html: parse doctype tokens; merge adjacent text nodes.

The test case input is "<!DOCTYPE html><span><button>foo</span>bar". The correct parse is: | <!DOCTYPE html> | <html> | <head> | <body> | <span> | <button> | "foobar" R=gri CC=golang-dev https://golang.org/cl/4794063

html: parse doctype tokens; merge adjacent text nodes.
1d0c141d · Nigel Tao · 4d2766e9 · 1d0c141d · 1d0c141d · 1d0c141d
Commit 1d0c141d authored Aug 01, 2011 by Nigel Tao
4 changed files
--- a/src/pkg/html/node.go
+++ b/src/pkg/html/node.go
@@ -13,6 +13,7 @@ const (
 	DocumentNode
 	ElementNode
 	CommentNode
+	DoctypeNode
 	scopeMarkerNode
 )


--- a/src/pkg/html/parse.go
+++ b/src/pkg/html/parse.go
@@ -81,8 +81,8 @@ func (p *parser) popUntil(stopTags []string, matchTags ...string) bool {
 	return false
 }

-// addChild adds a child node n to the top element, and pushes n if it is an
-// element node (text nodes are not part of the stack of open elements).
+// addChild adds a child node n to the top element, and pushes n onto the stack
+// of open elements if it is an element node.
 func (p *parser) addChild(n *Node) {
 	p.top().Add(n)
 	if n.Type == ElementNode {
@@ -90,10 +90,15 @@ func (p *parser) addChild(n *Node) {
 	}
 }

-// addText calls addChild with a text node.
+// addText adds text to the preceding node if it is a text node, or else it
+// calls addChild with a new text node.
 func (p *parser) addText(text string) {
-	// TODO: merge s with previous text, if the preceding node is a text node.
 	// TODO: distinguish whitespace text from others.
+	t := p.top()
+	if i := len(t.Child); i > 0 && t.Child[i-1].Type == TextNode {
+		t.Child[i-1].Data += text
+		return
+	}
 	p.addChild(&Node{
 		Type: TextNode,
 		Data: text,
@@ -201,7 +206,15 @@ func useTheRulesFor(p *parser, actual, delegate insertionMode) (insertionMode, b

 // Section 11.2.5.4.1.
 func initialIM(p *parser) (insertionMode, bool) {
-	// TODO: check p.tok for DOCTYPE.
+	if p.tok.Type == DoctypeToken {
+		p.addChild(&Node{
+			Type: DoctypeNode,
+			Data: p.tok.Data,
+		})
+		return beforeHTMLIM, true
+	}
+	// TODO: set "quirks mode"? It's defined in the DOM spec instead of HTML5 proper,
+	// and so switching on "quirks mode" might belong in a different package.
 	return beforeHTMLIM, false
 }


--- a/src/pkg/html/parse_test.go
+++ b/src/pkg/html/parse_test.go
@@ -85,6 +85,8 @@ func dumpLevel(w io.Writer, n *Node, level int) os.Error {
 		fmt.Fprintf(w, "%q", EscapeString(n.Data))
 	case CommentNode:
 		return os.NewError("COMMENT")
+	case DoctypeNode:
+		fmt.Fprintf(w, "<!DOCTYPE %s>", EscapeString(n.Data))
 	case scopeMarkerNode:
 		return os.NewError("unexpected scopeMarkerNode")
 	default:
@@ -121,7 +123,7 @@ func TestParser(t *testing.T) {
 		rc := make(chan io.Reader)
 		go readDat(filename, rc)
 		// TODO(nigeltao): Process all test cases, not just a subset.
-		for i := 0; i < 23; i++ {
+		for i := 0; i < 25; i++ {
 			// Parse the #data section.
 			b, err := ioutil.ReadAll(<-rc)
 			if err != nil {

--- a/src/pkg/html/token.go
+++ b/src/pkg/html/token.go