script: Implement Element::GetHTML and ShadowRoot::GetHTML (#36106)

* Serialize html fragments without going through html5ever

Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>

* Implement ShadowRoot::GetHtml / Element::GetHtml

Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>

* Update WPT expectations

Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>

* Propagate CanGc annotations

Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>

---------

Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>

components/script/dom/element.rs

@@ -17,6 +17,7 @@ use devtools_traits::AttrInfo;
 use dom_struct::dom_struct;
 use embedder_traits::InputMethodType;
 use euclid::default::{Rect, Size2D};
+use html5ever::serialize::TraversalScope;
 use html5ever::serialize::TraversalScope::{ChildrenOnly, IncludeNode};
 use html5ever::{
     LocalName, Namespace, Prefix, QualName, local_name, namespace_prefix, namespace_url, ns,
@@ -67,7 +68,9 @@ use crate::dom::attr::{Attr, AttrHelpersForLayout};
 use crate::dom::bindings::cell::{DomRefCell, Ref, RefMut, ref_filter_map};
 use crate::dom::bindings::codegen::Bindings::AttrBinding::AttrMethods;
 use crate::dom::bindings::codegen::Bindings::DocumentBinding::DocumentMethods;
-use crate::dom::bindings::codegen::Bindings::ElementBinding::{ElementMethods, ShadowRootInit};
+use crate::dom::bindings::codegen::Bindings::ElementBinding::{
+    ElementMethods, GetHTMLOptions, ShadowRootInit,
+};
 use crate::dom::bindings::codegen::Bindings::FunctionBinding::Function;
 use crate::dom::bindings::codegen::Bindings::HTMLTemplateElementBinding::HTMLTemplateElementMethods;
 use crate::dom::bindings::codegen::Bindings::NodeBinding::NodeMethods;
@@ -2951,17 +2954,31 @@ impl ElementMethods<crate::DomTypeHolder> for Element {
         Node::replace_all(Some(frag.upcast()), &target);
     }
 
+    /// <https://html.spec.whatwg.org/multipage/#dom-element-gethtml>
+    fn GetHTML(&self, options: &GetHTMLOptions, can_gc: CanGc) -> DOMString {
+        // > Element's getHTML(options) method steps are to return the result of HTML fragment serialization
+        // > algorithm with this, options["serializableShadowRoots"], and options["shadowRoots"].
+        self.upcast::<Node>().html_serialize(
+            TraversalScope::ChildrenOnly(None),
+            options.serializableShadowRoots,
+            options.shadowRoots.clone(),
+            can_gc,
+        )
+    }
+
     /// <https://html.spec.whatwg.org/multipage/#dom-element-innerhtml>
-    fn GetInnerHTML(&self) -> Fallible<DOMString> {
+    fn GetInnerHTML(&self, can_gc: CanGc) -> Fallible<DOMString> {
         let qname = QualName::new(
             self.prefix().clone(),
             self.namespace().clone(),
             self.local_name().clone(),
         );
 
+        // FIXME: This should use the fragment serialization algorithm, which takes
+        // care of distinguishing between html/xml documents
         let result = if self.owner_document().is_html_document() {
             self.upcast::<Node>()
-                .html_serialize(ChildrenOnly(Some(qname)))
+                .html_serialize(ChildrenOnly(Some(qname)), false, vec![], can_gc)
         } else {
             self.upcast::<Node>()
                 .xml_serialize(XmlChildrenOnly(Some(qname)))
@@ -3001,9 +3018,12 @@ impl ElementMethods<crate::DomTypeHolder> for Element {
     }
 
