exp/template/html: autoescape actions in HTML style attributes.

This does not wire up <style> elements as that is pending support
for raw text content in CL https://golang.org/cl/4964045/

This CL allows actions to appear in contexts like

selectors:        {{.Tag}}{{.Class}}{{.Id}}
property names:   border-{{.BidiLeadingEdge}}
property values:  color: {{.Color}}
strings:          font-family: "{{font-name}}"
URL strings:      background: "/foo?image={{.ImgQuery}}"
URL literals:     background: url("{{.Image}}")

but disallows actions inside CSS comments and disallows
embedding of JS in CSS entirely.

It is based on the CSS3 lexical grammar with affordances for
common browser extensions including line comments.

R=nigeltao
CC=golang-dev
https://golang.org/cl/4968058

src/pkg/exp/template/html/Makefile

@@ -7,7 +7,10 @@ include ../../../../Make.inc
 TARG=exp/template/html
 GOFILES=\
 	context.go\
+	css.go\
 	escape.go\
+	html.go\
 	js.go\
+	url.go\
 
 include ../../../../Make.pkg

src/pkg/exp/template/html/context.go

@@ -63,6 +63,22 @@ const (
 	stateJSBlockCmt
 	// stateJSLineCmt occurs inside a JavaScript // line comment.
 	stateJSLineCmt
+	// stateCSS occurs inside a <style> element or style attribute.
+	stateCSS
+	// stateCSSDqStr occurs inside a CSS double quoted string.
+	stateCSSDqStr
+	// stateCSSSqStr occurs inside a CSS single quoted string.
+	stateCSSSqStr
+	// stateCSSDqURL occurs inside a CSS double quoted url("...").
+	stateCSSDqURL
+	// stateCSSSqURL occurs inside a CSS single quoted url('...').
+	stateCSSSqURL
+	// stateCSSURL occurs inside a CSS unquoted url(...).
+	stateCSSURL
+	// stateCSSBlockCmt occurs inside a CSS /* block comment */.
+	stateCSSBlockCmt
+	// stateCSSLineCmt occurs inside a CSS // line comment.
+	stateCSSLineCmt
 	// stateError is an infectious error state outside any valid
 	// HTML/CSS/JS construct.
 	stateError
@@ -79,6 +95,14 @@ var stateNames = [...]string{
 	stateJSRegexp:    "stateJSRegexp",
 	stateJSBlockCmt:  "stateJSBlockCmt",
 	stateJSLineCmt:   "stateJSLineCmt",
+	stateCSS:         "stateCSS",
+	stateCSSDqStr:    "stateCSSDqStr",
+	stateCSSSqStr:    "stateCSSSqStr",
+	stateCSSDqURL:    "stateCSSDqURL",
+	stateCSSSqURL:    "stateCSSSqURL",
+	stateCSSURL:      "stateCSSURL",
+	stateCSSBlockCmt: "stateCSSBlockCmt",
+	stateCSSLineCmt:  "stateCSSLineCmt",
 	stateError:       "stateError",
 }
 
@@ -132,8 +156,8 @@ const (
 	// urlPartQueryOrFrag occurs in the query portion between the ^s in
 	// "http://auth/path?^k=v#frag^".
 	urlPartQueryOrFrag
-	// urlPartUnknown occurs due to joining of contexts both before and after
-	// the query separator.
+	// urlPartUnknown occurs due to joining of contexts both before and
+	// after the query separator.
 	urlPartUnknown
 )
 

