Bloodnok has asked for the wisdom of the Perl Monks concerning the following question:

Time again to prostrate myself before you highly-esteemed fellow monks concerning a problem I've encountered with Object::InsideOut ... and more specifically, the type validation of fields therein.

If a field type validation is declared as follows: 

sub validate_max_size {
    my $val = shift;

    ! defined $val || $val > 0;
}
.
.
# Then later ...
.
my @data :Field :Type( \&package::validate_max_size );

[download]
the following error is seen when OIO calls the sub with a value of undef (aka the constructor is called with type => undef) ... 
#   Failed test 'Valid max_size: undef'
#   at (eval 4) line 43.
# died: OIO::Code error: Problem with type check routine for initializ
+er 'max_size' for class 'IBM::Common::Collection
# Error: Argument "undef" isn't numeric in numeric gt (>)
# Package: main
# File: t/IBM-Common-Collection.t
# Line: 68

[download]
The problem is that no error unsues from running similar code as a one-liner e.g. 
perl -Mstrict -we 'sub f {my $v = shift; ! defined $v || $v > 0}; f(un
+def)'

[download]

Here's the rub: Has anybody else encountered (and hopefully, overcome) this problem ??

TIA ,

Update: Thanx, once again to ikegami who noticed a disparity between the use of or and ||. Having now tried all combinations of the 2 in the 2 circumstances, exasperatingly, the principle problem remains.

A user level that continues to overstate my experience :-))

Replies are listed 'Best First'.
Re: Object::InsideOut field type validation
by ikegami (Patriarch) on Feb 05, 2009 at 19:25 UTC

    Note that you used
    ! defined $val || $val > 0
    in one place and
    ! defined $v or $v > 0
    in the other.

    Argument "undef" isn't numeric in numeric gt (>)

    That's not a Perl message. Perl 5.6 to 5.10 all return 

    Use of uninitialized value in numeric gt (>)

    [download]

    Update: LOL! As oshalla points out, this is the warning you get for the string 'undef', as opposed to an undefined value.

[reply]
[d/l]
[select]
      TFT ikegami, I knew it would be something simple.

      Having said that, the problem can be removed by including no warnings; in the code ... 

      sub validate_max_size {
    my $val = shift;

    no warnings;
    ! defined $val || $val > 0;
}

      [download]
      gets rid of the problem ... or at least allows a different problem to surface in my test harness ;-)

      A user level that continues to overstate my experience :-))
[reply]
[d/l]
[select]

        I don't get it :-( I tried:

          use strict ;
  use warnings ;

  try() ;        #     none: 1
  try(undef) ;   #    undef: 1
  try(-1) ;      #       -1: 0
  try(0) ;       #        0: 0
  try(1) ;       #        1: 1

  sub try {
    my ($s) = @_ ;

    my $r = validate_max_size(@_) ;

    if    (@_ == 0)      { $s = "none" ;  }
    elsif (!defined($s)) { $s = "undef" ; } ;

    printf "%8s: %d\n", $s, $r+0 ;
  } ;

  sub validate_max_size {
    my $val = shift;

    ! defined $val || $val > 0;
  }

        [download]
        on perl 5.10.0, and got no warnings...

        I don't know if there's an ointment for bracket-alergy as extreme as this... I would write:

          !defined($val) || ($val > 0) ;

        [download]
        'cos I'm comfortable with unaries binding tighter than binaries... but the effect is the same as what you writ.

        Wait a tick... this gives the error you mentioned:

          try('undef') ;  # Argument "undef" isn't numeric in numeric gt (>) .
+..

        [download]
        could it be that somebody, somewhere, is mapping undef to 'undef' ?

[reply]
[d/l]
[select]

Back to Seekers of Perl Wisdom