fix: make doctype id spans encoding-independent

2 files changed, 73 insertions, 14 deletions

diff --git a/src/emitter.rs b/src/emitter.rs
index 341d335..9fdf967 100644
--- a/src/emitter.rs
+++ b/src/emitter.rs
@@ -158,6 +158,11 @@ pub trait Emitter<O> {
     /// If the current token is not a doctype, this method may panic.
     fn init_doctype_public_id(&mut self, offset: O);
 
+    /// Called after the last [`push_doctype_public_id`] call for a DOCTYPE public identifier.
+    ///
+    /// [`push_doctype_public_id`]: Self::push_doctype_public_id
+    fn terminate_doctype_public_id(&mut self, offset: O) {}
+
     /// Assuming the _current token_ is a doctype, set its "system identifier" to the empty string.
     ///
     /// If the current token is not a doctype, this method may panic.
@@ -172,6 +177,11 @@ pub trait Emitter<O> {
     ///
     /// If the current token is not a doctype, this method may panic.
     fn push_doctype_system_id(&mut self, s: &str);
+
+    /// Called after the last [`push_doctype_system_id`] call for a DOCTYPE system identifier.
+    ///
+    /// [`push_doctype_system_id`]: Self::push_doctype_system_id
+    fn terminate_doctype_system_id(&mut self, offset: O) {}
 }
 
 /// The default implementation of [`Emitter`], used to produce tokens.
@@ -410,8 +420,8 @@ impl<O: Offset> Emitter<O> for DefaultEmitter<O> {
             public_id: None,
             system_id: None,
             span: offset..O::default(),
-            public_id_offset: O::default(),
-            system_id_offset: O::default(),
+            public_id_span: O::default()..O::default(),
+            system_id_span: O::default()..O::default(),
         }));
     }
 
@@ -459,7 +469,7 @@ impl<O: Offset> Emitter<O> for DefaultEmitter<O> {
             return;
         };
         doctype.public_id = Some("".to_owned());
-        doctype.public_id_offset = offset;
+        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 {
@@ -467,7 +477,7 @@ impl<O: Offset> Emitter<O> for DefaultEmitter<O> {
             return;
         };
         doctype.system_id = Some("".to_owned());
-        doctype.system_id_offset = offset;
+        doctype.system_id_span.start = offset;
     }
     fn push_doctype_public_id(&mut self, s: &str) {
         if let Some(Token::Doctype(Doctype {
@@ -480,6 +490,19 @@ impl<O: Offset> Emitter<O> for DefaultEmitter<O> {
             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),
@@ -491,6 +514,18 @@ impl<O: Offset> Emitter<O> for DefaultEmitter<O> {
             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);
+        }
+    }
 }
 
 /// An HTML start tag, such as `<p>` or `<a>`.
@@ -573,24 +608,24 @@ pub struct Doctype<O> {
     /// The source code span of the doctype.
     pub span: Range<O>,
 
-    /// The source offset of the public identifier.
-    public_id_offset: O,
+    /// The span of the public identifier.
+    public_id_span: Range<O>,
 
-    /// The source offset of the system identifier.
-    system_id_offset: O,
+    /// The span of the system identifier.
+    system_id_span: Range<O>,
 }
 
 impl<O: Offset> Doctype<O> {
-    /// Calculates the span of the public identifier and returns it.
+    /// Returns the span of the public identifier.
     pub fn public_id_span(&self) -> Option<Range<O>> {
-        let public_id = self.public_id.as_ref()?;
-        Some(self.public_id_offset..self.public_id_offset + public_id.len())
+        self.public_id.as_ref()?;
+        Some(self.public_id_span.clone())
     }
 
-    /// Calculates the span of the system identifier and returns it.
+    /// Returns the span of the system identifier.
     pub fn system_id_span(&self) -> Option<Range<O>> {
-        let system_id = self.system_id.as_ref()?;
-        Some(self.system_id_offset..self.system_id_offset + system_id.len())
+        self.system_id.as_ref()?;
+        Some(self.system_id_span.clone())
     }
 }
 
