Bug #76712 [Opn]: Assignment of empty string creates extraneous text node

Edit report at https://bugs.php.net/bug.php?id=76712&edit=1 ID: 76712 Updated by: cmb@php.net Reported by: frank dot mohaupt at brainworx dot audio -Summary: Empty Tags are preserved if node value is overwritten with empty value +Summary: Assignment of empty string creates extraneous text node Status: Open Type: Bug Package: SimpleXML related Operating System: Linux d45011cc91b7 4.9.93-boot2d PHP Version: 7.0.31 Block user comment: N Private report: N New Comment: Ah, that's interesting! Apparently, xmlNodeSetContent() with an empty content string does not create a text child node, while xmlNodeSetContentLen() does. That's pretty inconsistent. Anyhow, using xmlNodeSetContent() here would solve the reported issue, and wouldn't make a difference otherwise, from what I can tell. I have submitted <https://github.com/php/php-src/pull/3431>. Previous Comments: ------------------------------------------------------------------------ [2018-08-06 23:24:38] requinix@php.net The XML spec explicitly says empty elements (no content) can be either a start/end tag pair or a self-closing tag. They are definitely equivalent, though it does later say one "SHOULD" use self-closed tags when the element is defined (like in a spec) to be empty. For the record. For comparison, in DOMNode (a) Setting ->textContent uses xmlNodeSetContent("") + xmlNodeAddContent and produces a self-closed element (b) Setting ->nodeValue uses xmlNodeSetContentLen and produces an element with no content https://3v4l.org/Of5p1 I see SimpleXML is a library on top of libxml, one that goes beyond being a simple abstraction layer that PHP often does, and libraries regularly do things like normalize behavior from the underlying API. Personally I would go ahead and patch it. ------------------------------------------------------------------------ [2018-08-06 22:09:02] cmb@php.net While I can confirm the reported behavior[1], I don't think this qualifies as bug, since both representations are equivalent. It rather seems to me that we're hitting a peculiarity of libxml here, which treats an empty string as content differently in xmlNewChild() and xmlNodeSetContentLen(). In the latter case a text child node is created; in the former this doesn't happen. We could work around this (see the attached patch[2]), but I'm not convinced that we should. [1] <https://3v4l.org/mQaLd> [2] <https://bugs.php.net/patch-display.php?bug=76712&patch=workaround&revision=1533593319> ------------------------------------------------------------------------ [2018-08-06 22:08:39] cmb@php.net The following patch has been added/updated: Patch Name: workaround Revision: 1533593319 URL: https://bugs.php.net/patch-display.php?bug=76712&patch=workaround&revision=1533593319 ------------------------------------------------------------------------ [2018-08-06 14:22:14] frank dot mohaupt at brainworx dot audio Description: ------------ If the value of node "foo" is empty while adding the node, its output (SimpleXMLElement::asXML()) is "<foo/>". If the value of node "foo" is overwritten with an empty value ($sxe->foo = ''), its output is "<foo></foo>" instead of "<foo/>". Test script: --------------- $sxe = new SimpleXMLElement('<foo></foo>'); $sxe->addChild('bar'); $arrResults['empty node'] = $sxe->asXML(); $sxe = new SimpleXMLElement('<foo></foo>'); $sxe->addChild('bar', ''); $arrResults['empty string'] = $sxe->asXML(); $sxe = new SimpleXMLElement('<foo></foo>'); $sxe->addChild('bar'); $sxe->bar = ''; $arrResults['overwritten empty string'] = $sxe->asXML(); foreach($arrResults as $key => $strResult){ echo $key . ': ' . $strResult; } // output // empty node: <?xml version="1.0"?>\n<foo><bar/></foo>\n // empty string: <?xml version="1.0"?>\n<foo><bar/></foo>\n // overwritten empty string: <?xml version="1.0"?>\n<foo><bar></bar></foo>\n Expected result: ---------------- I expect to get the same result for all three cases. ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=76712&edit=1