justhtml 0.12.0__py3-none-any.whl

justhtml/treebuilder_modes.py

@@ -0,0 +1,2016 @@
+# ruff: noqa: S101, RUF012
+# mypy: disable-error-code="attr-defined, has-type, var-annotated, assignment"
+
+from __future__ import annotations
+
+from typing import Any
+
+from .constants import (
+    FORMATTING_ELEMENTS,
+    HEADING_ELEMENTS,
+)
+from .node import SimpleDomNode, TemplateNode
+from .tokens import CharacterTokens, CommentToken, EOFToken, Tag, TokenSinkResult
+from .treebuilder_utils import (
+    InsertionMode,
+    doctype_error_and_quirks,
+    is_all_whitespace,
+)
+
+
+class TreeBuilderModesMixin:
+    def _handle_doctype(self, token: Any) -> Any:
+        if self.mode != InsertionMode.INITIAL:
+            self._parse_error("unexpected-doctype")
+            return TokenSinkResult.Continue
+
+        doctype = token.doctype
+        parse_error, quirks_mode = doctype_error_and_quirks(doctype, self.iframe_srcdoc)
+
+        node = SimpleDomNode("!doctype", data=doctype)
+        self.document.append_child(node)
+
+        if parse_error:
+            self._parse_error("unknown-doctype")
+
+        self._set_quirks_mode(quirks_mode)
+        self.mode = InsertionMode.BEFORE_HTML
+        return TokenSinkResult.Continue
+
+    def _mode_initial(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            if is_all_whitespace(token.data):
+                return None
+            self._parse_error("expected-doctype-but-got-chars")
+            self._set_quirks_mode("quirks")
+            return ("reprocess", InsertionMode.BEFORE_HTML, token)
+        if isinstance(token, CommentToken):
+            self._append_comment_to_document(token.data)
+            return None
+        if isinstance(token, EOFToken):
+            self._parse_error("expected-doctype-but-got-eof")
+            self._set_quirks_mode("quirks")
+            self.mode = InsertionMode.BEFORE_HTML
+            return ("reprocess", InsertionMode.BEFORE_HTML, token)
+        # Only Tags remain - no DOCTYPE seen, so quirks mode
+        if token.kind == Tag.START:
+            self._parse_error("expected-doctype-but-got-start-tag", tag_name=token.name, token=token)
+        else:
+            self._parse_error("expected-doctype-but-got-end-tag", tag_name=token.name, token=token)
+        self._set_quirks_mode("quirks")
+        return ("reprocess", InsertionMode.BEFORE_HTML, token)
+
+    def _mode_before_html(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens) and is_all_whitespace(token.data):
+            return None
+        if isinstance(token, CommentToken):
+            self._append_comment_to_document(token.data)
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                self._create_root(token.attrs)
+                self.mode = InsertionMode.BEFORE_HEAD
+                return None
+            if token.kind == Tag.END and token.name in {"head", "body", "html", "br"}:
+                self._create_root({})
+                self.mode = InsertionMode.BEFORE_HEAD
+                return ("reprocess", InsertionMode.BEFORE_HEAD, token)
+            if token.kind == Tag.END:
+                # Ignore other end tags
+                self._parse_error("unexpected-end-tag-before-html", tag_name=token.name)
+                return None
+        if isinstance(token, EOFToken):
+            self._create_root({})
+            self.mode = InsertionMode.BEFORE_HEAD
+            return ("reprocess", InsertionMode.BEFORE_HEAD, token)
+
+        if isinstance(token, CharacterTokens):
+            stripped = token.data.lstrip("\t\n\f\r ")
+            if len(stripped) != len(token.data):
+                token = CharacterTokens(stripped)
+
+        self._create_root({})
+        self.mode = InsertionMode.BEFORE_HEAD
+        return ("reprocess", InsertionMode.BEFORE_HEAD, token)
+
+    def _mode_before_head(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            data = token.data or ""
+            if "\x00" in data:
+                self._parse_error("invalid-codepoint-before-head")
+                data = data.replace("\x00", "")
+                if not data:
+                    return None
+            if is_all_whitespace(data):
+                return None
+            token = CharacterTokens(data)
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                # Duplicate html tag - add attributes to existing html element
+                # Note: open_elements[0] is always html at this point (created in BEFORE_HTML mode)
+                html = self.open_elements[0]
+                self._add_missing_attributes(html, token.attrs)
+                return None
+            if token.kind == Tag.START and token.name == "head":
+                head = self._insert_element(token, push=True)
+                self.head_element = head
+                self.mode = InsertionMode.IN_HEAD
+                return None
+            if token.kind == Tag.END and token.name in {"head", "body", "html", "br"}:
+                self.head_element = self._insert_phantom("head")
+                self.mode = InsertionMode.IN_HEAD
+                return ("reprocess", InsertionMode.IN_HEAD, token)
+            if token.kind == Tag.END:
+                # Ignore other end tags
+                self._parse_error("unexpected-end-tag-before-head", tag_name=token.name)
+                return None
+        if isinstance(token, EOFToken):
+            self.head_element = self._insert_phantom("head")
+            self.mode = InsertionMode.IN_HEAD
+            return ("reprocess", InsertionMode.IN_HEAD, token)
+
+        self.head_element = self._insert_phantom("head")
+        self.mode = InsertionMode.IN_HEAD
+        return ("reprocess", InsertionMode.IN_HEAD, token)
+
+    def _mode_in_head(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            if is_all_whitespace(token.data):
+                self._append_text(token.data)
+                return None
+            data = token.data or ""
+            i = 0
+            while i < len(data) and data[i] in "\t\n\f\r ":
+                i += 1
+            leading_ws = data[:i]
+            remaining = data[i:]
+            if leading_ws:
+                current = self.open_elements[-1] if self.open_elements else None
+                if current is not None and current.has_child_nodes():
+                    self._append_text(leading_ws)
+            self._pop_current()
+            self.mode = InsertionMode.AFTER_HEAD
+            return ("reprocess", InsertionMode.AFTER_HEAD, CharacterTokens(remaining))
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                # Pop head and transition to AFTER_HEAD, then reprocess
+                self._pop_current()
+                self.mode = InsertionMode.AFTER_HEAD
+                return ("reprocess", InsertionMode.AFTER_HEAD, token)
+            if token.kind == Tag.START and token.name in {"base", "basefont", "bgsound", "link", "meta"}:
+                self._insert_element(token, push=False)
+                return None
+            if token.kind == Tag.START and token.name == "template":
+                self._insert_element(token, push=True)
+                self._push_formatting_marker()
+                self.frameset_ok = False
+                self.mode = InsertionMode.IN_TEMPLATE
+                self.template_modes.append(InsertionMode.IN_TEMPLATE)
+                return None
+            if token.kind == Tag.END and token.name == "template":
+                # Check if template is on the stack (don't use scope check as table blocks it)
+                has_template = any(node.name == "template" for node in self.open_elements)
+                if not has_template:
+                    return None
+                self._generate_implied_end_tags()
+                self._pop_until_inclusive("template")
+                self._clear_active_formatting_up_to_marker()
+                # template_modes always non-empty here since we passed has_template check
+                self.template_modes.pop()
+                self._reset_insertion_mode()
+                return None
+            if token.kind == Tag.START and token.name in {"title", "style", "script", "noframes"}:
+                self._insert_element(token, push=True)
+                self.original_mode = self.mode
+                self.mode = InsertionMode.TEXT
+                return None
+            if token.kind == Tag.START and token.name == "noscript":
+                # Scripting is disabled: parse noscript content as HTML
+                self._insert_element(token, push=True)
+                self.mode = InsertionMode.IN_HEAD_NOSCRIPT
+                return None
+            if token.kind == Tag.END and token.name == "head":
+                self._pop_current()
+                self.mode = InsertionMode.AFTER_HEAD
+                return None
+            if token.kind == Tag.END and token.name in {"body", "html", "br"}:
+                self._pop_current()
+                self.mode = InsertionMode.AFTER_HEAD
+                return ("reprocess", InsertionMode.AFTER_HEAD, token)
+        if isinstance(token, EOFToken):
+            self._pop_current()
+            self.mode = InsertionMode.AFTER_HEAD
+            return ("reprocess", InsertionMode.AFTER_HEAD, token)
+
+        self._pop_current()
+        self.mode = InsertionMode.AFTER_HEAD
+        return ("reprocess", InsertionMode.AFTER_HEAD, token)
+
+    def _mode_in_head_noscript(self, token: Any) -> Any:
+        """Handle tokens in 'in head noscript' insertion mode (scripting disabled)."""