diff --git a/src/machine.rs b/src/machine.rs
index 5c5c533..f00af0a 100644
--- a/src/machine.rs
+++ b/src/machine.rs
@@ -1359,6 +1359,8 @@ where
         },
         State::DoctypePublicIdentifierDoubleQuoted => match slf.read_char()? {
             Some('"') => {
+                slf.emitter
+                    .terminate_doctype_public_id(slf.position_before_match);
                 slf.state = State::AfterDoctypePublicIdentifier;
                 Ok(ControlToken::Continue)
             }
@@ -1368,6 +1370,8 @@ where
                 Ok(ControlToken::Continue)
             }
             Some('>') => {
+                slf.emitter
+                    .terminate_doctype_public_id(slf.position_before_match);
                 slf.emit_error(Error::AbruptDoctypePublicIdentifier);
                 slf.emitter.set_force_quirks();
                 slf.state = State::Data;
@@ -1375,6 +1379,8 @@ where
                 Ok(ControlToken::Continue)
             }
             None => {
+                slf.emitter
+                    .terminate_doctype_public_id(slf.reader.position());
                 slf.emit_error(Error::EofInDoctype);
                 slf.emitter.set_force_quirks();
                 slf.emitter.emit_current_doctype(slf.reader.position());
@@ -1387,6 +1393,8 @@ where
         },
         State::DoctypePublicIdentifierSingleQuoted => match slf.read_char()? {
             Some('\'') => {
+                slf.emitter
+                    .terminate_doctype_public_id(slf.position_before_match);
                 slf.state = State::AfterDoctypePublicIdentifier;
                 Ok(ControlToken::Continue)
             }
@@ -1396,6 +1404,8 @@ where
                 Ok(ControlToken::Continue)
             }
             Some('>') => {
+                slf.emitter
+                    .terminate_doctype_public_id(slf.position_before_match);
                 slf.emit_error(Error::AbruptDoctypePublicIdentifier);
                 slf.emitter.set_force_quirks();
                 slf.state = State::Data;
@@ -1403,6 +1413,8 @@ where
                 Ok(ControlToken::Continue)
             }
             None => {
+                slf.emitter
+                    .terminate_doctype_public_id(slf.reader.position());
                 slf.emit_error(Error::EofInDoctype);
                 slf.emitter.set_force_quirks();
                 slf.emitter.emit_current_doctype(slf.reader.position());
@@ -1553,6 +1565,8 @@ where
         },
         State::DoctypeSystemIdentifierDoubleQuoted => match slf.read_char()? {
             Some('"') => {
+                slf.emitter
+                    .terminate_doctype_system_id(slf.position_before_match);
                 slf.state = State::AfterDoctypeSystemIdentifier;
                 Ok(ControlToken::Continue)
             }
@@ -1562,6 +1576,8 @@ where
                 Ok(ControlToken::Continue)
             }
             Some('>') => {
+                slf.emitter
+                    .terminate_doctype_system_id(slf.position_before_match);
                 slf.emit_error(Error::AbruptDoctypeSystemIdentifier);
                 slf.emitter.set_force_quirks();
                 slf.state = State::Data;
@@ -1569,6 +1585,8 @@ where
                 Ok(ControlToken::Continue)
             }
             None => {
+                slf.emitter
+                    .terminate_doctype_system_id(slf.reader.position());
                 slf.emit_error(Error::EofInDoctype);
                 slf.emitter.set_force_quirks();
                 slf.emitter.emit_current_doctype(slf.reader.position());
@@ -1581,6 +1599,8 @@ where
         },
         State::DoctypeSystemIdentifierSingleQuoted => match slf.read_char()? {
             Some('\'') => {
+                slf.emitter
+                    .terminate_doctype_system_id(slf.position_before_match);
                 slf.state = State::AfterDoctypeSystemIdentifier;
                 Ok(ControlToken::Continue)
             }
@@ -1590,6 +1610,8 @@ where
                 Ok(ControlToken::Continue)
             }
             Some('>') => {
+                slf.emitter
+                    .terminate_doctype_system_id(slf.position_before_match);
                 slf.emit_error(Error::AbruptDoctypeSystemIdentifier);
                 slf.emitter.set_force_quirks();
                 slf.state = State::Data;
@@ -1597,6 +1619,8 @@ where
                 Ok(ControlToken::Continue)
             }
             None => {
+                slf.emitter
+                    .terminate_doctype_system_id(slf.reader.position());
                 slf.emit_error(Error::EofInDoctype);
                 slf.emitter.set_force_quirks();
                 slf.emitter.emit_current_doctype(slf.reader.position());