According to the docs for $node->cloneNode, it's behaviour required by the DOM spec (for $node->cloneNode, the function $reader->copyCurrentNode mirrors).

I've had similar experiences with XML::LibXML, see my node here for a potential workaround (basically manually copying and re-creating the nodes instead of cloning).

Updated wording slightly for clarification.

Hi,

Its just another object property

https://metacpan.org/pod/XML::LibXML::Element#setNamespaceDeclURI

$node->setNamespaceDeclURI(undef, undef);
[download]

Thanks, despite being marked "experimental" it does appear to work in the OP's case:

use warnings;
use strict;
use XML::LibXML::Reader;

my $xml = <<'ENDXML';
<foo xmlns="http://www.example.com"><bar/></foo>
ENDXML

my $reader = XML::LibXML::Reader->new(string => $xml);
while ($reader->read) {
    my $node = $reader->copyCurrentNode;
    $node->setNamespaceDeclURI(undef, undef)
        if $node->nodeName eq 'bar';
    print "$node";
}

__END__

<foo xmlns="http://www.example.com"/><bar/><foo xmlns="http://www.exam
+ple.com"/>
[download]

Unfortunately I might still be missing something, because applied to choroba's code here it doesn't seem to change the namespace of the nodes.

Hi,

Why copy? If you print the node ( eg toString) instead of a copy, the output will not contain the xmlns.

Thank you all for the reply. Seems the namespace is forcely written out and there is no way to change that behavior.

I don't want to do go through each attribute and construct a new element neither.

Currently I still use the copy and then do a string replacement  s/ xmlns="[^"]"// to remove it

Hi,

So why are you cloning the nodes?  $node->setNamespaceDeclURI(undef, undef); will unset the namespace