chore: move DefaultEmitter to own module

1 files changed, 427 insertions, 0 deletions

diff --git a/src/default_emitter.rs b/src/default_emitter.rs
new file mode 100644
index 0000000..c957b20
--- /dev/null
+++ b/src/default_emitter.rs
@@ -0,0 +1,427 @@
+use std::collections::btree_map::Entry;
+use std::collections::BTreeSet;
+use std::collections::VecDeque;
+use std::mem;
+use std::ops::Range;
+
+use crate::offset::NoopOffset;
+use crate::offset::Offset;
+use crate::token::{AttrValueSyntax, Comment, Doctype, EndTag, StartTag, Token};
+use crate::Emitter;
+use crate::Error;
+
+/// The default implementation of [`Emitter`], used to produce tokens.
+///
+/// # Warning
+///
+/// * Using the DefaultEmitter without calling [`Tokenizer::set_state`]
+///   results in wrong state transitions:
+///
+///   ```
+///   # use html5tokenizer::{DefaultEmitter, Event, Tokenizer, Token};
+///   let emitter = DefaultEmitter::default();
+///   let html = "<script><b>";
+///   let mut tokens = Tokenizer::new(html, emitter).flatten();
+///   assert!(matches!(tokens.next().unwrap(), Event::Token(Token::StartTag(tag)) if tag.name == "script"));
+///   assert!(matches!(tokens.next().unwrap(), Event::Token(Token::StartTag(tag)) if tag.name == "b"));
+///   ```
+///
+/// [`Tokenizer::set_state`]: crate::Tokenizer::set_state
+pub struct DefaultEmitter<O = NoopOffset> {
+    current_characters: String,
+    current_token: Option<Token<O>>,
+    current_attribute: Option<(String, crate::token::AttrInternal<O>)>,
+    seen_attributes: BTreeSet<String>,
+    emitted_tokens: VecDeque<Token<O>>,
+    attr_in_end_tag_span: Option<Range<O>>,
+}
+
+impl<O> Default for DefaultEmitter<O> {
+    fn default() -> Self {
+        DefaultEmitter {
+            current_characters: String::new(),
+            current_token: None,
+            current_attribute: None,
+            seen_attributes: BTreeSet::new(),
+            emitted_tokens: VecDeque::new(),
+            attr_in_end_tag_span: None,
+        }
+    }
+}
+
+impl<O: Offset> Emitter<O> for DefaultEmitter<O> {
+    type Token = Token<O>;
+
+    fn emit_eof(&mut self) {
+        self.flush_current_characters();
+    }
+
+    fn emit_error(&mut self, error: Error, span: Range<O>) {
+        self.push_error(error, span);
+    }
+
+    fn pop_token(&mut self) -> Option<Self::Token> {
+        self.emitted_tokens.pop_back()
+    }
+
+    fn emit_string(&mut self, s: &str) {
+        self.current_characters.push_str(s);
+    }
+
+    fn init_start_tag(&mut self, tag_offset: O, name_offset: O) {
+        self.current_token = Some(Token::StartTag(StartTag {
+            span: tag_offset..O::default(),
+            self_closing: false,
+            name: String::new(),
+            attributes: Default::default(),
+            name_span: name_offset..O::default(),
+        }));
+    }
+    fn init_end_tag(&mut self, tag_offset: O, name_offset: O) {
+        self.current_token = Some(Token::EndTag(EndTag {
+            span: tag_offset..O::default(),
+            name: String::new(),
+            name_span: name_offset..O::default(),
+        }));
+        self.seen_attributes.clear();
+    }
+
+    fn init_comment(&mut self, data_start_offset: O) {
+        self.current_token = Some(Token::Comment(Comment {
+            data: String::new(),
+            data_span: data_start_offset..O::default(),
+        }));
+    }
+    fn emit_current_tag(&mut self, offset: O) {
+        self.flush_current_attribute();
+        let mut token = self.current_token.take().unwrap();
+        match &mut token {
+            Token::EndTag(tag) => {
+                if !self.seen_attributes.is_empty() {
+                    let span = self.attr_in_end_tag_span.take().unwrap();
+                    self.push_error(Error::EndTagWithAttributes, span);
+                }
+                self.seen_attributes.clear();
+                tag.span.end = offset;
+            }
+            Token::StartTag(tag) => {
+                tag.span.end = offset;
+            }
+            _ => debug_assert!(false),
+        }
+        self.emit_token(token);
+    }
+    fn emit_current_comment(&mut self, data_end_offset: O) {
+        let mut token = self.current_token.take().unwrap();
+        if let Token::Comment(comment) = &mut token {
+            comment.data_span.end = data_end_offset;
+        } else {
+            debug_assert!(false);
+        }
+        self.emit_token(token);
+    }
+
+    fn emit_current_doctype(&mut self, offset: O) {
+        let Some(Token::Doctype(mut doctype)) = self.current_token.take() else {
+            debug_assert!(false);
+            return;
+        };
+        doctype.span.end = offset;
+        self.emit_token(Token::Doctype(doctype));
+    }
+
+    fn set_self_closing(&mut self, slash_span: Range<O>) {
+        let tag = self.current_token.as_mut().unwrap();
+        match tag {
+            Token::StartTag(StartTag {
+                ref mut self_closing,
+                ..
