All the Perl that's Practical to Extract and Report

How Can I Write This Better?

Ovid on 2008-08-19T12:32:47

Using XML::LibXML, I'm trying to determine if two nodes are different. I have the following humiliating code, but it works.

sub _node_differs {
    my ($self,$left,$right) = @_;

    # toStringC14N ensures "canonization" and allows '' to match
    # '<title>'
    my ($left_string, $right_string) = map {
        defined($_) ? ( eval { $_->toStringC14N } || $_->toString ) : ''
    } ( $left, $right );

    if ($left_string ne $right_string) {
        return [ $left, $right ];
    }
    return;
}

What's a "proper" way of doing that? I know I'm missing something obvious.

Looks proper enough&#226;&#8364;&#166;

dakkar on 2008-08-19T13:36:51

Apart from the weird 'eval' (toStringC14N is defined in XML::LibXML::Node, so anything that's part of a XML::LibXML document should have it), it doesn't look "humiliating".

You can create a quotient of all XML (text) representations by putting in the same class those that produce the same DOM (internal) representation; you want to see if two nodes fall into the same equivalence class. Canonicalization is a way to produce a representative of the class, so comparing two canonicalized representation of DOM subtrees seems right enough to me.

In other words: it's a way to define equivalence, and probably saner than most.

Re:Looks proper enough&#226;&#8364;&#166;

Ovid on 2008-08-19T15:50:08

The eval is there because I kept hitting nodes which would have that method defined, but I get this error:

Failed to convert doc to string in doc->toStringC14N at /opt/csw/lib/perl/site_perl/XML/LibXML.pm line 955

I don't know what's causing that, but the eval protects against it.

Re:Looks proper enough...

dakkar on 2008-08-19T16:00:13

Looking through the source code of libxml2, it looks like it's caused by nodes that have no textual content: the xmlC14NDocDumpMemory function (used by toStringC14N) does not set its "output buffer" parameter if the buffer would be empty.

Strange behaviour, I'd call it a bug, or at least a documentation deficiency.

Re:Looks proper enough...

Ovid on 2008-08-22T09:18:05

Thanks for digging into this. I've included the URL of this discussion in our code base so people have a clue what that weird eval is for.