src/pkg/exp/template/html/css.go

+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package html
+
+import (
+	"bytes"
+	"fmt"
+	"unicode"
+	"utf8"
+)
+
+// endsWithCSSKeyword returns whether b ends with an ident that
+// case-insensitively matches the lower-case kw.
+func endsWithCSSKeyword(b []byte, kw string) bool {
+	i := len(b) - len(kw)
+	if i < 0 {
+		// Too short.
+		return false
+	}
+	if i != 0 {
+		r, _ := utf8.DecodeLastRune(b[:i])
+		if isCSSNmchar(r) {
+			// Too long.
+			return false
+		}
+	}
+	// Many CSS keywords, such as "!important" can have characters encoded,
+	// but the URI production does not allow that according to
+	// http://www.w3.org/TR/css3-syntax/#TOK-URI
+	// This does not attempt to recognize encoded keywords. For example,
+	// given "\75\72\6c" and "url" this return false.
+	return string(bytes.ToLower(b[i:])) == kw
+}
+
+// isCSSNmchar returns whether rune is allowed anywhere in a CSS identifier.
+func isCSSNmchar(rune int) bool {
+	// Based on the CSS3 nmchar production but ignores multi-rune escape
+	// sequences.
+	// http://www.w3.org/TR/css3-syntax/#SUBTOK-nmchar
+	return 'a' <= rune && rune <= 'z' ||
+		'A' <= rune && rune <= 'Z' ||
+		'0' <= rune && rune <= '9' ||
+		'-' == rune ||
+		'_' == rune ||
+		// Non-ASCII cases below.
+		0x80 <= rune && rune <= 0xd7ff ||
+		0xe000 <= rune && rune <= 0xfffd ||
+		0x10000 <= rune && rune <= 0x10ffff
+}
+
+// decodeCSS decodes CSS3 escapes given a sequence of stringchars.
+// If there is no change, it returns the input, otherwise it returns a slice
+// backed by a new array.
+// http://www.w3.org/TR/css3-syntax/#SUBTOK-stringchar defines stringchar.
+func decodeCSS(s []byte) []byte {
+	i := bytes.IndexByte(s, '\\')
+	if i == -1 {
+		return s
+	}
+	// The UTF-8 sequence for a codepoint is never longer than 1 + the
+	// number hex digits need to represent that codepoint, so len(s) is an
+	// upper bound on the output length.
+	b := make([]byte, 0, len(s))
+	for len(s) != 0 {
+		i := bytes.IndexByte(s, '\\')
+		if i == -1 {
+			i = len(s)
+		}
+		b, s = append(b, s[:i]...), s[i:]
+		if len(s) < 2 {
+			break
+		}
+		// http://www.w3.org/TR/css3-syntax/#SUBTOK-escape
+		// escape ::= unicode | '\' [#x20-#x7E#x80-#xD7FF#xE000-#xFFFD#x10000-#x10FFFF]
+		if isHex(s[1]) {
+			// http://www.w3.org/TR/css3-syntax/#SUBTOK-unicode
+			//   unicode ::= '\' [0-9a-fA-F]{1,6} wc?
+			j := 2
+			for j < len(s) && j < 7 && isHex(s[j]) {
+				j++
+			}
+			rune := hexDecode(s[1:j])
+			if rune > unicode.MaxRune {
+				rune, j = rune/16, j-1
+			}
+			n := utf8.EncodeRune(b[len(b):cap(b)], rune)
+			// The optional space at the end allows a hex
+			// sequence to be followed by a literal hex.
+			// string(decodeCSS([]byte(`\A B`))) == "\nB"
+			b, s = b[:len(b)+n], skipCSSSpace(s[j:])
+		} else {
+			// `\\` decodes to `\` and `\"` to `"`.
+			_, n := utf8.DecodeRune(s[1:])
+			b, s = append(b, s[1:1+n]...), s[1+n:]
+		}
+	}
+	return b
+}
+
+// isHex returns whether the given character is a hex digit.
+func isHex(c byte) bool {
+	return '0' <= c && c <= '9' || 'a' <= c && c <= 'f' || 'A' <= c && c <= 'F'
+}
+
+// hexDecode decodes a short hex digit sequence: "10" -> 16.
+func hexDecode(s []byte) int {
+	n := 0
+	for _, c := range s {
+		n <<= 4
+		switch {
+		case '0' <= c && c <= '9':
+			n |= int(c - '0')
+		case 'a' <= c && c <= 'f':
+			n |= int(c-'a') + 10
+		case 'A' <= c && c <= 'F':
+			n |= int(c-'A') + 10
+		default:
+			panic(fmt.Sprintf("Bad hex digit in %q", s))
+		}
+	}
+	return n
+}
+
+// skipCSSSpace returns a suffix of c, skipping over a single space.
+func skipCSSSpace(c []byte) []byte {
+	if len(c) == 0 {
+		return c
+	}
+	// wc ::= #x9 | #xA | #xC | #xD | #x20
+	switch c[0] {
+	case '\t', '\n', '\f', ' ':
+		return c[1:]
+	case '\r':
+		// This differs from CSS3's wc production because it contains a
+		// probable spec error whereby wc contains all the single byte
+		// sequences in nl (newline) but not CRLF.
+		if len(c) >= 2 && c[1] == '\n' {
+			return c[2:]
+		}
+		return c[1:]
+	}
+	return c
+}
+
+// cssEscaper escapes HTML and CSS special characters using \<hex>+ escapes.
+func cssEscaper(args ...interface{}) string {
+	s := stringify(args...)
+	var b bytes.Buffer
+	written := 0
+	for i, r := range s {
+		var repl string
+		switch r {
+		case 0:
+			repl = `\0`
+		case '\t':
+			repl = `\9`
+		case '\n':
+			repl = `\a`
+		case '\f':
+			repl = `\c`
+		case '\r':
+			repl = `\d`
+		// Encode HTML specials as hex so the output can be embedded
+		// in HTML attributes without further encoding.
+		case '"':
+			repl = `\22`
+		case '&':
+			repl = `\26`
+		case '\'':
+			repl = `\27`
+		case '(':
+			repl = `\28`
+		case ')':
+			repl = `\29`
+		case '+':
+			repl = `\2b`
+		case '/':
+			repl = `\2f`
+		case ':':
+			repl = `\3a`
+		case ';':
+			repl = `\3b`
+		case '<':
+			repl = `\3c`
+		case '>':
+			repl = `\3e`
+		case '\\':
+			repl = `\\`
+		case '{':
+			repl = `\7b`
+		case '}':
+			repl = `\7d`
+		default:
+			continue
+		}
+		b.WriteString(s[written:i])
+		b.WriteString(repl)
+		written = i + utf8.RuneLen(r)
+		if repl != `\\` && (written == len(s) || isHex(s[written])) {
+			b.WriteByte(' ')
+		}
+	}
+	if written == 0 {
+		return s
+	}
+	b.WriteString(s[written:])
+	return b.String()
+}
+
+var expressionBytes = []byte("expression")
+var mozBindingBytes = []byte("mozbinding")
+
+// cssValueFilter allows innocuous CSS values in the output including CSS
+// quantities (10px or 25%), ID or class literals (#foo, .bar), keyword values
+// (inherit, blue), and colors (#888).
+// It filters out unsafe values, such as those that affect token boundaries,
+// and anything that might execute scripts.
+func cssValueFilter(args ...interface{}) string {
+	s, id := decodeCSS([]byte(stringify(args...))), make([]byte, 0, 64)
+
+	// CSS3 error handling is specified as honoring string boundaries per
+	// http://www.w3.org/TR/css3-syntax/#error-handling :
+	//     Malformed declarations. User agents must handle unexpected
+	//     tokens encountered while parsing a declaration by reading until
+	//     the end of the declaration, while observing the rules for
+	//     matching pairs of (), [], {}, "", and '', and correctly handling
+	//     escapes. For example, a malformed declaration may be missing a
+	//     property, colon (:) or value.
+	// So we need to make sure that values do not have mismatched bracket
+	// or quote characters to prevent the browser from restarting parsing
+	// inside a string that might embed JavaScript source.
+	for i, c := range s {
+		switch c {
+		case 0, '"', '\'', '(', ')', '/', ';', '@', '[', '\\', ']', '`', '{', '}':
+			return filterFailsafe
+		case '-':
+			// Disallow <!-- or -->.
+			// -- should not appear in valid identifiers.
+			if i != 0 && '-' == s[i-1] {
+				return filterFailsafe
+			}
+		default:
+			if c < 0x80 && isCSSNmchar(int(c)) {
+				id = append(id, c)
+			}
+		}
+	}
+	id = bytes.ToLower(id)
+	if bytes.Index(id, expressionBytes) != -1 || bytes.Index(id, mozBindingBytes) != -1 {
+		return filterFailsafe
+	}
+	return string(s)
+}