+        if isinstance(token, CharacterTokens):
+            data = token.data or ""
+            # Whitespace: process using in head rules
+            if is_all_whitespace(data):
+                return self._mode_in_head(token)
+            # Non-whitespace: parse error, pop noscript, reprocess in head
+            self._parse_error("unexpected-start-tag", tag_name="text")
+            self._pop_current()  # Pop noscript
+            self.mode = InsertionMode.IN_HEAD
+            return ("reprocess", InsertionMode.IN_HEAD, token)
+        if isinstance(token, CommentToken):
+            return self._mode_in_head(token)
+        if isinstance(token, Tag):
+            if token.kind == Tag.START:
+                if token.name == "html":
+                    return self._mode_in_body(token)
+                if token.name in {"basefont", "bgsound", "link", "meta", "noframes", "style"}:
+                    return self._mode_in_head(token)
+                if token.name in {"head", "noscript"}:
+                    self._parse_error("unexpected-start-tag", tag_name=token.name)
+                    return None  # Ignore
+                # Any other start tag: parse error, pop noscript, reprocess in head
+                self._parse_error("unexpected-start-tag", tag_name=token.name)
+                self._pop_current()  # Pop noscript
+                self.mode = InsertionMode.IN_HEAD
+                return ("reprocess", InsertionMode.IN_HEAD, token)
+            # token.kind == Tag.END
+            if token.name == "noscript":
+                self._pop_current()  # Pop noscript
+                self.mode = InsertionMode.IN_HEAD
+                return None
+            if token.name == "br":
+                self._parse_error("unexpected-end-tag", tag_name=token.name)
+                self._pop_current()  # Pop noscript
+                self.mode = InsertionMode.IN_HEAD
+                return ("reprocess", InsertionMode.IN_HEAD, token)
+            # Any other end tag: parse error, ignore
+            self._parse_error("unexpected-end-tag", tag_name=token.name)
+            return None
+        if isinstance(token, EOFToken):
+            self._parse_error("expected-closing-tag-but-got-eof", tag_name="noscript")
+            self._pop_current()  # Pop noscript
+            self.mode = InsertionMode.IN_HEAD
+            return ("reprocess", InsertionMode.IN_HEAD, token)
+        # All token types are handled above - CharacterTokens, CommentToken, Tag, EOFToken
+        return None  # pragma: no cover
+
+    def _mode_after_head(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            data = token.data or ""
+            if "\x00" in data:
+                self._parse_error("invalid-codepoint-in-body")
+                data = data.replace("\x00", "")
+            if "\x0c" in data:
+                self._parse_error("invalid-codepoint-in-body")
+                data = data.replace("\x0c", "")
+            if not data or is_all_whitespace(data):
+                if data:
+                    self._append_text(data)
+                return None
+            self._insert_body_if_missing()
+            return ("reprocess", InsertionMode.IN_BODY, CharacterTokens(data))
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                self._insert_body_if_missing()
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.START and token.name == "body":
+                self._insert_element(token, push=True)
+                self.mode = InsertionMode.IN_BODY
+                self.frameset_ok = False
+                return None
+            if token.kind == Tag.START and token.name == "frameset":
+                self._insert_element(token, push=True)
+                self.mode = InsertionMode.IN_FRAMESET
+                return None
+            # Special handling: input type="hidden" doesn't create body or affect frameset_ok
+            if token.kind == Tag.START and token.name == "input":
+                input_type = None
+                for name, value in token.attrs.items():
+                    if name == "type":
+                        input_type = (value or "").lower()
+                        break
+                if input_type == "hidden":
+                    # Parse error but ignore - don't create body, don't insert element
+                    self._parse_error("unexpected-hidden-input-after-head")
+                    return None
+                # Non-hidden input creates body
+                self._insert_body_if_missing()
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.START and token.name in {
+                "base",
+                "basefont",
+                "bgsound",
+                "link",
+                "meta",
+                "title",
+                "style",
+                "script",
+                "noscript",
+            }:
+                self.open_elements.append(self.head_element)
+                result = self._mode_in_head(token)
+                # Remove the head element from wherever it is in the stack
+                # (it might not be at the end if we inserted other elements like <title>)
+                self.open_elements.remove(self.head_element)
+                return result
+            if token.kind == Tag.START and token.name == "template":
+                # Template in after-head needs special handling:
+                # Process in IN_HEAD mode, which will switch to IN_TEMPLATE
+                # Don't remove head from stack - let normal processing continue
+                self.open_elements.append(self.head_element)
+                self.mode = InsertionMode.IN_HEAD
+                return ("reprocess", InsertionMode.IN_HEAD, token)
+            if token.kind == Tag.END and token.name == "template":
+                return self._mode_in_head(token)
+            if token.kind == Tag.END and token.name == "body":
+                self._insert_body_if_missing()
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.END and token.name in {"html", "br"}:
+                self._insert_body_if_missing()
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.END:
+                # Ignore other end tags
+                self._parse_error("unexpected-end-tag-after-head", tag_name=token.name)
+                return None
+        if isinstance(token, EOFToken):
+            self._insert_body_if_missing()
+            self.mode = InsertionMode.IN_BODY
+            return ("reprocess", InsertionMode.IN_BODY, token)
+
+        self._insert_body_if_missing()
+        return ("reprocess", InsertionMode.IN_BODY, token)
+
+    def _mode_text(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            self._append_text(token.data)
+            return None
+        if isinstance(token, EOFToken):
+            # Get the tag name of the unclosed element
+            tag_name = self.open_elements[-1].name if self.open_elements else None
+            self._parse_error("expected-named-closing-tag-but-got-eof", tag_name=tag_name)
+            self._pop_current()
+            self.mode = self.original_mode or InsertionMode.IN_BODY
+            return ("reprocess", self.mode, token)
+        # End tag
+        self._pop_current()
+        self.mode = self.original_mode or InsertionMode.IN_BODY
+        return None
+
+    def _mode_in_body(self, token: Any) -> Any:
+        handler = self._BODY_TOKEN_HANDLERS.get(type(token))
+        return handler(self, token) if handler else None
+
+    def _handle_characters_in_body(self, token: Any) -> Any:
+        data = token.data or ""
+        if "\x00" in data:
+            self._parse_error("invalid-codepoint")
+            data = data.replace("\x00", "")
+        if is_all_whitespace(data):
+            self._reconstruct_active_formatting_elements()
+            self._append_text(data)
+            return
+        self._reconstruct_active_formatting_elements()
+        self.frameset_ok = False
+        self._append_text(data)
+        return
+
+    def _handle_comment_in_body(self, token: Any) -> Any:
+        self._append_comment(token.data)
+        return
+
+    def _handle_tag_in_body(self, token: Any) -> Any:
+        if token.kind == Tag.START:
+            handler = self._BODY_START_HANDLERS.get(token.name)
+            if handler:
+                return handler(self, token)
+            return self._handle_body_start_default(token)
+        name = token.name
+
+        # Special case: </br> end tag is treated as <br> start tag
+        if name == "br":
+            self._parse_error("unexpected-end-tag", tag_name=name, token=token)
+            br_tag = Tag(Tag.START, "br", {}, False)
+            return self._mode_in_body(br_tag)
+
+        if name in FORMATTING_ELEMENTS:
+            self._adoption_agency(name)
+            return None
+        handler = self._BODY_END_HANDLERS.get(name)
+        if handler:
+            return handler(self, token)
+        # Any other end tag
+        self._any_other_end_tag(token.name)
+        return None
+
+    def _handle_eof_in_body(self, token: Any) -> Any:
+        # If we're in a template, handle EOF in template mode first
+        if self.template_modes:
+            return self._mode_in_template(token)
+        # Check for unclosed elements (excluding html, body, head which are implicit)
+        for node in self.open_elements:
+            if node.name not in {
+                "dd",
+                "dt",
+                "li",
+                "optgroup",
+                "option",
+                "p",
+                "rb",
+                "rp",
+                "rt",
+                "rtc",
+                "tbody",
+                "td",
+                "tfoot",
+                "th",
+                "thead",
+                "tr",
+                "body",
+                "html",
+            }:
+                self._parse_error("expected-closing-tag-but-got-eof", tag_name=node.name)
+                break
+        self.mode = InsertionMode.AFTER_BODY
+        return ("reprocess", InsertionMode.AFTER_BODY, token)
+
+    # ---------------------
+    # Body mode start tag handlers
+    # ---------------------
+
+    def _handle_body_start_html(self, token: Any) -> Any:
+        if self.template_modes:
+            self._parse_error("unexpected-start-tag", tag_name=token.name)
+            return
+        # In IN_BODY mode, html element is always at open_elements[0]
+        if self.open_elements:  # pragma: no branch
+            html = self.open_elements[0]
+            self._add_missing_attributes(html, token.attrs)
+        return
+
+    def _handle_body_start_body(self, token: Any) -> Any:
+        if self.template_modes:
+            self._parse_error("unexpected-start-tag", tag_name=token.name)
+            return
+        if len(self.open_elements) > 1:
+            self._parse_error("unexpected-start-tag", tag_name=token.name)
+            body = self.open_elements[1] if len(self.open_elements) > 1 else None
+            if body and body.name == "body":
+                self._add_missing_attributes(body, token.attrs)
+            self.frameset_ok = False
+            return
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_head(self, token: Any) -> Any:
+        self._parse_error("unexpected-start-tag", tag_name=token.name)
+        return
+
+    def _handle_body_start_in_head(self, token: Any) -> Any:
+        return self._mode_in_head(token)
+
+    def _handle_body_start_block_with_p(self, token: Any) -> Any:
+        self._close_p_element()
+        self._insert_element(token, push=True)
+        return
+
+    def _handle_body_start_heading(self, token: Any) -> Any:
+        self._close_p_element()
+        if self.open_elements and self.open_elements[-1].name in HEADING_ELEMENTS:
+            self._parse_error("unexpected-start-tag", tag_name=token.name)
+            self._pop_current()
+        self._insert_element(token, push=True)
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_pre_listing(self, token: Any) -> Any:
+        self._close_p_element()
+        self._insert_element(token, push=True)
+        self.ignore_lf = True
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_form(self, token: Any) -> Any:
+        if self.form_element is not None:
+            self._parse_error("unexpected-start-tag", tag_name=token.name)
+            return
+        self._close_p_element()
+        node = self._insert_element(token, push=True)
+        self.form_element = node
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_button(self, token: Any) -> Any:
+        if self._has_in_scope("button"):
+            self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=token.name)
+            self._close_element_by_name("button")
+        self._insert_element(token, push=True)
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_paragraph(self, token: Any) -> Any:
+        self._close_p_element()
+        self._insert_element(token, push=True)
+        return
+
+    def _handle_body_start_math(self, token: Any) -> Any:
+        self._reconstruct_active_formatting_elements()
+        attrs = self._prepare_foreign_attributes("math", token.attrs)
+        new_tag = Tag(Tag.START, token.name, attrs, token.self_closing)
+        self._insert_element(new_tag, push=not token.self_closing, namespace="math")
+        return
+
+    def _handle_body_start_svg(self, token: Any) -> Any:
+        self._reconstruct_active_formatting_elements()
+        adjusted_name = self._adjust_svg_tag_name(token.name)
+        attrs = self._prepare_foreign_attributes("svg", token.attrs)
+        new_tag = Tag(Tag.START, adjusted_name, attrs, token.self_closing)
+        self._insert_element(new_tag, push=not token.self_closing, namespace="svg")
+        return
+
+    def _handle_body_start_li(self, token: Any) -> Any:
+        self.frameset_ok = False
+        self._close_p_element()
+        if self._has_in_list_item_scope("li"):
+            self._pop_until_any_inclusive({"li"})
+        self._insert_element(token, push=True)
+        return
+
+    def _handle_body_start_dd_dt(self, token: Any) -> Any:
+        self.frameset_ok = False
+        self._close_p_element()
+        name = token.name
+        if name == "dd":
+            if self._has_in_definition_scope("dd"):
+                self._pop_until_any_inclusive({"dd"})
+            if self._has_in_definition_scope("dt"):
+                self._pop_until_any_inclusive({"dt"})
+        else:
+            if self._has_in_definition_scope("dt"):
+                self._pop_until_any_inclusive({"dt"})
+            if self._has_in_definition_scope("dd"):
+                self._pop_until_any_inclusive({"dd"})
+        self._insert_element(token, push=True)
+        return
+
+    def _adoption_agency(self, subject: Any) -> None:
+        # 1. If the current node is the subject, and it is not in the active formatting elements list...
