAD magic to provide selected_file_KEY getters that read $self->[0]->{KEY} on an iterator. The explicit selected_file method is therefore an alias for selected_file_name with this approach.

      OK, I think I see what you are saying. I will play with it some more tomorrow when I'm fresh. Thanks.

      $PM = "Perl Monk's";
      $MCF = "Most Clueless Friar Abbot Bishop Pontiff Deacon Curate Priest Vicar";
      $nysus = $PM . ' ' . $MCF;
      Click here if you love Perl Monks