src/pkg/exp/template/html/css_test.go

+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package html
+
+import (
+	"strconv"
+	"strings"
+	"testing"
+)
+
+func TestEndsWithCSSKeyword(t *testing.T) {
+	tests := []struct {
+		css, kw string
+		want    bool
+	}{
+		{"", "url", false},
+		{"url", "url", true},
+		{"URL", "url", true},
+		{"Url", "url", true},
+		{"url", "important", false},
+		{"important", "important", true},
+		{"image-url", "url", false},
+		{"imageurl", "url", false},
+		{"image url", "url", true},
+	}
+	for _, test := range tests {
+		got := endsWithCSSKeyword([]byte(test.css), test.kw)
+		if got != test.want {
+			t.Errorf("want %t but got %t for css=%v, kw=%v", test.want, got, test.css, test.kw)
+		}
+	}
+}
+
+func TestIsCSSNmchar(t *testing.T) {
+	tests := []struct {
+		rune int
+		want bool
+	}{
+		{0, false},
+		{'0', true},
+		{'9', true},
+		{'A', true},
+		{'Z', true},
+		{'a', true},
+		{'z', true},
+		{'_', true},
+		{'-', true},
+		{':', false},
+		{';', false},
+		{' ', false},
+		{0x7f, false},
+		{0x80, true},
+		{0x1234, true},
+		{0xd800, false},
+		{0xdc00, false},
+		{0xfffe, false},
+		{0x10000, true},
+		{0x110000, false},
+	}
+	for _, test := range tests {
+		got := isCSSNmchar(test.rune)
+		if got != test.want {
+			t.Errorf("%q: want %t but got %t", string(test.rune), test.want, got)
+		}
+	}
+}
+
+func TestDecodeCSS(t *testing.T) {
+	tests := []struct {
+		css, want string
+	}{
+		{``, ``},
+		{`foo`, `foo`},
+		{`foo\`, `foo`},
+		{`foo\\`, `foo\`},
+		{`\`, ``},
+		{`\A`, "\n"},
+		{`\a`, "\n"},
+		{`\0a`, "\n"},
+		{`\00000a`, "\n"},
+		{`\000000a`, "\u0000a"},
+		{`\1234 5`, "\u1234" + "5"},
+		{`\1234\20 5`, "\u1234" + " 5"},
+		{`\1234\A 5`, "\u1234" + "\n5"},
+		{"\\1234\t5", "\u1234" + "5"},
+		{"\\1234\n5", "\u1234" + "5"},
+		{"\\1234\r\n5", "\u1234" + "5"},
+		{`\12345`, "\U00012345"},
+		{`\\`, `\`},
+		{`\\ `, `\ `},
+		{`\"`, `"`},
+		{`\'`, `'`},
+		{`\.`, `.`},
+		{`\. .`, `. .`},
+		{
+			`The \3c i\3equick\3c/i\3e,\d\A\3cspan style=\27 color:brown\27\3e brown\3c/span\3e  fox jumps\2028over the \3c canine class=\22lazy\22 \3e dog\3c/canine\3e`,
+			"The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>",
+		},
+	}
+	for _, test := range tests {
+		got := string(decodeCSS([]byte(test.css)))
+		if got != test.want {
+			t.Errorf("%q: want\n\t%q\nbut got\n\t%q", test.css, test.want, got)
+		}
+	}
+}
+
+func TestHexDecode(t *testing.T) {
+	for i := 0; i < 0x200000; i += 101 /* coprime with 16 */ {
+		s := strconv.Itob(i, 16)
+		if got := hexDecode([]byte(s)); got != i {
+			t.Errorf("%s: want %d but got %d", s, i, got)
+		}
+		s = strings.ToUpper(s)
+		if got := hexDecode([]byte(s)); got != i {
+			t.Errorf("%s: want %d but got %d", s, i, got)
+		}
+	}
+}
+
+func TestSkipCSSSpace(t *testing.T) {
+	tests := []struct {
+		css, want string
+	}{
+		{"", ""},
+		{"foo", "foo"},
+		{"\n", ""},
+		{"\r\n", ""},
+		{"\r", ""},
+		{"\t", ""},
+		{" ", ""},
+		{"\f", ""},
+		{" foo", "foo"},
+		{"  foo", " foo"},
+		{`\20`, `\20`},
+	}
+	for _, test := range tests {
+		got := string(skipCSSSpace([]byte(test.css)))
+		if got != test.want {
+			t.Errorf("%q: want %q but got %q", test.css, test.want, got)
+		}
+	}
+}
+
+func TestCSSEscaper(t *testing.T) {
+	input := ("\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f" +
+		"\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" +
+		` !"#$%&'()*+,-./` +
+		`0123456789:;<=>?` +
+		`@ABCDEFGHIJKLMNO` +
+		`PQRSTUVWXYZ[\]^_` +
+		"`abcdefghijklmno" +
+		"pqrstuvwxyz{|}~\x7f" +
+		"\u00A0\u0100\u2028\u2029\ufeff\U0001D11E")
+
+	want := ("\\0\x01\x02\x03\x04\x05\x06\x07" +
+		"\x08\\9\\a\x0b\\c\\d\x0E\x0F" +
+		"\x10\x11\x12\x13\x14\x15\x16\x17" +
+		"\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" +
+		` !\22#$%\26\27\28\29*\2b,-.\2f ` +
+		`0123456789\3a\3b\3c=\3e?` +
+		`@ABCDEFGHIJKLMNO` +
+		`PQRSTUVWXYZ[\\]^_` +
+		"`abcdefghijklmno" +
+		`pqrstuvwxyz\7b|\7d~` + "\u007f" +
+		"\u00A0\u0100\u2028\u2029\ufeff\U0001D11E")
+
+	got := cssEscaper(input)
+	if got != want {
+		t.Errorf("encode: want\n\t%q\nbut got\n\t%q", want, got)
+	}
+
+	got = string(decodeCSS([]byte(got)))
+	if input != got {
+		t.Errorf("decode: want\n\t%q\nbut got\n\t%q", input, got)
+	}
+}
+
+func TestCSSValueFilter(t *testing.T) {
+	tests := []struct {
+		css, want string
+	}{
+		{"", ""},
+		{"foo", "foo"},
+		{"0", "0"},
+		{"0px", "0px"},
+		{"-5px", "-5px"},
+		{"1.25in", "1.25in"},
+		{"+.33em", "+.33em"},
+		{"100%", "100%"},
+		{"12.5%", "12.5%"},
+		{".foo", ".foo"},
+		{"#bar", "#bar"},
+		{"corner-radius", "corner-radius"},
+		{"-moz-corner-radius", "-moz-corner-radius"},
+		{"#000", "#000"},
+		{"#48f", "#48f"},
+		{"#123456", "#123456"},
+		{"U+00-FF, U+980-9FF", "U+00-FF, U+980-9FF"},
+		{"color: red", "color: red"},
+		{"<!--", "ZgotmplZ"},
+		{"-->", "ZgotmplZ"},
+		{"<![CDATA[", "ZgotmplZ"},
+		{"]]>", "ZgotmplZ"},
+		{"</style", "ZgotmplZ"},
+		{`"`, "ZgotmplZ"},
+		{`'`, "ZgotmplZ"},
+		{"`", "ZgotmplZ"},
+		{"\x00", "ZgotmplZ"},
+		{"/* foo */", "ZgotmplZ"},
+		{"//", "ZgotmplZ"},
+		{"[href=~", "ZgotmplZ"},
+		{"expression(alert(1337))", "ZgotmplZ"},
+		{"-expression(alert(1337))", "ZgotmplZ"},
+		{"expression", "ZgotmplZ"},
+		{"Expression", "ZgotmplZ"},
+		{"EXPRESSION", "ZgotmplZ"},
+		{"-moz-binding", "ZgotmplZ"},
+		{"-expr\x00ession(alert(1337))", "ZgotmplZ"},
+		{`-expr\0ession(alert(1337))`, "ZgotmplZ"},
+		{`-express\69on(alert(1337))`, "ZgotmplZ"},
+		{`-express\69 on(alert(1337))`, "ZgotmplZ"},
+		{`-exp\72 ession(alert(1337))`, "ZgotmplZ"},
+		{`-exp\52 ession(alert(1337))`, "ZgotmplZ"},
+		{`-exp\000052 ession(alert(1337))`, "ZgotmplZ"},
+		{`-expre\0000073sion`, "-expre\x073sion"},
+		{`@import url evil.css`, "ZgotmplZ"},
+	}
+	for _, test := range tests {
+		got := cssValueFilter(test.css)
+		if got != test.want {
+			t.Errorf("%q: want %q but got %q", test.css, test.want, got)
+		}
+	}
+}
+
+func BenchmarkCSSEscaper(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		cssEscaper("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>")
+	}
+}
+
+func BenchmarkCSSEscaperNoSpecials(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		cssEscaper("The quick, brown fox jumps over the lazy dog.")
+	}
+}
+
+func BenchmarkDecodeCSS(b *testing.B) {
+	s := []byte(`The \3c i\3equick\3c/i\3e,\d\A\3cspan style=\27 color:brown\27\3e brown\3c/span\3e fox jumps\2028over the \3c canine class=\22lazy\22 \3edog\3c/canine\3e`)
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		decodeCSS(s)
+	}
+}
+
+func BenchmarkDecodeCSSNoSpecials(b *testing.B) {
+	s := []byte("The quick, brown fox jumps over the lazy dog.")
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		decodeCSS(s)
+	}
+}
+
+func BenchmarkCSSValueFilter(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		cssValueFilter(`  e\78preS\0Sio/**/n(alert(1337))`)
+	}
+}
+
+func BenchmarkCSSValueFilterOk(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		cssValueFilter(`Times New Roman`)
+	}
+}

src/pkg/exp/template/html/escape.go

@@ -33,12 +33,24 @@ func Escape(t *template.Template) (*template.Template, os.Error) {
 
 // funcMap maps command names to functions that render their inputs safe.
 var funcMap = template.FuncMap{
-	"exp_template_html_urlfilter":       urlFilter,
-	"exp_template_html_jsvalescaper":    jsValEscaper,
-	"exp_template_html_jsstrescaper":    jsStrEscaper,
+	"exp_template_html_cssescaper":      cssEscaper,
+	"exp_template_html_cssvaluefilter":  cssValueFilter,
 	"exp_template_html_jsregexpescaper": jsRegexpEscaper,
+	"exp_template_html_jsstrescaper":    jsStrEscaper,
+	"exp_template_html_jsvalescaper":    jsValEscaper,
+	"exp_template_html_nospaceescaper":  htmlNospaceEscaper,
+	"exp_template_html_urlescaper":      urlEscaper,
+	"exp_template_html_urlfilter":       urlFilter,
+	"exp_template_html_urlnormalizer":   urlNormalizer,
 }
 