+        if self.open_elements and self.open_elements[-1].name == subject:
+            if not self._has_active_formatting_entry(subject):
+                self._pop_until_inclusive(subject)
+                return
+
+        # 2. Outer loop
+        for _ in range(8):
+            # 3. Find formatting element
+            formatting_element_index = self._find_active_formatting_index(subject)
+            if formatting_element_index is None:
+                return
+
+            formatting_element_entry = self.active_formatting[formatting_element_index]
+            formatting_element = formatting_element_entry["node"]
+
+            # 4. If formatting element is not in open elements
+            if formatting_element not in self.open_elements:
+                self._parse_error("adoption-agency-1.3")
+                self._remove_formatting_entry(formatting_element_index)
+                return
+
+            # 5. If formatting element is in open elements but not in scope
+            if not self._has_element_in_scope(formatting_element.name):
+                self._parse_error("adoption-agency-1.3")
+                return
+
+            # 6. If formatting element is not the current node
+            if formatting_element is not self.open_elements[-1]:
+                self._parse_error("adoption-agency-1.3")
+
+            # 7. Find furthest block
+            furthest_block = None
+            formatting_element_in_open_index = self.open_elements.index(formatting_element)
+
+            for i in range(formatting_element_in_open_index + 1, len(self.open_elements)):
+                node = self.open_elements[i]
+                if self._is_special_element(node):
+                    furthest_block = node
+                    break
+
+            if furthest_block is None:
+                # formatting_element is known to be on the stack
+                while True:
+                    popped = self.open_elements.pop()
+                    if popped is formatting_element:
+                        break
+                self._remove_formatting_entry(formatting_element_index)
+                return
+
+            # 8. Bookmark
+            bookmark = formatting_element_index + 1
+
+            # 9. Node and Last Node
+            node = furthest_block
+            last_node = furthest_block
+
+            # 10. Inner loop
+            inner_loop_counter = 0
+            while True:
+                inner_loop_counter += 1
+
+                # 10.1 Node = element above node
+                node_index = self.open_elements.index(node)
+                node = self.open_elements[node_index - 1]
+
+                # 10.2 If node is formatting element, break
+                if node is formatting_element:
+                    break
+
+                # 10.3 Find active formatting entry for node
+                node_formatting_index = self._find_active_formatting_index_by_node(node)
+
+                if inner_loop_counter > 3 and node_formatting_index is not None:
+                    self._remove_formatting_entry(node_formatting_index)
+                    if node_formatting_index < bookmark:
+                        bookmark -= 1
+                    node_formatting_index = None
+
+                if node_formatting_index is None:
+                    node_index = self.open_elements.index(node)
+                    self.open_elements.remove(node)
+                    node = self.open_elements[node_index]
+                    continue
+
+                # 10.4 Replace entry with new element
+                entry = self.active_formatting[node_formatting_index]
+                new_element = self._create_element(entry["name"], entry["node"].namespace, entry["attrs"])
+                entry["node"] = new_element
+                self.open_elements[self.open_elements.index(node)] = new_element
+                node = new_element
+
+                # 10.5 If last node is furthest block, update bookmark
+                if last_node is furthest_block:
+                    bookmark = node_formatting_index + 1
+
+                # 10.6 Reparent last_node
+                if last_node.parent:
+                    last_node.parent.remove_child(last_node)
+                node.append_child(last_node)
+
+                # 10.7
+                last_node = node
+
+            # 11. Insert last_node into common ancestor
+            common_ancestor = self.open_elements[formatting_element_in_open_index - 1]
+            if last_node.parent:
+                last_node.parent.remove_child(last_node)
+
+            if self._should_foster_parenting(common_ancestor, for_tag=last_node.name):
+                parent, position = self._appropriate_insertion_location(common_ancestor, foster_parenting=True)
+                self._insert_node_at(parent, position, last_node)
+            else:
+                if type(common_ancestor) is TemplateNode and common_ancestor.template_content:
+                    common_ancestor.template_content.append_child(last_node)
+                else:
+                    common_ancestor.append_child(last_node)
+
+            # 12. Create new formatting element
+            entry = self.active_formatting[formatting_element_index]
+            new_formatting_element = self._create_element(entry["name"], entry["node"].namespace, entry["attrs"])
+            entry["node"] = new_formatting_element
+
+            # 13. Move children of furthest block
+            while furthest_block.has_child_nodes():
+                child = furthest_block.children[0]
+                furthest_block.remove_child(child)
+                new_formatting_element.append_child(child)
+
+            furthest_block.append_child(new_formatting_element)
+
+            # 14. Remove formatting element from active formatting and insert new at bookmark
+            # Per spec, bookmark is always > formatting_element_index (starts at fmt_idx+1,
+            # can only be set to higher values or decremented when entries above fmt_idx are removed)
+            self._remove_formatting_entry(formatting_element_index)
+            bookmark -= 1
+            self.active_formatting.insert(bookmark, entry)
+
+            # 15. Remove formatting element from open elements and insert new one
+            self.open_elements.remove(formatting_element)
+            furthest_block_index = self.open_elements.index(furthest_block)
+            self.open_elements.insert(furthest_block_index + 1, new_formatting_element)
+
+    def _handle_body_start_a(self, token: Any) -> Any:
+        if self._has_active_formatting_entry("a"):
+            self._adoption_agency("a")
+            self._remove_last_active_formatting_by_name("a")
+            self._remove_last_open_element_by_name("a")
+        self._reconstruct_active_formatting_elements()
+        node = self._insert_element(token, push=True)
+        self._append_active_formatting_entry("a", token.attrs, node)
+        return
+
+    def _handle_body_start_formatting(self, token: Any) -> Any:
+        name = token.name
+        if name == "nobr" and self._in_scope("nobr"):
+            self._adoption_agency("nobr")
+            self._remove_last_active_formatting_by_name("nobr")
+            self._remove_last_open_element_by_name("nobr")
+        self._reconstruct_active_formatting_elements()
+        duplicate_index = self._find_active_formatting_duplicate(name, token.attrs)
+        if duplicate_index is not None:
+            self._remove_formatting_entry(duplicate_index)
+        node = self._insert_element(token, push=True)
+        self._append_active_formatting_entry(name, token.attrs, node)
+        return
+
+    def _handle_body_start_applet_like(self, token: Any) -> Any:
+        self._reconstruct_active_formatting_elements()
+        self._insert_element(token, push=True)
+        self._push_formatting_marker()
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_br(self, token: Any) -> Any:
+        self._close_p_element()
+        self._reconstruct_active_formatting_elements()
+        self._insert_element(token, push=False)
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_frameset(self, token: Any) -> Any:
+        if not self.frameset_ok:
+            self._parse_error("unexpected-start-tag-ignored", tag_name=token.name)
+            return
+        # Find body element on the stack (may not exist if already in frameset)
+        body_index = None
+        for i, elem in enumerate(self.open_elements):
+            if elem.name == "body":
+                body_index = i
+                break
+        if body_index is None:
+            # No body on stack (e.g., nested frameset after mode reset), ignore
+            self._parse_error("unexpected-start-tag-ignored", tag_name=token.name)
+            return
+        body_elem = self.open_elements[body_index]
+        body_elem.parent.remove_child(body_elem)
+        self.open_elements = self.open_elements[:body_index]
+        self._insert_element(token, push=True)
+        self.mode = InsertionMode.IN_FRAMESET
+        return
+
+    # ---------------------
+    # Body mode end tag handlers
+    # ---------------------
+
+    def _handle_body_end_body(self, token: Any) -> Any:
+        if self._in_scope("body"):
+            self.mode = InsertionMode.AFTER_BODY
+        return
+
+    def _handle_body_end_html(self, token: Any) -> Any:
+        if self._in_scope("body"):
+            return ("reprocess", InsertionMode.AFTER_BODY, token)
+        return None
+
+    def _handle_body_end_p(self, token: Any) -> Any:
+        if not self._close_p_element():
+            self._parse_error("unexpected-end-tag", tag_name=token.name)
+            phantom = Tag(Tag.START, "p", {}, False)
+            self._insert_element(phantom, push=True)
+            self._close_p_element()
+        return
+
+    def _handle_body_end_li(self, token: Any) -> Any:
+        if not self._has_in_list_item_scope("li"):
+            self._parse_error("unexpected-end-tag", tag_name=token.name)
+            return
+        self._pop_until_any_inclusive({"li"})
+        return
+
+    def _handle_body_end_dd_dt(self, token: Any) -> Any:
+        name = token.name
+        if not self._has_in_definition_scope(name):
+            self._parse_error("unexpected-end-tag", tag_name=name)
+            return
+        self._pop_until_any_inclusive({"dd", "dt"})
+
+    def _handle_body_end_form(self, token: Any) -> Any:
+        if self.form_element is None:
+            self._parse_error("unexpected-end-tag", tag_name=token.name)
+            return
+        removed = self._remove_from_open_elements(self.form_element)
+        self.form_element = None
+        if not removed:
+            self._parse_error("unexpected-end-tag", tag_name=token.name)
+        return
+
+    def _handle_body_end_applet_like(self, token: Any) -> Any:
+        name = token.name
+        if not self._in_scope(name):
+            self._parse_error("unexpected-end-tag", tag_name=name)
+            return
+        # Element verified in scope above
+        while self.open_elements:  # pragma: no branch
+            popped = self.open_elements.pop()
+            if popped.name == name:
+                break
+        self._clear_active_formatting_up_to_marker()
+        return
+
+    def _handle_body_end_heading(self, token: Any) -> Any:
+        name = token.name
+        if not self._has_any_in_scope(HEADING_ELEMENTS):
+            self._parse_error("unexpected-end-tag", tag_name=name)
+            return
+        self._generate_implied_end_tags()
+        if self.open_elements and self.open_elements[-1].name != name:
+            self._parse_error("end-tag-too-early", tag_name=name)
+        # Heading verified in scope by caller
+        while self.open_elements:  # pragma: no branch
+            popped = self.open_elements.pop()
+            if popped.name in HEADING_ELEMENTS:
+                break
+        return
+
+    def _handle_body_end_block(self, token: Any) -> Any:
