`, + `
`, + }, + { + "styleBidiPropNamePassed", + `
`, + }, + { + "styleExpressionBlocked", + `
`, + }, + { + "styleTagSelectorPassed", + ``, + ``, + }, + { + "styleIDPassed", + ``, + ``, + }, + { + "styleClassPassed", + ``, + ``, + }, + { + "styleQuantityPassed", + ``, + ``, + }, + { + "stylePctPassed", + `
`, + }, + { + "styleObfuscatedExpressionBlocked", + `
`, + }, + { + "styleMozBindingBlocked", + `
`, + }, + { + "styleObfuscatedMozBindingBlocked", + `
`, + }, + { + "styleFontNameString", + `
`, + }, + { + "styleFontNameUnquoted", + `
`, + }, + { + "styleURLQueryEncoded", + `
.png"}})">`, + `
`, + }, + { + "styleQuotedURLQueryEncoded", + `
.png"}}')">`, + `
`, + }, + { + "styleStrQueryEncoded", + `
.png"}}'">`, + `
`, + }, + { + "styleURLBadProtocolBlocked", + ``, + ``, + }, + { + "styleStrBadProtocolBlocked", + ``, + ``, + }, + { + "styleStrEncodedProtocolEncoded", + ``, + // The CSS string 'javascript\\3a alert(1337)' does not contains a colon. + ``, + }, + { + "styleURLGoodProtocolPassed", + `;{}.html"}}')">`, + ``, + }, + { + "styleStrGoodProtocolPassed", + `;{}.html"}}'">`, + ``, + }, + { + "styleURLEncodedForHTMLInAttr", + ``, + ``, + }, + { + "styleURLNotEncodedForHTMLInCdata", + ``, + ``, + }, + { + "styleURLMixedCase", + ``, + ``, + }, + { + "stylePropertyPairPassed", + ``, + ``, + }, + { + "styleStrSpecialsEncoded", + ``, + ``, + }, + { + "styleURLSpecialsEncoded", + ``, + ``, + }, + { + "HTML comment", + "Hello, {{.C}}", + "Hello, <Cincinatti>", + }, + { + "HTML comment not first < in text node.", + "<!--", + "<!--", + }, + { + "HTML normalization 1", + "a < b", + "a < b", + }, + { + "HTML normalization 2", + "a << b", + "a << b", + }, + { + "HTML normalization 3", + "a<b", + "a<b", + }, + { + "HTML doctype not normalized", + "Hello, World!", + "Hello, World!", + }, + { + "HTML doctype not case-insensitive", + "Hello, World!", + "Hello, World!", + }, + { + "No doctype injection", + `Hello, {{.C}}{{else}}world -->{{.W}}{{end}}", + "Hello, <Cincinatti>", + }, + { + "JS line comment", + "", + "", + }, + { + "JS multiline block comment", + "", + // Newline separates break from call. If newline + // removed, then break will consume label leaving + // code invalid. + "", + }, + { + "JS single-line block comment", + "", + // Newline separates break from call. If newline + // removed, then break will consume label leaving + // code invalid. + "", + }, + { + "JS block comment flush with mathematical division", + "", + "", + }, + { + "JS mixed comments", + "", + "", + }, + { + "CSS comments", + "`, + "", + }, + { + "JS attr block comment", + ``, + // Attribute comment tests should pass if the comments + // are successfully elided. + ``, + }, + { + "JS attr line comment", + ``, + ``, + }, + { + "CSS attr block comment", + ``, + ``, + }, + { + "CSS attr line comment", + ``, + ``, + }, + { + "HTML substitution commented out", + "", + "", + }, + { + "Comment ends flush with start", + "", + "", + }, + { + "typed HTML in text", + `{{.W}}`, + `¡Hello, O'World!`, + }, + { + "typed HTML in attribute", + ``, + ``, + }, + { + "typed HTML in script", + ``, + ``, + }, + { + "typed HTML in RCDATA", + `{{.W}}`, + `¡<b class="foo">Hello</b>, <textarea>O'World</textarea>!`, + }, + { + "range in textarea", + "{{range .A}}{{.}}{{end}}", + "<a><b>", + }, + { + "No tag injection", + `{{"10$"}}<{{"script src,evil.org/pwnd.js"}}...`, + `10$<script src,evil.org/pwnd.js...`, + }, + { + "No comment injection", + `<{{"!--"}}`, + `<!--`, + }, + { + "No RCDATA end tag injection", + `<{{"/textarea "}}...`, + `</textarea ...`, + }, + { + "optional attrs", + `"}}"{{end}}` + + // Double quotes inside if/else. + ` src=` + + `{{if .T}}"?{{""}}"` + + `{{else}}"images/cleardot.gif"{{end}}` + + // Missing space before title, but it is not a + // part of the src attribute. + `{{if .T}}title="{{""}}"{{end}}` + + // Quotes outside if/else. + ` alt="` + + `{{if .T}}{{"<alt>"}}` + + `{{else}}{{if .F}}{{"<title>"}}{{end}}` + + `{{end}}"` + + `>`, + `<img class="iconClass" id="<iconId>" src="?%3ciconPath%3e"title="<title>" alt="<alt>">`, + }, + { + "conditional valueless attr name", + `<input{{if .T}} checked{{end}} name=n>`, + `<input checked name=n>`, + }, + { + "conditional dynamic valueless attr name 1", + `<input{{if .T}} {{"checked"}}{{end}} name=n>`, + `<input checked name=n>`, + }, + { + "conditional dynamic valueless attr name 2", + `<input {{if .T}}{{"checked"}} {{end}}name=n>`, + `<input checked name=n>`, + }, + { + "dynamic attribute name", + `<img on{{"load"}}="alert({{"loaded"}})">`, + // Treated as JS since quotes are inserted. + `<img onload="alert("loaded")">`, + }, + { + "bad dynamic attribute name 1", + // Allow checked, selected, disabled, but not JS or + // CSS attributes. + `<input {{"onchange"}}="{{"doEvil()"}}">`, + `<input ZgotmplZ="doEvil()">`, + }, + { + "bad dynamic attribute name 2", + `<div {{"sTyle"}}="{{"color: expression(alert(1337))"}}">`, + `<div ZgotmplZ="color: expression(alert(1337))">`, + }, + { + "bad dynamic attribute name 3", + // Allow title or alt, but not a URL. + `<img {{"src"}}="{{"javascript:doEvil()"}}">`, + `<img ZgotmplZ="javascript:doEvil()">`, + }, + { + "bad dynamic attribute name 4", + // Structure preservation requires values to associate + // with a consistent attribute. + `<input checked {{""}}="Whose value am I?">`, + `<input checked ZgotmplZ="Whose value am I?">`, + }, + { + "dynamic element name", + `<h{{3}}><table><t{{"head"}}>...</h{{3}}>`, + `<h3><table><thead>...</h3>`, + }, + { + "bad dynamic element name", + // Dynamic element names are typically used to switch + // between (thead, tfoot, tbody), (ul, ol), (th, td), + // and other replaceable sets. + // We do not currently easily support (ul, ol). + // If we do change to support that, this test should + // catch failures to filter out special tag names which + // would violate the structure preservation property -- + // if any special tag name could be substituted, then + // the content could be raw text/RCDATA for some inputs + // and regular HTML content for others. + `<{{"script"}}>{{"doEvil()"}}</{{"script"}}>`, + `<script>doEvil()</script>`, + }, + } + + for _, test := range tests { + tmpl := New(test.name) + tmpl = Must(tmpl.Parse(test.input)) + // Check for bug 6459: Tree field was not set in Parse. + if tmpl.Tree != tmpl.text.Tree { + t.Errorf("%s: tree not set properly", test.name) + continue + } + b := new(bytes.Buffer) + if err := tmpl.Execute(b, data); err != nil { + t.Errorf("%s: template execution failed: %s", test.name, err) + continue + } + if w, g := test.output, b.String(); w != g { + t.Errorf("%s: escaped output: want\n\t%q\ngot\n\t%q", test.name, w, g) + continue + } + b.Reset() + if err := tmpl.Execute(b, pdata); err != nil { + t.Errorf("%s: template execution failed for pointer: %s", test.name, err) + continue + } + if w, g := test.output, b.String(); w != g { + t.Errorf("%s: escaped output for pointer: want\n\t%q\ngot\n\t%q", test.name, w, g) + continue + } + if tmpl.Tree != tmpl.text.Tree { + t.Errorf("%s: tree mismatch", test.name) + continue + } + } +} + +func TestEscapeSet(t *testing.T) { + type dataItem struct { + Children []*dataItem + X string + } + + data := dataItem{ + Children: []*dataItem{ + {X: "foo"}, + {X: "<bar>"}, + { + Children: []*dataItem{ + {X: "baz"}, + }, + }, + }, + } + + tests := []struct { + inputs map[string]string + want string + }{ + // The trivial set. + { + map[string]string{ + "main": ``, + }, + ``, + }, + // A template called in the start context. + { + map[string]string{ + "main": `Hello, {{template "helper"}}!`, + // Not a valid top level HTML template. + // "<b" is not a full tag. + "helper": `{{"<World>"}}`, + }, + `Hello, <World>!`, + }, + // A template called in a context other than the start. + { + map[string]string{ + "main": `<a onclick='a = {{template "helper"}};'>`, + // Not a valid top level HTML template. + // "<b" is not a full tag. + "helper": `{{"<a>"}}<b`, + }, + `<a onclick='a = "\u003ca\u003e"<b;'>`, + }, + // A recursive template that ends in its start context. + { + map[string]string{ + "main": `{{range .Children}}{{template "main" .}}{{else}}{{.X}} {{end}}`, + }, + `foo <bar> baz `, + }, + // A recursive helper template that ends in its start context. + { + map[string]string{ + "main": `{{template "helper" .}}`, + "helper": `{{if .Children}}<ul>{{range .Children}}<li>{{template "main" .}}</li>{{end}}</ul>{{else}}{{.X}}{{end}}`, + }, + `<ul><li>foo</li><li><bar></li><li><ul><li>baz</li></ul></li></ul>`, + }, + // Co-recursive templates that end in its start context. + { + map[string]string{ + "main": `<blockquote>{{range .Children}}{{template "helper" .}}{{end}}</blockquote>`, + "helper": `{{if .Children}}{{template "main" .}}{{else}}{{.X}}<br>{{end}}`, + }, + `<blockquote>foo<br><bar><br><blockquote>baz<br></blockquote></blockquote>`, + }, + // A template that is called in two different contexts. + { + map[string]string{ + "main": `<button onclick="title='{{template "helper"}}'; ...">{{template "helper"}}</button>`, + "helper": `{{11}} of {{"<100>"}}`, + }, + `<button onclick="title='11 of \x3c100\x3e'; ...">11 of <100></button>`, + }, + // A non-recursive template that ends in a different context. + // helper starts in jsCtxRegexp and ends in jsCtxDivOp. + { + map[string]string{ + "main": `<script>var x={{template "helper"}}/{{"42"}};</script>`, + "helper": "{{126}}", + }, + `<script>var x= 126 /"42";</script>`, + }, + // A recursive template that ends in a similar context. + { + map[string]string{ + "main": `<script>var x=[{{template "countdown" 4}}];</script>`, + "countdown": `{{.}}{{if .}},{{template "countdown" . | pred}}{{end}}`, + }, + `<script>var x=[ 4 , 3 , 2 , 1 , 0 ];</script>`, + }, + // A recursive template that ends in a different context. + /* + { + map[string]string{ + "main": `<a href="/foo{{template "helper" .}}">`, + "helper": `{{if .Children}}{{range .Children}}{{template "helper" .}}{{end}}{{else}}?x={{.X}}{{end}}`, + }, + `<a href="/foo?x=foo?x=%3cbar%3e?x=baz">`, + }, + */ + } + + // pred is a template function that returns the predecessor of a + // natural number for testing recursive templates. + fns := FuncMap{"pred": func(a ...interface{}) (interface{}, error) { + if len(a) == 1 { + if i, _ := a[0].