+// filterFailsafe is an innocuous word that is emitted in place of unsafe values
+// by sanitizer functions.  It is not a keyword in any programming language,
+// contains no special characters, is not empty, and when it appears in output
+// it is distinct enough that a developer can find the source of the problem
+// via a search engine.
+const filterFailsafe = "ZgotmplZ"
+
 // escape escapes a template node.
 func escape(c context, n parse.Node) context {
 	switch n := n.(type) {
@@ -61,16 +73,22 @@ func escape(c context, n parse.Node) context {
 
 // escapeAction escapes an action template node.
 func escapeAction(c context, n *parse.ActionNode) context {
-	s := make([]string, 0, 2)
+	s := make([]string, 0, 3)
 	switch c.state {
-	case stateURL:
+	case stateURL, stateCSSDqStr, stateCSSSqStr, stateCSSDqURL, stateCSSSqURL, stateCSSURL:
 		switch c.urlPart {
 		case urlPartNone:
 			s = append(s, "exp_template_html_urlfilter")
-		case urlPartQueryOrFrag:
-			s = append(s, "urlquery")
+			fallthrough
 		case urlPartPreQuery:
-			s = append(s, "html")
+			switch c.state {
+			case stateCSSDqStr, stateCSSSqStr:
+				s = append(s, "exp_template_html_cssescaper")
+			case stateCSSDqURL, stateCSSSqURL, stateCSSURL:
+				s = append(s, "exp_template_html_urlnormalizer")
+			}
+		case urlPartQueryOrFrag:
+			s = append(s, "exp_template_html_urlescaper")
 		case urlPartUnknown:
 			return context{
 				state:   stateError,
@@ -82,19 +100,26 @@ func escapeAction(c context, n *parse.ActionNode) context {
 		}
 	case stateJS:
 		s = append(s, "exp_template_html_jsvalescaper")
-		if c.delim != delimNone {
-			s = append(s, "html")
-		}
 	case stateJSDqStr, stateJSSqStr:
 		s = append(s, "exp_template_html_jsstrescaper")
 	case stateJSRegexp:
 		s = append(s, "exp_template_html_jsregexpescaper")
-	case stateJSBlockCmt, stateJSLineCmt:
+	case stateJSBlockCmt, stateJSLineCmt, stateCSSBlockCmt, stateCSSLineCmt:
 		return context{
 			state:   stateError,
 			errLine: n.Line,
 			errStr:  fmt.Sprintf("%s appears inside a comment", n),
 		}
+	case stateCSS:
+		s = append(s, "exp_template_html_cssvaluefilter")
+	case stateText:
+		s = append(s, "html")
+	}
+	switch c.delim {
+	case delimNone:
+		// No extra-escaping needed for raw text content.
+	case delimSpaceOrTagEnd:
+		s = append(s, "exp_template_html_nospaceescaper")
 	default:
 		s = append(s, "html")
 	}
@@ -282,14 +307,22 @@ func escapeText(c context, s []byte) context {
 var transitionFunc = [...]func(context, []byte) (context, []byte){
 	stateText:        tText,
 	stateTag:         tTag,
+	stateAttr:        tAttr,
 	stateURL:         tURL,
 	stateJS:          tJS,
 	stateJSDqStr:     tJSStr,
 	stateJSSqStr:     tJSStr,
 	stateJSRegexp:    tJSRegexp,
-	stateJSBlockCmt: tJSBlockCmt,
-	stateJSLineCmt:  tJSLineCmt,
-	stateAttr:       tAttr,
+	stateJSBlockCmt:  tBlockCmt,
+	stateJSLineCmt:   tLineCmt,
+	stateCSS:         tCSS,
+	stateCSSDqStr:    tCSSStr,
+	stateCSSSqStr:    tCSSStr,
+	stateCSSDqURL:    tCSSStr,
+	stateCSSSqURL:    tCSSStr,
+	stateCSSURL:      tCSSStr,
+	stateCSSBlockCmt: tBlockCmt,
+	stateCSSLineCmt:  tLineCmt,
 	stateError:       tError,
 }
 
@@ -337,6 +370,8 @@ func tTag(c context, s []byte) (context, []byte) {
 		state = stateURL
 	} else if strings.HasPrefix(canonAttrName, "on") {
 		state = stateJS
+	} else if canonAttrName == "style" {
+		state = stateCSS
 	}
 
 	// Look for the start of the value.
@@ -376,7 +411,7 @@ func tAttr(c context, s []byte) (context, []byte) {
 func tURL(c context, s []byte) (context, []byte) {
 	if bytes.IndexAny(s, "#?") >= 0 {
 		c.urlPart = urlPartQueryOrFrag
-	} else if c.urlPart == urlPartNone {
+	} else if len(s) != 0 && c.urlPart == urlPartNone {
 		c.urlPart = urlPartPreQuery
 	}
 	return c, nil
@@ -499,29 +534,50 @@ func tJSRegexp(c context, s []byte) (context, []byte) {
 
 var blockCommentEnd = []byte("*/")
 
-// tJSBlockCmt is the context transition function for the JS /*comment*/ state.
-func tJSBlockCmt(c context, s []byte) (context, []byte) {
-	// TODO: delegate to tSpecialTagEnd to find any </script> once that CL
-	// has been merged.
-
+// tBlockCmt is the context transition function for /*comment*/ states.
+func tBlockCmt(c context, s []byte) (context, []byte) {
+	// TODO: look for </script or </style end tags.
 	i := bytes.Index(s, blockCommentEnd)
 	if i == -1 {
 		return c, nil
 	}
+	switch c.state {
+	case stateJSBlockCmt:
 		c.state = stateJS
+	case stateCSSBlockCmt:
+		c.state = stateCSS
+	default:
+		panic(c.state.String())
+	}
 	return c, s[i+2:]
 }
 
-// tJSLineCmt is the context transition function for the JS //comment state.
-func tJSLineCmt(c context, s []byte) (context, []byte) {
-	// TODO: delegate to tSpecialTagEnd to find any </script> once that CL
-	// has been merged.
+// tLineCmt is the context transition function for //comment states.
+func tLineCmt(c context, s []byte) (context, []byte) {
+	// TODO: look for </script or </style end tags.
+	var lineTerminators string
+	var endState state
+	switch c.state {
+	case stateJSLineCmt:
+		lineTerminators, endState = "\n\r\u2028\u2029", stateJS
+	case stateCSSLineCmt:
+		lineTerminators, endState = "\n\f\r", stateCSS
+		// Line comments are not part of any published CSS standard but
+		// are supported by the 4 major browsers.
+		// This defines line comments as
+		//     LINECOMMENT ::= "//" [^\n\f\d]*
+		// since http://www.w3.org/TR/css3-syntax/#SUBTOK-nl defines
+		// newlines:
+		//     nl ::= #xA | #xD #xA | #xD | #xC
+	default:
+		panic(c.state.String())
+	}
 
-	i := bytes.IndexAny(s, "\r\n\u2028\u2029")
+	i := bytes.IndexAny(s, lineTerminators)
 	if i == -1 {
 		return c, nil
 	}
-	c.state = stateJS
+	c.state = endState
 	// Per section 7.4 of EcmaScript 5 : http://es5.github.com/#x7.4
 	// "However, the LineTerminator at the end of the line is not
 	// considered to be part of the single-line comment; it is recognised
@@ -530,6 +586,124 @@ func tJSLineCmt(c context, s []byte) (context, []byte) {
 	return c, s[i:]
 }
 
