ps_blitz_html/
html_sink.rs

1//! An implementation for Html5ever's sink trait, allowing us to parse HTML into a DOM.
2
3use html5ever::ParseOpts;
4use html5ever::tokenizer::TokenizerOpts;
5use html5ever::tree_builder::TreeBuilderOpts;
6use std::borrow::Cow;
7use std::cell::{Cell, Ref, RefCell, RefMut};
8
9use blitz_dom::node::Attribute;
10use blitz_dom::{DocumentMutator, HtmlParserProvider};
11use html5ever::{
12    QualName,
13    tendril::{StrTendril, TendrilSink},
14    tree_builder::{ElementFlags, NodeOrText, QuirksMode, TreeSink},
15};
16
17/// Convert an html5ever Attribute which uses tendril for its value to a blitz Attribute
18/// which uses String.
19fn html5ever_to_blitz_attr(attr: html5ever::Attribute) -> Attribute {
20    Attribute {
21        name: attr.name,
22        value: attr.value.to_string(),
23    }
24}
25
26#[derive(Copy, Clone, Default, Debug)]
27pub struct HtmlProvider;
28
29impl HtmlParserProvider for HtmlProvider {
30    fn parse_inner_html<'m2, 'doc2>(
31        &self,
32        mutr: &'m2 mut DocumentMutator<'doc2>,
33        element_id: usize,
34        html: &str,
35    ) {
36        DocumentHtmlParser::parse_inner_html_into_mutator(mutr, element_id, html);
37    }
38}
39
40pub struct DocumentHtmlParser<'m, 'doc> {
41    document_mutator: RefCell<&'m mut DocumentMutator<'doc>>,
42
43    /// Errors that occurred during parsing.
44    pub errors: RefCell<Vec<Cow<'static, str>>>,
45
46    /// The document's quirks mode.
47    pub quirks_mode: Cell<QuirksMode>,
48    pub is_xml: bool,
49}
50
51impl<'m, 'doc> DocumentHtmlParser<'m, 'doc> {
52    #[track_caller]
53    /// Get a mutable borrow of the DocumentMutator
54    fn mutr(&self) -> RefMut<'_, &'m mut DocumentMutator<'doc>> {
55        self.document_mutator.borrow_mut()
56    }
57}
58
59impl<'m, 'doc> DocumentHtmlParser<'m, 'doc> {
60    pub fn new(mutr: &'m mut DocumentMutator<'doc>) -> DocumentHtmlParser<'m, 'doc> {
61        DocumentHtmlParser {
62            document_mutator: RefCell::new(mutr),
63            errors: RefCell::new(Vec::new()),
64            quirks_mode: Cell::new(QuirksMode::NoQuirks),
65            is_xml: false,
66        }
67    }
68
69    /// Detects documents without an XML or DOCTYPE declaration whose root `<html>` element
70    /// declares the XHTML namespace (e.g. `<html xmlns="http://www.w3.org/1999/xhtml">`)
71    fn root_element_has_xhtml_namespace(html: &str) -> bool {
72        let rest = html.trim_start_matches('\u{feff}').trim_start();
73        let Some(rest) = rest.strip_prefix("<html") else {
74            return false;
75        };
76        let Some(tag_end) = rest.find('>') else {
77            return false;
78        };
79        rest[..tag_end].contains("xmlns=\"http://www.w3.org/1999/xhtml\"")
80            || rest[..tag_end].contains("xmlns='http://www.w3.org/1999/xhtml'")
81    }
82
83    pub fn parse_into_mutator<'a, 'd>(mutr: &'a mut DocumentMutator<'d>, html: &str) {
84        let mut sink = DocumentHtmlParser::new(mutr);
85
86        let is_xhtml_doc = html.starts_with("<?xml")
87            || html.starts_with("<!DOCTYPE") && {
88                let first_line = html.lines().next().unwrap();
89                first_line.contains("XHTML") || first_line.contains("xhtml")
90            }
91            || Self::root_element_has_xhtml_namespace(html);
92
93        if is_xhtml_doc {
94            // Parse as XHTML
95            sink.is_xml = true;
96            xml5ever::driver::parse_document(sink, Default::default())
97                .from_utf8()
98                .read_from(&mut html.as_bytes())
99                .unwrap();
100        } else {
101            // Parse as HTML
102            sink.is_xml = false;
103            let opts = ParseOpts {
104                tokenizer: TokenizerOpts::default(),
