tl/parser/
base.rs

1use super::{
2    constants,
3    handle::NodeHandle,
4    tag::{Attributes, HTMLTag, Node},
5};
6use crate::InnerNodeHandle;
7use crate::{bytes::Bytes, inline::vec::InlineVec, simd, ParseError};
8use crate::{stream::Stream, ParserOptions};
9use std::collections::HashMap;
10
11/// A list of HTML nodes
12pub type Tree<'a> = Vec<Node<'a>>;
13
14/// Inline class vector
15pub type ClassVec = InlineVec<NodeHandle, 2>;
16
17/// HTML Version (<!DOCTYPE>)
18#[derive(Debug, Copy, Clone, PartialEq)]
19#[repr(C)]
20pub enum HTMLVersion {
21    /// HTML Version 5
22    HTML5,
23    /// Strict HTML 4.01
24    StrictHTML401,
25    /// Transitional HTML 4.01
26    TransitionalHTML401,
27    /// Frameset HTML 4.01:
28    FramesetHTML401,
29}
30/// The main HTML parser
31///
32/// Users of this library are not supposed to directly construct this struct.
33/// Instead, users must call `tl::parse()` and use the returned `VDom`.
34#[derive(Debug)]
35pub struct Parser<'a> {
36    /// The inner stream that is used to iterate through the HTML source
37    pub(crate) stream: Stream<'a, u8>,
38    pub(crate) stack: Vec<NodeHandle>,
39    /// Specified options for this HTML parser
40    pub(crate) options: ParserOptions,
41    /// A global collection of all HTML tags that appear in the source code
42    ///
43    /// HTML Nodes contain indicies into this vector
44    pub(crate) tags: Tree<'a>,
45    /// The topmost HTML nodes
46    pub(crate) ast: Vec<NodeHandle>,
47    /// A HashMap that maps Tag ID to a Node ID
48    pub(crate) ids: HashMap<Bytes<'a>, NodeHandle>,
49    /// A HashMap that maps Tag Class to a Node ID
50    pub(crate) classes: HashMap<Bytes<'a>, ClassVec>,
51    /// The current HTML version, if set
52    pub(crate) version: Option<HTMLVersion>,
53}
54
55impl<'a> Parser<'a> {
56    pub(crate) fn new(input: &str, options: ParserOptions) -> Parser<'_> {
57        Parser {
58            stack: Vec::with_capacity(4),
59            options,
60            tags: Vec::new(),
61            stream: Stream::new(input.as_bytes()),
62            ast: Vec::new(),
63            ids: HashMap::new(),
64            classes: HashMap::new(),
65            version: None,
66        }
67    }
68
69    #[inline(always)]
70    fn register_tag(&mut self, node: Node<'a>) -> NodeHandle {
71        self.tags.push(node);
72        NodeHandle::new((self.tags.len() - 1) as u32)
73    }
74
75    #[inline(always)]
76    fn skip_whitespaces(&mut self) {
77        self.read_while2(b' ', b'\n');
78    }
79
80    fn read_to(&mut self, needle: u8) -> &'a [u8] {
81        let start = self.stream.idx;
82        let bytes = &self.stream.data()[start..];
83
84        let end = simd::find(bytes, needle).unwrap_or_else(|| self.stream.len() - start);
85
86        self.stream.idx += end;
87        self.stream.slice(start, start + end)
88    }
89
90    fn read_to3(&mut self, needle: [u8; 3]) -> &'a [u8] {
91        let start = self.stream.idx;
92        let bytes = &self.stream.data()[start..];
93
94        let end = simd::find3(bytes, needle).unwrap_or_else(|| self.stream.len() - start);
95
96        self.stream.idx += end;
97        self.stream.slice(start, start + end)
98    }
99
100    fn read_while2(&mut self, needle1: u8, needle2: u8) -> Option<()> {
101        loop {
102            let ch = self.stream.current_cpy()?;
103
104            let eq1 = ch == needle1;
105            let eq2 = ch == needle2;
106
107            if !eq1 & !eq2 {
108                return Some(());
109            }
110
111            self.stream.advance();
112        }
113    }
114
115    fn read_ident(&mut self) -> Option<&'a [u8]> {
116        let start = self.stream.idx;
117        let bytes = &self.stream.data()[start..];
118
119        // If we do not find any characters that are not identifiers
120        // then we are probably at the end of the stream
121        let end = simd::search_non_ident(bytes).unwrap_or_else(|| self.stream.len() - start);
122
123        self.stream.idx += end;
124        Some(self.stream.slice(start, start + end))
125    }
126
127    fn skip_comment_with_start(&mut self, start: usize) -> &'a [u8] {
128        while !self.stream.is_eof() {
129            let idx = self.stream.idx;
130
131            if self
132                .stream
133                .slice_len(idx, constants::COMMENT.len())
134                .eq(constants::COMMENT)
135            {
136                self.stream.advance_by(constants::COMMENT.len());
137
138                let is_end_of_comment = self.stream.expect_and_skip_cond(b'>');