+// tCSS is the context transition function for the CSS state.
+func tCSS(c context, s []byte) (context, []byte) {
+	// TODO: look for </style
+
+	// 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"
+	// (3) List separators in content values as in inline-lists:
+	//    <style>
+	//    ul.inlineList { list-style: none; padding:0 }
+	//    ul.inlineList > li { display: inline }
+	//    ul.inlineList > li:before { content: ", " }
+	//    ul.inlineList > li:first-child:before { content: "" }
+	//    </style>
+	//    <ul class=inlineList><li>One<li>Two<li>Three</ul>
+	// (4) Attribute value selectors as in a[href="http://example.com/"]
+	//
+	// We conservatively treat all strings as URLs, but make some
+	// allowances to avoid confusion.
+	//
+	// In (1), our conservative assumption is justified.
+	// In (2), valid font names do not contain ':', '?', or '#', so our
+	// conservative assumption is fine since we will never transition past
+	// urlPartPreQuery.
+	// In (3), our protocol heuristic should not be tripped, and there
+	// should not be non-space content after a '?' or '#', so as long as
+	// we only %-encode RFC 3986 reserved characters we are ok.
+	// In (4), we should URL escape for URL attributes, and for others we
+	// have the attribute name available if our conservative assumption
+	// proves problematic for real code.
+
+	for {
+		i := bytes.IndexAny(s, `("'/`)
+		if i == -1 {
+			return c, nil
+		}
+		switch s[i] {
+		case '(':
+			// Look for url to the left.
+			p := bytes.TrimRight(s[:i], "\t\n\f\r ")
+			if endsWithCSSKeyword(p, "url") {
+				q := bytes.TrimLeft(s[i+1:], "\t\n\f\r ")
+				switch {
+				case len(q) != 0 && q[0] == '"':
+					c.state, s = stateCSSDqURL, q[1:]
+				case len(q) != 0 && q[0] == '\'':
+					c.state, s = stateCSSSqURL, q[1:]
+
+				default:
+					c.state, s = stateCSSURL, q
+				}
+				return c, s
+			}
+		case '/':
+			if i+1 < len(s) {
+				switch s[i+1] {
+				case '/':
+					c.state = stateCSSLineCmt
+					return c, s[i+2:]
+				case '*':
+					c.state = stateCSSBlockCmt
+					return c, s[i+2:]
+				}
+			}
+		case '"':
+			c.state = stateCSSDqStr
+			return c, s[i+1:]
+		case '\'':
+			c.state = stateCSSSqStr
+			return c, s[i+1:]
+		}
+		s = s[i+1:]
+	}
+	panic("unreachable")
+}
+
+// tCSSStr is the context transition function for the CSS string and URL states.
+func tCSSStr(c context, s []byte) (context, []byte) {
+	// TODO: look for </style
+
+	var endAndEsc string
+	switch c.state {
+	case stateCSSDqStr, stateCSSDqURL:
+		endAndEsc = `\"`
+	case stateCSSSqStr, stateCSSSqURL:
+		endAndEsc = `\'`
+	case stateCSSURL:
+		// Unquoted URLs end with a newline or close parenthesis.
+		// The below includes the wc (whitespace character) and nl.
+		endAndEsc = "\\\t\n\f\r )"
+	default:
+		panic(c.state.String())
+	}
+
+	b := s
+	for {
+		i := bytes.IndexAny(b, endAndEsc)
+		if i == -1 {
+			return tURL(c, decodeCSS(b))
+		}
+		if b[i] == '\\' {
+			i++
+			if i == len(b) {
+				return context{
+					state:  stateError,
+					errStr: fmt.Sprintf("unfinished escape sequence in CSS string: %q", s),
+				}, nil
+			}
+		} else {
+			c.state = stateCSS
+			return c, b[i+1:]
+		}
+		c, _ = tURL(c, decodeCSS(b[:i+1]))
+		b = b[i+1:]
+	}
+	panic("unreachable")
+}
+
 // tError is the context transition function for the error state.
 func tError(c context, s []byte) (context, []byte) {
 	return c, nil
@@ -612,28 +786,3 @@ var urlAttr = map[string]bool{
 	"src":        true,
 	"usemap":     true,
 }
-
-// urlFilter returns the HTML equivalent of its input unless it contains an
-// unsafe protocol in which case it defangs the entire URL.
-func urlFilter(args ...interface{}) string {
-	ok := false
-	var s string
-	if len(args) == 1 {
-		s, ok = args[0].(string)
-	}
-	if !ok {
-		s = fmt.Sprint(args...)
-	}
-	i := strings.IndexRune(s, ':')
-	if i >= 0 && strings.IndexRune(s[:i], '/') < 0 {
-		protocol := strings.ToLower(s[:i])
-		if protocol != "http" && protocol != "https" && protocol != "mailto" {
-			// Return a value that someone investigating a bug
-			// report can put into a search engine.
-			return "#ZgotmplZ"
-		}
-	}
-	// TODO: Once we handle <style>#id { background: url({{.Img}}) }</style>
-	// we will need to stop this from HTML escaping and pipeline sanitizers.
-	return template.HTMLEscapeString(s)
-}