(int); i > 0 { + return i - 1, nil + } + } + return nil, fmt.Errorf("undefined pred(%v)", a) + }} + + for _, test := range tests { + source := "" + for name, body := range test.inputs { + source += fmt.Sprintf("{{define %q}}%s{{end}} ", name, body) + } + tmpl, err := New("root").Funcs(fns).Parse(source) + if err != nil { + t.Errorf("error parsing %q: %v", source, err) + continue + } + var b bytes.Buffer + + if err := tmpl.ExecuteTemplate(&b, "main", data); err != nil { + t.Errorf("%q executing %v", err.Error(), tmpl.Lookup("main")) + continue + } + if got := b.String(); test.want != got { + t.Errorf("want\n\t%q\ngot\n\t%q", test.want, got) + } + } + +} + +func TestErrors(t *testing.T) { + tests := []struct { + input string + err string + }{ + // Non-error cases. + { + "{{if .Cond}}<a>{{else}}<b>{{end}}", + "", + }, + { + "{{if .Cond}}<a>{{end}}", + "", + }, + { + "{{if .Cond}}{{else}}<b>{{end}}", + "", + }, + { + "{{with .Cond}}<div>{{end}}", + "", + }, + { + "{{range .Items}}<a>{{end}}", + "", + }, + { + "<a href='/foo?{{range .Items}}&{{.K}}={{.V}}{{end}}'>", + "", + }, + // Error cases. + { + "{{if .Cond}}<a{{end}}", + "z:1:5: {{if}} branches", + }, + { + "{{if .Cond}}\n{{else}}\n<a{{end}}", + "z:1:5: {{if}} branches", + }, + { + // Missing quote in the else branch. + `{{if .Cond}}<a href="foo">{{else}}<a href="bar>{{end}}`, + "z:1:5: {{if}} branches", + }, + { + // Different kind of attribute: href implies a URL. + "<a {{if .Cond}}href='{{else}}title='{{end}}{{.X}}'>", + "z:1:8: {{if}} branches", + }, + { + "\n{{with .X}}<a{{end}}", + "z:2:7: {{with}} branches", + }, + { + "\n{{with .X}}<a>{{else}}<a{{end}}", + "z:2:7: {{with}} branches", + }, + { + "{{range .Items}}<a{{end}}", + `z:1: on range loop re-entry: "<" in attribute name: "<a"`, + }, + { + "\n{{range .Items}} x='<a{{end}}", + "z:2:8: on range loop re-entry: {{range}} branches", + }, + { + "<a b=1 c={{.H}}", + "z: ends in a non-text context: {stateAttr delimSpaceOrTagEnd", + }, + { + "<script>foo();", + "z: ends in a non-text context: {stateJS", + }, + { + `<a href="{{if .F}}/foo?a={{else}}/bar/{{end}}{{.H}}">`, + "z:1:47: {{.H}} appears in an ambiguous URL context", + }, + { + `<a onclick="alert('Hello \`, + `unfinished escape sequence in JS string: "Hello \\"`, + }, + { + `<a onclick='alert("Hello\, World\`, + `unfinished escape sequence in JS string: "Hello\\, World\\"`, + }, + { + `<a onclick='alert(/x+\`, + `unfinished escape sequence in JS string: "x+\\"`, + }, + { + `<a onclick="/foo[\]/`, + `unfinished JS regexp charset: "foo[\\]/"`, + }, + { + // It is ambiguous whether 1.5 should be 1\.5 or 1.5. + // Either `var x = 1/- 1.5 /i.test(x)` + // where `i.test(x)` is a method call of reference i, + // or `/-1\.5/i.test(x)` which is a method call on a + // case insensitive regular expression. + `<script>{{if false}}var x = 1{{end}}/-{{"1.5"}}/i.test(x)</script>`, + `'/' could start a division or regexp: "/-"`, + }, + { + `{{template "foo"}}`, + "z:1:11: no such template \"foo\"", + }, + { + `<div{{template "y"}}>` + + // Illegal starting in stateTag but not in stateText. + `{{define "y"}} foo<b{{end}}`, + `"<" in attribute name: " foo<b"`, + }, + { + `<script>reverseList = [{{template "t"}}]</script>` + + // Missing " after recursive call. + `{{define "t"}}{{if .Tail}}{{template "t" .Tail}}{{end}}{{.Head}}",{{end}}`, + `: cannot compute output context for template t$htmltemplate_stateJS_elementScript`, + }, + { + `<input type=button value=onclick=>`, + `html/template:z: "=" in unquoted attr: "onclick="`, + }, + { + `<input type=button value= onclick=>`, + `html/template:z: "=" in unquoted attr: "onclick="`, + }, + { + `<input type=button value= 1+1=2>`, + `html/template:z: "=" in unquoted attr: "1+1=2"`, + }, + { + "<a class=`foo>", + "html/template:z: \"`\" in unquoted attr: \"`foo\"", + }, + { + `<a style=font:'Arial'>`, + `html/template:z: "'" in unquoted attr: "font:'Arial'"`, + }, + { + `<a=foo>`, + `: expected space, attr name, or end of tag, but got "=foo>"`, + }, + } + + for _, test := range tests { + buf := new(bytes.Buffer) + tmpl, err := New("z").Parse(test.input) + if err != nil { + t.Errorf("input=%q: unexpected parse error %s\n", test.input, err) + continue + } + err = tmpl.Execute(buf, nil) + var got string + if err != nil { + got = err.Error() + } + if test.err == "" { + if got != "" { + t.Errorf("input=%q: unexpected error %q", test.input, got) + } + continue + } + if strings.Index(got, test.err) == -1 { + t.Errorf("input=%q: error\n\t%q\ndoes not contain expected string\n\t%q", test.input, got, test.err) + continue + } + // Check that we get the same error if we call Execute again. + if err := tmpl.Execute(buf, nil); err == nil || err.Error() != got { + t.Errorf("input=%q: unexpected error on second call %q", test.input, err) + + } + } +} + +func TestEscapeText(t *testing.T) { + tests := []struct { + input string + output context + }{ + { + ``, + context{}, + }, + { + `Hello, World!`, + context{}, + }, + { + // An orphaned "<" is OK. + `I <3 Ponies!`, + context{}, + }, + { + `<a`, + context{state: stateTag}, + }, + { + `<a `, + context{state: stateTag}, + }, + { + `<a>`, + context{state: stateText}, + }, + { + `<a href`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a on`, + context{state: stateAttrName, attr: attrScript}, + }, + { + `<a href `, + context{state: stateAfterName, attr: attrURL}, + }, + { + `<a style = `, + context{state: stateBeforeValue, attr: attrStyle}, + }, + { + `<a href=`, + context{state: stateBeforeValue, attr: attrURL}, + }, + { + `<a href=x`, + context{state: stateURL, delim: delimSpaceOrTagEnd, urlPart: urlPartPreQuery}, + }, + { + `<a href=x `, + context{state: stateTag}, + }, + { + `<a href=>`, + context{state: stateText}, + }, + { + `<a href=x>`, + context{state: stateText}, + }, + { + `<a href ='`, + context{state: stateURL, delim: delimSingleQuote}, + }, + { + `<a href=''`, + context{state: stateTag}, + }, + { + `<a href= "`, + context{state: stateURL, delim: delimDoubleQuote}, + }, + { + `<a href=""`, + context{state: stateTag}, + }, + { + `<a title="`, + context{state: stateAttr, delim: delimDoubleQuote}, + }, + { + `<a HREF='http:`, + context{state: stateURL, delim: delimSingleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a Href='/`, + context{state: stateURL, delim: delimSingleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href='"`, + context{state: stateURL, delim: delimSingleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href="'`, + context{state: stateURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href=''`, + context{state: stateURL, delim: delimSingleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href=""`, + context{state: stateURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href=""`, + context{state: stateURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href="`, + context{state: stateURL, delim: delimSpaceOrTagEnd, urlPart: urlPartPreQuery}, + }, + { + `<img alt="1">`, + context{state: stateText}, + }, + { + `<img alt="1>"`, + context{state: stateTag}, + }, + { + `<img alt="1>">`, + context{state: stateText}, + }, + { + `<input checked type="checkbox"`, + context{state: stateTag}, + }, + { + `<a onclick="`, + context{state: stateJS, delim: delimDoubleQuote}, + }, + { + `<a onclick="//foo`, + context{state: stateJSLineCmt, delim: delimDoubleQuote}, + }, + { + "<a onclick='//\n", + context{state: stateJS, delim: delimSingleQuote}, + }, + { + "<a onclick='//\r\n", + context{state: stateJS, delim: delimSingleQuote}, + }, + { + "<a onclick='//\u2028", + context{state: stateJS, delim: delimSingleQuote}, + }, + { + `<a onclick="/*`, + context{state: stateJSBlockCmt, delim: delimDoubleQuote}, + }, + { + `<a onclick="/*/`, + context{state: stateJSBlockCmt, delim: delimDoubleQuote}, + }, + { + `<a onclick="/**/`, + context{state: stateJS, delim: delimDoubleQuote}, + }, + { + `<a onkeypress=""`, + context{state: stateJSDqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick='"foo"`, + context{state: stateJS, delim: delimSingleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick='foo'`, + context{state: stateJS, delim: delimSpaceOrTagEnd, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick='foo`, + context{state: stateJSSqStr, delim: delimSpaceOrTagEnd}, + }, + { + `<a onclick=""foo'`, + context{state: stateJSDqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick="'foo"`, + context{state: stateJSSqStr, delim: delimDoubleQuote}, + }, + { + `<A ONCLICK="'`, + context{state: stateJSSqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick="/`, + context{state: stateJSRegexp, delim: delimDoubleQuote}, + }, + { + `<a onclick="'foo'`, + context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick="'foo\'`, + context{state: stateJSSqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick="'foo\'`, + context{state: stateJSSqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick="/foo/`, + context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<script>/foo/ /=`, + context{state: stateJS, element: elementScript}, + }, + { + `<a onclick="1 /foo`, + context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick="1 /*c*/ /foo`, + context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick="/foo[/]`, + context{state: stateJSRegexp, delim: delimDoubleQuote}, + }, + { + `<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: "`, + 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("/`, + 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}, + }, + { + `<!-- foo`, + context{state: stateHTMLCmt}, + }, + { + `<!-->`, + context{state: stateHTMLCmt}, + }, + { + `<!--->`, + context{state: stateHTMLCmt}, + }, + { + `<!-- foo -->`, + context{state: stateText}, + }, + { + `<script`, + context{state: stateTag, element: elementScript}, + }, + { + `<script `, + context{state: stateTag, element: elementScript}, + }, + { + `<script src="foo.js" `, + context{state: stateTag, element: elementScript}, + }, + { + `<script src='foo.js' `, + context{state: stateTag, element: elementScript}, + }, + { + `<script type=text/javascript `, + context{state: stateTag, element: elementScript}, + }, + { + `<script>foo`, + context{state: stateJS, jsCtx: jsCtxDivOp, element: elementScript}, + }, + { + `<script>foo</script>`, + context{state: stateText}, + }, + { + `<script>foo</script><!