139
140                if is_end_of_comment {
141                    return self.stream.slice(start, self.stream.idx);
142                }
143            }
144
145            self.stream.advance();
146        }
147
148        &[]
149    }
150
151    fn parse_attribute(&mut self) -> Option<(&'a [u8], Option<&'a [u8]>)> {
152        let name = self.read_ident()?;
153        self.skip_whitespaces();
154
155        let has_value = self.stream.expect_and_skip_cond(b'=');
156        if !has_value {
157            return Some((name, None));
158        }
159
160        self.skip_whitespaces();
161
162        let value = if let Some(quote) = self.stream.expect_oneof_and_skip(b"\"'") {
163            self.read_to(quote)
164        } else {
165            self.read_to3([b' ', b'\n', b'>'])
166        };
167
168        Some((name, Some(value)))
169    }
170
171    fn parse_attributes(&mut self) -> Option<Attributes<'a>> {
172        let mut attributes = Attributes::new();
173
174        loop {
175            self.skip_whitespaces();
176
177            let cur = self.stream.current_cpy()?;
178
179            if simd::is_closing(cur) {
180                break;
181            }
182
183            if let Some((key, value)) = self.parse_attribute() {
184                let value: Option<Bytes<'a>> = value.map(Into::into);
185
186                match key {
187                    b"id" => attributes.id = value,
188                    b"class" => attributes.class = value,
189                    _ => attributes.raw.insert(key.into(), value),
190                };
191            }
192
193            if !simd::is_closing(self.stream.current_cpy()?) {
194                self.stream.advance();
195            }
196        }
197
198        Some(attributes)
199    }
200
201    #[inline]
202    fn add_to_parent(&mut self, handle: NodeHandle) {
203        if let Some(last) = self.stack.last() {
204            let last = self
205                .tags
206                .get_mut(last.get_inner() as usize)
207                .unwrap()
208                .as_tag_mut()
209                .unwrap();
210
211            last._children.push(handle);
212        } else {
213            self.ast.push(handle);
214        }
215    }
216
217    fn read_end(&mut self) {
218        self.stream.advance();
219
220        let closing_tag_name = self.read_to(b'>');
221
222        self.stream.expect_and_skip_cond(b'>');
223
224        let closing_tag_matches_parent = self
225            .stack
226            .last()
227            .and_then(|last_handle| last_handle.get(self))
228            .and_then(|last_item| last_item.as_tag())
229            .is_some_and(|last_tag| last_tag.name() == closing_tag_name);
230
231        if !closing_tag_matches_parent {
232            return;
233        }
234
235        if let Some(handle) = self.stack.pop() {
236            let tag = self
237                .tags
238                .get_mut(handle.get_inner() as usize)
239                .unwrap()
240                .as_tag_mut()
241                .unwrap();
242
243            let ptr = self.stream.data().as_ptr() as usize;
244            let offset = tag._raw.as_ptr() as usize;
245            let offset = offset - ptr;
246
247            tag._raw = self.stream.slice(offset, self.stream.idx).into();
248
249            let (track_classes, track_ids) = (
250                self.options.is_tracking_classes(),
251                self.options.is_tracking_ids(),
252            );
253
254            if let (true, Some(bytes)) = (track_classes, &tag._attributes.class) {
255                let s = bytes
256                    .as_bytes_borrowed()
257                    .and_then(|x| std::str::from_utf8(x).ok())
258                    .map(|x| x.split_ascii_whitespace());
259
260                if let Some(s) = s {
261                    for class in s {
262                        self.classes
263                            .entry(class.into())
264                            .or_insert_with(InlineVec::new)
265                            .push(handle);
266                    }
267                }
268            }
269
270            if let (true, Some(bytes)) = (track_ids, &tag._attributes.id) {
271                self.ids.insert(bytes.clone(), handle);
272            }
273        }
274    }
275
276    #[cold]
277    #[inline(never)]
278    fn read_markdown(&mut self) -> Option<()> {
279        let start = self.stream.idx - 1; // position of the < which is needed when registering the comment
280
281        self.stream.advance(); // skip !
