From: Mike Samuel <mikesamuel@gmail.com>
Date: Thu, 22 Sep 2011 02:04:41 +0000 (-0700)
Subject: exp/template/html: simplify transition functions
X-Git-Tag: weekly.2011-09-21~3
X-Git-Url: http://www.git.cypherpunks.su/?a=commitdiff_plain;h=1f577d26d7c7a935c792dacc6cc50826cfd77af0;p=gostls13.git

exp/template/html: simplify transition functions

This simplifies transition functions to make it easier to reliably
elide comments in a later CL.

Before:
- transition functions are responsible for detecting special end tags.
After:
- the code to detect special end tags is done in one place.

We were relying on end tags being skipped which meant we were
not noticing comments inside script/style elements that contain no
substitutions.
This change means we will notice all such comments where necessary,
but stripTags will notice none since it does not need to.  This speeds
up stripTags.

R=nigeltao
CC=golang-dev
https://golang.org/cl/5074041
---

diff --git a/src/pkg/exp/template/html/context.go b/src/pkg/exp/template/html/context.go
index 57d44938ca..de073f134a 100644
--- a/src/pkg/exp/template/html/context.go
+++ b/src/pkg/exp/template/html/context.go
@@ -175,6 +175,15 @@ func isComment(s state) bool {
 	return false
 }
 
+// isInTag return whether s occurs solely inside an HTML tag.
+func isInTag(s state) bool {
+	switch s {
+	case stateTag, stateAttrName, stateAfterName, stateBeforeValue, stateAttr:
+		return true
+	}
+	return false
+}
+
 // delim is the delimiter that will end the current HTML attribute.
 type delim uint8
 
diff --git a/src/pkg/exp/template/html/escape.go b/src/pkg/exp/template/html/escape.go
index 050746c1b2..28019f2525 100644
--- a/src/pkg/exp/template/html/escape.go
+++ b/src/pkg/exp/template/html/escape.go
@@ -583,7 +583,14 @@ func (e *escaper) escapeText(c context, n *parse.TextNode) context {
 // s, then returns the context after those tokens and the unprocessed suffix.
 func contextAfterText(c context, s []byte) (context, int) {
 	if c.delim == delimNone {
-		return transitionFunc[c.state](c, s)
+		c1, i := tSpecialTagEnd(c, s)
+		if i == 0 {
+			// A special end tag (`</script>`) has been seen and
+			// all content preceding it has been consumed.
+			return c1, 0
+		}
+		// Consider all content up to any end tag.
+		return transitionFunc[c.state](c, s[:i])
 	}
 
 	i := bytes.IndexAny(s, delimEnds[c.delim])
diff --git a/src/pkg/exp/template/html/escape_test.go b/src/pkg/exp/template/html/escape_test.go
index cf1c828002..84bf6b7a4a 100644
--- a/src/pkg/exp/template/html/escape_test.go
+++ b/src/pkg/exp/template/html/escape_test.go
@@ -814,7 +814,7 @@ func TestErrors(t *testing.T) {
 		},
 		{
 			`<a onclick='alert(/x+\`,
-			`unfinished escape sequence in JS regexp: "x+\\"`,
+			`unfinished escape sequence in JS string: "x+\\"`,
 		},
 		{
 			`<a onclick="/foo[\]/`,
diff --git a/src/pkg/exp/template/html/html.go b/src/pkg/exp/template/html/html.go
index 27ef65229a..3924b193db 100644
--- a/src/pkg/exp/template/html/html.go
+++ b/src/pkg/exp/template/html/html.go
@@ -165,12 +165,17 @@ func htmlReplacer(s string, replacementTable []string, badRunes bool) string {
 // For example, `<b>&iexcl;Hi!</b> <script>...</script>` -> `&iexcl;Hi! `.
 func stripTags(html string) string {
 	var b bytes.Buffer
-	s, c, i := []byte(html), context{}, 0
+	s, c, i, allText := []byte(html), context{}, 0, true
 	// Using the transition funcs helps us avoid mangling
 	// `<div title="1>2">` or `I <3 Ponies!`.
 	for i != len(s) {
 		if c.delim == delimNone {
-			d, nread := transitionFunc[c.state](c, s[i:])
+			st := c.state
+			// Use RCDATA instead of parsing into JS or CSS styles.
+			if c.element != elementNone && !isInTag(st) {
+				st = stateRCDATA
+			}
+			d, nread := transitionFunc[st](c, s[i:])
 			i1 := i + nread
 			if c.state == stateText || c.state == stateRCDATA {
 				// Emit text up to the start of the tag or comment.
@@ -184,6 +189,8 @@ func stripTags(html string) string {
 					}
 				}
 				b.Write(s[i:j])
+			} else {
+				allText = false
 			}
 			c, i = d, i1
 			continue
@@ -198,10 +205,9 @@ func stripTags(html string) string {
 		}
 		c, i = context{state: stateTag, element: c.element}, i1
 	}
-	if c.state == stateText {
-		if b.Len() == 0 {
-			return html
-		}
+	if allText {
+		return html
+	} else if c.state == stateText || c.state == stateRCDATA {
 		b.Write(s[i:])
 	}
 	return b.String()
diff --git a/src/pkg/exp/template/html/html_test.go b/src/pkg/exp/template/html/html_test.go
index 2866fdd0ce..e178d0f27e 100644
--- a/src/pkg/exp/template/html/html_test.go
+++ b/src/pkg/exp/template/html/html_test.go
@@ -59,6 +59,7 @@ func TestStripTags(t *testing.T) {
 		{`Foo<script type="text/javascript">alert(1337)</script>Bar`, "FooBar"},
 		{`Foo<div title="1>2">Bar`, "FooBar"},
 		{`I <3 Ponies!`, `I <3 Ponies!`},
+		{`<script>foo()</script>`, ``},
 	}
 