src/pkg/exp/template/html/escape_test.go

@@ -82,14 +82,9 @@ func TestEscape(t *testing.T) {
 			"true",
 		},
 		{
-			// TODO: Make sure the URL escaper escapes single quotes so it can
-			// be embedded in single quoted URI attributes and CSS url(...)
-			// constructs. Single quotes are reserved in URLs, but are only used
-			// in the obsolete "mark" rule in an appendix in RFC 3986 so can be
-			// safely encoded.
 			"constant",
 			`<a href="/search?q={{"'a<b'"}}">`,
-			`<a href="/search?q='a%3Cb'">`,
+			`<a href="/search?q=%27a%3cb%27">`,
 		},
 		{
 			"multipleAttrs",
@@ -121,6 +116,11 @@ func TestEscape(t *testing.T) {
 			`<a href='{{"javascript:alert(%22pwned%22)"}}'>`,
 			`<a href='#ZgotmplZ'>`,
 		},
+		{
+			"nonHierURL",
+			`<a href={{"mailto:Muhammed \"The Greatest\" Ali <m.ali@example.com>"}}>`,
+			`<a href=mailto:Muhammed&#32;&#34;The&#32;Greatest&#34;&#32;Ali&#32;&lt;m.ali@example.com&gt;>`,
+		},
 		{
 			"urlPath",
 			`<a href='http://{{"javascript:80"}}/foo'>`,
@@ -129,12 +129,12 @@ func TestEscape(t *testing.T) {
 		{
 			"urlQuery",
 			`<a href='/search?q={{.H}}'>`,
-			`<a href='/search?q=%3CHello%3E'>`,
+			`<a href='/search?q=%3cHello%3e'>`,
 		},
 		{
 			"urlFragment",
 			`<a href='/faq#{{.H}}'>`,
-			`<a href='/faq#%3CHello%3E'>`,
+			`<a href='/faq#%3cHello%3e'>`,
 		},
 		{
 			"urlBranch",
@@ -144,7 +144,7 @@ func TestEscape(t *testing.T) {
 		{
 			"urlBranchConflictMoot",
 			`<a href="{{if .T}}/foo?a={{else}}/bar#{{end}}{{.C}}">`,
-			`<a href="/foo?a=%3CCincinatti%3E">`,
+			`<a href="/foo?a=%3cCincinatti%3e">`,
 		},
 		{
 			"jsStrValue",
@@ -192,6 +192,138 @@ func TestEscape(t *testing.T) {
 			"<button onclick='alert(&quot;{{.H}}&quot;)'>",
 			`<button onclick='alert(&quot;\x3cHello\x3e&quot;)'>`,
 		},
+		{
+			"styleBidiKeywordPassed",
+			`<p style="dir: {{"ltr"}}">`,
+			`<p style="dir: ltr">`,
+		},
+		{
+			"styleBidiPropNamePassed",
+			`<p style="border-{{"left"}}: 0; border-{{"right"}}: 1in">`,
+			`<p style="border-left: 0; border-right: 1in">`,
+		},
+		{
+			"styleExpressionBlocked",
+			`<p style="width: {{"expression(alert(1337))"}}">`,
+			`<p style="width: ZgotmplZ">`,
+		},
+		{
+			"styleTagSelectorPassed",
+			`<style>{{"p"}} { color: pink }</style>`,
+			`<style>p { color: pink }</style>`,
+		},
+		{
+			"styleIDPassed",
+			`<style>p{{"#my-ID"}} { font: Arial }`,
+			`<style>p#my-ID { font: Arial }`,
+		},
+		{
+			"styleClassPassed",
+			`<style>p{{".my_class"}} { font: Arial }`,
+			`<style>p.my_class { font: Arial }`,
+		},
+		{
+			"styleQuantityPassed",
+			`<a style="left: {{"2em"}}; top: {{0}}">`,
+			`<a style="left: 2em; top: 0">`,
+		},
+		{
+			"stylePctPassed",
+			`<table style=width:{{"100%"}}>`,
+			`<table style=width:100%>`,
+		},
+		{
+			"styleColorPassed",
+			`<p style="color: {{"#8ff"}}; background: {{"#000"}}">`,
+			`<p style="color: #8ff; background: #000">`,
+		},
+		{
+			"styleObfuscatedExpressionBlocked",
+			`<p style="width: {{"  e\78preS\0Sio/**/n(alert(1337))"}}">`,
+			`<p style="width: ZgotmplZ">`,
+		},
+		{
+			"styleMozBindingBlocked",
+			`<p style="{{"-moz-binding(alert(1337))"}}: ...">`,
+			`<p style="ZgotmplZ: ...">`,
+		},
+		{
+			"styleObfuscatedMozBindingBlocked",
+			`<p style="{{"  -mo\7a-B\0I/**/nding(alert(1337))"}}: ...">`,
+			`<p style="ZgotmplZ: ...">`,
+		},
+		{
+			"styleFontNameString",
+			`<p style='font-family: "{{"Times New Roman"}}"'>`,
+			`<p style='font-family: "Times New Roman"'>`,
+		},
+		{
+			"styleFontNameString",
+			`<p style='font-family: "{{"Times New Roman"}}", "{{"sans-serif"}}"'>`,
+			`<p style='font-family: "Times New Roman", "sans-serif"'>`,
+		},
+		{
+			"styleFontNameUnquoted",
+			`<p style='font-family: {{"Times New Roman"}}'>`,
+			`<p style='font-family: Times New Roman'>`,
+		},
+		{
+			"styleURLQueryEncoded",
+			`<p style="background: url(/img?name={{"O'Reilly Animal(1)<2>.png"}})">`,
+			`<p style="background: url(/img?name=O%27Reilly%20Animal%281%29%3c2%3e.png)">`,
+		},
+		{
+			"styleQuotedURLQueryEncoded",
+			`<p style="background: url('/img?name={{"O'Reilly Animal(1)<2>.png"}}')">`,
+			`<p style="background: url('/img?name=O%27Reilly%20Animal%281%29%3c2%3e.png')">`,
+		},
+		{
+			"styleStrQueryEncoded",
+			`<p style="background: '/img?name={{"O'Reilly Animal(1)<2>.png"}}'">`,
+			`<p style="background: '/img?name=O%27Reilly%20Animal%281%29%3c2%3e.png'">`,
+		},
+		{
+			"styleURLBadProtocolBlocked",
+			`<a style="background: url('{{"javascript:alert(1337)"}}')">`,
+			`<a style="background: url('#ZgotmplZ')">`,
+		},
+		{
+			"styleStrBadProtocolBlocked",
+			`<a style="background: '{{"javascript:alert(1337)"}}'">`,
+			`<a style="background: '#ZgotmplZ'">`,
+		},
+		{
+			"styleURLGoodProtocolPassed",
+			`<a style="background: url('{{"http://oreilly.com/O'Reilly Animals(1)<2>;{}.html"}}')">`,
+			`<a style="background: url('http://oreilly.com/O%27Reilly%20Animals%281%29%3c2%3e;%7b%7d.html')">`,
+		},
+		{
+			"styleStrGoodProtocolPassed",
+			`<a style="background: '{{"http://oreilly.com/O'Reilly Animals(1)<2>;{}.html"}}'">`,
+			`<a style="background: 'http\3a\2f\2foreilly.com\2fO\27Reilly Animals\28 1\29\3c 2\3e\3b\7b\7d.html'">`,
+		},
+		{
+			"styleURLMixedCase",
+			`<p style="background: URL(#{{.H}})">`,
+			`<p style="background: URL(#%3cHello%3e)">`,
+		},
+		{
+			"stylePropertyPairPassed",
+			`<a style='{{"color: red"}}'>`,
+			`<a style='color: red'>`,
+		},
+		{
+			"styleStrSpecialsEncoded",
+			`<a style="font-family: '{{"/**/'\";:// \\"}}', &quot;{{"/**/'\";:// \\"}}&quot;">`,
+			`<a style="font-family: '\2f**\2f\27\22\3b\3a\2f\2f \\', &quot;\2f**\2f\27\22\3b\3a\2f\2f \\&quot;">`,
+		},
+		{
+			"styleURLSpecialsEncoded",
+			// TODO: Find out what IE does with url(/*foo*/bar)
+			// FF, Chrome, and Safari seem to treat it as a URL.
+			`<a style="border-image: url({{"/**/'\";:// \\"}}), url(&quot;{{"/**/'\";:// \\"}}&quot;), url('{{"/**/'\";:// \\"}}'), 'http://www.example.com/?q={{"/**/'\";:// \\"}}''">`,
+			`<a style="border-image: url(/**/%27%22;://%20%5c), url(&quot;/**/%27%22;://%20%5c&quot;), url('/**/%27%22;://%20%5c'), 'http://www.example.com/?q=%2f%2a%2a%2f%27%22%3b%3a%2f%2f%20%5c''">`,
+		},
 	}
 
 	for _, test := range tests {
@@ -299,11 +431,19 @@ func TestErrors(t *testing.T) {
 			`unfinished JS regexp charset: "foo[\\]/"`,
 		},
 		{
-			`<a onclick="/* alert({{.X}} */">`,
+			`<a onclick="/* alert({{.X}}) */">`,
 			`z:1: (action: [(command: [F=[X]])]) appears inside a comment`,
 		},
 		{
-			`<a onclick="// alert({{.X}}">`,
+			`<a onclick="// alert({{.X}})">`,
+			`z:1: (action: [(command: [F=[X]])]) appears inside a comment`,
+		},
+		{
+			`<a style="/* color: {{.X}} */">`,
+			`z:1: (action: [(command: [F=[X]])]) appears inside a comment`,
+		},
+		{
+			`<a style="// color: {{.X}}">`,
 			`z:1: (action: [(command: [F=[X]])]) appears inside a comment`,
 		},
 	}
@@ -533,6 +673,98 @@ func TestEscapeText(t *testing.T) {
 			`<a onclick="/foo\/`,
 			context{state: stateJSRegexp, delim: delimDoubleQuote},
 		},
+		{
+			`<a onclick="/foo/`,
+			context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp},
+		},
+		{
+			`<input checked style="`,
+			context{state: stateCSS, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="//`,
+			context{state: stateCSSLineCmt, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="//</script>`,
+			context{state: stateCSSLineCmt, delim: delimDoubleQuote},
+		},
+		{
+			"<a style='//\n",
+			context{state: stateCSS, delim: delimSingleQuote},
+		},
+		{
+			"<a style='//\r",
+			context{state: stateCSS, delim: delimSingleQuote},
+		},
+		{
+			`<a style="/*`,
+			context{state: stateCSSBlockCmt, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="/*/`,
+			context{state: stateCSSBlockCmt, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="/**/`,
+			context{state: stateCSS, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="background: '`,
+			context{state: stateCSSSqStr, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="background: &quot;`,
+			context{state: stateCSSDqStr, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="background: '/foo?img=`,
+			context{state: stateCSSSqStr, delim: delimDoubleQuote, urlPart: urlPartQueryOrFrag},
+		},
+		{
+			`<a style="background: '/`,
+			context{state: stateCSSSqStr, delim: delimDoubleQuote, urlPart: urlPartPreQuery},
+		},
+		{
+			`<a style="background: url(&#x22;/`,
+			context{state: stateCSSDqURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery},
+		},
+		{
+			`<a style="background: url('/`,
+			context{state: stateCSSSqURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery},
+		},
+		{
+			`<a style="background: url('/)`,
+			context{state: stateCSSSqURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery},
+		},
+		{
+			`<a style="background: url('/ `,
+			context{state: stateCSSSqURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery},
+		},
+		{
+			`<a style="background: url(/`,
+			context{state: stateCSSURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery},
+		},
+		{
+			`<a style="background: url( `,
+			context{state: stateCSSURL, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="background: url( /image?name=`,
+			context{state: stateCSSURL, delim: delimDoubleQuote, urlPart: urlPartQueryOrFrag},
+		},
+		{
+			`<a style="background: url(x)`,
+			context{state: stateCSS, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="background: url('x'`,
+			context{state: stateCSS, delim: delimDoubleQuote},
+		},
+		{
+			`<a style="background: url( x `,
+			context{state: stateCSS, delim: delimDoubleQuote},
+		},
 	}
 