+        name = token.name
+        if not self._in_scope(name):
+            self._parse_error("unexpected-end-tag", tag_name=name)
+            return
+        self._generate_implied_end_tags()
+        if self.open_elements and self.open_elements[-1].name != name:
+            self._parse_error("end-tag-too-early", tag_name=name)
+        self._pop_until_any_inclusive({name})
+        return
+
+    def _handle_body_end_template(self, token: Any) -> Any:
+        has_template = any(node.name == "template" for node in self.open_elements)
+        if not has_template:
+            return
+        self._generate_implied_end_tags()
+        self._pop_until_inclusive("template")
+        self._clear_active_formatting_up_to_marker()
+        # Pop template mode if available
+        if self.template_modes:  # pragma: no branch
+            self.template_modes.pop()
+        self._reset_insertion_mode()
+        return
+
+    def _handle_body_start_structure_ignored(self, token: Any) -> Any:
+        self._parse_error("unexpected-start-tag-ignored", tag_name=token.name)
+        return
+
+    def _handle_body_start_col_or_frame(self, token: Any) -> Any:
+        if self.fragment_context is None:
+            self._parse_error("unexpected-start-tag-ignored", tag_name=token.name)
+            return
+        self._insert_element(token, push=False)
+        return
+
+    def _handle_body_start_image(self, token: Any) -> Any:
+        self._parse_error("image-start-tag", tag_name=token.name)
+        img_token = Tag(Tag.START, "img", token.attrs, token.self_closing)
+        self._reconstruct_active_formatting_elements()
+        self._insert_element(img_token, push=False)
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_void_with_formatting(self, token: Any) -> Any:
+        self._reconstruct_active_formatting_elements()
+        self._insert_element(token, push=False)
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_simple_void(self, token: Any) -> Any:
+        self._insert_element(token, push=False)
+        return
+
+    def _handle_body_start_input(self, token: Any) -> Any:
+        input_type = None
+        for name, value in token.attrs.items():
+            if name == "type":
+                input_type = (value or "").lower()
+                break
+        self._insert_element(token, push=False)
+        if input_type != "hidden":
+            self.frameset_ok = False
+        return
+
+    def _handle_body_start_table(self, token: Any) -> Any:
+        if self.quirks_mode != "quirks":
+            self._close_p_element()
+        self._insert_element(token, push=True)
+        self.frameset_ok = False
+        self.mode = InsertionMode.IN_TABLE
+        return
+
+    def _handle_body_start_plaintext_xmp(self, token: Any) -> Any:
+        self._close_p_element()
+        self._insert_element(token, push=True)
+        self.frameset_ok = False
+        if token.name == "plaintext":
+            self.tokenizer_state_override = TokenSinkResult.Plaintext
+        else:
+            # xmp, iframe, noembed, noframes, noscript (scripting disabled)
+            self.original_mode = self.mode
+            self.mode = InsertionMode.TEXT
+        return
+
+    def _handle_body_start_textarea(self, token: Any) -> Any:
+        self._insert_element(token, push=True)
+        self.ignore_lf = True
+        self.frameset_ok = False
+        return
+
+    def _handle_body_start_select(self, token: Any) -> Any:
+        self._reconstruct_active_formatting_elements()
+        self._insert_element(token, push=True)
+        self.frameset_ok = False
+        self._reset_insertion_mode()
+        return
+
+    def _handle_body_start_option(self, token: Any) -> Any:
+        if self.open_elements and self.open_elements[-1].name == "option":
+            self.open_elements.pop()
+        self._reconstruct_active_formatting_elements()
+        self._insert_element(token, push=True)
+        return
+
+    def _handle_body_start_optgroup(self, token: Any) -> Any:
+        if self.open_elements and self.open_elements[-1].name == "option":
+            self.open_elements.pop()
+        self._reconstruct_active_formatting_elements()
+        self._insert_element(token, push=True)
+        return
+
+    def _handle_body_start_rp_rt(self, token: Any) -> Any:
+        self._generate_implied_end_tags(exclude="rtc")
+        self._insert_element(token, push=True)
+        return
+
+    def _handle_body_start_rb_rtc(self, token: Any) -> Any:
+        if self.open_elements and self.open_elements[-1].name in {"rb", "rp", "rt", "rtc"}:
+            self._generate_implied_end_tags()
+        self._insert_element(token, push=True)
+        return
+
+    def _handle_body_start_table_parse_error(self, token: Any) -> Any:
+        self._parse_error("unexpected-start-tag", tag_name=token.name)
+        return
+
+    def _handle_body_start_default(self, token: Any) -> Any:
+        self._reconstruct_active_formatting_elements()
+        self._insert_element(token, push=True)
+        if token.self_closing:
+            self._parse_error("non-void-html-element-start-tag-with-trailing-solidus", tag_name=token.name)
+        # Elements reaching here have no handler - never in FRAMESET_NEUTRAL/FORMATTING_ELEMENTS
+        self.frameset_ok = False
+        return
+
+    def _mode_in_table(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            data = token.data or ""
+            if "\x00" in data:
+                self._parse_error("unexpected-null-character")
+                data = data.replace("\x00", "")
+                if not data:
+                    return None
+                token = CharacterTokens(data)
+            self.pending_table_text = []
+            self.table_text_original_mode = self.mode
+            self.mode = InsertionMode.IN_TABLE_TEXT
+            return ("reprocess", InsertionMode.IN_TABLE_TEXT, token)
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            name = token.name
+            if token.kind == Tag.START:
+                if name == "caption":
+                    self._clear_stack_until({"table", "template", "html"})
+                    self._push_formatting_marker()
+                    self._insert_element(token, push=True)
+                    self.mode = InsertionMode.IN_CAPTION
+                    return None
+                if name == "colgroup":
+                    self._clear_stack_until({"table", "template", "html"})
+                    self._insert_element(token, push=True)
+                    self.mode = InsertionMode.IN_COLUMN_GROUP
+                    return None
+                if name == "col":
+                    self._clear_stack_until({"table", "template", "html"})
+                    implied = Tag(Tag.START, "colgroup", {}, False)
+                    self._insert_element(implied, push=True)
+                    self.mode = InsertionMode.IN_COLUMN_GROUP
+                    return ("reprocess", InsertionMode.IN_COLUMN_GROUP, token)
+                if name in {"tbody", "tfoot", "thead"}:
+                    self._clear_stack_until({"table", "template", "html"})
+                    self._insert_element(token, push=True)
+                    self.mode = InsertionMode.IN_TABLE_BODY
+                    return None
+                if name in {"td", "th", "tr"}:
+                    self._clear_stack_until({"table", "template", "html"})
+                    implied = Tag(Tag.START, "tbody", {}, False)
+                    self._insert_element(implied, push=True)
+                    self.mode = InsertionMode.IN_TABLE_BODY
+                    return ("reprocess", InsertionMode.IN_TABLE_BODY, token)
+                if name == "table":
+                    self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=name)
+                    closed = self._close_table_element()
+                    if closed:
+                        return ("reprocess", self.mode, token)
+                    return None
+                if name in {"style", "script"}:
+                    # Per HTML5 spec: style and script are inserted directly into the table
+                    # (not processed as in-head which would move them)
+                    self._insert_element(token, push=True)
+                    self.original_mode = self.mode
+                    self.mode = InsertionMode.TEXT
+                    return None
+                if name == "template":
+                    # Template is handled by delegating to IN_HEAD
+                    return self._mode_in_head(token)
+                if name == "input":
+                    input_type = None
+                    for attr_name, attr_value in token.attrs.items():
+                        if attr_name == "type":
+                            input_type = (attr_value or "").lower()
+                            break
+                    if input_type == "hidden":
+                        self._parse_error("unexpected-hidden-input-in-table")
+                        self._insert_element(token, push=True)
+                        self.open_elements.pop()  # push=True always adds to stack
+                        return None
+                if name == "form":
+                    self._parse_error("unexpected-form-in-table")
+                    if self.form_element is None:
+                        node = self._insert_element(token, push=True)
+                        self.form_element = node
+                        self.open_elements.pop()  # push=True always adds to stack
+                    return None
+                self._parse_error("unexpected-start-tag-implies-table-voodoo", tag_name=name)
+                previous = self.insert_from_table
+                self.insert_from_table = True
+                try:
+                    return self._mode_in_body(token)
+                finally:
+                    self.insert_from_table = previous
+            else:
+                if name == "table":
+                    self._close_table_element()
+                    return None
+                if name in {"body", "caption", "col", "colgroup", "html", "tbody", "td", "tfoot", "th", "thead", "tr"}:
+                    self._parse_error("unexpected-end-tag", tag_name=name)
+                    return None
+                self._parse_error("unexpected-end-tag-implies-table-voodoo", tag_name=name)
+                previous = self.insert_from_table
+                self.insert_from_table = True
+                try:
+                    return self._mode_in_body(token)
+                finally:
+                    self.insert_from_table = previous
+        # Per spec, only CharacterTokens, CommentToken, Tag, and EOFToken exist
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        # If we're in a template, handle EOF in template mode first
+        if self.template_modes:
+            return self._mode_in_template(token)
+        if self._has_in_table_scope("table"):
+            self._parse_error("expected-closing-tag-but-got-eof", tag_name="table")
+        return None
+
+    def _mode_in_table_text(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            # IN_TABLE mode guarantees non-empty data
+            data = token.data
+            if "\x0c" in data:
+                self._parse_error("invalid-codepoint-in-table-text")
+                data = data.replace("\x0c", "")
+            if data:
+                self.pending_table_text.append(data)
+            return None
+        self._flush_pending_table_text()
+        original = self.table_text_original_mode or InsertionMode.IN_TABLE
+        self.table_text_original_mode = None
+        self.mode = original
+        return ("reprocess", original, token)
+
+    def _mode_in_caption(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            return self._mode_in_body(token)
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            name = token.name
+            if token.kind == Tag.START:
+                if name in {"caption", "col", "colgroup", "tbody", "tfoot", "thead", "tr", "td", "th"}:
+                    self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=name)
+                    if self._close_caption_element():
+                        return ("reprocess", InsertionMode.IN_TABLE, token)
+                    # Fragment parsing with caption context: caption not on stack, ignore table structure elements
+                    return None
+                if name == "table":
+                    self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=name)
+                    if self._close_caption_element():
+                        return ("reprocess", InsertionMode.IN_TABLE, token)
+                    # Fragment parsing: no caption on stack - handle in body mode
+                    return self._mode_in_body(token)
+                return self._mode_in_body(token)
+            if name == "caption":
+                if not self._close_caption_element():
+                    return None
+                return None
+            if name == "table":
+                if self._close_caption_element():
+                    return ("reprocess", InsertionMode.IN_TABLE, token)
+                return None
+            if name in {"tbody", "tfoot", "thead"}:
+                # These elements are never in table scope when in caption -
+                # caption closes any open tbody/tfoot/thead when created
+                self._parse_error("unexpected-end-tag", tag_name=name)
+                return None
+            return self._mode_in_body(token)
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        return self._mode_in_body(token)
+
+    def _close_caption_element(self) -> bool:
+        if not self._has_in_table_scope("caption"):
+            self._parse_error("unexpected-end-tag", tag_name="caption")
+            return False
+        self._generate_implied_end_tags()
+        # Caption verified in scope above
+        while self.open_elements:  # pragma: no branch
+            node = self.open_elements.pop()
+            if node.name == "caption":
+                break
+        self._clear_active_formatting_up_to_marker()
+        self.mode = InsertionMode.IN_TABLE
+        return True
+
+    def _mode_in_column_group(self, token: Any) -> Any:
+        current = self.open_elements[-1] if self.open_elements else None
+        if isinstance(token, CharacterTokens):
+            data = token.data or ""
+            # Find first non-whitespace character
+            stripped = data.lstrip(" \t\n\r\f")
+
+            if len(stripped) < len(data):
+                # Has leading whitespace - insert it
+                ws = data[: len(data) - len(stripped)]
+                self._append_text(ws)
+
+            # Continue processing non-whitespace with a new token
+            non_ws_token = CharacterTokens(stripped)
+            if current and current.name == "html":
+                # Fragment parsing with colgroup context: drop non-whitespace characters
+                # (This is the only way html can be current in IN_COLUMN_GROUP mode)
+                self._parse_error("unexpected-characters-in-column-group")
+                return None
+            # In a template, non-whitespace characters are parse errors - ignore them
+            if current and current.name == "template":
+                self._parse_error("unexpected-characters-in-template-column-group")
+                return None
+            self._parse_error("unexpected-characters-in-column-group")
+            self._pop_current()
+            self.mode = InsertionMode.IN_TABLE
+            return ("reprocess", InsertionMode.IN_TABLE, non_ws_token)
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            name = token.name
+            if token.kind == Tag.START:
+                if name == "html":
+                    return self._mode_in_body(token)
+                if name == "col":
+                    self._insert_element(token, push=True)
+                    self.open_elements.pop()  # push=True always adds to stack
+                    return None
+                if name == "template":
+                    # Template is handled by delegating to IN_HEAD
+                    return self._mode_in_head(token)
+                if name == "colgroup":
+                    self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=name)
+                    # Don't pop template element - only pop actual colgroup
+                    if current and current.name == "colgroup":
+                        self._pop_current()
+                        self.mode = InsertionMode.IN_TABLE
+                        return ("reprocess", InsertionMode.IN_TABLE, token)
+                    return None
+                if (
+                    self.fragment_context
+                    and self.fragment_context.tag_name.lower() == "colgroup"
+                    and not self._has_in_table_scope("table")
+                ):
+                    self._parse_error("unexpected-start-tag-in-column-group", tag_name=name)
+                    return None
+                # Anything else: if we're in a colgroup, pop it and switch to IN_TABLE
+                if current and current.name == "colgroup":
+                    self._pop_current()
+                    self.mode = InsertionMode.IN_TABLE
+                    return ("reprocess", InsertionMode.IN_TABLE, token)
+                # In template column group context (via <col> in template), ignore non-column content
+                # At this point current is template - the only other case after colgroup fragment
+                # and colgroup element are handled
+                self._parse_error("unexpected-start-tag-in-template-column-group", tag_name=name)
+                return None
+            if name == "colgroup":
+                if current and current.name == "colgroup":
+                    self._pop_current()
+                    self.mode = InsertionMode.IN_TABLE
+                else:
+                    self._parse_error("unexpected-end-tag", tag_name=token.name)
+                return None
+            if name == "col":
+                self._parse_error("unexpected-end-tag", tag_name=name)
+                return None
+            if name == "template":
+                # Template end tag needs proper handling
+                return self._mode_in_head(token)
+            if current and current.name != "html":  # pragma: no branch
+                self._pop_current()
+                self.mode = InsertionMode.IN_TABLE
+            return ("reprocess", InsertionMode.IN_TABLE, token)
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        if current and current.name == "colgroup":
+            self._pop_current()
+            self.mode = InsertionMode.IN_TABLE
+            return ("reprocess", InsertionMode.IN_TABLE, token)
+        if current and current.name == "template":
+            # In template, delegate EOF handling to IN_TEMPLATE
+            return self._mode_in_template(token)
+        return None
+        # Per spec: EOF when current is html - implicit None return
+
+    def _mode_in_table_body(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens) or isinstance(token, CommentToken):
+            return self._mode_in_table(token)
+        if isinstance(token, Tag):
+            name = token.name
+            if token.kind == Tag.START:
+                if name == "tr":
+                    self._clear_stack_until({"tbody", "tfoot", "thead", "template", "html"})
+                    self._insert_element(token, push=True)
+                    self.mode = InsertionMode.IN_ROW
+                    return None
+                if name in {"td", "th"}:
+                    self._parse_error("unexpected-cell-in-table-body")
+                    self._clear_stack_until({"tbody", "tfoot", "thead", "template", "html"})
+                    implied = Tag(Tag.START, "tr", {}, False)
+                    self._insert_element(implied, push=True)
+                    self.mode = InsertionMode.IN_ROW
+                    return ("reprocess", InsertionMode.IN_ROW, token)
+                if name in {"caption", "col", "colgroup", "tbody", "tfoot", "thead", "table"}:
+                    current = self.open_elements[-1] if self.open_elements else None
+                    # When in a template, these tags create invalid structure - treat as "anything else"
+                    if current and current.name == "template":
+                        self._parse_error("unexpected-start-tag-in-template-table-context", tag_name=name)
+                        return None
+                    # In fragment parsing with tbody/tfoot/thead context and no tbody on stack, ignore these tags
+                    if (
+                        self.fragment_context
+                        and current
+                        and current.name == "html"
+                        and self.fragment_context.tag_name.lower() in {"tbody", "tfoot", "thead"}
+                    ):
+                        self._parse_error("unexpected-start-tag")
+                        return None
+                    # Pop tbody/tfoot/thead (stack always has elements here in normal parsing)
+                    if self.open_elements:
+                        self.open_elements.pop()
+                        self.mode = InsertionMode.IN_TABLE
+                        return ("reprocess", InsertionMode.IN_TABLE, token)
+                    # Empty stack edge case - go directly to IN_TABLE without reprocess
+                    self.mode = InsertionMode.IN_TABLE  # pragma: no cover
+                    return None  # pragma: no cover
+                return self._mode_in_table(token)
+            if name in {"tbody", "tfoot", "thead"}:
+                if not self._has_in_table_scope(name):
+                    self._parse_error("unexpected-end-tag", tag_name=name)
+                    return None
+                self._clear_stack_until({"tbody", "tfoot", "thead", "template", "html"})
+                self._pop_current()
+                self.mode = InsertionMode.IN_TABLE
+                return None
+            if name == "table":
+                current = self.open_elements[-1] if self.open_elements else None
+                # In a template, reject </table> as there's no table element
+                if current and current.name == "template":
+                    self._parse_error("unexpected-end-tag", tag_name=token.name)
+                    return None
+                # In fragment parsing with tbody/tfoot/thead context and no tbody on stack, ignore </table>
+                if (
+                    self.fragment_context
+                    and current
+                    and current.name == "html"
+                    and self.fragment_context.tag_name.lower() in {"tbody", "tfoot", "thead"}
+                ):
+                    self._parse_error("unexpected-end-tag", tag_name=token.name)
+                    return None
+                if current and current.name in {"tbody", "tfoot", "thead"}:
+                    self.open_elements.pop()
+                self.mode = InsertionMode.IN_TABLE
+                return ("reprocess", InsertionMode.IN_TABLE, token)
+            if name in {"caption", "col", "colgroup", "td", "th", "tr"}:
+                self._parse_error("unexpected-end-tag", tag_name=name)
+                return None
+            return self._mode_in_table(token)
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        return self._mode_in_table(token)
+
+    def _mode_in_row(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens) or isinstance(token, CommentToken):
+            return self._mode_in_table(token)
+        if isinstance(token, Tag):
+            name = token.name
+            if token.kind == Tag.START:
+                if name in {"td", "th"}:
+                    self._clear_stack_until({"tr", "template", "html"})
