xml5ever drops xml attributes with matching namespace and local names

[noahbald]

Hey there, I've run into this issue recently where only one attribute of a given namespace and local name will remain in the document. Consider the following test

#[test]
fn test_weird() -> anyhow::Result<()> {
    use rcdom::SerializableHandle;
    use xml5ever::{
        driver::{parse_document, XmlParseOpts},
        serialize::{serialize, SerializeOpts},
        tendril::TendrilSink,
    };

    let src = r##"<svg xmlns="http://www.w3.org/2000/svg" xmlns:x="http://www.w3.org/1999/xlink">
    <defs>
        <g id="mid-line"/>
        <g id="line-plus">
            <use href="#mid-line"/>
            <use href="#plus"/>
        </g>
        <g id="plus"/>
        <g id="line-circle">
            <use x:href="#mid-line"/>
        </g>
    </defs>
    <path d="M0 0" id="a"/>
    <use x:href="#a" x="50" y="50"/>
    <use x:href="#line-plus"/>
    <use x:href="#line-circle"/>
</svg>"##;
    println!("src:\n{src}");

    let dom: rcdom::RcDom =
        parse_document(rcdom::RcDom::default(), XmlParseOpts::default()).one(src);
    let mut sink: std::io::BufWriter<_> = std::io::BufWriter::new(Vec::new());
    serialize(
        &mut sink,
        &std::convert::Into::<SerializableHandle>::into(dom.document),
        SerializeOpts::default(),
    )?;

    let sink: Vec<_> = sink.into_inner()?;
    println!("\noutput:\n{}", String::from_utf8_lossy(&sink));
    Ok(())
}

The output for me is as follows

<svg xmlns="http://www.w3.org/2000/svg" xmlns:x="http://www.w3.org/1999/xlink">
    <defs>
        <g id="mid-line"/>
        <g id="line-plus">
            <use x:href="#mid-line"/>
            <use x:href="#plus"/>
        </g>
        <g id="plus"/>
        <g id="line-circle">
            <use x:href="#mid-line"/>
        </g>
    </defs>
    <path d="M0 0" id="a"/>
    <use x:href="#a" x="50" y="50"/>
    <use x:href="#line-plus"/>
    <use x:href="#line-circle"/>
</svg>

output:
<svg xmlns="http://www.w3.org/2000/svg">
    <defs>
        <g id="mid-line"></g>
        <g id="line-plus">
            <use x:href="#mid-line"></use>
            <use></use>
        </g>
        <g id="plus"></g>
        <g id="line-circle">
            <use></use>
        </g>
    </defs>
    <path d="M0 0" id="a"></path>
    <use x="50" y="50"></use>
    <use></use>
    <use></use>
</svg>

As you can see, all the x:href attributes, except for the first, seem to be dropped completely.
This doesn't seem like expected behaviour unless I'm missing something.

If needed, this are the versions I'm using

rcdom = { package = "markup5ever_rcdom", version = "0.3" }
xml5ever = "0.18.0"

[demurgos]

I've just hit this exact issue when parsing this PECL API response:

<?xml version="1.0" encoding="UTF-8"?>
<r xmlns="http://pear.php.net/dtd/rest.release" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink" xsi:schemaLocation="http://pear.php.net/dtd/rest.release     http://pear.php.net/dtd/rest.release.xsd">
 <p xlink:href="/rest/p/protobuf">protobuf</p>
 <c>pecl.php.net</c>
 <v>4.27.0</v>
 <st>stable</st>
 <l>BSD-3-Clause</l>
 <m>protobufpackages</m>
 <s>Google's language-neutral, platform-neutral, extensible mechanism for serializing structured data.</s>
 <d>https://developers.google.com/protocol-buffers/</d>
 <da>2024-05-23 14:46:50</da>
 <n>* See github.com/protocolbuffers/protobuf/releases/tag/v27.0 for release notes.</n>
 <f>243961</f>
 <g>https://pecl.php.net/get/protobuf-4.27.0</g>
 <x xlink:href="package.4.27.0.xml"/>
</r>

The xlink:href appears in the tree for the <p> node at the start, but not in the <x> node at the end (empty attribute list).