+            }) => {
+                *self_closing = true;
+            }
+            Token::EndTag(_) => {
+                self.emit_error(Error::EndTagWithTrailingSolidus, slash_span);
+            }
+            _ => {
+                debug_assert!(false);
+            }
+        }
+    }
+    fn set_force_quirks(&mut self) {
+        match self.current_token {
+            Some(Token::Doctype(ref mut doctype)) => doctype.force_quirks = true,
+            _ => debug_assert!(false),
+        }
+    }
+    fn push_tag_name(&mut self, s: &str) {
+        match self.current_token {
+            Some(Token::StartTag(StartTag { ref mut name, .. })) => {
+                name.push_str(s);
+            }
+            Some(Token::EndTag(EndTag { ref mut name, .. })) => {
+                name.push_str(s);
+            }
+            _ => debug_assert!(false),
+        }
+    }
+
+    fn terminate_tag_name(&mut self, offset: O) {
+        match self.current_token {
+            Some(Token::StartTag(StartTag {
+                ref mut name_span, ..
+            })) => {
+                name_span.end = offset;
+            }
+            Some(Token::EndTag(EndTag {
+                ref mut name_span, ..
+            })) => {
+                name_span.end = offset;
+            }
+            _ => debug_assert!(false),
+        }
+    }
+
+    fn push_comment(&mut self, s: &str) {
+        match self.current_token {
+            Some(Token::Comment(Comment { ref mut data, .. })) => data.push_str(s),
+            _ => debug_assert!(false),
+        }
+    }
+
+    fn init_doctype_name(&mut self, offset: O) {
+        let Some(Token::Doctype(doctype)) = &mut self.current_token else {
+            debug_assert!(false);
+            return;
+        };
+        doctype.name = Some("".into());
+        doctype.name_span.start = offset;
+    }
+
+    fn push_doctype_name(&mut self, s: &str) {
+        match self.current_token {
+            Some(Token::Doctype(Doctype {
+                name: Some(ref mut name),
+                ..
+            })) => name.push_str(s),
+            _ => debug_assert!(false),
+        }
+    }
+
+    fn terminate_doctype_name(&mut self, offset: O) {
+        let Some(Token::Doctype(doctype)) = &mut self.current_token else {
+            debug_assert!(false);
+            return;
+        };
+        doctype.name_span.end = offset;
+    }
+
+    fn init_doctype(&mut self, offset: O) {
+        self.current_token = Some(Token::Doctype(Doctype {
+            name: None,
+            force_quirks: false,
+            public_id: None,
+            system_id: None,
+            span: offset..O::default(),
+            name_span: O::default()..O::default(),
+            public_id_span: O::default()..O::default(),
+            system_id_span: O::default()..O::default(),
+        }));
+    }
+
+    fn init_attribute_name(&mut self, offset: O) {
+        self.flush_current_attribute();
+        self.current_attribute = Some((
+            String::new(),
+            crate::token::AttrInternal {
+                name_span: offset..O::default(),
+                value: String::new(),
+                value_span: O::default()..O::default(),
+                value_syntax: None,
+            },
+        ));
+    }
+    fn init_attribute_value(&mut self, syntax: AttrValueSyntax, offset: O) {
+        let (_, current_attribute) = self.current_attribute.as_mut().unwrap();
+        current_attribute.value_span.start = offset;
+        current_attribute.value_syntax = Some(syntax);
+    }
+
+    fn push_attribute_name(&mut self, s: &str) {
+        let current_attr = self.current_attribute.as_mut().unwrap();
+        current_attr.0.push_str(s);
+    }
+
+    fn terminate_attribute_name(&mut self, offset: O) {
+        let current_attr = self.current_attribute.as_mut().unwrap();
+        current_attr.1.name_span.end = offset;
+    }
+
+    fn push_attribute_value(&mut self, s: &str) {
+        let current_attr = self.current_attribute.as_mut().unwrap();
+        current_attr.1.value.push_str(s);
+    }
+
+    fn terminate_attribute_value(&mut self, offset: O) {
+        let current_attr = self.current_attribute.as_mut().unwrap();
+        current_attr.1.value_span.end = offset;
+    }
+
+    fn init_doctype_public_id(&mut self, offset: O) {
+        let Some(Token::Doctype(doctype)) = &mut self.current_token else {
+            debug_assert!(false);
+            return;
+        };
+        doctype.public_id = Some("".to_owned());
+        doctype.public_id_span.start = offset;
+    }
+    fn init_doctype_system_id(&mut self, offset: O) {
+        let Some(Token::Doctype(doctype)) = &mut self.current_token else {
+            debug_assert!(false);
+            return;
+        };
+        doctype.system_id = Some("".to_owned());
+        doctype.system_id_span.start = offset;
+    }
+    fn push_doctype_public_id(&mut self, s: &str) {
+        if let Some(Token::Doctype(Doctype {
+            public_id: Some(ref mut id),
+            ..