+                    self._insert_element(token, push=True)
+                    self._push_formatting_marker()
+                    self.mode = InsertionMode.IN_CELL
+                    return None
+                if name in {"caption", "col", "colgroup", "tbody", "tfoot", "thead", "tr", "table"}:
+                    if not self._has_in_table_scope("tr"):
+                        self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=name)
+                        return None
+                    self._end_tr_element()
+                    return ("reprocess", self.mode, token)
+                previous = self.insert_from_table
+                self.insert_from_table = True
+                try:
+                    return self._mode_in_body(token)
+                finally:
+                    self.insert_from_table = previous
+            else:
+                if name == "tr":
+                    if not self._has_in_table_scope("tr"):
+                        self._parse_error("unexpected-end-tag", tag_name=name)
+                        return None
+                    self._end_tr_element()
+                    return None
+                if name in {"table", "tbody", "tfoot", "thead"}:
+                    if self._has_in_table_scope(name):
+                        self._end_tr_element()
+                        return ("reprocess", self.mode, token)
+                    self._parse_error("unexpected-end-tag", tag_name=name)
+                    return None
+                if name in {"caption", "col", "group", "td", "th"}:
+                    self._parse_error("unexpected-end-tag", tag_name=name)
+                    return None
+                previous = self.insert_from_table
+                self.insert_from_table = True
+                try:
+                    return self._mode_in_body(token)
+                finally:
+                    self.insert_from_table = previous
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        return self._mode_in_table(token)
+
+    def _end_tr_element(self) -> None:
+        self._clear_stack_until({"tr", "template", "html"})
+        # Pop tr if on top (may not be if stack was exhausted)
+        if self.open_elements and self.open_elements[-1].name == "tr":
+            self.open_elements.pop()
+        # When in a template, restore template mode; otherwise use IN_TABLE_BODY
+        if self.template_modes:
+            self.mode = self.template_modes[-1]
+        else:
+            self.mode = InsertionMode.IN_TABLE_BODY
+
+    def _mode_in_cell(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            previous = self.insert_from_table
+            self.insert_from_table = False
+            try:
+                return self._mode_in_body(token)
+            finally:
+                self.insert_from_table = previous
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            name = token.name
+            if token.kind == Tag.START:
+                if name in {"caption", "col", "colgroup", "tbody", "td", "tfoot", "th", "thead", "tr"}:
+                    if self._close_table_cell():
+                        return ("reprocess", self.mode, token)
+                    # Per spec: if we reach here in IN_CELL mode with no cell to close,
+                    # we're in a fragment context with td/th as context element and no table structure.
+                    # Issue parse error and ignore the token.
+                    self._parse_error("unexpected-start-tag-in-cell-fragment", tag_name=name)
+                    return None
+                previous = self.insert_from_table
+                self.insert_from_table = False
+                try:
+                    return self._mode_in_body(token)
+                finally:
+                    self.insert_from_table = previous
+            else:
+                if name in {"td", "th"}:
+                    if not self._has_in_table_scope(name):
+                        self._parse_error("unexpected-end-tag", tag_name=name)
+                        return None
+                    self._end_table_cell(name)
+                    return None
+                if name in {"table", "tbody", "tfoot", "thead", "tr"}:
+                    # Per HTML5 spec: only close cell if the element is actually in scope
+                    # Otherwise it's a parse error and we ignore the token
+                    if not self._has_in_table_scope(name):
+                        self._parse_error("unexpected-end-tag", tag_name=name)
+                        return None
+                    self._close_table_cell()
+                    return ("reprocess", self.mode, token)
+                previous = self.insert_from_table
+                self.insert_from_table = False
+                try:
+                    return self._mode_in_body(token)
+                finally:
+                    self.insert_from_table = previous
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        if self._close_table_cell():
+            return ("reprocess", self.mode, token)
+        return self._mode_in_table(token)
+
+    def _mode_in_select(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            data = token.data or ""
+            if "\x00" in data:
+                self._parse_error("invalid-codepoint-in-select")
+                data = data.replace("\x00", "")
+            if "\x0c" in data:
+                self._parse_error("invalid-codepoint-in-select")
+                data = data.replace("\x0c", "")
+            if data:
+                self._reconstruct_active_formatting_elements()
+                self._append_text(data)
+            return None
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            name = token.name
+            if token.kind == Tag.START:
+                if name == "html":
+                    return ("reprocess", InsertionMode.IN_BODY, token)
+                if name == "option":
+                    if self.open_elements and self.open_elements[-1].name == "option":
+                        self.open_elements.pop()
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=True)
+                    return None
+                if name == "optgroup":
+                    if self.open_elements and self.open_elements[-1].name == "option":
+                        self.open_elements.pop()
+                    if self.open_elements and self.open_elements[-1].name == "optgroup":
+                        self.open_elements.pop()
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=True)
+                    return None
+                if name == "select":
+                    self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=name)
+                    # select is always in scope in IN_SELECT mode
+                    self._pop_until_any_inclusive({"select"})
+                    self._reset_insertion_mode()
+                    return None
+                if name in {"input", "textarea"}:
+                    self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=name)
+                    # select is always in scope in IN_SELECT mode
+                    self._pop_until_any_inclusive({"select"})
+                    self._reset_insertion_mode()
+                    return ("reprocess", self.mode, token)
+                if name == "keygen":
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=False)
+                    return None
+                if name in {"caption", "col", "colgroup", "tbody", "td", "tfoot", "th", "thead", "tr", "table"}:
+                    self._parse_error("unexpected-start-tag-implies-end-tag", tag_name=name)
+                    # select is always in scope in IN_SELECT mode
+                    self._pop_until_any_inclusive({"select"})
+                    self._reset_insertion_mode()
+                    return ("reprocess", self.mode, token)
+                if name in {"script", "template"}:
+                    return self._mode_in_head(token)
+                if name in {"svg", "math"}:
+                    # For foreign elements, honor the self-closing flag
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=not token.self_closing, namespace=name)
+                    return None
+                if name in FORMATTING_ELEMENTS:
+                    self._reconstruct_active_formatting_elements()
+                    node = self._insert_element(token, push=True)
+                    self._append_active_formatting_entry(name, token.attrs, node)
+                    return None
+                if name == "hr":
+                    # Per spec: pop option and optgroup before inserting hr (makes hr sibling, not child)
+                    if self.open_elements and self.open_elements[-1].name == "option":
+                        self.open_elements.pop()
+                    if self.open_elements and self.open_elements[-1].name == "optgroup":
+                        self.open_elements.pop()
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=False)
+                    return None
+                if name == "menuitem":
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=True)
+                    return None
+                # Allow common HTML elements in select (newer spec)
+                if name in {"p", "div", "span", "button", "datalist", "selectedcontent"}:
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=not token.self_closing)
+                    return None
+                if name in {"br", "img"}:
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=False)
+                    return None
+                if name == "plaintext":
+                    # Per spec: plaintext element is inserted in select (consumes all remaining text)
+                    self._reconstruct_active_formatting_elements()
+                    self._insert_element(token, push=True)
+                return None
+            if name == "optgroup":
+                if self.open_elements and self.open_elements[-1].name == "option":
+                    self.open_elements.pop()
+                if self.open_elements and self.open_elements[-1].name == "optgroup":
+                    self.open_elements.pop()
+                else:
+                    self._parse_error("unexpected-end-tag", tag_name=token.name)
+                return None
+            if name == "option":
+                if self.open_elements and self.open_elements[-1].name == "option":
+                    self.open_elements.pop()
+                else:
+                    self._parse_error("unexpected-end-tag", tag_name=token.name)
+                return None
+            if name == "select":
+                # In IN_SELECT mode, select is always in scope - pop to it
+                self._pop_until_any_inclusive({"select"})
+                self._reset_insertion_mode()
+                return None
+            # Handle end tags for allowed HTML elements in select
+            if name == "a" or name in FORMATTING_ELEMENTS:
+                # select is always on stack in IN_SELECT mode
+                select_node = self._find_last_on_stack("select")
+                fmt_index = self._find_active_formatting_index(name)
+                if fmt_index is not None:
+                    target = self.active_formatting[fmt_index]["node"]
+                    if target in self.open_elements:  # pragma: no branch
+                        select_index = self.open_elements.index(select_node)
+                        target_index = self.open_elements.index(target)
+                        if target_index < select_index:
+                            self._parse_error("unexpected-end-tag", tag_name=name)
+                            return None
+                self._adoption_agency(name)
+                return None
+            if name in {"p", "div", "span", "button", "datalist", "selectedcontent"}:
+                # Per HTML5 spec: these end tags in select mode close the element if it's on the stack.