105                tree_builder: TreeBuilderOpts {
106                    exact_errors: false,
107                    scripting_enabled: false, // Enables parsing of <noscript> tags
108                    iframe_srcdoc: false,
109                    drop_doctype: true,
110                    quirks_mode: QuirksMode::NoQuirks,
111                },
112            };
113            html5ever::parse_document(sink, opts)
114                .from_utf8()
115                .read_from(&mut html.as_bytes())
116                .unwrap();
117        }
118    }
119
120    pub fn parse_inner_html_into_mutator<'a, 'd>(
121        mutr: &'a mut DocumentMutator<'d>,
122        element_id: usize,
123        html: &str,
124    ) {
125        let sink = DocumentHtmlParser::new(mutr);
126
127        let opts = ParseOpts {
128            tokenizer: TokenizerOpts::default(),
129            tree_builder: TreeBuilderOpts {
130                exact_errors: false,
131                scripting_enabled: false, // Enables parsing of <noscript> tags
132                iframe_srcdoc: false,
133                drop_doctype: true,
134                quirks_mode: QuirksMode::NoQuirks,
135            },
136        };
137        html5ever::driver::parse_fragment_for_element(sink, opts, element_id, false, None)
138            .from_utf8()
139            .read_from(&mut html.as_bytes())
140            .unwrap();
141
142        // html5ever creates a new fragment root node under the document node and parses the nodes into that fragment root.
143        // So here we move the children of the fragment root to element_id and then remove the fragment root
144        let fragment_root_id = mutr.last_child_id(0).unwrap();
145        let child_ids = mutr.child_ids(fragment_root_id);
146        mutr.append_children(element_id, &child_ids);
147        mutr.remove_node(fragment_root_id);
148    }
149}
150
151impl<'m, 'doc> TreeSink for DocumentHtmlParser<'m, 'doc> {
152    type Output = ();
153
154    // we use the ID of the nodes in the tree as the handle
155    type Handle = usize;
156
157    type ElemName<'a>
158        = Ref<'a, QualName>
159    where
160        Self: 'a;
161
162    fn finish(self) -> Self::Output {
163        #[cfg(feature = "tracing")]
164        for error in self.errors.borrow().iter() {
165            tracing::error!("{error}");
166        }
167    }
168
169    fn parse_error(&self, msg: Cow<'static, str>) {
170        self.errors.borrow_mut().push(msg);
171    }
172
173    fn get_document(&self) -> Self::Handle {
174        0
175    }
176
177    fn elem_name<'a>(&'a self, target: &'a Self::Handle) -> Self::ElemName<'a> {
178        Ref::map(self.document_mutator.borrow(), |docm| {
179            docm.element_name(*target)
180                .expect("TreeSink::elem_name called on a node which is not an element!")
181        })
182    }
183
184    fn create_element(
185        &self,
186        name: QualName,
187        attrs: Vec<html5ever::Attribute>,
188        _flags: ElementFlags,
189    ) -> Self::Handle {
190        let attrs = attrs.into_iter().map(html5ever_to_blitz_attr).collect();
191        self.mutr().create_element(name, attrs)
192    }
193
194    fn create_comment(&self, _text: StrTendril) -> Self::Handle {
195        self.mutr().create_comment_node()
196    }
197
198    fn create_pi(&self, _target: StrTendril, _data: StrTendril) -> Self::Handle {
199        self.mutr().create_comment_node()
200    }
201
202    fn append(&self, parent_id: &Self::Handle, child: NodeOrText<Self::Handle>) {
203        match child {
204            NodeOrText::AppendNode(id) => self.mutr().append_children(*parent_id, &[id]),
205            // If content to append is text, first attempt to append it to the last child of parent.
