LibWeb: Implement fragment parsing and use it for Element.innerHTML

This patch implements most of the HTML fragment parsing algorithm and
ports Element::set_inner_html() to it. This was the last remaining user
of the old HTML parser. :^)

7 files changed, 91 insertions, 18 deletions

diff --git a/Libraries/LibWeb/DOM/Document.cpp b/Libraries/LibWeb/DOM/Document.cpp
index 4c9810f374..93f27062ee 100644
--- a/Libraries/LibWeb/DOM/Document.cpp
+++ b/Libraries/LibWeb/DOM/Document.cpp
@@ -468,4 +468,12 @@ NonnullRefPtrVector<HTMLScriptElement> Document::take_scripts_to_execute_as_soon
     return move(m_scripts_to_execute_as_soon_as_possible);
 }
 
+void Document::adopt_node(Node& subtree_root)
+{
+    subtree_root.for_each_in_subtree([&](auto& node) {
+        node.set_document({}, *this);
+        return IterationDecision::Continue;
+    });
+}
+
 }
diff --git a/Libraries/LibWeb/DOM/Document.h b/Libraries/LibWeb/DOM/Document.h
index ff03cd1329..f40f0f81cb 100644
--- a/Libraries/LibWeb/DOM/Document.h
+++ b/Libraries/LibWeb/DOM/Document.h
@@ -145,6 +145,8 @@ public:
     bool in_quirks_mode() const { return m_quirks_mode; }
     void set_quirks_mode(bool mode) { m_quirks_mode = mode; }
 
+    void adopt_node(Node&);
+
 private:
     virtual RefPtr<LayoutNode> create_layout_node(const StyleProperties* parent_style) override;
 
diff --git a/Libraries/LibWeb/DOM/Element.cpp b/Libraries/LibWeb/DOM/Element.cpp
index af7dd041a7..f6439ee87a 100644
--- a/Libraries/LibWeb/DOM/Element.cpp
+++ b/Libraries/LibWeb/DOM/Element.cpp
@@ -41,7 +41,7 @@
 #include <LibWeb/Layout/LayoutTableRow.h>
 #include <LibWeb/Layout/LayoutTableRowGroup.h>
 #include <LibWeb/Layout/LayoutTreeBuilder.h>
-#include <LibWeb/Parser/HTMLParser.h>
+#include <LibWeb/Parser/HTMLDocumentParser.h>
 
 namespace Web {
 
@@ -247,13 +247,10 @@ NonnullRefPtr<StyleProperties> Element::computed_style()
 
 void Element::set_inner_html(StringView markup)
 {
-    auto fragment = parse_html_fragment(document(), markup);
+    auto new_children = HTMLDocumentParser::parse_html_fragment(*this, markup);
     remove_all_children();
-    if (!fragment)
-        return;
-    while (RefPtr<Node> child = fragment->first_child()) {
-        fragment->remove_child(*child);
-        append_child(*child);
+    while (!new_children.is_empty()) {
+        append_child(new_children.take_first());
     }
 
     set_needs_style_update(true);
diff --git a/Libraries/LibWeb/DOM/Node.cpp b/Libraries/LibWeb/DOM/Node.cpp
index 80aa2248c6..9869de0e52 100644
--- a/Libraries/LibWeb/DOM/Node.cpp
+++ b/Libraries/LibWeb/DOM/Node.cpp
@@ -51,7 +51,7 @@
 namespace Web {
 
 Node::Node(Document& document, NodeType type)
-    : m_document(document)
+    : m_document(&document)
     , m_type(type)
 {
 }
@@ -212,4 +212,9 @@ RefPtr<Node> Node::insert_before(NonnullRefPtr<Node> node, RefPtr<Node> child, b
     return node;
 }
 
+void Node::set_document(Badge<Document>, Document& document)
+{
+    m_document = &document;
+}
+
 }
diff --git a/Libraries/LibWeb/DOM/Node.h b/Libraries/LibWeb/DOM/Node.h
index 987e0794dc..4e1da6b0e1 100644
--- a/Libraries/LibWeb/DOM/Node.h
+++ b/Libraries/LibWeb/DOM/Node.h
@@ -91,8 +91,8 @@ public:
 
     virtual String text_content() const;
 
-    Document& document() { return m_document; }
-    const Document& document() const { return m_document; }
+    Document& document() { return *m_document; }
+    const Document& document() const { return *m_document; }
 
     const HTMLAnchorElement* enclosing_link_element() const;
     const HTMLElement* enclosing_html_element() const;
@@ -140,10 +140,12 @@ public:
     virtual void document_did_attach_to_frame(Frame&) {}
     virtual void document_will_detach_from_frame(Frame&) {}
 
+    void set_document(Badge<Document>, Document&);
+
 protected:
     Node(Document&, NodeType);
 
-    Document& m_document;
+    Document* m_document { nullptr };
     mutable LayoutNode* m_layout_node { nullptr };
     NodeType m_type { NodeType::INVALID };
     bool m_needs_style_update { true };
diff --git a/Libraries/LibWeb/Parser/HTMLDocumentParser.cpp b/Libraries/LibWeb/Parser/HTMLDocumentParser.cpp
index 1c316e0d41..4d98e935d2 100644
--- a/Libraries/LibWeb/Parser/HTMLDocumentParser.cpp
+++ b/Libraries/LibWeb/Parser/HTMLDocumentParser.cpp
@@ -56,6 +56,7 @@ RefPtr<Document> parse_html_document(const StringView& data, const URL& url, con
 HTMLDocumentParser::HTMLDocumentParser(const StringView& input, const String& encoding)
     : m_tokenizer(input, encoding)
 {
+    m_document = adopt(*new Document);
 }
 