+                # But we must not pop across the select boundary (i.e., don't pop elements BEFORE select).
+                select_idx = None
+                target_idx = None
+                for i, node in enumerate(self.open_elements):
+                    if node.name == "select" and select_idx is None:
+                        select_idx = i
+                    if node.name == name:
+                        target_idx = i  # Track the LAST occurrence
+                # Only pop if target exists and is AFTER (or at same level as) select
+                # i.e., the target is inside the select or there's no select
+                if target_idx is not None and (select_idx is None or target_idx > select_idx):
+                    while True:
+                        popped = self.open_elements.pop()
+                        if popped.name == name:
+                            break
+                else:
+                    self._parse_error("unexpected-end-tag", tag_name=name)
+                return None
+            if name in {"caption", "col", "colgroup", "tbody", "td", "tfoot", "th", "thead", "tr", "table"}:
+                self._parse_error("unexpected-end-tag", tag_name=name)
+                # select is always in scope in IN_SELECT mode
+                self._pop_until_any_inclusive({"select"})
+                self._reset_insertion_mode()
+                return ("reprocess", self.mode, token)
+            # Any other end tag: parse error, ignore
+            self._parse_error("unexpected-end-tag", tag_name=name)
+            return None
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        return self._mode_in_body(token)
+
+    def _mode_in_template(self, token: Any) -> Any:
+        # § The "in template" insertion mode
+        # https://html.spec.whatwg.org/multipage/parsing.html#parsing-main-intemplate
+        if isinstance(token, CharacterTokens):
+            return self._mode_in_body(token)
+        if isinstance(token, CommentToken):
+            return self._mode_in_body(token)
+        if isinstance(token, Tag):
+            if token.kind == Tag.START:
+                # Table-related tags switch template mode
+                if token.name in {"caption", "colgroup", "tbody", "tfoot", "thead"}:
+                    self.template_modes.pop()
+                    self.template_modes.append(InsertionMode.IN_TABLE)
+                    self.mode = InsertionMode.IN_TABLE
+                    return ("reprocess", InsertionMode.IN_TABLE, token)
+                if token.name == "col":
+                    self.template_modes.pop()
+                    self.template_modes.append(InsertionMode.IN_COLUMN_GROUP)
+                    self.mode = InsertionMode.IN_COLUMN_GROUP
+                    return ("reprocess", InsertionMode.IN_COLUMN_GROUP, token)
+                if token.name == "tr":
+                    self.template_modes.pop()
+                    self.template_modes.append(InsertionMode.IN_TABLE_BODY)
+                    self.mode = InsertionMode.IN_TABLE_BODY
+                    return ("reprocess", InsertionMode.IN_TABLE_BODY, token)
+                if token.name in {"td", "th"}:
+                    self.template_modes.pop()
+                    self.template_modes.append(InsertionMode.IN_ROW)
+                    self.mode = InsertionMode.IN_ROW
+                    return ("reprocess", InsertionMode.IN_ROW, token)
+                # Default: pop template mode and push IN_BODY
+                if token.name not in {
+                    "base",
+                    "basefont",
+                    "bgsound",
+                    "link",
+                    "meta",
+                    "noframes",
+                    "script",
+                    "style",
+                    "template",
+                    "title",
+                }:
+                    self.template_modes.pop()
+                    self.template_modes.append(InsertionMode.IN_BODY)
+                    self.mode = InsertionMode.IN_BODY
+                    return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.END and token.name == "template":
+                return self._mode_in_head(token)
+            # Head-related tags process in InHead
+            if token.name in {
+                "base",
+                "basefont",
+                "bgsound",
+                "link",
+                "meta",
+                "noframes",
+                "script",
+                "style",
+                "template",
+                "title",
+            }:
+                return self._mode_in_head(token)
+        if isinstance(token, EOFToken):
+            # Check if template is on the stack (don't use _in_scope as table blocks it)
+            has_template = any(node.name == "template" for node in self.open_elements)
+            if not has_template:
+                return None
+            # Parse error for EOF in template
+            self._parse_error("expected-closing-tag-but-got-eof", tag_name="template")
+            # Pop until template, then handle EOF in reset mode
+            self._pop_until_inclusive("template")
+            self._clear_active_formatting_up_to_marker()
+            # template_modes is always non-empty when template is on stack
+            self.template_modes.pop()
+            self._reset_insertion_mode()
+            return ("reprocess", self.mode, token)
+        return None
+
+    def _mode_after_body(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            if is_all_whitespace(token.data):
+                # Whitespace is processed using InBody rules (appended to body)
+                # but we stay in AfterBody mode
+                self._mode_in_body(token)
+                return None
+            return ("reprocess", InsertionMode.IN_BODY, token)
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data, parent=self.open_elements[0])
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.END and token.name == "html":
+                self.mode = InsertionMode.AFTER_AFTER_BODY
+                return None
+            return ("reprocess", InsertionMode.IN_BODY, token)
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        return None
+
+    def _mode_after_after_body(self, token: Any) -> Any:
+        if isinstance(token, CharacterTokens):
+            if is_all_whitespace(token.data):
+                # Per spec: whitespace characters are inserted using the rules for the "in body" mode
+                # Process with InBody rules but stay in AfterAfterBody mode
+                self._mode_in_body(token)
+                return None
+            # Non-whitespace character: parse error, reprocess in IN_BODY
+            self._parse_error("unexpected-char-after-body")
+            return ("reprocess", InsertionMode.IN_BODY, token)
+        if isinstance(token, CommentToken):
+            if self.fragment_context is not None:
+                # html is always on stack in fragment parsing
+                html_node = self._find_last_on_stack("html")
+                html_node.append_child(SimpleDomNode("#comment", data=token.data))
+                return None
+            self._append_comment_to_document(token.data)
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            # Any other tag: parse error, reprocess in IN_BODY
+            self._parse_error("unexpected-token-after-body")
+            return ("reprocess", InsertionMode.IN_BODY, token)
+        assert isinstance(token, EOFToken), f"Unexpected token type: {type(token)}"
+        return None
+
+    def _mode_in_frameset(self, token: Any) -> Any:
+        # Per HTML5 spec §13.2.6.4.16: In frameset insertion mode
+        if isinstance(token, CharacterTokens):
+            # Only whitespace characters allowed; ignore all others
+            whitespace = "".join(ch for ch in token.data if ch in "\t\n\f\r ")
+            if whitespace:
+                self._append_text(whitespace)
+            return None
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.START and token.name == "frameset":
+                self._insert_element(token, push=True)
+                return None
+            if token.kind == Tag.END and token.name == "frameset":
+                if self.open_elements and self.open_elements[-1].name == "html":
+                    self._parse_error("unexpected-end-tag", tag_name=token.name)
+                    return None
+                self.open_elements.pop()
+                if self.open_elements and self.open_elements[-1].name != "frameset":
+                    self.mode = InsertionMode.AFTER_FRAMESET
+                return None
+            if token.kind == Tag.START and token.name == "frame":
+                self._insert_element(token, push=True)
+                self.open_elements.pop()
+                return None
+            if token.kind == Tag.START and token.name == "noframes":
+                # Per spec: use IN_HEAD rules but preserve current mode for TEXT restoration
+                self._insert_element(token, push=True)
+                self.original_mode = self.mode
+                self.mode = InsertionMode.TEXT
+                return None
+        if isinstance(token, EOFToken):
+            if self.open_elements and self.open_elements[-1].name != "html":
+                self._parse_error("expected-closing-tag-but-got-eof", tag_name=self.open_elements[-1].name)
+            return None
+        self._parse_error("unexpected-token-in-frameset")
+        return None
+
+    def _mode_after_frameset(self, token: Any) -> Any:
+        # Per HTML5 spec §13.2.6.4.17: After frameset insertion mode
+        if isinstance(token, CharacterTokens):
+            # Only whitespace characters allowed; ignore all others
+            whitespace = "".join(ch for ch in token.data if ch in "\t\n\f\r ")
+            if whitespace:
+                self._append_text(whitespace)
+            return None
+        if isinstance(token, CommentToken):
+            self._append_comment(token.data)
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.END and token.name == "html":