     /// <https://html.spec.whatwg.org/multipage/#dom-element-outerhtml>
-    fn GetOuterHTML(&self) -> Fallible<DOMString> {
+    fn GetOuterHTML(&self, can_gc: CanGc) -> Fallible<DOMString> {
+        // FIXME: This should use the fragment serialization algorithm, which takes
+        // care of distinguishing between html/xml documents
         let result = if self.owner_document().is_html_document() {
-            self.upcast::<Node>().html_serialize(IncludeNode)
+            self.upcast::<Node>()
+                .html_serialize(IncludeNode, false, vec![], can_gc)
         } else {
             self.upcast::<Node>().xml_serialize(XmlIncludeNode)
         };

components/script/dom/node.rs

@@ -21,6 +21,7 @@ use constellation_traits::{
 use devtools_traits::NodeInfo;
 use dom_struct::dom_struct;
 use euclid::default::{Rect, Size2D, Vector2D};
+use html5ever::serialize::HtmlSerializer;
 use html5ever::{Namespace, Prefix, QualName, namespace_url, ns, serialize as html_serialize};
 use js::jsapi::JSObject;
 use js::rust::HandleObject;
@@ -110,6 +111,7 @@ use crate::dom::nodelist::NodeList;
 use crate::dom::processinginstruction::ProcessingInstruction;
 use crate::dom::range::WeakRangeVec;
 use crate::dom::raredata::NodeRareData;
+use crate::dom::servoparser::serialize_html_fragment;
 use crate::dom::shadowroot::{IsUserAgentWidget, LayoutShadowRootHelpers, ShadowRoot};
 use crate::dom::stylesheetlist::StyleSheetListOwner;
 use crate::dom::svgsvgelement::{LayoutSVGSVGElementHelpers, SVGSVGElement};
@@ -2862,22 +2864,34 @@ impl Node {
     pub(crate) fn html_serialize(
         &self,
         traversal_scope: html_serialize::TraversalScope,
+        serialize_shadow_roots: bool,
+        shadow_roots: Vec<DomRoot<ShadowRoot>>,
+        can_gc: CanGc,
     ) -> DOMString {
         let mut writer = vec![];
-        html_serialize::serialize(
+        let mut serializer = HtmlSerializer::new(
             &mut writer,
-            &self,
             html_serialize::SerializeOpts {
-                traversal_scope,
+                traversal_scope: traversal_scope.clone(),
                 ..Default::default()
             },
+        );
+
+        serialize_html_fragment(
+            self,
+            &mut serializer,
+            traversal_scope,
+            serialize_shadow_roots,
+            shadow_roots,
+            can_gc,
         )
-        .expect("Cannot serialize node");
+        .expect("Serializing node failed");
 
         // FIXME(ajeffrey): Directly convert UTF8 to DOMString
         DOMString::from(String::from_utf8(writer).unwrap())
     }
 
+    /// <https://w3c.github.io/DOM-Parsing/#dfn-xml-serialization>
     pub(crate) fn xml_serialize(
         &self,
         traversal_scope: xml_serialize::TraversalScope,
@@ -2895,14 +2909,23 @@ impl Node {
     }
 
     /// <https://html.spec.whatwg.org/multipage/#fragment-serializing-algorithm-steps>
-    pub(crate) fn fragment_serialization_algorithm(&self, require_well_formed: bool) -> DOMString {
+    pub(crate) fn fragment_serialization_algorithm(
+        &self,
+        require_well_formed: bool,
+        can_gc: CanGc,
+    ) -> DOMString {
         // Step 1. Let context document be node's node document.
         let context_document = self.owner_document();
 
         // Step 2. If context document is an HTML document, return the result of HTML fragment serialization algorithm
         // with node, false, and « ».
         if context_document.is_html_document() {
-            return self.html_serialize(html_serialize::TraversalScope::ChildrenOnly(None));
+            return self.html_serialize(
+                html_serialize::TraversalScope::ChildrenOnly(None),
+                false,
+                vec![],
+                can_gc,
+            );
         }
 
         // Step 3. Return the XML serialization of node given require well-formed.

components/script/dom/servoparser/html.rs

@@ -7,16 +7,19 @@
 use std::cell::Cell;
 use std::io;
 