--`, + context{state: stateHTMLCmt}, + }, + { + `<script>document.write("<p>foo</p>");`, + context{state: stateJS, element: elementScript}, + }, + { + `<script>document.write("<p>foo<\/script>");`, + context{state: stateJS, element: elementScript}, + }, + { + `<script>document.write("<script>alert(1)</script>");`, + context{state: stateText}, + }, + { + `<Script>`, + context{state: stateJS, element: elementScript}, + }, + { + `<SCRIPT>foo`, + context{state: stateJS, jsCtx: jsCtxDivOp, element: elementScript}, + }, + { + `<textarea>value`, + context{state: stateRCDATA, element: elementTextarea}, + }, + { + `<textarea>value</TEXTAREA>`, + context{state: stateText}, + }, + { + `<textarea name=html><b`, + context{state: stateRCDATA, element: elementTextarea}, + }, + { + `<title>value`, + context{state: stateRCDATA, element: elementTitle}, + }, + { + `<style>value`, + context{state: stateCSS, element: elementStyle}, + }, + { + `<a xlink:href`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a xmlns`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a xmlns:foo`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a xmlnsxyz`, + context{state: stateAttrName}, + }, + { + `<a data-url`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a data-iconUri`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a data-urlItem`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a g:`, + context{state: stateAttrName}, + }, + { + `<a g:url`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a g:iconUri`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a g:urlItem`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a g:value`, + context{state: stateAttrName}, + }, + { + `<a svg:style='`, + context{state: stateCSS, delim: delimSingleQuote}, + }, + { + `<svg:font-face`, + context{state: stateTag}, + }, + { + `<svg:a svg:onclick="`, + context{state: stateJS, delim: delimDoubleQuote}, + }, + } + + for _, test := range tests { + b, e := []byte(test.input), newEscaper(nil) + c := e.escapeText(context{}, &parse.TextNode{NodeType: parse.NodeText, Text: b}) + if !test.output.eq(c) { + t.Errorf("input %q: want context\n\t%v\ngot\n\t%v", test.input, test.output, c) + continue + } + if test.input != string(b) { + t.Errorf("input %q: text node was modified: want %q got %q", test.input, test.input, b) + continue + } + } +} + +func TestEnsurePipelineContains(t *testing.T) { + tests := []struct { + input, output string + ids []string + }{ + { + "{{.X}}", + ".X", + []string{}, + }, + { + "{{.X | html}}", + ".X | html", + []string{}, + }, + { + "{{.X}}", + ".X | html", + []string{"html"}, + }, + { + "{{.X | html}}", + ".X | html | urlquery", + []string{"urlquery"}, + }, + { + "{{.X | html | urlquery}}", + ".X | html | urlquery", + []string{"urlquery"}, + }, + { + "{{.X | html | urlquery}}", + ".X | html | urlquery", + []string{"html", "urlquery"}, + }, + { + "{{.X | html | urlquery}}", + ".X | html | urlquery", + []string{"html"}, + }, + { + "{{.X | urlquery}}", + ".X | html | urlquery", + []string{"html", "urlquery"}, + }, + { + "{{.X | html | print}}", + ".X | urlquery | html | print", + []string{"urlquery", "html"}, + }, + { + "{{($).X | html | print}}", + "($).X | urlquery | html | print", + []string{"urlquery", "html"}, + }, + } + for i, test := range tests { + tmpl := template.Must(template.New("test").Parse(test.input)) + action, ok := (tmpl.Tree.Root.Nodes[0].(*parse.ActionNode)) + if !ok { + t.Errorf("#%d: First node is not an action: %s", i, test.input) + continue + } + pipe := action.Pipe + ensurePipelineContains(pipe, test.ids) + got := pipe.String() + if got != test.output { + t.Errorf("#%d: %s, %v: want\n\t%s\ngot\n\t%s", i, test.input, test.ids, test.output, got) + } + } +} + +func TestEscapeErrorsNotIgnorable(t *testing.T) { + var b bytes.Buffer + tmpl, _ := New("dangerous").Parse("<a") + err := tmpl.Execute(&b, nil) + if err == nil { + t.Errorf("Expected error") + } else if b.Len() != 0 { + t.Errorf("Emitted output despite escaping failure") + } +} + +func TestEscapeSetErrorsNotIgnorable(t *testing.T) { + var b bytes.Buffer + tmpl, err := New("root").Parse(`{{define "t"}}<a{{end}}`) + if err != nil { + t.Errorf("failed to parse set: %q", err) + } + err = tmpl.ExecuteTemplate(&b, "t", nil) + if err == nil { + t.Errorf("Expected error") + } else if b.Len() != 0 { + t.Errorf("Emitted output despite escaping failure") + } +} + +func TestRedundantFuncs(t *testing.T) { + inputs := []interface{}{ + "\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f" + + "\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !"#$%&'()*+,-./` + + `0123456789:;<=>?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ[\]^_` + + "`abcdefghijklmno" + + "pqrstuvwxyz{|}~\x7f" + + "\u00A0\u0100\u2028\u2029\ufeff\ufdec\ufffd\uffff\U0001D11E" + + "&%22\\", + CSS(`a[href =~ "//example.com"]#foo`), + HTML(`Hello, <b>World</b> &tc!`), + HTMLAttr(` dir="ltr"`), + JS(`c && alert("Hello, World!");`), + JSStr(`Hello, World & O'Reilly\x21`), + URL(`greeting=H%69&addressee=(World)`), + } + + for n0, m := range redundantFuncs { + f0 := funcMap[n0].(func(...interface{}) string) + for n1 := range m { + f1 := funcMap[n1].(func(...interface{}) string) + for _, input := range inputs { + want := f0(input) + if got := f1(want); want != got { + t.Errorf("%s %s with %T %q: want\n\t%q,\ngot\n\t%q", n0, n1, input, input, want, got) + } + } + } + } +} + +func TestIndirectPrint(t *testing.T) { + a := 3 + ap := &a + b := "hello" + bp := &b + bpp := &bp + tmpl := Must(New("t").Parse(`{{.}}`)) + var buf bytes.Buffer + err := tmpl.Execute(&buf, ap) + if err != nil { + t.Errorf("Unexpected error: %s", err) + } else if buf.String() != "3" { + t.Errorf(`Expected "3"; got %q`, buf.String()) + } + buf.Reset() + err = tmpl.Execute(&buf, bpp) + if err != nil { + t.Errorf("Unexpected error: %s", err) + } else if buf.String() != "hello" { + t.Errorf(`Expected "hello"; got %q`, buf.String()) + } +} + +// This is a test for issue 3272. +func TestEmptyTemplate(t *testing.T) { + page := Must(New("page").ParseFiles(os.DevNull)) + if err := page.ExecuteTemplate(os.Stdout, "page", "nothing"); err == nil { + t.Fatal("expected error") + } +} + +type Issue7379 int + +func (Issue7379) SomeMethod(x int) string { + return fmt.Sprintf("<%d>", x) +} + +// This is a test for issue 7379: type assertion error caused panic, and then +// the code to handle the panic breaks escaping. It's hard to see the second +// problem once the first is fixed, but its fix is trivial so we let that go. See +// the discussion for issue 7379. +func TestPipeToMethodIsEscaped(t *testing.T) { + tmpl := Must(New("x").Parse("<html>{{0 | .SomeMethod}}</html>\n")) + tryExec := func() string { + defer func() { + panicValue := recover() + if panicValue != nil { + t.Errorf("panicked: %v\n", panicValue) + } + }() + var b bytes.Buffer + tmpl.Execute(&b, Issue7379(0)) + return b.String() + } + for i := 0; i < 3; i++ { + str := tryExec() + const expect = "<html><0></html>\n" + if str != expect { + t.Errorf("expected %q got %q", expect, str) + } + } +} + +func BenchmarkEscapedExecute(b *testing.B) { + tmpl := Must(New("t").Parse(`<a onclick="alert('{{.}}')">{{.}}</a>`)) + var buf bytes.Buffer + b.ResetTimer() + for i := 0; i < b.N; i++ { + tmpl.Execute(&buf, "foo & 'bar' & baz") + buf.Reset() + } +} diff --git a/template/html.go b/template/html.go new file mode 100644 index 00000000..9c069efd --- /dev/null +++ b/template/html.go @@ -0,0 +1,257 @@ +// 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 template + +import ( + "bytes" + "fmt" + "strings" + "unicode/utf8" +) + +// htmlNospaceEscaper escapes for inclusion in unquoted attribute values. +func htmlNospaceEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTML { + return htmlReplacer(stripTags(s), htmlNospaceNormReplacementTable, false) + } + return htmlReplacer(s, htmlNospaceReplacementTable, false) +} + +// attrEscaper escapes for inclusion in quoted attribute values. +func attrEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTML { + return htmlReplacer(stripTags(s), htmlNormReplacementTable, true) + } + return htmlReplacer(s, htmlReplacementTable, true) +} + +// rcdataEscaper escapes for inclusion in an RCDATA element body. +func rcdataEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTML { + return htmlReplacer(s, htmlNormReplacementTable, true) + } + return htmlReplacer(s, htmlReplacementTable, true) +} + +// htmlEscaper escapes for inclusion in HTML text. +func htmlEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTML { + return s + } + return htmlReplacer(s, htmlReplacementTable, true) +} + +// htmlReplacementTable contains the runes that need to be escaped +// inside a quoted attribute value or in a text node. +var htmlReplacementTable = []string{ + // http://www.w3.org/TR/html5/syntax.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/syntax.html#before-attribute-value-state + 0: "\uFFFD", + '"': """, + '&': "&", + '\'': "'", + '+': "+", + '<': "<", + '>': ">", +} + +// htmlNormReplacementTable is like htmlReplacementTable but without '&' to +// avoid over-encoding existing entities. +var htmlNormReplacementTable = []string{ + 0: "\uFFFD", + '"': """, + '\'': "'", + '+': "+", + '<': "<", + '>': ">", +} + +// htmlNospaceReplacementTable contains the runes that need to be escaped +// inside an unquoted attribute value. +// 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 htmlNospaceReplacementTable = []string{ + 0: "�", + '\t': " ", + '\n': " ", + '\v': " ", + '\f': "", + '\r': " ", + ' ': " ", + '"': """, + '&': "&", + '\'': "'", + '+': "+", + '<': "<", + '=': "=", + '>': ">", + // A parse error in the attribute value (unquoted) and + // before attribute value states. + // Treated as a quoting character by IE. + '`': "`", +} + +// htmlNospaceNormReplacementTable is like htmlNospaceReplacementTable but +// without '&' to avoid over-encoding existing entities. +var htmlNospaceNormReplacementTable = []string{ + 0: "�", + '\t': " ", + '\n': " ", + '\v': " ", + '\f': "", + '\r': " ", + ' ': " ", + '"': """, + '\'': "'", + '+': "+", + '<': "<", + '=': "=", + '>': ">", + // A parse error in the attribute value (unquoted) and + // before attribute value states. + // Treated as a quoting character by IE. + '`': "`", +} + +// htmlReplacer returns s with runes replaced according to replacementTable +// and when badRunes is true, certain bad runes are allowed through unescaped. +func htmlReplacer(s string, replacementTable []string, badRunes bool) string { + written, b := 0, new(bytes.Buffer) + for i, r := range s { + if int(r) < len(replacementTable) { + if repl := replacementTable[r]; len(repl) != 0 { + b.WriteString(s[written:i]) + b.WriteString(repl) + // Valid as long as replacementTable doesn't + // include anything above 0x7f. + written = i + utf8.RuneLen(r) + } + } else if badRunes { + // No-op. + // IE does not allow these ranges in unquoted attrs. + } else if 0xfdd0 <= r && r <= 0xfdef || 0xfff0 <= r && r <= 0xffff { + fmt.Fprintf(b, "%s&#x%x;", s[written:i], r) + written = i + utf8.RuneLen(r) + } + } + if written == 0 { + return s + } + b.WriteString(s[written:]) + return b.String() +} + +// stripTags takes a snippet of HTML and returns only the text content. +// For example, `<b>¡Hi!