+        })) = self.current_token
+        {
+            id.push_str(s);
+        } else {
+            debug_assert!(false);
+        }
+    }
+
+    fn terminate_doctype_public_id(&mut self, offset: O) {
+        if let Some(Token::Doctype(Doctype {
+            ref mut public_id_span,
+            ..
+        })) = self.current_token
+        {
+            public_id_span.end = offset;
+        } else {
+            debug_assert!(false);
+        }
+    }
+
+    fn push_doctype_system_id(&mut self, s: &str) {
+        if let Some(Token::Doctype(Doctype {
+            system_id: Some(ref mut id),
+            ..
+        })) = self.current_token
+        {
+            id.push_str(s);
+        } else {
+            debug_assert!(false);
+        }
+    }
+
+    fn terminate_doctype_system_id(&mut self, offset: O) {
+        if let Some(Token::Doctype(Doctype {
+            ref mut system_id_span,
+            ..
+        })) = self.current_token
+        {
+            system_id_span.end = offset;
+        } else {
+            debug_assert!(false);
+        }
+    }
+}
+
+impl<O> DefaultEmitter<O> {
+    fn emit_token(&mut self, token: Token<O>) {
+        self.flush_current_characters();
+        self.emitted_tokens.push_front(token);
+    }
+
+    fn flush_current_attribute(&mut self)
+    where
+        O: Offset,
+    {
+        if let Some((name, map_val)) = self.current_attribute.take() {
+            match self.current_token {
+                Some(Token::StartTag(ref mut tag)) => match tag.attributes.inner.entry(name) {
+                    Entry::Vacant(vacant) => {
+                        vacant.insert(map_val);
+                    }
+                    Entry::Occupied(_) => {
+                        self.push_error(Error::DuplicateAttribute, map_val.name_span);
+                    }
+                },
+                Some(Token::EndTag(_)) => {
+                    self.attr_in_end_tag_span = Some(map_val.name_span.clone());
+                    if !self.seen_attributes.insert(name) {
+                        self.push_error(Error::DuplicateAttribute, map_val.name_span);
+                    }
+                }
+                _ => {
+                    debug_assert!(false);
+                }
+            }
+        }
+    }
+
+    fn flush_current_characters(&mut self) {
+        if self.current_characters.is_empty() {
+            return;
+        }
+
+        let s = mem::take(&mut self.current_characters);
+        self.emit_token(Token::String(s));
+    }
+
+    fn push_error(&mut self, error: Error, span: Range<O>) {
+        // bypass character flushing in self.emit_token: we don't need the error location to be
+        // that exact
+        self.emitted_tokens.push_front(Token::Error { error, span });
+    }
+}
+
+/// The majority of our testing of the [`DefaultEmitter`] is done against the
+/// html5lib-tests in the html5lib integration test. This module only tests
+/// details that aren't present in the html5lib test data.
+#[cfg(test)]
+mod tests {
+    use super::DefaultEmitter;
+    use crate::token::{AttrValueSyntax, Token};
+    use crate::{Event, Tokenizer};
+
+    #[test]
+    fn test_attribute_value_syntax() {
+        let mut tokenizer = Tokenizer::new(
+            "<div empty unquoted=foo single-quoted='foo' double-quoted=\"foo\">",
+            DefaultEmitter::default(),
+        )
+        .flatten();
+        let Event::Token(Token::StartTag(start_tag)) = tokenizer.next().unwrap() else {
+            panic!("expected start tag");
+        };
+        assert_eq!(
+            start_tag.attributes.get("empty").unwrap().value_syntax(),
+            None
+        );
+        assert_eq!(
+            start_tag.attributes.get("unquoted").unwrap().value_syntax(),
+            Some(AttrValueSyntax::Unquoted)
+        );
+        assert_eq!(
+            start_tag
+                .attributes
+                .get("single-quoted")
+                .unwrap()
+                .value_syntax(),
+            Some(AttrValueSyntax::SingleQuoted)
+        );
+        assert_eq!(
+            start_tag
+                .attributes
+                .get("double-quoted")
+                .unwrap()
+                .value_syntax(),
+            Some(AttrValueSyntax::DoubleQuoted)
+        );
+    }
+}