-use html5ever::QualName;
 use html5ever::buffer_queue::BufferQueue;
 use html5ever::serialize::TraversalScope::IncludeNode;
 use html5ever::serialize::{AttrRef, Serialize, Serializer, TraversalScope};
 use html5ever::tokenizer::{Tokenizer as HtmlTokenizer, TokenizerOpts, TokenizerResult};
 use html5ever::tree_builder::{TreeBuilder, TreeBuilderOpts};
+use html5ever::{QualName, local_name, namespace_url, ns};
 use script_bindings::trace::CustomTraceable;
 use servo_url::ServoUrl;
+use xml5ever::LocalName;
 
 use crate::dom::bindings::codegen::Bindings::HTMLTemplateElementBinding::HTMLTemplateElementMethods;
+use crate::dom::bindings::codegen::Bindings::ShadowRootBinding::ShadowRootMode;
+use crate::dom::bindings::codegen::GenericBindings::ShadowRootBinding::ShadowRoot_Binding::ShadowRootMethods;
 use crate::dom::bindings::inheritance::{Castable, CharacterDataTypeId, NodeTypeId};
 use crate::dom::bindings::root::{Dom, DomRoot};
 use crate::dom::characterdata::CharacterData;
@@ -29,6 +32,7 @@ use crate::dom::htmltemplateelement::HTMLTemplateElement;
 use crate::dom::node::Node;
 use crate::dom::processinginstruction::ProcessingInstruction;
 use crate::dom::servoparser::{ParsingAlgorithm, Sink};
+use crate::dom::shadowroot::ShadowRoot;
 use crate::script_runtime::CanGc;
 
 #[derive(JSTraceable, MallocSizeOf)]
@@ -121,54 +125,91 @@ fn start_element<S: Serializer>(node: &Element, serializer: &mut S) -> io::Resul
     Ok(())
 }
 
-fn end_element<S: Serializer>(node: &Element, serializer: &mut S) -> io::Result<()> {
-    let name = QualName::new(None, node.namespace().clone(), node.local_name().clone());
-    serializer.end_elem(name)
-}
-
 enum SerializationCommand {
     OpenElement(DomRoot<Element>),
-    CloseElement(DomRoot<Element>),
+    CloseElement(QualName),
     SerializeNonelement(DomRoot<Node>),
+    SerializeShadowRoot(DomRoot<ShadowRoot>),
 }
 
 struct SerializationIterator {
     stack: Vec<SerializationCommand>,
+
+    /// Whether or not shadow roots should be serialized
+    serialize_shadow_roots: bool,
+
+    /// List of shadow root objects that should be serialized
+    shadow_roots: Vec<DomRoot<ShadowRoot>>,
 }
 
-fn rev_children_iter(n: &Node, can_gc: CanGc) -> impl Iterator<Item = DomRoot<Node>> + use<'_> {
-    if n.downcast::<Element>().is_some_and(|e| e.is_void()) {
-        return Node::new_document_node().rev_children();
-    }
-
-    match n.downcast::<HTMLTemplateElement>() {
-        Some(t) => t.Content(can_gc).upcast::<Node>().rev_children(),
-        None => n.rev_children(),
-    }
+enum SerializationChildrenIterator<C, S> {
+    None,
+    Children(C),
+    ShadowContents(S),
 }
 
 impl SerializationIterator {
-    fn new(node: &Node, skip_first: bool, can_gc: CanGc) -> SerializationIterator {
-        let mut ret = SerializationIterator { stack: vec![] };
+    fn new(
+        node: &Node,
+        skip_first: bool,
+        serialize_shadow_roots: bool,
+        shadow_roots: Vec<DomRoot<ShadowRoot>>,
+        can_gc: CanGc,
+    ) -> SerializationIterator {
+        let mut ret = SerializationIterator {
+            stack: vec![],
+            serialize_shadow_roots,
+            shadow_roots,
+        };
         if skip_first || node.is::<DocumentFragment>() || node.is::<Document>() {
-            for c in rev_children_iter(node, can_gc) {
-                ret.push_node(&c);
-            }
+            ret.handle_node_contents(node, can_gc);
         } else {
             ret.push_node(node);
         }
         ret
     }
 