282
283        let is_comment = self
284            .stream
285            .slice_len(self.stream.idx, 2)
286            .eq(constants::COMMENT);
287
288        if is_comment {
289            let comment = self.skip_comment_with_start(start);
290            let comment = self.register_tag(Node::Comment(comment.into()));
291            self.add_to_parent(comment);
292        } else {
293            let tag = self.read_ident()?;
294
295            self.skip_whitespaces();
296
297            if simd::matches_case_insensitive(tag, *b"doctype") {
298                let doctype = self.read_ident()?;
299
300                let html5 = simd::matches_case_insensitive(doctype, *b"html");
301
302                if html5 {
303                    self.version = Some(HTMLVersion::HTML5);
304                }
305
306                self.skip_whitespaces();
307                self.stream.advance(); // skip >
308            }
309        }
310
311        Some(())
312    }
313
314    fn parse_tag(&mut self) -> Option<()> {
315        let start = self.stream.idx;
316
317        self.stream.advance();
318        self.skip_whitespaces();
319        let cur = self.stream.current_cpy()?;
320
321        match cur {
322            b'/' => self.read_end(),
323            b'!' => {
324                self.read_markdown();
325            }
326            _ => {
327                let name = self.read_ident()?;
328                self.skip_whitespaces();
329
330                let attr = self.parse_attributes()?;
331
332                let is_self_closing = self.stream.expect_and_skip_cond(b'/');
333
334                self.stream.expect_and_skip(b'>')?;
335
336                let this = self.register_tag(Node::Tag(HTMLTag::new(
337                    name.into(),
338                    attr,
339                    InlineVec::new(),
340                    self.stream.slice(start, self.stream.idx).into(),
341                )));
342
343                self.add_to_parent(this);
344
345                // some tags are self closing, so even though there might not be a /,
346                // we don't always want to push them to the stack
347                // e.g. <br><p>Hello</p>
348                // <p> should not be a subtag of <br>
349                if !is_self_closing && !constants::VOID_TAGS.contains(&name) {
350                    self.stack.push(this);
351                }
352            }
353        };
354
355        Some(())
356    }
357
358    pub(crate) fn parse_single(&mut self) -> Option<()> {
359        loop {
360            let cur = self.stream.current()?;
361
362            if *cur == b'<' {
363                self.parse_tag();
364            } else {
365                let raw = Node::Raw(self.read_to(b'<').into());
366                let handle = self.register_tag(raw);
367                self.add_to_parent(handle);
368            }
369        }
370    }
371
372    /// Resolves an internal Node ID obtained from a NodeHandle to a Node
373    #[inline]
374    pub fn resolve_node_id(&self, id: InnerNodeHandle) -> Option<&Node<'a>> {
375        self.tags.get(id as usize)
376    }
377
378    /// Resolves an internal Node ID obtained from a NodeHandle to a Node
379    #[inline]
380    pub fn resolve_node_id_mut(&mut self, id: InnerNodeHandle) -> Option<&mut Node<'a>> {
381        self.tags.get_mut(id as usize)
382    }
383
384    pub(crate) fn parse(&mut self) -> Result<(), ParseError> {
385        if self.stream.len() > u32::MAX as usize {
386            return Err(ParseError::InvalidLength);
387        }
388
389        while !self.stream.is_eof() {
390            self.parse_single();
391        }
392
393        Ok(())
394    }
395}
tl/parser/base.rs

tl/parser/
base.rs