 	for _, test := range tests {

src/pkg/exp/template/html/html.go

+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package html
+
+import (
+	"bytes"
+	"fmt"
+	"utf8"
+)
+
+// htmlNospaceEscaper escapes for inclusion in unquoted attribute values.
+func htmlNospaceEscaper(args ...interface{}) string {
+	s := stringify(args...)
+	// The set of runes escaped is the union of the HTML specials and
+	// those determined by running the JS below in browsers:
+
+	// <div id=d></div>
+	// <script>(function () {
+	// var a = [], d = document.getElementById("d"), i, c, s;
+	// for (i = 0; i < 0x10000; ++i) {
+	//   c = String.fromCharCode(i);
+	//   d.innerHTML = "<span title=" + c + "lt" + c + "></span>"
+	//   s = d.getElementsByTagName("SPAN")[0];
+	//   if (!s || s.title !== c + "lt" + c) { a.push(i.toString(16)); }
+	// }
+	// document.write(a.join(", "));
+	// })()</script>
+
+	var b bytes.Buffer
+	written := 0
+	for i, r := range s {
+		var repl string
+		switch r {
+		case 0:
+			// http://www.w3.org/TR/html5/tokenization.html#attribute-value-unquoted-state: "
+			// U+0000 NULL Parse error. Append a U+FFFD REPLACEMENT
+			// CHARACTER character to the current attribute's value.
+			// "
+			// and similarly
+			// http://www.w3.org/TR/html5/tokenization.html#before-attribute-value-state
+			repl = "\uFFFD"
+		case '\t':
+			repl = "&#9;"
+		case '\n':
+			repl = "&#10;"
+		case '\v':
+			repl = "&#11;"
+		case '\f':
+			repl = "&#12;"
+		case '\r':
+			repl = "&#13;"
+		case ' ':
+			repl = "&#32;"
+		case '"':
+			repl = "&#34;"
+		case '&':
+			repl = "&amp;"
+		case '\'':
+			repl = "&#39;"
+		case '+':
+			repl = "&#43;"
+		case '<':
+			repl = "&lt;"
+		case '=':
+			repl = "&#61;"
+		case '>':
+			repl = "&gt;"
+		case '`':
+			// A parse error in the attribute value (unquoted) and 
+			// before attribute value states.
+			// Treated as a quoting character by IE.
+			repl = "&#96;"
+		default:
+			// IE does not allow the ranges below raw in attributes.
+			if 0xfdd0 <= r && r <= 0xfdef || 0xfff0 <= r && r <= 0xffff {
+				b.WriteString(s[written:i])
+				b.WriteString("&#x")
+				b.WriteByte("0123456789abcdef"[r>>24])
+				b.WriteByte("0123456789abcdef"[r>>16&0xf])
+				b.WriteByte("0123456789abcdef"[r>>8&0xf])
+				b.WriteByte("0123456789abcdef"[r&0xf])
+				b.WriteByte(';')
+				fmt.Fprintf(&b, "&#x%x;", r)
+				written = i + utf8.RuneLen(r)
+			}
+			continue
+		}
+		b.WriteString(s[written:i])
+		b.WriteString(repl)
+		// Valid as long as we don't include any cases above in the
+		// 0x80-0xff range.
+		written = i + utf8.RuneLen(r)
+	}
+	if written == 0 {
+		return s
+	}
+	b.WriteString(s[written:])
+	return b.String()
+}

src/pkg/exp/template/html/html_test.go

+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package html
+
+import (
+	"html"
+	"strings"
+	"testing"
+)
+
+func TestHTMLNospaceEscaper(t *testing.T) {
+	input := ("\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f" +
+		"\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" +
+		` !"#$%&'()*+,-./` +
+		`0123456789:;<=>?` +
+		`@ABCDEFGHIJKLMNO` +
+		`PQRSTUVWXYZ[\]^_` +
+		"`abcdefghijklmno" +
+		"pqrstuvwxyz{|}~\x7f" +
+		"\u00A0\u0100\u2028\u2029\ufeff\U0001D11E")
+
+	want := ("\ufffd\x01\x02\x03\x04\x05\x06\x07" +
+		"\x08&#9;&#10;&#11;&#12;&#13;\x0E\x0F" +
+		"\x10\x11\x12\x13\x14\x15\x16\x17" +
+		"\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" +
+		`&#32;!&#34;#$%&amp;&#39;()*&#43;,-./` +
+		`0123456789:;&lt;&#61;&gt;?` +
+		`@ABCDEFGHIJKLMNO` +
+		`PQRSTUVWXYZ[\]^_` +
+		`&#96;abcdefghijklmno` +
+		`pqrstuvwxyz{|}~` + "\u007f" +
+		"\u00A0\u0100\u2028\u2029\ufeff\U0001D11E")
+
+	got := htmlNospaceEscaper(input)
+	if got != want {
+		t.Errorf("encode: want\n\t%q\nbut got\n\t%q", want, got)
+	}
+
+	got, want = html.UnescapeString(got), strings.Replace(input, "\x00", "\ufffd", 1)
+	if want != got {
+		t.Errorf("decode: want\n\t%q\nbut got\n\t%q", want, got)
+	}
+}
+
+func BenchmarkHTMLNospaceEscaper(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		htmlNospaceEscaper("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>")
+	}
+}
+
+func BenchmarkHTMLNospaceEscaperNoSpecials(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		htmlNospaceEscaper("The_quick,_brown_fox_jumps_over_the_lazy_dog.")
+	}
+}

src/pkg/exp/template/html/js.go

@@ -269,7 +269,7 @@ var jsRegexpReplacementTable = []string{
 	'}':  `\}`,
 }
 
-// isJSIdentPart is true if the given rune is a JS identifier part.
+// isJSIdentPart returns whether the given rune is a JS identifier part.
 // It does not handle all the non-Latin letters, joiners, and combining marks,
 // but it does handle every codepoint that can occur in a numeric literal or
 // a keyword.

src/pkg/exp/template/html/url.go

+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package html
+
+import (
+	"bytes"
+	"fmt"
+	"strings"
+)
+
+// urlFilter returns the HTML equivalent of its input unless it contains an
+// unsafe protocol in which case it defangs the entire URL.
+func urlFilter(args ...interface{}) string {
+	s := stringify(args...)
+	i := strings.IndexRune(s, ':')
+	if i >= 0 && strings.IndexRune(s[:i], '/') < 0 {
+		protocol := strings.ToLower(s[:i])
+		if protocol != "http" && protocol != "https" && protocol != "mailto" {
+			// Return a value that someone investigating a bug
+			// report can put into a search engine.
+			return "#" + filterFailsafe
+		}
+	}
+	// TODO: Once we handle <style>#id { background: url({{.Img}}) }</style>
+	// we will need to stop this from HTML escaping and pipeline sanitizers.
+	return s
+}
+
+// urlEscaper produces an output that can be embedded in a URL query.
+// The output can be embedded in an HTML attribute without further escaping.
+func urlEscaper(args ...interface{}) string {
+	return urlProcessor(false, args...)
+}
+
+// urlEscaper normalizes URL content so it can be embedded in a quote-delimited
+// string or parenthesis delimited url(...).
+// The normalizer does not encode all HTML specials.  Specifically, it does not
+// encode '&' so correct embedding in an HTML attribute requires escaping of
+// '&' to '&amp;'.
+func urlNormalizer(args ...interface{}) string {
+	return urlProcessor(true, args...)
+}
+
+// urlProcessor normalizes (when norm is true) or escapes its input to produce
+// a valid hierarchical or opaque URL part.
+func urlProcessor(norm bool, args ...interface{}) string {
+	s := stringify(args...)
+	var b bytes.Buffer
+	written := 0
+	// The byte loop below assumes that all URLs use UTF-8 as the
+	// content-encoding. This is similar to the URI to IRI encoding scheme
+	// defined in section 3.1 of  RFC 3987, and behaves the same as the
+	// EcmaScript builtin encodeURIComponent.
+	// It should not cause any misencoding of URLs in pages with
+	// Content-type: text/html;charset=UTF-8.
+	for i, n := 0, len(s); i < n; i++ {
+		c := s[i]
+		switch c {
+		// Single quote and parens are sub-delims in RFC 3986, but we
+		// escape them so the output can be embedded in in single
+		// quoted attributes and unquoted CSS url(...) constructs.
+		// Single quotes are reserved in URLs, but are only used in
+		// the obsolete "mark" rule in an appendix in RFC 3986
+		// so can be safely encoded.
+		case '!', '#', '$', '&', '*', '+', ',', '/', ':', ';', '=', '?', '@', '[', ']':
+			if norm {
+				continue
+			}
+		// Unreserved according to RFC 3986 sec 2.3
+		// "For consistency, percent-encoded octets in the ranges of
+		// ALPHA (%41-%5A and %61-%7A), DIGIT (%30-%39), hyphen (%2D),
+		// period (%2E), underscore (%5F), or tilde (%7E) should not be
+		// created by URI producers
+		case '-', '.', '_', '~':
+			continue
+		case '%':
+			// When normalizing do not re-encode valid escapes.
+			if norm && i+2 < len(s) && isHex(s[i+1]) && isHex(s[i+2]) {
+				continue
+			}
+		default:
+			// Unreserved according to RFC 3986 sec 2.3
+			if 'a' <= c && c <= 'z' {
+				continue
+			}
+			if 'A' <= c && c <= 'Z' {
+				continue
+			}
+			if '0' <= c && c <= '9' {
+				continue
+			}
+		}
+		b.WriteString(s[written:i])
+		fmt.Fprintf(&b, "%%%02x", c)
+		written = i + 1
+	}
+	if written == 0 {
+		return s
+	}
+	b.WriteString(s[written:])
+	return b.String()
+}

src/pkg/exp/template/html/url_test.go

+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package html
+
+import (
+	"testing"
+)
+
+func TestURLNormalizer(t *testing.T) {
+	tests := []struct {
+		url, want string
+	}{
+		{"", ""},
+		{
+			"http://example.com:80/foo/bar?q=foo%20&bar=x+y#frag",
+			"http://example.com:80/foo/bar?q=foo%20&bar=x+y#frag",
+		},
+		{" ", "%20"},
+		{"%7c", "%7c"},
+		{"%7C", "%7C"},
+		{"%2", "%252"},
+		{"%", "%25"},
+		{"%z", "%25z"},
+		{"/foo|bar/%5c\u1234", "/foo%7cbar/%5c%e1%88%b4"},
+	}
+	for _, test := range tests {
+		if got := urlNormalizer(test.url); test.want != got {
+			t.Errorf("%q: want\n\t%q\nbut got\n\t%q", test.url, test.want, got)
+		}
+		if test.want != urlNormalizer(test.want) {
+			t.Errorf("not idempotent: %q", test.want)
+		}
+	}
+}
+
+func TestURLFilters(t *testing.T) {
+	input := ("\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f" +
+		"\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" +
+		` !"#$%&'()*+,-./` +
+		`0123456789:;<=>?` +
+		`@ABCDEFGHIJKLMNO` +
+		`PQRSTUVWXYZ[\]^_` +
+		"`abcdefghijklmno" +
+		"pqrstuvwxyz{|}~\x7f" +
+		"\u00A0\u0100\u2028\u2029\ufeff\U0001D11E")
+
+	tests := []struct {
+		name    string
+		escaper func(...interface{}) string
+		escaped string
+	}{
+		{
+			"urlEscaper",
+			urlEscaper,
+			"%00%01%02%03%04%05%06%07%08%09%0a%0b%0c%0d%0e%0f" +
+				"%10%11%12%13%14%15%16%17%18%19%1a%1b%1c%1d%1e%1f" +
+				"%20%21%22%23%24%25%26%27%28%29%2a%2b%2c-.%2f" +
+				"0123456789%3a%3b%3c%3d%3e%3f" +
+				"%40ABCDEFGHIJKLMNO" +
+				"PQRSTUVWXYZ%5b%5c%5d%5e_" +
+				"%60abcdefghijklmno" +
+				"pqrstuvwxyz%7b%7c%7d~%7f" +
+				"%c2%a0%c4%80%e2%80%a8%e2%80%a9%ef%bb%bf%f0%9d%84%9e",
+		},
+		{
+			"urlNormalizer",
+			urlNormalizer,
+			"%00%01%02%03%04%05%06%07%08%09%0a%0b%0c%0d%0e%0f" +
+				"%10%11%12%13%14%15%16%17%18%19%1a%1b%1c%1d%1e%1f" +
+				"%20!%22#$%25&%27%28%29*+,-./" +
+				"0123456789:;%3c=%3e?" +
+				"@ABCDEFGHIJKLMNO" +
+				"PQRSTUVWXYZ[%5c]%5e_" +
+				"%60abcdefghijklmno" +
+				"pqrstuvwxyz%7b%7c%7d~%7f" +
+				"%c2%a0%c4%80%e2%80%a8%e2%80%a9%ef%bb%bf%f0%9d%84%9e",
+		},
+	}
+
+	for _, test := range tests {
+		if s := test.escaper(input); s != test.escaped {
+			t.Errorf("%s: want\n\t%q\ngot\n\t%q", test.name, test.escaped, s)
+			continue
+		}
+	}
+}
+
+func BenchmarkURLEscaper(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		urlEscaper("http://example.com:80/foo?q=bar%20&baz=x+y#frag")
+	}
+}
+
+func BenchmarkURLEscaperNoSpecials(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		urlEscaper("TheQuickBrownFoxJumpsOverTheLazyDog.")
+	}
+}
+
+func BenchmarkURLNormalizer(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		urlNormalizer("The quick brown fox jumps over the lazy dog.\n")
+	}
+}
+
+func BenchmarkURLNormalizerNoSpecials(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		urlNormalizer("http://example.com:80/foo?q=bar%20&baz=x+y#frag")
+	}
+}