-    fn push_node(&mut self, n: &Node) {
-        match n.downcast::<Element>() {
-            Some(e) => self
-                .stack
-                .push(SerializationCommand::OpenElement(DomRoot::from_ref(e))),
-            None => self.stack.push(SerializationCommand::SerializeNonelement(
-                DomRoot::from_ref(n),
-            )),
+    fn handle_node_contents(&mut self, node: &Node, can_gc: CanGc) {
+        if node.downcast::<Element>().is_some_and(Element::is_void) {
+            return;
         }
+
+        if let Some(template_element) = node.downcast::<HTMLTemplateElement>() {
+            for child in template_element
+                .Content(can_gc)
+                .upcast::<Node>()
+                .rev_children()
+            {
+                self.push_node(&child);
+            }
+        } else {
+            for child in node.rev_children() {
+                self.push_node(&child);
+            }
+        }
+
+        if let Some(shadow_root) = node.downcast::<Element>().and_then(Element::shadow_root) {
+            let should_be_serialized = (self.serialize_shadow_roots && shadow_root.Serializable()) ||
+                self.shadow_roots.contains(&shadow_root);
+            if !shadow_root.is_user_agent_widget() && should_be_serialized {
+                self.stack
+                    .push(SerializationCommand::SerializeShadowRoot(shadow_root));
+            }
+        }
+    }
+
+    fn push_node(&mut self, node: &Node) {
+        let Some(element) = node.downcast::<Element>() else {
+            self.stack.push(SerializationCommand::SerializeNonelement(
+                DomRoot::from_ref(node),
+            ));
+            return;
+        };
+
+        self.stack
+            .push(SerializationCommand::OpenElement(DomRoot::from_ref(
+                element,
+            )));
     }
 }
 
@@ -176,71 +217,139 @@ impl Iterator for SerializationIterator {
     type Item = SerializationCommand;
 
     fn next(&mut self) -> Option<SerializationCommand> {
-        let res = self.stack.pop();
+        let res = self.stack.pop()?;
 
-        if let Some(SerializationCommand::OpenElement(ref e)) = res {
-            self.stack
-                .push(SerializationCommand::CloseElement(e.clone()));
-            for c in rev_children_iter(e.upcast::<Node>(), CanGc::note()) {
-                self.push_node(&c);
-            }
+        match &res {
+            SerializationCommand::OpenElement(element) => {
+                let name = QualName::new(
+                    None,
+                    element.namespace().clone(),
+                    element.local_name().clone(),
+                );
+                self.stack.push(SerializationCommand::CloseElement(name));
+                self.handle_node_contents(element.upcast(), CanGc::note());
+            },
+            SerializationCommand::SerializeShadowRoot(shadow_root) => {
+                self.stack
+                    .push(SerializationCommand::CloseElement(QualName::new(
+                        None,
+                        ns!(),
+                        local_name!("template"),
+                    )));
+                self.handle_node_contents(shadow_root.upcast(), CanGc::note());
+            },
+            _ => {},
         }
 
-        res
+        Some(res)
     }
 }
 
+/// <https://html.spec.whatwg.org/multipage/#html-fragment-serialisation-algorithm>
+pub(crate) fn serialize_html_fragment<S: Serializer>(
+    node: &Node,
+    serializer: &mut S,
+    traversal_scope: TraversalScope,
+    serialize_shadow_roots: bool,
+    shadow_roots: Vec<DomRoot<ShadowRoot>>,
+    can_gc: CanGc,
+) -> io::Result<()> {
+    let iter = SerializationIterator::new(
+        node,
+        traversal_scope != IncludeNode,
+        serialize_shadow_roots,
+        shadow_roots,
+        can_gc,
+    );
+
+    for cmd in iter {
+        match cmd {
+            SerializationCommand::OpenElement(n) => {
+                start_element(&n, serializer)?;
+            },
+            SerializationCommand::CloseElement(name) => {
+                serializer.end_elem(name)?;
+            },
+            SerializationCommand::SerializeNonelement(n) => match n.type_id() {
+                NodeTypeId::DocumentType => {
+                    let doctype = n.downcast::<DocumentType>().unwrap();
+                    serializer.write_doctype(doctype.name())?;
+                },
+
+                NodeTypeId::CharacterData(CharacterDataTypeId::Text(_)) => {
+                    let cdata = n.downcast::<CharacterData>().unwrap();
+                    serializer.write_text(&cdata.data())?;
+                },
+
+                NodeTypeId::CharacterData(CharacterDataTypeId::Comment) => {
+                    let cdata = n.downcast::<CharacterData>().unwrap();
+                    serializer.write_comment(&cdata.data())?;
+                },
+
+                NodeTypeId::CharacterData(CharacterDataTypeId::ProcessingInstruction) => {
+                    let pi = n.downcast::<ProcessingInstruction>().unwrap();
+                    let data = pi.upcast::<CharacterData>().data();
+                    serializer.write_processing_instruction(pi.target(), &data)?;
+                },
+
+                NodeTypeId::DocumentFragment(_) => {},
+
+                NodeTypeId::Document(_) => panic!("Can't serialize Document node itself"),
+                NodeTypeId::Element(_) => panic!("Element shouldn't appear here"),
+                NodeTypeId::Attr => panic!("Attr shouldn't appear here"),
+            },
+            SerializationCommand::SerializeShadowRoot(shadow_root) => {
+                // Shadow roots are serialized as template elements with a fixed set of
+                // attributes. Because these template elements don't actually exist in the DOM
+                // we have to make up a vector of attributes ourselves.
+                let mut attributes = vec![];
+                let mut push_attribute = |name, value| {
+                    let qualified_name = QualName::new(None, ns!(), LocalName::from(name));
+                    attributes.push((qualified_name, value))
+                };
+
+                let mode = if shadow_root.Mode() == ShadowRootMode::Open {
+                    "open"
+                } else {
+                    "closed"
+                };
+                push_attribute("shadowrootmode", mode);
+
+                if shadow_root.DelegatesFocus() {
+                    push_attribute("shadowrootdelegatesfocus", "");
+                }
+
+                if shadow_root.Serializable() {
+                    push_attribute("shadowrootserializable", "");
+                }
+
+                if shadow_root.Clonable() {
+                    push_attribute("shadowrootclonable", "");
+                }
+
+                let name = QualName::new(None, ns!(), local_name!("template"));
+                serializer.start_elem(name, attributes.iter().map(|(a, b)| (a, *b)))?;
+            },
+        }
+    }
+
+    Ok(())
+}
+
+// TODO: This trait confuses the concepts of XML serialization and HTML serialization and
+// the impl should go away eventually
 impl Serialize for &Node {
-    fn serialize<S: Serializer>(
-        &self,
-        serializer: &mut S,
-        traversal_scope: TraversalScope,
-    ) -> io::Result<()> {
-        let node = *self;
-
-        let iter = SerializationIterator::new(node, traversal_scope != IncludeNode, CanGc::note());
-
-        for cmd in iter {
-            match cmd {
-                SerializationCommand::OpenElement(n) => {
-                    start_element(&n, serializer)?;
-                },
-
-                SerializationCommand::CloseElement(n) => {
-                    end_element(&n, serializer)?;
-                },
-
-                SerializationCommand::SerializeNonelement(n) => match n.type_id() {
-                    NodeTypeId::DocumentType => {
-                        let doctype = n.downcast::<DocumentType>().unwrap();
-                        serializer.write_doctype(doctype.name())?;
-                    },
-
-                    NodeTypeId::CharacterData(CharacterDataTypeId::Text(_)) => {
-                        let cdata = n.downcast::<CharacterData>().unwrap();
-                        serializer.write_text(&cdata.data())?;
-                    },
-
-                    NodeTypeId::CharacterData(CharacterDataTypeId::Comment) => {
-                        let cdata = n.downcast::<CharacterData>().unwrap();
-                        serializer.write_comment(&cdata.data())?;
-                    },
-
-                    NodeTypeId::CharacterData(CharacterDataTypeId::ProcessingInstruction) => {
-                        let pi = n.downcast::<ProcessingInstruction>().unwrap();
-                        let data = pi.upcast::<CharacterData>().data();
-                        serializer.write_processing_instruction(pi.target(), &data)?;
-                    },
-
-                    NodeTypeId::DocumentFragment(_) => {},
-
-                    NodeTypeId::Document(_) => panic!("Can't serialize Document node itself"),
-                    NodeTypeId::Element(_) => panic!("Element shouldn't appear here"),
-                    NodeTypeId::Attr => panic!("Attr shouldn't appear here"),
-                },
-            }
-        }
-
-        Ok(())
+    fn serialize<S>(&self, serializer: &mut S, traversal_scope: TraversalScope) -> io::Result<()>
+    where
+        S: Serializer,
+    {
+        serialize_html_fragment(
+            self,
+            serializer,
+            traversal_scope,
+            false,
+            vec![],
+            CanGc::note(),
+        )
     }
 }