</b> <script>...</script>` -> `¡Hi! `. +func stripTags(html string) string { + var b bytes.Buffer + 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 { + 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. + j := i1 + if d.state != c.state { + for j1 := j - 1; j1 >= i; j1-- { + if s[j1] == '<' { + j = j1 + break + } + } + } + b.Write(s[i:j]) + } else { + allText = false + } + c, i = d, i1 + continue + } + i1 := i + bytes.IndexAny(s[i:], delimEnds[c.delim]) + if i1 < i { + break + } + if c.delim != delimSpaceOrTagEnd { + // Consume any quote. + i1++ + } + c, i = context{state: stateTag, element: c.element}, i1 + } + if allText { + return html + } else if c.state == stateText || c.state == stateRCDATA { + b.Write(s[i:]) + } + return b.String() +} + +// htmlNameFilter accepts valid parts of an HTML attribute or tag name or +// a known-safe HTML attribute. +func htmlNameFilter(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTMLAttr { + return s + } + if len(s) == 0 { + // Avoid violation of structure preservation. + // <input checked {{.K}}={{.V}}>. + // Without this, if .K is empty then .V is the value of + // checked, but otherwise .V is the value of the attribute + // named .K. + return filterFailsafe + } + s = strings.ToLower(s) + if t := attrType(s); t != contentTypePlain { + // TODO: Split attr and element name part filters so we can whitelist + // attributes. + return filterFailsafe + } + for _, r := range s { + switch { + case '0' <= r && r <= '9': + case 'a' <= r && r <= 'z': + default: + return filterFailsafe + } + } + return s +} + +// commentEscaper returns the empty string regardless of input. +// Comment content does not correspond to any parsed structure or +// human-readable content, so the simplest and most secure policy is to drop +// content interpolated into comments. +// This approach is equally valid whether or not static comment content is +// removed from the template. +func commentEscaper(args ...interface{}) string { + return "" +} diff --git a/template/html_test.go b/template/html_test.go new file mode 100644 index 00000000..b9b97038 --- /dev/null +++ b/template/html_test.go @@ -0,0 +1,94 @@ +// 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 template + +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\ufdec\U0001D11E") + + want := ("�\x01\x02\x03\x04\x05\x06\x07" + + "\x08 \x0E\x0F" + + "\x10\x11\x12\x13\x14\x15\x16\x17" + + "\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !"#$%&'()*+,-./` + + `0123456789:;<=>?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ[\]^_` + + ``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 TestStripTags(t *testing.T) { + tests := []struct { + input, want string + }{ + {"", ""}, + {"Hello, World!", "Hello, World!"}, + {"foo&bar", "foo&bar"}, + {`Hello <a href="www.example.com/">World</a>!`, "Hello World!"}, + {"Foo <textarea>Bar</textarea> Baz", "Foo Bar Baz"}, + {"Foo <!-- Bar --> Baz", "Foo Baz"}, + {"<", "<"}, + {"foo < bar", "foo < bar"}, + {`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 { + if got := stripTags(test.input); got != test.want { + t.Errorf("%q: want %q, got %q", test.input, test.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.") + } +} + +func BenchmarkStripTags(b *testing.B) { + for i := 0; i < b.N; i++ { + stripTags("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>") + } +} + +func BenchmarkStripTagsNoSpecials(b *testing.B) { + for i := 0; i < b.N; i++ { + stripTags("The quick, brown fox jumps over the lazy dog.") + } +} diff --git a/template/js.go b/template/js.go new file mode 100644 index 00000000..999a61ed --- /dev/null +++ b/template/js.go @@ -0,0 +1,362 @@ +// 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 template + +import ( + "bytes" + "encoding/json" + "fmt" + "reflect" + "strings" + "unicode/utf8" +) + +// nextJSCtx returns the context that determines whether a slash after the +// given run of tokens starts a regular expression instead of a division +// operator: / or /=. +// +// This assumes that the token run does not include any string tokens, comment +// tokens, regular expression literal tokens, or division operators. +// +// This fails on some valid but nonsensical JavaScript programs like +// "x = ++/foo/i" which is quite different than "x++/foo/i", but is not known to +// fail on any known useful programs. It is based on the draft +// JavaScript 2.0 lexical grammar and requires one token of lookbehind: +// http://www.mozilla.org/js/language/js20-2000-07/rationale/syntax.html +func nextJSCtx(s []byte, preceding jsCtx) jsCtx { + s = bytes.TrimRight(s, "\t\n\f\r \u2028\u2029") + if len(s) == 0 { + return preceding + } + + // All cases below are in the single-byte UTF-8 group. + switch c, n := s[len(s)-1], len(s); c { + case '+', '-': + // ++ and -- are not regexp preceders, but + and - are whether + // they are used as infix or prefix operators. + start := n - 1 + // Count the number of adjacent dashes or pluses. + for start > 0 && s[start-1] == c { + start-- + } + if (n-start)&1 == 1 { + // Reached for trailing minus signs since "---" is the + // same as "-- -". + return jsCtxRegexp + } + return jsCtxDivOp + case '.': + // Handle "42." + if n != 1 && '0' <= s[n-2] && s[n-2] <= '9' { + return jsCtxDivOp + } + return jsCtxRegexp + // Suffixes for all punctuators from section 7.7 of the language spec + // that only end binary operators not handled above. + case ',', '<', '>', '=', '*', '%', '&', '|', '^', '?': + return jsCtxRegexp + // Suffixes for all punctuators from section 7.7 of the language spec + // that are prefix operators not handled above. + case '!', '~': + return jsCtxRegexp + // Matches all the punctuators from section 7.7 of the language spec + // that are open brackets not handled above. + case '(', '[': + return jsCtxRegexp + // Matches all the punctuators from section 7.7 of the language spec + // that precede expression starts. + case ':', ';', '{': + return jsCtxRegexp + // CAVEAT: the close punctuators ('}', ']', ')') precede div ops and + // are handled in the default except for '}' which can precede a + // division op as in + // ({ valueOf: function () { return 42 } } / 2 + // which is valid, but, in practice, developers don't divide object + // literals, so our heuristic works well for code like + // function () { ... } /foo/.test(x) && sideEffect(); + // The ')' punctuator can precede a regular expression as in + // if (b) /foo/.test(x) && ... + // but this is much less likely than + // (a + b) / c + case '}': + return jsCtxRegexp + default: + // Look for an IdentifierName and see if it is a keyword that + // can precede a regular expression. + j := n + for j > 0 && isJSIdentPart(rune(s[j-1])) { + j-- + } + if regexpPrecederKeywords[string(s[j:])] { + return jsCtxRegexp + } + } + // Otherwise is a punctuator not listed above, or + // a string which precedes a div op, or an identifier + // which precedes a div op. + return jsCtxDivOp +} + +// regexpPrecederKeywords is a set of reserved JS keywords that can precede a +// regular expression in JS source. +var regexpPrecederKeywords = map[string]bool{ + "break": true, + "case": true, + "continue": true, + "delete": true, + "do": true, + "else": true, + "finally": true, + "in": true, + "instanceof": true, + "return": true, + "throw": true, + "try": true, + "typeof": true, + "void": true, +} + +var jsonMarshalType = reflect.TypeOf((*json.Marshaler)(nil)).Elem() + +// indirectToJSONMarshaler returns the value, after dereferencing as many times +// as necessary to reach the base type (or nil) or an implementation of json.Marshal. +func indirectToJSONMarshaler(a interface{}) interface{} { + v := reflect.ValueOf(a) + for !v.Type().Implements(jsonMarshalType) && v.Kind() == reflect.Ptr && !v.IsNil() { + v = v.Elem() + } + return v.Interface() +} + +// jsValEscaper escapes its inputs to a JS Expression (section 11.14) that has +// neither side-effects nor free variables outside (NaN, Infinity). +func jsValEscaper(args ...interface{}) string { + var a interface{} + if len(args) == 1 { + a = indirectToJSONMarshaler(args[0]) + switch t := a.(type) { + case JS: + return string(t) + case JSStr: + // TODO: normalize quotes. + return `"` + string(t) + `"` + case json.Marshaler: + // Do not treat as a Stringer. + case fmt.Stringer: + a = t.String() + } + } else { + for i, arg := range args { + args[i] = indirectToJSONMarshaler(arg) + } + a = fmt.Sprint(args...) + } + // TODO: detect cycles before calling Marshal which loops infinitely on + // cyclic data. This may be an unacceptable DoS risk. + + b, err := json.Marshal(a) + if err != nil { + // Put a space before comment so that if it is flush against + // a division operator it is not turned into a line comment: + // x/{{y}} + // turning into + // x//* error marshalling y: + // second line of error message */null + return fmt.Sprintf(" /* %s */null ", strings.Replace(err.Error(), "*/", "* /", -1)) + } + + // TODO: maybe post-process output to prevent it from containing + // "<!--", "-->", "<![CDATA[", "]]>", or "</script" + // in case custom marshallers produce output containing those. + + // TODO: Maybe abbreviate \u00ab to \xab to produce more compact output. + if len(b) == 0 { + // In, `x=y/{{.}}*z` a json.Marshaler that produces "" should + // not cause the output `x=y/*z`. + return " null " + } + first, _ := utf8.DecodeRune(b) + last, _ := utf8.DecodeLastRune(b) + var buf bytes.Buffer + // Prevent IdentifierNames and NumericLiterals from running into + // keywords: in, instanceof, typeof, void + pad := isJSIdentPart(first) || isJSIdentPart(last) + if pad { + buf.WriteByte(' ') + } + written := 0 + // Make sure that json.Marshal escapes codepoints U+2028 & U+2029 + // so it falls within the subset of JSON which is valid JS. + for i := 0; i < len(b); { + rune, n := utf8.DecodeRune(b[i:]) + repl := "" + if rune == 0x2028 { + repl = `\u2028` + } else if rune == 0x2029 { + repl = `\u2029` + } + if repl != "" { + buf.Write(b[written:i]) + buf.WriteString(repl) + written = i + n + } + i += n + } + if buf.Len() != 0 { + buf.Write(b[written:]) + if pad { + buf.WriteByte(' ') + } + b = buf.Bytes() + } + return string(b) +} + +// jsStrEscaper produces a string that can be included between quotes in +// JavaScript source, in JavaScript embedded in an HTML5 <script> element, +// or in an HTML5 event handler attribute such as onclick. +func jsStrEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeJSStr { + return replace(s, jsStrNormReplacementTable) + } + return replace(s, jsStrReplacementTable) +} + +// jsRegexpEscaper behaves like jsStrEscaper but escapes regular expression +// specials so the result is treated literally when included in a regular +// expression literal. /foo{{.X}}bar/ matches the string "foo" followed by +// the literal text of {{.X}} followed by the string "bar". +func jsRegexpEscaper(args ...interface{}) string { + s, _ := stringify(args...) + s = replace(s, jsRegexpReplacementTable) + if s == "" { + // /{{.X}}/ should not produce a line comment when .X == "". + return "(?:)" + } + return s +} + +// replace replaces each rune r of s with replacementTable[r], provided that +// r < len(replacementTable). If replacementTable[r] is the empty string then +// no replacement is made. +// It also replaces runes U+2028 and U+2029 with the raw strings `\u2028` and +// `\u2029`. +func replace(s string, replacementTable []string) string { + var b bytes.Buffer + written := 0 + for i, r := range s { + var repl string + switch { + case int(r) < len(replacementTable) && replacementTable[r] != "": + repl = replacementTable[r] + case r == '\u2028': + repl = `\u2028` + case r == '\u2029': + repl = `\u2029` + default: + continue + } + b.WriteString(s[written:i]) + b.WriteString(repl) + written = i + utf8.RuneLen(r) + } + if written == 0 { + return s + } + b.WriteString(s[written:]) + return b.String() +} + +var jsStrReplacementTable = []string{ + 0: `\0`, + '\t': `\t`, + '\n': `\n`, + '\v': `\x0b`, // "\v" == "v" on IE 6. + '\f': `\f`, + '\r': `\r`, + // Encode HTML specials as hex so the output can be embedded + // in HTML attributes without further encoding. + '"': `\x22`, + '&': `\x26`, + '\'': `\x27`, + '+': `\x2b`, + '/': `\/`, + '<': `\x3c`, + '>': `\x3e`, + '\\': `\\`, +} + +// jsStrNormReplacementTable is like jsStrReplacementTable but does not +// overencode existing escapes since this table has no entry for `\`. +var jsStrNormReplacementTable = []string{ + 0: `\0`, + '\t': `\t`, + '\n': `\n`, + '\v': `\x0b`, // "\v" == "v" on IE 6. + '\f': `\f`, + '\r': `\r`, + // Encode HTML specials as hex so the output can be embedded + // in HTML attributes without further encoding. + '"': `\x22`, + '&': `\x26`, + '\'': `\x27`, + '+': `\x2b`, + '/': `\/`, + '<': `\x3c`, + '>': `\x3e`, +} + +var jsRegexpReplacementTable = []string{ + 0: `\0`, + '\t': `\t`, + '\n': `\n`, + '\v': `\x0b`, // "\v" == "v" on IE 6. + '\f': `\f`, + '\r': `\r`, + // Encode HTML specials as hex so the output can be embedded + // in HTML attributes without further encoding. + '"': `\x22`, + '$': `\$`, + '&': `\x26`, + '\'': `\x27`, + '(': `\(`, + ')': `\)`, + '*': `\*`, + '+': `\x2b`, + '-': `\-`, + '.': `\.`, + '/': `\/`, + '<': `\x3c`, + '>': `\x3e`, + '?': `\?`, + '[': `\[`, + '\\': `\\`, + ']': `\]`, + '^': `\^`, + '{': `\{`, + '|': `\|`, + '}': `\}`, +} + +// isJSIdentPart reports 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. +func isJSIdentPart(r rune) bool { + switch { + case r == '$': + return true + case '0' <= r && r <= '9': + return true + case 'A' <= r && r <= 'Z': + return true + case r == '_': + return true + case 'a' <= r && r <= 'z': + return true + } + return false +} diff --git a/template/js_test.go b/template/js_test.go new file mode 100644 index 00000000..7af7997d --- /dev/null +++ b/template/js_test.go @@ -0,0 +1,401 @@ +// 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 template + +import ( + "bytes" + "math" + "strings" + "testing" +) + +func TestNextJsCtx(t *testing.T) { + tests := []struct { + jsCtx jsCtx + s string + }{ + // Statement terminators precede regexps. + {jsCtxRegexp, ";"}, + // This is not airtight. + // ({ valueOf: function () { return 1 } } / 2) + // is valid JavaScript but in practice, devs do not do this. + // A block followed by a statement starting with a RegExp is + // much more common: + // while (x) {...} /foo/.test(x) || panic() + {jsCtxRegexp, "}"}, + // But member, call, grouping, and array expression terminators + // precede div ops. + {jsCtxDivOp, ")"}, + {jsCtxDivOp, "]"}, + // At the start of a primary expression, array, or expression + // statement, expect a regexp. + {jsCtxRegexp, "("}, + {jsCtxRegexp, "["}, + {jsCtxRegexp, "{"}, + // Assignment operators precede regexps as do all exclusively + // prefix and binary operators. + {jsCtxRegexp, "="}, + {jsCtxRegexp, "+="}, + {jsCtxRegexp, "*="}, + {jsCtxRegexp, "*"}, + {jsCtxRegexp, "!"}, + // Whether the + or - is infix or prefix, it cannot precede a + // div op. + {jsCtxRegexp, "+"}, + {jsCtxRegexp, "-"}, + // An incr/decr op precedes a div operator. + // This is not airtight. In (g = ++/h/i) a regexp follows a + // pre-increment operator, but in practice devs do not try to + // increment or decrement regular expressions. + // (g++/h/i) where ++ is a postfix operator on g is much more + // common. + {jsCtxDivOp, "--"}, + {jsCtxDivOp, "++"}, + {jsCtxDivOp, "x--"}, + // When we have many dashes or pluses, then they are grouped + // left to right. + {jsCtxRegexp, "x---"}, // A postfix -- then a -. + // return followed by a slash returns the regexp literal or the + // slash starts a regexp literal in an expression statement that + // is dead code. + {jsCtxRegexp, "return"}, + {jsCtxRegexp, "return "}, + {jsCtxRegexp, "return\t"}, + {jsCtxRegexp, "return\n"}, + {jsCtxRegexp, "return\u2028"}, + // Identifiers can be divided and cannot validly be preceded by + // a regular expressions. Semicolon insertion cannot happen + // between an identifier and a regular expression on a new line + // because the one token lookahead for semicolon insertion has + // to conclude that it could be a div binary op and treat it as + // such. + {jsCtxDivOp, "x"}, + {jsCtxDivOp, "x "}, + {jsCtxDivOp, "x\t"}, + {jsCtxDivOp, "x\n"}, + {jsCtxDivOp, "x\u2028"}, + {jsCtxDivOp, "preturn"}, + // Numbers precede div ops. + {jsCtxDivOp, "0"}, + // Dots that are part of a number are div preceders. + {jsCtxDivOp, "0."}, + } + + for _, test := range tests { + if nextJSCtx([]byte(test.s), jsCtxRegexp) != test.jsCtx { + t.Errorf("want %s got %q", test.jsCtx, test.s) + } + if nextJSCtx([]byte(test.s), jsCtxDivOp) != test.jsCtx { + t.Errorf("want %s got %q", test.jsCtx, test.s) + } + } + + if nextJSCtx([]byte(" "), jsCtxRegexp) != jsCtxRegexp { + t.Error("Blank tokens") + } + + if nextJSCtx([]byte(" "), jsCtxDivOp) != jsCtxDivOp { + t.Error("Blank tokens") + } +} + +func TestJSValEscaper(t *testing.T) { + tests := []struct { + x interface{} + js string + }{ + {int(42), " 42 "}, + {uint(42), " 42 "}, + {int16(42), " 42 "}, + {uint16(42), " 42 "}, + {int32(-42), " -42 "}, + {uint32(42), " 42 "}, + {int16(-42), " -42 "}, + {uint16(42), " 42 "}, + {int64(-42), " -42 "}, + {uint64(42), " 42 "}, + {uint64(1) << 53, " 9007199254740992 "}, + // ulp(1 << 53) > 1 so this loses precision in JS + // but it is still a representable integer literal. + {uint64(1)<<53 + 1, " 9007199254740993 "}, + {float32(1.0), " 1 "}, + {float32(-1.0), " -1 "}, + {float32(0.5), " 0.5 "}, + {float32(-0.5), " -0.5 "}, + {float32(1.0) / float32(256), " 0.00390625 "}, + {float32(0), " 0 "}, + {math.Copysign(0, -1), " -0 "}, + {float64(1.0), " 1 "}, + {float64(-1.0), " -1 "}, + {float64(0.5), " 0.5 "}, + {float64(-0.5), " -0.5 "}, + {float64(0), " 0 "}, + {math.Copysign(0, -1), " -0 "}, + {"", `""`}, + {"foo", `"foo"`}, + // Newlines. + {"\r\n\u2028\u2029", `"\r\n\u2028\u2029"`}, + // "\v" == "v" on IE 6 so use "\x0b" instead. + {"\t\x0b", `"\t\u000b"`}, + {struct{ X, Y int }{1, 2}, `{"X":1,"Y":2}`}, + {[]interface{}{}, "[]"}, + {[]interface{}{42, "foo", nil}, `[42,"foo",null]`}, + {[]string{"<!--", "</script>", "-->"}, `["\u003c!--","\u003c/script\u003e","--\u003e"]`}, + {"<!--", `"\u003c!--"`}, + {"-->", `"--\u003e"`}, + {"<![CDATA[", `"\u003c![CDATA["`}, + {"]]>", `"]]\u003e"`}, + {"</script", `"\u003c/script"`}, + {"\U0001D11E", "\"\U0001D11E\""}, // or "\uD834\uDD1E" + } + + for _, test := range tests { + if js := jsValEscaper(test.x); js != test.js { + t.Errorf("%+v: want\n\t%q\ngot\n\t%q", test.x, test.js, js) + } + // Make sure that escaping corner cases are not broken + // by nesting. + a := []interface{}{test.x} + want := "[" + strings.TrimSpace(test.js) + "]" + if js := jsValEscaper(a); js != want { + t.Errorf("%+v: want\n\t%q\ngot\n\t%q", a, want, js) + } + } +} + +func TestJSStrEscaper(t *testing.T) { + tests := []struct { + x interface{} + esc string + }{ + {"", ``}, + {"foo", `foo`}, + {"\u0000", `\0`}, + {"\t", `\t`}, + {"\n", `\n`}, + {"\r", `\r`}, + {"\u2028", `\u2028`}, + {"\u2029", `\u2029`}, + {"\\", `\\`}, + {"\\n", `\\n`}, + {"foo\r\nbar", `foo\r\nbar`}, + // Preserve attribute boundaries. + {`"`, `\x22`}, + {`'`, `\x27`}, + // Allow embedding in HTML without further escaping. + {`&`, `\x26amp;`}, + // Prevent breaking out of text node and element boundaries. + {"</script>", `\x3c\/script\x3e`}, + {"<![CDATA[", `\x3c![CDATA[`}, + {"]]>", `]]\x3e`}, + // http://dev.w3.org/html5/markup/aria/syntax.html#escaping-text-span + // "The text in style, script, title, and textarea elements + // must not have an escaping text span start that is not + // followed by an escaping text span end." + // Furthermore, spoofing an escaping text span end could lead + // to different interpretation of a </script> sequence otherwise + // masked by the escaping text span, and spoofing a start could + // allow regular text content to be interpreted as script + // allowing script execution via a combination of a JS string + // injection followed by an HTML text injection. + {"<!--", `\x3c!--`}, + {"-->", `--\x3e`}, + // From http://code.google.com/p/doctype/wiki/ArticleUtf7 + {"+ADw-script+AD4-alert(1)+ADw-/script+AD4-", + `\x2bADw-script\x2bAD4-alert(1)\x2bADw-\/script\x2bAD4-`, + }, + // Invalid UTF-8 sequence + {"foo\xA0bar", "foo\xA0bar"}, + // Invalid unicode scalar value. + {"foo\xed\xa0\x80bar", "foo\xed\xa0\x80bar"}, + } + + for _, test := range tests { + esc := jsStrEscaper(test.x) + if esc != test.esc { + t.Errorf("%q: want %q got %q", test.x, test.esc, esc) + } + } +} + +func TestJSRegexpEscaper(t *testing.T) { + tests := []struct { + x interface{} + esc string + }{ + {"", `(?:)`}, + {"foo", `foo`}, + {"\u0000", `\0`}, + {"\t", `\t`}, + {"\n", `\n`}, + {"\r", `\r`}, + {"\u2028", `\u2028`}, + {"\u2029", `\u2029`}, + {"\\", `\\`}, + {"\\n", `\\n`}, + {"foo\r\nbar", `foo\r\nbar`}, + // Preserve attribute boundaries. + {`"`, `\x22`}, + {`'`, `\x27`}, + // Allow embedding in HTML without further escaping. + {`&`, `\x26amp;`}, + // Prevent breaking out of text node and element boundaries. + {"</script>", `\x3c\/script\x3e`}, + {"<![CDATA[", `\x3c!\[CDATA\[`}, + {"]]>", `\]\]\x3e`}, + // Escaping text spans. + {"<!--", `\x3c!\-\-`}, + {"-->", `\-\-\x3e`}, + {"*", `\*`}, + {"+", `\x2b`}, + {"?", `\?`}, + {"[](){}", `\[\]\(\)\{\}`}, + {"$foo|x.y", `\$foo\|x\.y`}, + {"x^y", `x\^y`}, + } + + for _, test := range tests { + esc := jsRegexpEscaper(test.x) + if esc != test.esc { + t.Errorf("%q: want %q got %q", test.x, test.esc, esc) + } + } +} + +func TestEscapersOnLower7AndSelectHighCodepoints(t *testing.T) { + input := ("\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f" + + "\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !"#$%&'()*+,-./` + + `0123456789:;<=>?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ[\]^_` + + "`abcdefghijklmno" + + "pqrstuvwxyz{|}~\x7f" + + "\u00A0\u0100\u2028\u2029\ufeff\U0001D11E") + + tests := []struct { + name string + escaper func(...interface{}) string + escaped string + }{ + { + "jsStrEscaper", + jsStrEscaper, + "\\0\x01\x02\x03\x04\x05\x06\x07" + + "\x08\\t\\n\\x0b\\f\\r\x0E\x0F" + + "\x10\x11\x12\x13\x14\x15\x16\x17" + + "\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !\x22#$%\x26\x27()*\x2b,-.\/` + + `0123456789:;\x3c=\x3e?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ[\\]^_` + + "`abcdefghijklmno" + + "pqrstuvwxyz{|}~\x7f" + + "\u00A0\u0100\\u2028\\u2029\ufeff\U0001D11E", + }, + { + "jsRegexpEscaper", + jsRegexpEscaper, + "\\0\x01\x02\x03\x04\x05\x06\x07" + + "\x08\\t\\n\\x0b\\f\\r\x0E\x0F" + + "\x10\x11\x12\x13\x14\x15\x16\x17" + + "\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !\x22#\$%\x26\x27\(\)\*\x2b,\-\.\/` + + `0123456789:;\x3c=\x3e\?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ\[\\\]\^_` + + "`abcdefghijklmno" + + `pqrstuvwxyz\{\|\}~` + "\u007f" + + "\u00A0\u0100\\u2028\\u2029\ufeff\U0001D11E", + }, + } + + for _, test := range tests { + if s := test.escaper(input); s != test.escaped { + t.Errorf("%s once: want\n\t%q\ngot\n\t%q", test.name, test.escaped, s) + continue + } + + // Escape it rune by rune to make sure that any + // fast-path checking does not break escaping. + var buf bytes.Buffer + for _, c := range input { + buf.WriteString(test.escaper(string(c))) + } + + if s := buf.String(); s != test.escaped { + t.Errorf("%s rune-wise: want\n\t%q\ngot\n\t%q", test.name, test.escaped, s) + continue + } + } +} + +func BenchmarkJSValEscaperWithNum(b *testing.B) { + for i := 0; i < b.N; i++ { + jsValEscaper(3.141592654) + } +} + +func BenchmarkJSValEscaperWithStr(b *testing.B) { + for i := 0; i < b.N; i++ { + jsValEscaper("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>") + } +} + +func BenchmarkJSValEscaperWithStrNoSpecials(b *testing.B) { + for i := 0; i < b.N; i++ { + jsValEscaper("The quick, brown fox jumps over the lazy dog") + } +} + +func BenchmarkJSValEscaperWithObj(b *testing.B) { + o := struct { + S string + N int + }{ + "The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>\u2028", + 42, + } + for i := 0; i < b.N; i++ { + jsValEscaper(o) + } +} + +func BenchmarkJSValEscaperWithObjNoSpecials(b *testing.B) { + o := struct { + S string + N int + }{ + "The quick, brown fox jumps over the lazy dog", + 42, + } + for i := 0; i < b.N; i++ { + jsValEscaper(o) + } +} + +func BenchmarkJSStrEscaperNoSpecials(b *testing.B) { + for i := 0; i < b.N; i++ { + jsStrEscaper("The quick, brown fox jumps over the lazy dog.") + } +} + +func BenchmarkJSStrEscaper(b *testing.B) { + for i := 0; i < b.N; i++ { + jsStrEscaper("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>") + } +} + +func BenchmarkJSRegexpEscaperNoSpecials(b *testing.B) { + for i := 0; i < b.N; i++ { + jsRegexpEscaper("The quick, brown fox jumps over the lazy dog") + } +} + +func BenchmarkJSRegexpEscaper(b *testing.B) { + for i := 0; i < b.N; i++ { + jsRegexpEscaper("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>") + } +} diff --git a/template/template.go b/template/template.go new file mode 100644 index 00000000..65b5bff5 --- /dev/null +++ b/template/template.go @@ -0,0 +1,390 @@ +// 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 template + +import ( + "fmt" + "io" + "io/ioutil" + "path/filepath" + "sync" + "text/template" + "text/template/parse" +) + +// Template is a specialized Template from "text/template" that produces a safe +// HTML document fragment. +type Template struct { + // Sticky error if escaping fails. + escapeErr error + // We could embed the text/template field, but it's safer not to because + // we need to keep our version of the name space and the underlying + // template's in sync. + text *template.Template + // The underlying template's parse tree, updated to be HTML-safe. + Tree *parse.Tree + *nameSpace // common to all associated templates +} + +// escapeOK is a sentinel value used to indicate valid escaping. +var escapeOK = fmt.Errorf("template escaped correctly") + +// nameSpace is the data structure shared by all templates in an association. +type nameSpace struct { + mu sync.Mutex + set map[string]*Template +} + +// Templates returns a slice of the templates associated with t, including t +// itself. +func (t *Template) Templates() []*Template { + ns := t.nameSpace + ns.mu.Lock() + defer ns.mu.Unlock() + // Return a slice so we don't expose the map. + m := make([]*Template, 0, len(ns.set)) + for _, v := range ns.set { + m = append(m, v) + } + return m +} + +// escape escapes all associated templates. +func (t *Template) escape() error { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + if t.escapeErr == nil { + if err := escapeTemplate(t, t.text.Root, t.Name()); err != nil { + return err + } + } else if t.escapeErr != escapeOK { + return t.escapeErr + } + return nil +} + +// Execute applies a parsed template to the specified data object, +// writing the output to wr. +// If an error occurs executing the template or writing its output, +// execution stops, but partial results may already have been written to +// the output writer. +// A template may be executed safely in parallel. +func (t *Template) Execute(wr io.Writer, data interface{}) error { + if err := t.escape(); err != nil { + return err + } + return t.text.Execute(wr, data) +} + +// ExecuteTemplate applies the template associated with t that has the given +// name to the specified data object and writes the output to wr. +// If an error occurs executing the template or writing its output, +// execution stops, but partial results may already have been written to +// the output writer. +// A template may be executed safely in parallel. +func (t *Template) ExecuteTemplate(wr io.Writer, name string, data interface{}) error { + tmpl, err := t.lookupAndEscapeTemplate(name) + if err != nil { + return err + } + return tmpl.text.Execute(wr, data) +} + +// lookupAndEscapeTemplate guarantees that the template with the given name +// is escaped, or returns an error if it cannot be. It returns the named +// template. +func (t *Template) lookupAndEscapeTemplate(name string) (tmpl *Template, err error) { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + tmpl = t.set[name] + if tmpl == nil { + return nil, fmt.Errorf("html/template: %q is undefined", name) + } + if tmpl.escapeErr != nil && tmpl.escapeErr != escapeOK { + return nil, tmpl.escapeErr + } + if tmpl.text.Tree == nil || tmpl.text.Root == nil { + return nil, fmt.Errorf("html/template: %q is an incomplete template", name) + } + if t.text.Lookup(name) == nil { + panic("html/template internal error: template escaping out of sync") + } + if tmpl.escapeErr == nil { + err = escapeTemplate(tmpl, tmpl.text.Root, name) + } + return tmpl, err +} + +// Parse parses a string into a template. Nested template definitions +// will be associated with the top-level template t. Parse may be +// called multiple times to parse definitions of templates to associate +// with t. It is an error if a resulting template is non-empty (contains +// content other than template definitions) and would replace a +// non-empty template with the same name. (In multiple calls to Parse +// with the same receiver template, only one call can contain text +// other than space, comments, and template definitions.) +func (t *Template) Parse(src string) (*Template, error) { + t.nameSpace.mu.Lock() + t.escapeErr = nil + t.nameSpace.mu.Unlock() + ret, err := t.text.Parse(src) + if err != nil { + return nil, err + } + // In general, all the named templates might have changed underfoot. + // Regardless, some new ones may have been defined. + // The template.Template set has been updated; update ours. + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + for _, v := range ret.Templates() { + name := v.Name() + tmpl := t.set[name] + if tmpl == nil { + tmpl = t.new(name) + } + // Restore our record of this text/template to its unescaped original state. + tmpl.escapeErr = nil + tmpl.text = v + tmpl.Tree = v.Tree + } + return t, nil +} + +// AddParseTree creates a new template with the name and parse tree +// and associates it with t. +// +// It returns an error if t has already been executed. +func (t *Template) AddParseTree(name string, tree *parse.Tree) (*Template, error) { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + if t.escapeErr != nil { + return nil, fmt.Errorf("html/template: cannot AddParseTree to %q after it has executed", t.Name()) + } + text, err := t.text.AddParseTree(name, tree) + if err != nil { + return nil, err + } + ret := &Template{ + nil, + text, + text.Tree, + t.nameSpace, + } + t.set[name] = ret + return ret, nil +} + +// Clone returns a duplicate of the template, including all associated +// templates. The actual representation is not copied, but the name space of +// associated templates is, so further calls to Parse in the copy will add +// templates to the copy but not to the original. Clone can be used to prepare +// common templates and use them with variant definitions for other templates +// by adding the variants after the clone is made. +// +// It returns an error if t has already been executed. +func (t *Template) Clone() (*Template, error) { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + if t.escapeErr != nil { + return nil, fmt.Errorf("html/template: cannot Clone %q after it has executed", t.Name()) + } + textClone, err := t.text.Clone() + if err != nil { + return nil, err + } + ret := &Template{ + nil, + textClone, + textClone.Tree, + &nameSpace{ + set: make(map[string]*Template), + }, + } + for _, x := range textClone.Templates() { + name := x.Name() + src := t.set[name] + if src == nil || src.escapeErr != nil { + return nil, fmt.Errorf("html/template: cannot Clone %q after it has executed", t.Name()) + } + x.Tree = x.Tree.Copy() + ret.set[name] = &Template{ + nil, + x, + x.Tree, + ret.nameSpace, + } + } + return ret, nil +} + +// New allocates a new HTML template with the given name. +func New(name string) *Template { + tmpl := &Template{ + nil, + template.New(name), + nil, + &nameSpace{ + set: make(map[string]*Template), + }, + } + tmpl.set[name] = tmpl + return tmpl +} + +// New allocates a new HTML template associated with the given one +// and with the same delimiters. The association, which is transitive, +// allows one template to invoke another with a {{template}} action. +func (t *Template) New(name string) *Template { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + return t.new(name) +} + +// new is the implementation of New, without the lock. +func (t *Template) new(name string) *Template { + tmpl := &Template{ + nil, + t.text.New(name), + nil, + t.nameSpace, + } + tmpl.set[name] = tmpl + return tmpl +} + +// Name returns the name of the template. +func (t *Template) Name() string { + return t.text.Name() +} + +// FuncMap is the type of the map defining the mapping from names to +// functions. Each function must have either a single return value, or two +// return values of which the second has type error. In that case, if the +// second (error) argument evaluates to non-nil during execution, execution +// terminates and Execute returns that error. FuncMap has the same base type +// as FuncMap in "text/template", copied here so clients need not import +// "text/template". +type FuncMap map[string]interface{} + +// Funcs adds the elements of the argument map to the template's function map. +// It panics if a value in the map is not a function with appropriate return +// type. However, it is legal to overwrite elements of the map. The return +// value is the template, so calls can be chained. +func (t *Template) Funcs(funcMap FuncMap) *Template { + t.text.Funcs(template.FuncMap(funcMap)) + return t +} + +// Delims sets the action delimiters to the specified strings, to be used in +// subsequent calls to Parse, ParseFiles, or ParseGlob. Nested template +// definitions will inherit the settings. An empty delimiter stands for the +// corresponding default: {{ or }}. +// The return value is the template, so calls can be chained. +func (t *Template) Delims(left, right string) *Template { + t.text.Delims(left, right) + return t +} + +// Lookup returns the template with the given name that is associated with t, +// or nil if there is no such template. +func (t *Template) Lookup(name string) *Template { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + return t.set[name] +} + +// Must is a helper that wraps a call to a function returning (*Template, error) +// and panics if the error is non-nil. It is intended for use in variable initializations +// such as +// var t = template.Must(template.New("name").Parse("html")) +func Must(t *Template, err error) *Template { + if err != nil { + panic(err) + } + return t +} + +// ParseFiles creates a new Template and parses the template definitions from +// the named files. The returned template's name will have the (base) name and +// (parsed) contents of the first file. There must be at least one file. +// If an error occurs, parsing stops and the returned *Template is nil. +func ParseFiles(filenames ...string) (*Template, error) { + return parseFiles(nil, filenames...) +} + +// ParseFiles parses the named files and associates the resulting templates with +// t. If an error occurs, parsing stops and the returned template is nil; +// otherwise it is t. There must be at least one file. +func (t *Template) ParseFiles(filenames ...string) (*Template, error) { + return parseFiles(t, filenames...) +} + +// parseFiles is the helper for the method and function. If the argument +// template is nil, it is created from the first file. +func parseFiles(t *Template, filenames ...string) (*Template, error) { + if len(filenames) == 0 { + // Not really a problem, but be consistent. + return nil, fmt.Errorf("html/template: no files named in call to ParseFiles") + } + for _, filename := range filenames { + b, err := ioutil.ReadFile(filename) + if err != nil { + return nil, err + } + s := string(b) + // name := filepath.Base(filename) + name := filename + // First template becomes return value if not already defined, + // and we use that one for subsequent New calls to associate + // all the templates together. Also, if this file has the same name + // as t, this file becomes the contents of t, so + // t, err := New(name).Funcs(xxx).ParseFiles(name) + // works. Otherwise we create a new template associated with t. + var tmpl *Template + if t == nil { + t = New(name) + } + if name == t.Name() { + tmpl = t + } else { + tmpl = t.New(name) + } + _, err = tmpl.Parse(s) + if err != nil { + return nil, err + } + } + return t, nil +} + +// ParseGlob creates a new Template and parses the template definitions from the +// files identified by the pattern, which must match at least one file. The +// returned template will have the (base) name and (parsed) contents of the +// first file matched by the pattern. ParseGlob is equivalent to calling +// ParseFiles with the list of files matched by the pattern. +func ParseGlob(pattern string) (*Template, error) { + return parseGlob(nil, pattern) +} + +// ParseGlob parses the template definitions in the files identified by the +// pattern and associates the resulting templates with t. The pattern is +// processed by filepath.Glob and must match at least one file. ParseGlob is +// equivalent to calling t.ParseFiles with the list of files matched by the +// pattern. +func (t *Template) ParseGlob(pattern string) (*Template, error) { + return parseGlob(t, pattern) +} + +// parseGlob is the implementation of the function and method ParseGlob. +func parseGlob(t *Template, pattern string) (*Template, error) { + filenames, err := filepath.Glob(pattern) + if err != nil { + return nil, err + } + if len(filenames) == 0 { + return nil, fmt.Errorf("html/template: pattern matches no files: %#q", pattern) + } + return parseFiles(t, filenames...) +} diff --git a/template/transition.go b/template/transition.go new file mode 100644 index 00000000..b486fcd2 --- /dev/null +++ b/template/transition.go @@ -0,0 +1,550 @@ +// 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 template + +import ( + "bytes" + "strings" +) + +// transitionFunc is the array of context transition functions for text nodes. +// A transition function takes a context and template text input, and returns +// the updated context and the number of bytes consumed from the front of the +// input. +var transitionFunc = [...]func(context, []byte) (context, int){ + stateText: tText, + stateTag: tTag, + stateAttrName: tAttrName, + stateAfterName: tAfterName, + stateBeforeValue: tBeforeValue, + stateHTMLCmt: tHTMLCmt, + stateRCDATA: tSpecialTagEnd, + stateAttr: tAttr, + stateURL: tURL, + stateJS: tJS, + stateJSDqStr: tJSDelimited, + stateJSSqStr: tJSDelimited, + stateJSRegexp: tJSDelimited, + stateJSBlockCmt: tBlockCmt, + stateJSLineCmt: tLineCmt, + stateCSS: tCSS, + stateCSSDqStr: tCSSStr, + stateCSSSqStr: tCSSStr, + stateCSSDqURL: tCSSStr, + stateCSSSqURL: tCSSStr, + stateCSSURL: tCSSStr, + stateCSSBlockCmt: tBlockCmt, + stateCSSLineCmt: tLineCmt, + stateError: tError, +} + +var commentStart = []byte("<!--") +var commentEnd = []byte("-->") + +// tText is the context transition function for the text state. +func tText(c context, s []byte) (context, int) { + k := 0 + for { + i := k + bytes.IndexByte(s[k:], '<') + if i < k || i+1 == len(s) { + return c, len(s) + } else if i+4 <= len(s) && bytes.Equal(commentStart, s[i:i+4]) { + return context{state: stateHTMLCmt}, i + 4 + } + i++ + end := false + if s[i] == '/' { + if i+1 == len(s) { + return c, len(s) + } + 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 + } + k = j + } +} + +var elementContentType = [...]state{ + elementNone: stateText, + elementScript: stateJS, + elementStyle: stateCSS, + elementTextarea: stateRCDATA, + elementTitle: stateRCDATA, +} + +// tTag is the context transition function for the tag state. +func tTag(c context, s []byte) (context, int) { + // Find the attribute name. + i := eatWhiteSpace(s, 0) + if i == len(s) { + return c, len(s) + } + if s[i] == '>' { + return context{ + state: elementContentType[c.element], + element: c.element, + }, i + 1 + } + j, err := eatAttrName(s, i) + if err != nil { + return context{state: stateError, err: err}, len(s) + } + state, attr := stateTag, attrNone + if i == j { + return context{ + state: stateError, + err: errorf(ErrBadHTML, nil, 0, "expected space, attr name, or end of tag, but got %q", s[i:]), + }, len(s) + } + switch attrType(string(s[i:j])) { + case contentTypeURL: + attr = attrURL + case contentTypeCSS: + attr = attrStyle + case contentTypeJS: + attr = attrScript + } + if j == len(s) { + state = stateAttrName + } else { + state = stateAfterName + } + return context{state: state, element: c.element, attr: attr}, j +} + +// tAttrName is the context transition function for stateAttrName. +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) { + c.state = stateAfterName + } + return c, i +} + +// tAfterName is the context transition function for stateAfterName. +func tAfterName(c context, s []byte) (context, int) { + // Look for the start of the value. + i := eatWhiteSpace(s, 0) + if i == len(s) { + return c, len(s) + } else if s[i] != '=' { + // Occurs due to tag ending '>', and valueless attribute. + c.state = stateTag + return c, i + } + c.state = stateBeforeValue + // Consume the "=". + return c, i + 1 +} + +var attrStartStates = [...]state{ + attrNone: stateAttr, + attrScript: stateJS, + attrStyle: stateCSS, + attrURL: stateURL, +} + +// tBeforeValue is the context transition function for stateBeforeValue. +func tBeforeValue(c context, s []byte) (context, int) { + i := eatWhiteSpace(s, 0) + if i == len(s) { + return c, len(s) + } + // Find the attribute delimiter. + delim := delimSpaceOrTagEnd + switch s[i] { + case '\'': + delim, i = delimSingleQuote, i+1 + case '"': + delim, i = delimDoubleQuote, i+1 + } + c.state, c.delim, c.attr = attrStartStates[c.attr], delim, attrNone + return c, i +} + +// tHTMLCmt is the context transition function for stateHTMLCmt. +func tHTMLCmt(c context, s []byte) (context, int) { + if i := bytes.Index(s, commentEnd); i != -1 { + return context{}, i + 3 + } + return c, len(s) +} + +// specialTagEndMarkers maps element types to the character sequence that +// case-insensitively