 	for _, test := range tests {
diff --git a/src/pkg/exp/template/html/transition.go b/src/pkg/exp/template/html/transition.go
index 0274d9e710..3be3a01a8a 100644
--- a/src/pkg/exp/template/html/transition.go
+++ b/src/pkg/exp/template/html/transition.go
@@ -27,9 +27,9 @@ var transitionFunc = [...]func(context, []byte) (context, int){
 	stateAttr:        tAttr,
 	stateURL:         tURL,
 	stateJS:          tJS,
-	stateJSDqStr:     tJSStr,
-	stateJSSqStr:     tJSStr,
-	stateJSRegexp:    tJSRegexp,
+	stateJSDqStr:     tJSDelimited,
+	stateJSSqStr:     tJSDelimited,
+	stateJSRegexp:    tJSDelimited,
 	stateJSBlockCmt:  tBlockCmt,
 	stateJSLineCmt:   tLineCmt,
 	stateCSS:         tCSS,
@@ -57,14 +57,18 @@ func tText(c context, s []byte) (context, int) {
 			return context{state: stateHTMLCmt}, i + 4
 		}
 		i++
+		end := false
 		if s[i] == '/' {
 			if i+1 == len(s) {
 				return c, len(s)
 			}
-			i++
+			end, i = true, i+1
 		}
 		j, e := eatTagName(s, i)
 		if j != i {
+			if end {
+				e = elementNone
+			}
 			// We've found an HTML tag.
 			return context{state: stateTag, element: e}, j
 		}
@@ -122,10 +126,9 @@ func tAttrName(c context, s []byte) (context, int) {
 	i, err := eatAttrName(s, 0)
 	if err != nil {
 		return context{state: stateError, err: err}, len(s)
-	} else if i == len(s) {
-		return c, len(s)
+	} else if i != len(s) {
+		c.state = stateAfterName
 	}
-	c.state = stateAfterName
 	return c, i
 }
 
@@ -172,8 +175,7 @@ func tBeforeValue(c context, s []byte) (context, int) {
 
 // tHTMLCmt is the context transition function for stateHTMLCmt.
 func tHTMLCmt(c context, s []byte) (context, int) {
-	i := bytes.Index(s, commentEnd)
-	if i != -1 {
+	if i := bytes.Index(s, commentEnd); i != -1 {
 		return context{}, i + 3
 	}
 	return c, len(s)
@@ -192,10 +194,8 @@ var specialTagEndMarkers = [...]string{
 // element states.
 func tSpecialTagEnd(c context, s []byte) (context, int) {
 	if c.element != elementNone {
-		end := specialTagEndMarkers[c.element]
-		i := strings.Index(strings.ToLower(string(s)), end)
-		if i != -1 {
-			return context{state: stateTag}, i + len(end)
+		if i := strings.Index(strings.ToLower(string(s)), specialTagEndMarkers[c.element]); i != -1 {
+			return context{}, i
 		}
 	}
 	return c, len(s)
@@ -220,10 +220,6 @@ func tURL(c context, s []byte) (context, int) {
 
 // tJS is the context transition function for the JS state.
 func tJS(c context, s []byte) (context, int) {
-	if d, i := tSpecialTagEnd(c, s); i != len(s) {
-		return d, i
-	}
-
 	i := bytes.IndexAny(s, `"'/`)
 	if i == -1 {
 		// Entire input is non string, comment, regexp tokens.
@@ -258,64 +254,30 @@ func tJS(c context, s []byte) (context, int) {
 	return c, i + 1
 }
 