components/script/dom/servoparser/mod.rs

@@ -81,6 +81,8 @@ mod html;
 mod prefetch;
 mod xml;
 
+pub(crate) use html::serialize_html_fragment;
+
 #[dom_struct]
 /// The parser maintains two input streams: one for input from script through
 /// document.write(), and one for input from network.

components/script/dom/shadowroot.rs

@@ -7,6 +7,7 @@ use std::collections::HashMap;
 use std::collections::hash_map::Entry;
 
 use dom_struct::dom_struct;
+use html5ever::serialize::TraversalScope;
 use servo_arc::Arc;
 use style::author_styles::AuthorStyles;
 use style::dom::TElement;
@@ -17,6 +18,7 @@ use stylo_atoms::Atom;
 
 use crate::conversions::Convert;
 use crate::dom::bindings::cell::DomRefCell;
+use crate::dom::bindings::codegen::Bindings::ElementBinding::GetHTMLOptions;
 use crate::dom::bindings::codegen::Bindings::HTMLSlotElementBinding::HTMLSlotElement_Binding::HTMLSlotElementMethods;
 use crate::dom::bindings::codegen::Bindings::ShadowRootBinding::ShadowRoot_Binding::ShadowRootMethods;
 use crate::dom::bindings::codegen::Bindings::ShadowRootBinding::{
@@ -405,11 +407,24 @@ impl ShadowRootMethods<crate::DomTypeHolder> for ShadowRoot {
         })
     }
 
+    /// <https://html.spec.whatwg.org/multipage/#dom-shadowroot-gethtml>
+    fn GetHTML(&self, options: &GetHTMLOptions, can_gc: CanGc) -> DOMString {
+        // > ShadowRoot's getHTML(options) method steps are to return the result of HTML fragment serialization
+        // >  algorithm with this, options["serializableShadowRoots"], and options["shadowRoots"].
+        self.upcast::<Node>().html_serialize(
+            TraversalScope::ChildrenOnly(None),
+            options.serializableShadowRoots,
+            options.shadowRoots.clone(),
+            can_gc,
+        )
+    }
+
     /// <https://html.spec.whatwg.org/multipage/#dom-shadowroot-innerhtml>
-    fn InnerHTML(&self) -> DOMString {
+    fn InnerHTML(&self, can_gc: CanGc) -> DOMString {
         // ShadowRoot's innerHTML getter steps are to return the result of running fragment serializing
         // algorithm steps with this and true.
-        self.upcast::<Node>().fragment_serialization_algorithm(true)
+        self.upcast::<Node>()
+            .fragment_serialization_algorithm(true, can_gc)
     }
 
     /// <https://html.spec.whatwg.org/multipage/#dom-shadowroot-innerhtml>