 HTMLDocumentParser::~HTMLDocumentParser()
@@ -64,7 +65,6 @@ HTMLDocumentParser::~HTMLDocumentParser()
 
 void HTMLDocumentParser::run(const URL& url)
 {
-    m_document = adopt(*new Document);
     m_document->set_url(url);
     m_document->set_source(m_tokenizer.source());
 
@@ -2518,10 +2518,17 @@ void HTMLDocumentParser::handle_after_after_frameset(HTMLToken& token)
     PARSE_ERROR();
 }
 
-void HTMLDocumentParser::reset_the_insertion_mode_appropriately()
+void HTMLDocumentParser::reset_the_insertion_mode_appropriately(const Element* context_element)
 {
     for (ssize_t i = m_stack_of_open_elements.elements().size() - 1; i >= 0; --i) {
-        RefPtr<Element> node = m_stack_of_open_elements.elements().at(i);
+
+        // NOTE: When parsing fragments, we substitute the context element for the root of the stack of open elements.
+        RefPtr<Element> node;
+        if (i == 0 && context_element) {
+            node = context_element;
+        } else {
+            node = m_stack_of_open_elements.elements().at(i);
+        }
 
         if (node->tag_name() == HTML::TagNames::select) {
             TODO();
@@ -2580,9 +2587,6 @@ void HTMLDocumentParser::reset_the_insertion_mode_appropriately()
     }
 
     m_insertion_mode = InsertionMode::InBody;
-    if (m_parsing_fragment) {
-        TODO();
-    }
 }
 
 const char* HTMLDocumentParser::insertion_mode_name() const
@@ -2601,4 +2605,57 @@ Document& HTMLDocumentParser::document()
 {
     return *m_document;
 }
+
+NonnullRefPtrVector<Node> HTMLDocumentParser::parse_html_fragment(Element& context_element, const StringView& markup)
+{
+    HTMLDocumentParser parser(markup, "utf-8");
+    parser.m_parsing_fragment = true;
+    parser.document().set_quirks_mode(context_element.document().in_quirks_mode());
+
+    if (context_element.tag_name().is_one_of(HTML::TagNames::title, HTML::TagNames::textarea)) {
+        parser.m_tokenizer.switch_to({}, HTMLTokenizer::State::RCDATA);
+    } else if (context_element.tag_name().is_one_of(HTML::TagNames::style, HTML::TagNames::xmp, HTML::TagNames::iframe, HTML::TagNames::noembed, HTML::TagNames::noframes)) {
+        parser.m_tokenizer.switch_to({}, HTMLTokenizer::State::RAWTEXT);
+    } else if (context_element.tag_name().is_one_of(HTML::TagNames::script)) {
+        parser.m_tokenizer.switch_to({}, HTMLTokenizer::State::ScriptData);
+    } else if (context_element.tag_name().is_one_of(HTML::TagNames::noscript)) {
+        if (context_element.document().is_scripting_enabled())
+            parser.m_tokenizer.switch_to({}, HTMLTokenizer::State::RAWTEXT);
+    } else if (context_element.tag_name().is_one_of(HTML::TagNames::noscript)) {
+        if (context_element.document().is_scripting_enabled())
+            parser.m_tokenizer.switch_to({}, HTMLTokenizer::State::RAWTEXT);
+    } else if (context_element.tag_name().is_one_of(HTML::TagNames::plaintext)) {
+        parser.m_tokenizer.switch_to({}, HTMLTokenizer::State::PLAINTEXT);
+    }
+
+    auto root = create_element(context_element.document(), HTML::TagNames::html);
+    parser.document().append_child(root);
+    parser.m_stack_of_open_elements.push(root);
+
+    if (context_element.tag_name() == HTML::TagNames::template_) {
+        TODO();
+    }
+
+    // FIXME: Create a start tag token whose name is the local name of context and whose attributes are the attributes of context.
+
+    parser.reset_the_insertion_mode_appropriately(&context_element);
+
+    for (auto* form_candidate = &context_element; form_candidate; form_candidate = form_candidate->parent_element()) {
+        if (is<HTMLFormElement>(*form_candidate)) {
+            parser.m_form_element = to<HTMLFormElement>(*form_candidate);
+            break;
+        }
+    }
+
+    parser.run(context_element.document().url());
+
+    NonnullRefPtrVector<Node> children;
+    while (RefPtr<Node> child = root->first_child()) {
+        root->remove_child(*child);
+        context_element.document().adopt_node(*child);
+        children.append(*child);
+    }
+    return children;
+}
+
 }
diff --git a/Libraries/LibWeb/Parser/HTMLDocumentParser.h b/Libraries/LibWeb/Parser/HTMLDocumentParser.h
index fa1bb2c33f..4f33776b62 100644
--- a/Libraries/LibWeb/Parser/HTMLDocumentParser.h
+++ b/Libraries/LibWeb/Parser/HTMLDocumentParser.h
@@ -70,6 +70,8 @@ public:
 
     Document& document();
 
+    static NonnullRefPtrVector<Node> parse_html_fragment(Element& context_element, const StringView&);
+
     enum class InsertionMode {
 #define __ENUMERATE_INSERTION_MODE(mode) mode,
         ENUMERATE_INSERTION_MODES
@@ -134,7 +136,7 @@ private:
     void increment_script_nesting_level();
     void decrement_script_nesting_level();
     size_t script_nesting_level() const { return m_script_nesting_level; }
-    void reset_the_insertion_mode_appropriately();
+    void reset_the_insertion_mode_appropriately(const Element* context_element = nullptr);
 
     void adjust_mathml_attributes(HTMLToken&);
     void adjust_svg_attributes(HTMLToken&);