+                self.mode = InsertionMode.AFTER_AFTER_FRAMESET
+                return None
+            if token.kind == Tag.START and token.name == "noframes":
+                # Insert noframes element directly and switch to TEXT mode
+                self._insert_element(token, push=True)
+                self.original_mode = self.mode
+                self.mode = InsertionMode.TEXT
+                return None
+        if isinstance(token, EOFToken):
+            return None
+        self._parse_error("unexpected-token-after-frameset")
+        self.mode = InsertionMode.IN_FRAMESET
+        return ("reprocess", InsertionMode.IN_FRAMESET, token)
+
+    def _mode_after_after_frameset(self, token: Any) -> Any:
+        # Per HTML5 spec §13.2.6.4.18: After after frameset insertion mode
+        if isinstance(token, CharacterTokens):
+            # Whitespace is processed using InBody rules
+            # but we stay in AfterAfterFrameset mode
+            if is_all_whitespace(token.data):
+                self._mode_in_body(token)
+                return None
+            # Non-whitespace falls through to "Anything else"
+        if isinstance(token, CommentToken):
+            self._append_comment_to_document(token.data)
+            return None
+        if isinstance(token, Tag):
+            if token.kind == Tag.START and token.name == "html":
+                return ("reprocess", InsertionMode.IN_BODY, token)
+            if token.kind == Tag.START and token.name == "noframes":
+                # Insert noframes element directly and switch to TEXT mode
+                self._insert_element(token, push=True)
+                self.original_mode = self.mode
+                self.mode = InsertionMode.TEXT
+                return None
+            # Other tags fall through to "Anything else"
+        if isinstance(token, EOFToken):
+            return None
+        # Anything else: parse error, reprocess in IN_FRAMESET
+        self._parse_error("unexpected-token-after-after-frameset")
+        self.mode = InsertionMode.IN_FRAMESET
+        return ("reprocess", InsertionMode.IN_FRAMESET, token)
+
+    # Helpers ----------------------------------------------------------------
+
+    _MODE_HANDLERS = [
+        _mode_initial,
+        _mode_before_html,
+        _mode_before_head,
+        _mode_in_head,
+        _mode_in_head_noscript,
+        _mode_after_head,
+        _mode_text,
+        _mode_in_body,
+        _mode_after_body,
+        _mode_after_after_body,
+        _mode_in_table,
+        _mode_in_table_text,
+        _mode_in_caption,
+        _mode_in_column_group,
+        _mode_in_table_body,
+        _mode_in_row,
+        _mode_in_cell,
+        _mode_in_frameset,
+        _mode_after_frameset,
+        _mode_after_after_frameset,
+        _mode_in_select,
+        _mode_in_template,
+    ]
+
+    _BODY_TOKEN_HANDLERS = {
+        CharacterTokens: _handle_characters_in_body,
+        CommentToken: _handle_comment_in_body,
+        Tag: _handle_tag_in_body,
+        EOFToken: _handle_eof_in_body,
+    }
+
+    _BODY_START_HANDLERS = {
+        "a": _handle_body_start_a,
+        "address": _handle_body_start_block_with_p,
+        "applet": _handle_body_start_applet_like,
+        "area": _handle_body_start_void_with_formatting,
+        "article": _handle_body_start_block_with_p,
+        "aside": _handle_body_start_block_with_p,
+        "b": _handle_body_start_formatting,
+        "base": _handle_body_start_in_head,
+        "basefont": _handle_body_start_in_head,
+        "bgsound": _handle_body_start_in_head,
+        "big": _handle_body_start_formatting,
+        "blockquote": _handle_body_start_block_with_p,
+        "body": _handle_body_start_body,
+        "br": _handle_body_start_br,
+        "button": _handle_body_start_button,
+        "caption": _handle_body_start_table_parse_error,
+        "center": _handle_body_start_block_with_p,
+        "code": _handle_body_start_formatting,
+        "col": _handle_body_start_col_or_frame,
+        "colgroup": _handle_body_start_structure_ignored,
+        "dd": _handle_body_start_dd_dt,
+        "details": _handle_body_start_block_with_p,
+        "dialog": _handle_body_start_block_with_p,
+        "dir": _handle_body_start_block_with_p,
+        "div": _handle_body_start_block_with_p,
+        "dl": _handle_body_start_block_with_p,
+        "dt": _handle_body_start_dd_dt,
+        "em": _handle_body_start_formatting,
+        "embed": _handle_body_start_void_with_formatting,
+        "fieldset": _handle_body_start_block_with_p,
+        "figcaption": _handle_body_start_block_with_p,
+        "figure": _handle_body_start_block_with_p,
+        "font": _handle_body_start_formatting,
+        "footer": _handle_body_start_block_with_p,
+        "form": _handle_body_start_form,
+        "frame": _handle_body_start_col_or_frame,
+        "frameset": _handle_body_start_frameset,
+        "h1": _handle_body_start_heading,
+        "h2": _handle_body_start_heading,
+        "h3": _handle_body_start_heading,
+        "h4": _handle_body_start_heading,
+        "h5": _handle_body_start_heading,
+        "h6": _handle_body_start_heading,
+        "head": _handle_body_start_head,
+        "header": _handle_body_start_block_with_p,
+        "hgroup": _handle_body_start_block_with_p,
+        "html": _handle_body_start_html,
+        "i": _handle_body_start_formatting,
+        "image": _handle_body_start_image,
+        "img": _handle_body_start_void_with_formatting,
+        "input": _handle_body_start_input,
+        "keygen": _handle_body_start_void_with_formatting,
+        "li": _handle_body_start_li,
+        "link": _handle_body_start_in_head,
+        "listing": _handle_body_start_pre_listing,
+        "main": _handle_body_start_block_with_p,
+        "marquee": _handle_body_start_applet_like,
+        "math": _handle_body_start_math,
+        "menu": _handle_body_start_block_with_p,
+        "meta": _handle_body_start_in_head,
+        "nav": _handle_body_start_block_with_p,
+        "nobr": _handle_body_start_formatting,
+        "noframes": _handle_body_start_in_head,
+        "object": _handle_body_start_applet_like,
+        "ol": _handle_body_start_block_with_p,
+        "optgroup": _handle_body_start_optgroup,
+        "option": _handle_body_start_option,
+        "p": _handle_body_start_paragraph,
+        "param": _handle_body_start_simple_void,
+        "plaintext": _handle_body_start_plaintext_xmp,
+        "pre": _handle_body_start_pre_listing,
+        "rb": _handle_body_start_rb_rtc,
+        "rp": _handle_body_start_rp_rt,
+        "rt": _handle_body_start_rp_rt,
+        "rtc": _handle_body_start_rb_rtc,
+        "s": _handle_body_start_formatting,
+        "script": _handle_body_start_in_head,
+        "search": _handle_body_start_block_with_p,
+        "section": _handle_body_start_block_with_p,
+        "select": _handle_body_start_select,
+        "small": _handle_body_start_formatting,
+        "source": _handle_body_start_simple_void,
+        "strike": _handle_body_start_formatting,
+        "strong": _handle_body_start_formatting,
+        "style": _handle_body_start_in_head,
+        "summary": _handle_body_start_block_with_p,
+        "svg": _handle_body_start_svg,
+        "table": _handle_body_start_table,
+        "tbody": _handle_body_start_structure_ignored,
+        "td": _handle_body_start_structure_ignored,
+        "template": _handle_body_start_in_head,
+        "textarea": _handle_body_start_textarea,
+        "tfoot": _handle_body_start_structure_ignored,
+        "th": _handle_body_start_structure_ignored,
+        "thead": _handle_body_start_structure_ignored,
+        "title": _handle_body_start_in_head,
+        "tr": _handle_body_start_structure_ignored,
+        "track": _handle_body_start_simple_void,
+        "tt": _handle_body_start_formatting,
+        "u": _handle_body_start_formatting,
+        "ul": _handle_body_start_block_with_p,
+        "wbr": _handle_body_start_void_with_formatting,
+        "xmp": _handle_body_start_plaintext_xmp,
+    }
+    _BODY_END_HANDLERS = {
+        "address": _handle_body_end_block,
+        "applet": _handle_body_end_applet_like,
+        "article": _handle_body_end_block,
+        "aside": _handle_body_end_block,
+        "blockquote": _handle_body_end_block,
+        "body": _handle_body_end_body,
+        "button": _handle_body_end_block,
+        "center": _handle_body_end_block,
+        "dd": _handle_body_end_dd_dt,
+        "details": _handle_body_end_block,
+        "dialog": _handle_body_end_block,
+        "dir": _handle_body_end_block,
+        "div": _handle_body_end_block,
+        "dl": _handle_body_end_block,
+        "dt": _handle_body_end_dd_dt,
+        "fieldset": _handle_body_end_block,
+        "figcaption": _handle_body_end_block,
+        "figure": _handle_body_end_block,
+        "footer": _handle_body_end_block,
+        "form": _handle_body_end_form,
+        "h1": _handle_body_end_heading,
+        "h2": _handle_body_end_heading,
+        "h3": _handle_body_end_heading,
+        "h4": _handle_body_end_heading,
+        "h5": _handle_body_end_heading,
+        "h6": _handle_body_end_heading,
+        "header": _handle_body_end_block,
+        "hgroup": _handle_body_end_block,
+        "html": _handle_body_end_html,
+        "li": _handle_body_end_li,
+        "listing": _handle_body_end_block,
+        "main": _handle_body_end_block,
+        "marquee": _handle_body_end_applet_like,
+        "menu": _handle_body_end_block,
+        "nav": _handle_body_end_block,
+        "object": _handle_body_end_applet_like,
+        "ol": _handle_body_end_block,
+        "p": _handle_body_end_p,
+        "pre": _handle_body_end_block,
+        "search": _handle_body_end_block,
+        "section": _handle_body_end_block,
+        "summary": _handle_body_end_block,
+        "table": _handle_body_end_block,
+        "template": _handle_body_end_template,
+        "ul": _handle_body_end_block,
+    }