[jdm]

I've stepped through the implementation on the provided example and https://github.com/servo/html5ever/blob/main/xml5ever/src/tree_builder/mod.rs#L355-L364 is the code that is causing the unexpected behaviour. The namespace is reset after the first use, and this causes the filter in https://github.com/servo/html5ever/blob/main/xml5ever/src/tree_builder/mod.rs#L341-L349 to drop the attribute from subsequent elements.

[jdm]

https://github.com/servo/html5ever/blob/main/xml5ever/src/tree_builder/mod.rs#L341-L349 is the check that filters out subsequent attributes. This happens because of this duplicate attribute check: https://github.com/servo/html5ever/blob/main/xml5ever/src/tree_builder/mod.rs#L316

[jdm]

This was introduced in baeacb4 but the original repo's related issue is no longer visible: https://github.com/Ygg01/xml5ever/issues/30

[demurgos]

Thank you @jdm I confirm that it worked. Patching my local dependencies to use the commit in #539, I can properly parse the response, including the last link:

Release {
    package: ReleasePackage {
        name: "protobuf",
        link: "/rest/p/protobuf",
    },
    channel: "pecl.php.net",
    version: "4.27.0",
    status: "stable",
    license: "BSD-3-Clause",
    maintainer: "protobufpackages",
    summary: "Google's language-neutral, platform-neutral, extensible mechanism for serializing structured data.",
    description: "https://developers.google.com/protocol-buffers/",
    time: "2024-05-23 14:46:50",
    release_notes: "* See github.com/protocolbuffers/protobuf/releases/tag/v27.0 for release notes.",
    archive: ReleaseArchive {
        size: 243961,
        link: "https://pecl.php.net/get/protobuf-4.27.0",
    },
    extracted_link: "package.4.27.0.xml",
}

[noahbald]

Hey @jdm are there any plans to update the dependencies of markup5ever_rcdom to include the fixes?

---

edit:

Just found there's a verion =0.5.0-unnoficial -- which works better but it still seems to be moving/removing the xmlns-prefixed attributes

<!-- before -->
<svg xmlns="http://www.w3.org/2000/svg" xmlns:sodipodi="http://inkscape.sourceforge.net/DTD/sodipodi-0.dtd">
    <sodipodi:namedview>
        ...
    </sodipodi:namedview>

    <path d="..." sodipodi:nodetypes="cccc"/>
</svg>

<!-- after: xmlns:sodipodi moved to child -->
<svg xmlns="http://www.w3.org/2000/svg">
    <sodipodi:namedview xmlns:sodipodi="http://inkscape.sourceforge.net/DTD/sodipodi-0.dtd">
        ...
    </sodipodi:namedview>

    <path d="..." sodipodi:nodetypes="cccc"></path>
</svg>

<!-- before -->
<svg xmlns="http://www.w3.org/2000/svg" xmlns:sodipodi="http://inkscape.sourceforge.net/DTD/sodipodi-0.dtd">
    <path d="..." sodipodi:nodetypes="cccc"/>
</svg>

<!-- after: xmlns:sodipodi removed entirely -->
<svg xmlns="http://www.w3.org/2000/svg">
    <path d="..." sodipodi:nodetypes="cccc"></path>
</svg>

[jdm]

The published rcdom crates are not owned by the html5ever developers.

[noahbald]

Oh that's good to know. I think this issue may be coming from the parsing step, which in the example I gave uses xml5ever 0.20.0 and markup5ever 0.14

the example is based similarly on the original comment (#538 (comment))

[noahbald]

@jdm will this issue be re-opened this issue based on my last comment? I don't think the root xmlns'd attributes should be removed if the specified namespace is used as a prefix by descendants.