DOMDocument formatOutput fails when node contains both child text and element nodes

[billynoah]

Description

This behaves as expected:

$dom = new DOMDocument('1.0', 'UTF-8');
$dom->preserveWhiteSpace = false;
$dom->formatOutput = true;
$root = $dom->createElement('root');
$dom->appendChild($root);
$root->appendChild($dom->createElement('child', 'nodevalue B'));
echo $dom->saveXml();

Output:

<root>
  <child>nodevalue B</child>
</root>

But when a node contains both a child text node and a child element node the output is collapsed. The following code:

$dom = new DOMDocument('1.0', 'UTF-8');
$dom->preserveWhiteSpace = false;
$dom->formatOutput = true;
$root = $dom->createElement('root', 'nodeValue A');
$dom->appendChild($root);
$root->appendChild($dom->createElement('child', 'nodevalue B'));
echo $dom->saveXml();

Resulted in this output:

<root>nodeValue A<child>nodevalue B</child></root

But I expected this output instead:

<root>nodeValue A
  <child>nodevalue B</child>
</root>

PHP Version

PHP 8.3.12

Operating System

Ubuntu 24.04.01

[damianwadley]

Seems correct to me: whitespace is significant, and adding formatting would change the text node to be nodeValue A\n··.

[billynoah]

@damianwadley - I get what you're saying but it seems odd to collapse the whitespace of all descendants in the tree because of this.

$dom = new DOMDocument('1.0', 'UTF-8');
$dom->preserveWhiteSpace = false;
$dom->formatOutput = true;
$root = $dom->createElement('root', 'nodeValue A');
$dom->appendChild($root);
$root->appendChild($dom->createElement('child', 'nodevalue B'));
$root->appendChild($dom->createElement('child', 'nodevalue C'));
$child = $root->appendChild($dom->createElement('child'));
$child->appendChild($dom->createElement('grandChild', 'nodeValue E'));
$child->appendChild($dom->createElement('grandChild', 'nodeValue F'));
echo $dom->saveXml();

Gets you:

<root>nodeValue A<child>nodevalue B</child><child>nodevalue C</child><child><grandChild>nodeValue E</grandChild><grandChild>nodeValue F</grandChild></child></root>

But shouldn't it be something like this?

<root>nodeValue A<child>nodevalue B</child>
	<child>nodevalue C</child>
	<child>
		<grandChild>nodeValue E</grandChild>
		<grandChild>nodeValue F</grandChild>
	</child>
</root>

After all, we are asking for formatted output, which by it's very nature is going to be adding whitespace only text nodes. My example above would preserve the value of the first text node but still give you some semblance of formatted output.

[nielsdos]

I suppose this isn't done because the formatting could become inconsistent if you have inserted your own whitespace text nodes. In any case, changing this functionality now would be a BC break. Furthermore, this functionality actually comes from libxml2 not from PHP. So if you want this behaviour change then you have to report it over at https://gitlab.gnome.org/GNOME/libxml2

[billynoah]

Thanks @nielsdos - It makes sense, just counterintuitive for me.