-// tJSStr is the context transition function for the JS string states.
-func tJSStr(c context, s []byte) (context, int) {
-	if d, i := tSpecialTagEnd(c, s); i != len(s) {
-		return d, i
-	}
-
-	quoteAndEsc := `\"`
-	if c.state == stateJSSqStr {
-		quoteAndEsc = `\'`
-	}
-
-	k := 0
-	for {
-		i := k + bytes.IndexAny(s[k:], quoteAndEsc)
-		if i < k {
-			return c, len(s)
-		}
-		if s[i] == '\\' {
-			i++
-			if i == len(s) {
-				return context{
-					state: stateError,
-					err:   errorf(ErrPartialEscape, 0, "unfinished escape sequence in JS string: %q", s),
-				}, len(s)
-			}
-		} else {
-			c.state, c.jsCtx = stateJS, jsCtxDivOp
-			return c, i + 1
-		}
-		k = i + 1
-	}
-	panic("unreachable")
-}
-
-// tJSRegexp is the context transition function for the /RegExp/ literal state.
-func tJSRegexp(c context, s []byte) (context, int) {
-	if d, i := tSpecialTagEnd(c, s); i != len(s) {
-		return d, i
+// tJSDelimited is the context transition function for the JS string and regexp
+// states.
+func tJSDelimited(c context, s []byte) (context, int) {
+	specials := `\"`
+	switch c.state {
+	case stateJSSqStr:
+		specials = `\'`
+	case stateJSRegexp:
+		specials = `\/[]`
 	}
 
 	k, inCharset := 0, false
 	for {
-		i := k + bytes.IndexAny(s[k:], `\/[]`)
+		i := k + bytes.IndexAny(s[k:], specials)
 		if i < k {
 			break
 		}
 		switch s[i] {
-		case '/':
-			if !inCharset {
-				c.state, c.jsCtx = stateJS, jsCtxDivOp
-				return c, i + 1
-			}
 		case '\\':
 			i++
 			if i == len(s) {
 				return context{
 					state: stateError,
-					err:   errorf(ErrPartialEscape, 0, "unfinished escape sequence in JS regexp: %q", s),
+					err:   errorf(ErrPartialEscape, 0, "unfinished escape sequence in JS string: %q", s),
 				}, len(s)
 			}
 		case '[':
@@ -323,7 +285,11 @@ func tJSRegexp(c context, s []byte) (context, int) {
 		case ']':
 			inCharset = false
 		default:
-			panic("unreachable")
+			// end delimiter
+			if !inCharset {
+				c.state, c.jsCtx = stateJS, jsCtxDivOp
+				return c, i + 1
+			}
 		}
 		k = i + 1
 	}
@@ -344,9 +310,6 @@ var blockCommentEnd = []byte("*/")
 
 // tBlockCmt is the context transition function for /*comment*/ states.
 func tBlockCmt(c context, s []byte) (context, int) {
-	if d, i := tSpecialTagEnd(c, s); i != len(s) {
-		return d, i
-	}
 	i := bytes.Index(s, blockCommentEnd)
 	if i == -1 {
 		return c, len(s)
@@ -364,9 +327,6 @@ func tBlockCmt(c context, s []byte) (context, int) {
 
 // tLineCmt is the context transition function for //comment states.
 func tLineCmt(c context, s []byte) (context, int) {
-	if d, i := tSpecialTagEnd(c, s); i != len(s) {
-		return d, i
-	}
 	var lineTerminators string
 	var endState state
 	switch c.state {
@@ -400,10 +360,6 @@ func tLineCmt(c context, s []byte) (context, int) {
 
 // tCSS is the context transition function for the CSS state.
 func tCSS(c context, s []byte) (context, int) {
-	if d, i := tSpecialTagEnd(c, s); i != len(s) {
-		return d, i
-	}
-
 	// CSS quoted strings are almost never used except for:
 	// (1) URLs as in background: "/foo.png"
 	// (2) Multiword font-names as in font-family: "Times New Roman"
@@ -478,10 +434,6 @@ func tCSS(c context, s []byte) (context, int) {
 
 // tCSSStr is the context transition function for the CSS string and URL states.
 func tCSSStr(c context, s []byte) (context, int) {
-	if d, i := tSpecialTagEnd(c, s); i != len(s) {
-		return d, i
-	}
-
 	var endAndEsc string
 	switch c.state {
 	case stateCSSDqStr, stateCSSDqURL: