diff incorrect for identical scalars

[abraxxa]

A diff for a hashref containing scalars which values are identical but their default freezer (Storable) return value is different are reported as different:

For example:

my $diff = diff({ port => "1434" }, { port => "1434" }, noU => 1);

print Dumper($diff);

outputs:

{
  D => {
    port => {
      N => "1434",
      O => "1434",
    }
  }
}

I'm still trying to get a failing test case but haven't succeeded so far.

Adding these lines after https://github.com/mr-mixas/Struct-Diff.pm/blob/master/lib/Struct/Diff.pm#L233

print "CHECKING $k\n";
print "\tx: '" . $x->{$k} . "' " . $opts{freezer}($x->{$k}) . "\n";
print "\ty: '" . $y->{$k} . "' " . $opts{freezer}($y->{$k}) . "\n";

for example outputs this:

CHECKING port
	x: '1434' 
                 12345678
1434
	y: '1434' 
                 123451434

Line https://github.com/mr-mixas/Struct-Diff.pm/blob/master/lib/Struct/Diff.pm#L252 causes the failure as it doesn't compare scalars ($type undefined).

[abraxxa]

Just as I expected, one var has the UTF8 flag set and one doesn't:

CHECKING port
	x: '10022'
SV = PV(0x564789fbea70) at 0x564789fc3778
  REFCNT = 2
  FLAGS = (POK,pPOK)
  PV = 0x564789fc4de0 "10022"\0
  CUR = 5
  LEN = 10
	y: '10022'
SV = PV(0x564789ffa9e0) at 0x56478a015490
  REFCNT = 2
  FLAGS = (POK,pPOK,UTF8)
  PV = 0x564789ff4890 "10022"\0 [UTF8 "10022"]
  CUR = 5
  LEN = 10
DIFFERENT! x: 10022, y: 10022

I added a failing test case in my fork.

The big question is: should strings with different encoding but identical characters handled as diff or not?

[mr-mixas]

Hi,

I think diff() behave correctly, data is different after all.

If you need no diff in this case, just redefine freezer and drop (or set) utf flag for scalars in it, see https://github.com/mr-mixas/Struct-Diff.pm#configuration-variables

[abraxxa]

That's almost exactly what I did after I found out what the difference is. My custom freezer just returns scalars as such without invoking Storable which results in string comparison.

I'm still wondering what this line is supposed to do? https://metacpan.org/release/Struct-Diff/source/lib/Struct/Diff.pm#L252

I'd suggest adding my tests but changing the ascii vs. utf8 test to pass when different as this tells someone like me what (s)he should expect from the module and add documentation for it.

Comparing scalars, whether they are strings or numbers, as their internal representation is not what most if not all Perl 5 developers would expect. Maybe you should think about a config option for the module which you might want to change with the next major version bump (1.0).

[mr-mixas]

I'd suggest adding my tests but changing the ascii vs. utf8 test to pass when different as this tells someone like me what (s)he should expect from the module and add documentation for it.

Good thought, could you make a PR?

I'm still wondering what this line is supposed to do? https://metacpan.org/release/Struct-Diff/source/lib/Struct/Diff.pm#L252

Literally: refs compared by addresses, all the rest by value using freezer.

Comparing scalars, whether they are strings or numbers, as their internal representation is not what most if not all Perl 5 developers would expect.

Within the scope of perl it is. But when it comes to operations with systems written in other languages this become critical. Struct::Diff do it best to provide maximum accuracy, which can bee loosen when needed (this case, presicion for floats, '0' vs 0 and so on). I think I should add this to docs =)