206            // Else create a new text node and append it to the parent
207            NodeOrText::AppendText(text) => {
208                let last_child_id = self.mutr().last_child_id(*parent_id);
209                let has_appended = if let Some(id) = last_child_id {
210                    self.mutr().append_text_to_node(id, &text).is_ok()
211                } else {
212                    false
213                };
214                if !has_appended {
215                    let new_child_id = self.mutr().create_text_node(&text);
216                    self.mutr().append_children(*parent_id, &[new_child_id]);
217                }
218            }
219        }
220    }
221
222    // Note: The tree builder promises we won't have a text node after the insertion point.
223    // https://github.com/servo/html5ever/blob/main/rcdom/lib.rs#L338
224    fn append_before_sibling(&self, sibling_id: &Self::Handle, new_node: NodeOrText<Self::Handle>) {
225        match new_node {
226            NodeOrText::AppendNode(id) => self.mutr().insert_nodes_before(*sibling_id, &[id]),
227            // If content to append is text, first attempt to append it to the node before sibling_node
228            // Else create a new text node and insert it before sibling_node
229            NodeOrText::AppendText(text) => {
230                let previous_sibling_id = self.mutr().previous_sibling_id(*sibling_id);
231                let has_appended = if let Some(id) = previous_sibling_id {
232                    self.mutr().append_text_to_node(id, &text).is_ok()
233                } else {
234                    false
235                };
236                if !has_appended {
237                    let new_child_id = self.mutr().create_text_node(&text);
238                    self.mutr()
239                        .insert_nodes_before(*sibling_id, &[new_child_id]);
240                }
241            }
242        };
243    }
244
245    fn append_based_on_parent_node(
246        &self,
247        element: &Self::Handle,
248        prev_element: &Self::Handle,
249        child: NodeOrText<Self::Handle>,
250    ) {
251        if self.mutr().node_has_parent(*element) {
252            self.append_before_sibling(element, child);
253        } else {
254            self.append(prev_element, child);
255        }
256    }
257
258    fn append_doctype_to_document(
259        &self,
260        _name: StrTendril,
261        _public_id: StrTendril,
262        _system_id: StrTendril,
263    ) {
264        // Ignore. We don't care about the DOCTYPE for now.
265    }
266
267    fn get_template_contents(&self, target: &Self::Handle) -> Self::Handle {
268        // TODO: implement templates properly. This should allow to function like regular elements.
269        *target
270    }
271
272    fn same_node(&self, x: &Self::Handle, y: &Self::Handle) -> bool {
273        x == y
274    }
275
276    fn set_quirks_mode(&self, mode: QuirksMode) {
277        self.quirks_mode.set(mode);
278    }
279
280    fn add_attrs_if_missing(&self, target: &Self::Handle, attrs: Vec<html5ever::Attribute>) {
281        let attrs = attrs.into_iter().map(html5ever_to_blitz_attr).collect();
282        self.mutr().add_attrs_if_missing(*target, attrs);
283    }
284
285    fn remove_from_parent(&self, target: &Self::Handle) {
286        self.mutr().remove_node(*target);
287    }
288
289    fn reparent_children(&self, old_parent_id: &Self::Handle, new_parent_id: &Self::Handle) {
290        self.mutr()
291            .reparent_children(*old_parent_id, *new_parent_id);
292    }
293}
294
295#[test]
296fn parses_some_html() {
297    use blitz_dom::{BaseDocument, DocumentConfig};
298
299    let html = "<!DOCTYPE html><html><body><h1>hello world</h1></body></html>";
300    let mut doc = BaseDocument::new(DocumentConfig::default());
301    let mut mutr = doc.mutate();
302    let sink = DocumentHtmlParser::new(&mut mutr);
303
304    html5ever::parse_document(sink, Default::default())
305        .from_utf8()
306        .read_from(&mut html.as_bytes())
307        .unwrap();
308
309    drop(mutr);
310    doc.print_tree()
311
312    // Now our tree should have some nodes in it
313}
ps_blitz_html/html_sink.rs

ps_blitz_html/
html_sink.rs
