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_to4(&mut self, needle: [u8; 4]) -> &'a [u8] {
91        let start = self.stream.idx;
92        let bytes = &self.stream.data()[start..];
93
94        let end = simd::find4(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)
122            .unwrap_or_else(|| self.stream.len() - start);
123
124        self.stream.idx += end;
125        Some(self.stream.slice(start, start + end))
126    }
127
128    fn skip_comment_with_start(&mut self, start: usize) -> &'a [u8] {
129        while !self.stream.is_eof() {
130            let idx = self.stream.idx;
131
132            if self
133                .stream
134                .slice_len(idx, constants::COMMENT.len())
135                .eq(constants::COMMENT)
136            {
137                self.stream.advance_by(constants::COMMENT.len());
138
139                let is_end_of_comment = self.stream.expect_and_skip_cond(b'>');
140
141                if is_end_of_comment {
142                    return self.stream.slice(start, self.stream.idx);
143                }
144            }
145
146            self.stream.advance();
147        }
148
149        &[]
150    }
151
152    fn parse_attribute(&mut self) -> Option<(&'a [u8], Option<&'a [u8]>)> {
153        let name = self.read_ident()?;
154        self.skip_whitespaces();
155
156        let has_value = self.stream.expect_and_skip_cond(b'=');
157        if !has_value {
158            return Some((name, None));
159        }
160
161        self.skip_whitespaces();
162
163        let value = if let Some(quote) = self.stream.expect_oneof_and_skip(&[b'"', b'\'']) {
164            self.read_to(quote)
165        } else {
166            self.read_to4([b' ', b'\n', b'/', b'>'])
167        };
168
169        Some((name, Some(value)))
170    }
171
172    fn parse_attributes(&mut self) -> Option<Attributes<'a>> {
173        let mut attributes = Attributes::new();
174
175        loop {
176            self.skip_whitespaces();
177
178            let cur = self.stream.current_cpy()?;
179
180            if simd::is_closing(cur) {
181                break;
182            }
183
184            if let Some((key, value)) = self.parse_attribute() {
185                let value: Option<Bytes<'a>> = value.map(Into::into);
186
187                match key {
188                    b"id" => attributes.id = value,
189                    b"class" => attributes.class = value,
190                    _ => attributes.raw.insert(key.into(), value),
191                };
192            }
193
194            if !simd::is_closing(self.stream.current_cpy()?) {
195                self.stream.advance();
196            }
197        }
198
199        Some(attributes)
200    }
201
202    #[inline]
203    fn add_to_parent(&mut self, handle: NodeHandle) {
204        if let Some(last) = self.stack.last() {
205            let last = self
206                .tags
207                .get_mut(last.get_inner() as usize)
208                .unwrap()
209                .as_tag_mut()
210                .unwrap();
211
212            last._children.push(handle);
213        } else {
214            self.ast.push(handle);
215        }
216    }
217
218    fn read_end(&mut self) {
219        self.stream.advance();
220
221        let closing_tag_name = self.read_to(b'>');
222        
223        self.stream.expect_and_skip_cond(b'>');
224
225        let closing_tag_matches_parent = self.stack.last()
226            .and_then(|last_handle| last_handle.get(self))
227            .and_then(|last_item| last_item.as_tag())
228            .map_or(false, |last_tag| last_tag.name() == closing_tag_name);
229
230        if !closing_tag_matches_parent {
231            return;
232        }
233
234        if let Some(handle) = self.stack.pop() {
235            let tag = self
236                .tags
237                .get_mut(handle.get_inner() as usize)
238                .unwrap()
239                .as_tag_mut()
240                .unwrap();
241
242            let ptr = self.stream.data().as_ptr() as usize;
243            let offset = tag._raw.as_ptr() as usize;
244            let offset = offset - ptr;
245
246            tag._raw = self.stream.slice(offset, self.stream.idx).into();
247
248            let (track_classes, track_ids) = (
249                self.options.is_tracking_classes(),
250                self.options.is_tracking_ids(),
251            );
252
253            if let (true, Some(bytes)) = (track_classes, &tag._attributes.class) {
254                let s = bytes
255                    .as_bytes_borrowed()
256                    .and_then(|x| std::str::from_utf8(x).ok())
257                    .map(|x| x.split_ascii_whitespace());
258
259                if let Some(s) = s {
260                    for class in s {
261                        self.classes
262                            .entry(class.into())
263                            .or_insert_with(InlineVec::new)
264                            .push(handle);
265                    }
266                }
267            }
268
269            if let (true, Some(bytes)) = (track_ids, &tag._attributes.id) {
270                self.ids.insert(bytes.clone(), handle);
271            }
272        }
273    }
274
275    #[cold]
276    #[inline(never)]
277    fn read_markdown(&mut self) -> Option<()> {
278        let start = self.stream.idx - 1; // position of the < which is needed when registering the comment
279
280        self.stream.advance(); // skip !
281
282        let is_comment = self
283            .stream
284            .slice_len(self.stream.idx, 2)
285            .eq(constants::COMMENT);
286
287        if is_comment {
288            let comment = self.skip_comment_with_start(start);
289            let comment = self.register_tag(Node::Comment(comment.into()));
290            self.add_to_parent(comment);
291        } else {
292            let tag = self.read_ident()?;
293
294            self.skip_whitespaces();
295
296            if simd::matches_case_insensitive(tag, *b"doctype") {
297                let doctype = self.read_ident()?;
298
299                let html5 = simd::matches_case_insensitive(doctype, *b"html");
300
301                if html5 {
302                    self.version = Some(HTMLVersion::HTML5);
303                }
304
305                self.skip_whitespaces();
306                self.stream.advance(); // skip >
307            }
308        }
309
310        Some(())
311    }
312
313    fn parse_tag(&mut self) -> Option<()> {
314        let start = self.stream.idx;
315
316        self.stream.advance();
317        self.skip_whitespaces();
318        let cur = self.stream.current_cpy()?;
319
320        match cur {
321            b'/' => self.read_end(),
322            b'!' => {
323                self.read_markdown();
324            }
325            _ => {
326                let name = self.read_ident()?;
327                self.skip_whitespaces();
328
329                let attr = self.parse_attributes()?;
330
331                let is_self_closing = self.stream.expect_and_skip_cond(b'/');
332
333                self.stream.expect_and_skip(b'>')?;
334
335                let this = self.register_tag(Node::Tag(HTMLTag::new(
336                    name.into(),
337                    attr,
338                    InlineVec::new(),
339                    self.stream.slice(start, self.stream.idx).into(),
340                )));
341
342                self.add_to_parent(this);
343
344                // some tags are self closing, so even though there might not be a /,
345                // we don't always want to push them to the stack
346                // e.g. <br><p>Hello</p>
347                // <p> should not be a subtag of <br>
348                if !is_self_closing && !constants::VOID_TAGS.contains(&name) {
349                    self.stack.push(this);
350                }
351            }
352        };
353
354        Some(())
355    }
356
357    pub(crate) fn parse_single(&mut self) -> Option<()> {
358        loop {
359            let cur = self.stream.current()?;
360
361            if *cur == b'<' {
362                self.parse_tag();
363            } else {
364                let raw = Node::Raw(self.read_to(b'<').into());
365                let handle = self.register_tag(raw);
366                self.add_to_parent(handle);
367            }
368        }
369    }
370
371    /// Resolves an internal Node ID obtained from a NodeHandle to a Node
372    #[inline]
373    pub fn resolve_node_id(&self, id: InnerNodeHandle) -> Option<&Node<'a>> {
374        self.tags.get(id as usize)
375    }
376
377    /// Resolves an internal Node ID obtained from a NodeHandle to a Node
378    #[inline]
379    pub fn resolve_node_id_mut(&mut self, id: InnerNodeHandle) -> Option<&mut Node<'a>> {
380        self.tags.get_mut(id as usize)
381    }
382
383    pub(crate) fn parse(&mut self) -> Result<(), ParseError> {
384        if self.stream.len() > u32::MAX as usize {
385            return Err(ParseError::InvalidLength);
386        }
387
388        while !self.stream.is_eof() {
389            self.parse_single();
390        }
391
392        Ok(())
393    }
394}
tl/parser/base.rs

tl/parser/
base.rs