signals the end of the special tag body. +var specialTagEndMarkers = [...]string{ + elementScript: "</script", + elementStyle: "</style", + elementTextarea: "</textarea", + elementTitle: "</title", +} + +// tSpecialTagEnd is the context transition function for raw text and RCDATA +// element states. +func tSpecialTagEnd(c context, s []byte) (context, int) { + if c.element != elementNone { + if i := strings.Index(strings.ToLower(string(s)), specialTagEndMarkers[c.element]); i != -1 { + return context{}, i + } + } + return c, len(s) +} + +// tAttr is the context transition function for the attribute state. +func tAttr(c context, s []byte) (context, int) { + return c, len(s) +} + +// tURL is the context transition function for the URL state. +func tURL(c context, s []byte) (context, int) { + if bytes.IndexAny(s, "#?") >= 0 { + c.urlPart = urlPartQueryOrFrag + } else if len(s) != eatWhiteSpace(s, 0) && c.urlPart == urlPartNone { + // HTML5 uses "Valid URL potentially surrounded by spaces" for + // attrs: http://www.w3.org/TR/html5/index.html#attributes-1 + c.urlPart = urlPartPreQuery + } + return c, len(s) +} + +// tJS is the context transition function for the JS state. +func tJS(c context, s []byte) (context, int) { + i := bytes.IndexAny(s, `"'/`) + if i == -1 { + // Entire input is non string, comment, regexp tokens. + c.jsCtx = nextJSCtx(s, c.jsCtx) + return c, len(s) + } + c.jsCtx = nextJSCtx(s[:i], c.jsCtx) + switch s[i] { + case '"': + c.state, c.jsCtx = stateJSDqStr, jsCtxRegexp + case '\'': + c.state, c.jsCtx = stateJSSqStr, jsCtxRegexp + case '/': + switch { + case i+1 < len(s) && s[i+1] == '/': + c.state, i = stateJSLineCmt, i+1 + case i+1 < len(s) && s[i+1] == '*': + c.state, i = stateJSBlockCmt, i+1 + case c.jsCtx == jsCtxRegexp: + c.state = stateJSRegexp + case c.jsCtx == jsCtxDivOp: + c.jsCtx = jsCtxRegexp + default: + return context{ + state: stateError, + err: errorf(ErrSlashAmbig, nil, 0, "'/' could start a division or regexp: %.32q", s[i:]), + }, len(s) + } + default: + panic("unreachable") + } + return c, i + 1 +} + +// 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:], specials) + if i < k { + break + } + switch s[i] { + case '\\': + i++ + if i == len(s) { + return context{ + state: stateError, + err: errorf(ErrPartialEscape, nil, 0, "unfinished escape sequence in JS string: %q", s), + }, len(s) + } + case '[': + inCharset = true + case ']': + inCharset = false + default: + // end delimiter + if !inCharset { + c.state, c.jsCtx = stateJS, jsCtxDivOp + return c, i + 1 + } + } + k = i + 1 + } + + if inCharset { + // This can be fixed by making context richer if interpolation + // into charsets is desired. + return context{ + state: stateError, + err: errorf(ErrPartialCharset, nil, 0, "unfinished JS regexp charset: %q", s), + }, len(s) + } + + return c, len(s) +} + +var blockCommentEnd = []byte("*/") + +// tBlockCmt is the context transition function for /*comment*/ states. +func tBlockCmt(c context, s []byte) (context, int) { + i := bytes.Index(s, blockCommentEnd) + if i == -1 { + return c, len(s) + } + switch c.state { + case stateJSBlockCmt: + c.state = stateJS + case stateCSSBlockCmt: + c.state = stateCSS + default: + panic(c.state.String()) + } + return c, i + 2 +} + +// tLineCmt is the context transition function for //comment states. +func tLineCmt(c context, s []byte) (context, int) { + 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, lineTerminators) + if i == -1 { + return c, len(s) + } + 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 + // recognized separately by the lexical grammar and becomes part + // of the stream of input elements for the syntactic grammar." + return c, i +} + +// tCSS is the context transition function for the CSS state. +func tCSS(c context, s []byte) (context, int) { + // 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. + + k := 0 + for { + i := k + bytes.IndexAny(s[k:], `("'/`) + if i < k { + return c, len(s) + } + switch s[i] { + case '(': + // Look for url to the left. + p := bytes.TrimRight(s[:i], "\t\n\f\r ") + if endsWithCSSKeyword(p, "url") { + j := len(s) - len(bytes.TrimLeft(s[i+1:], "\t\n\f\r ")) + switch { + case j != len(s) && s[j] == '"': + c.state, j = stateCSSDqURL, j+1 + case j != len(s) && s[j] == '\'': + c.state, j = stateCSSSqURL, j+1 + default: + c.state = stateCSSURL + } + return c, j + } + case '/': + if i+1 < len(s) { + switch s[i+1] { + case '/': + c.state = stateCSSLineCmt + return c, i + 2 + case '*': + c.state = stateCSSBlockCmt + return c, i + 2 + } + } + case '"': + c.state = stateCSSDqStr + return c, i + 1 + case '\'': + c.state = stateCSSSqStr + return c, i + 1 + } + k = i + 1 + } +} + +// tCSSStr is the context transition function for the CSS string and URL states. +func tCSSStr(c context, s []byte) (context, int) { + 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()) + } + + k := 0 + for { + i := k + bytes.IndexAny(s[k:], endAndEsc) + if i < k { + c, nread := tURL(c, decodeCSS(s[k:])) + return c, k + nread + } + if s[i] == '\\' { + i++ + if i == len(s) { + return context{ + state: stateError, + err: errorf(ErrPartialEscape, nil, 0, "unfinished escape sequence in CSS string: %q", s), + }, len(s) + } + } else { + c.state = stateCSS + return c, i + 1 + } + c, _ = tURL(c, decodeCSS(s[:i+1])) + k = i + 1 + } +} + +// tError is the context transition function for the error state. +func tError(c context, s []byte) (context, int) { + return c, len(s) +} + +// eatAttrName returns the largest j such that s[i:j] is an attribute name. +// It returns an error if s[i:] does not look like it begins with an +// attribute name, such as encountering a quote mark without a preceding +// equals sign. +func eatAttrName(s []byte, i int) (int, *Error) { + for j := i; j < len(s); j++ { + switch s[j] { + case ' ', '\t', '\n', '\f', '\r', '=', '>': + return j, nil + case '\'', '"', '<': + // These result in a parse warning in HTML5 and are + // indicative of serious problems if seen in an attr + // name in a template. + return -1, errorf(ErrBadHTML, nil, 0, "%q in attribute name: %.32q", s[j:j+1], s) + default: + // No-op. + } + } + return len(s), nil +} + +var elementNameMap = map[string]element{ + "script": elementScript, + "style": elementStyle, + "textarea": elementTextarea, + "title": elementTitle, +} + +// asciiAlpha reports whether c is an ASCII letter. +func asciiAlpha(c byte) bool { + return 'A' <= c && c <= 'Z' || 'a' <= c && c <= 'z' +} + +// asciiAlphaNum reports whether c is an ASCII letter or digit. +func asciiAlphaNum(c byte) bool { + return asciiAlpha(c) || '0' <= c && c <= '9' +} + +// eatTagName returns the largest j such that s[i:j] is a tag name and the tag type. +func eatTagName(s []byte, i int) (int, element) { + if i == len(s) || !asciiAlpha(s[i]) { + return i, elementNone + } + j := i + 1 + for j < len(s) { + x := s[j] + if asciiAlphaNum(x) { + j++ + continue + } + // Allow "x-y" or "x:y" but not "x-", "-y", or "x--y". + if (x == ':' || x == '-') && j+1 < len(s) && asciiAlphaNum(s[j+1]) { + j += 2 + continue + } + break + } + return j, elementNameMap[strings.ToLower(string(s[i:j]))] +} + +// eatWhiteSpace returns the largest j such that s[i:j] is white space. +func eatWhiteSpace(s []byte, i int) int { + for j := i; j < len(s); j++ { + switch s[j] { + case ' ', '\t', '\n', '\f', '\r': + // No-op. + default: + return j + } + } + return len(s) +} diff --git a/template/url.go b/template/url.go new file mode 100644 index 00000000..2ca76bf3 --- /dev/null +++ b/template/url.go @@ -0,0 +1,105 @@ +// 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 template + +import ( + "bytes" + "fmt" + "strings" +) + +// urlFilter returns its input unless it contains an unsafe protocol in which +// case it defangs the entire URL. +func urlFilter(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeURL { + return s + } + if i := strings.IndexRune(s, ':'); i >= 0 && strings.IndexRune(s[:i], '/') < 0 { + protocol := strings.ToLower(s[:i]) + if protocol != "http" && protocol != "https" && protocol != "mailto" { + return "#" + filterFailsafe + } + } + 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 '&'. +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, t := stringify(args...) + if t == contentTypeURL { + norm = true + } + 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 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() +} diff --git a/template/url_test.go b/template/url_test.go new file mode 100644 index 00000000..5182e9d7 --- /dev/null +++ b/template/url_test.go @@ -0,0 +1,112 @@ +// 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 template + +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") + } +}
`, + }, + { + "stylePropertyPairPassed", + ``, + ``, + }, + { + "styleStrSpecialsEncoded", + ``, + ``, + }, + { + "styleURLSpecialsEncoded", + ``, + ``, + }, + { + "HTML comment", + "Hello, {{.C}}", + "Hello, <Cincinatti>", + }, + { + "HTML comment not first < in text node.", + "<!--", + "<!--", + }, + { + "HTML normalization 1", + "a < b", + "a < b", + }, + { + "HTML normalization 2", + "a << b", + "a << b", + }, + { + "HTML normalization 3", + "a<b", + "a<b", + }, + { + "HTML doctype not normalized", + "Hello, World!", + "Hello, World!", + }, + { + "HTML doctype not case-insensitive", + "Hello, World!", + "Hello, World!", + }, + { + "No doctype injection", + `Hello, {{.C}}{{else}}world -->{{.W}}{{end}}", + "Hello, <Cincinatti>", + }, + { + "JS line comment", + "", + "", + }, + { + "JS multiline block comment", + "", + // Newline separates break from call. If newline + // removed, then break will consume label leaving + // code invalid. + "", + }, + { + "JS single-line block comment", + "", + // Newline separates break from call. If newline + // removed, then break will consume label leaving + // code invalid. + "", + }, + { + "JS block comment flush with mathematical division", + "", + "", + }, + { + "JS mixed comments", + "", + "", + }, + { + "CSS comments", + "`, + "", + }, + { + "JS attr block comment", + ``, + // Attribute comment tests should pass if the comments + // are successfully elided. + ``, + }, + { + "JS attr line comment", + ``, + ``, + }, + { + "CSS attr block comment", + ``, + ``, + }, + { + "CSS attr line comment", + ``, + ``, + }, + { + "HTML substitution commented out", + "", + "", + }, + { + "Comment ends flush with start", + "", + "", + }, + { + "typed HTML in text", + `{{.W}}`, + `¡Hello, O'World!`, + }, + { + "typed HTML in attribute", + ``, + ``, + }, + { + "typed HTML in script", + ``, + ``, + }, + { + "typed HTML in RCDATA", + `{{.W}}`, + `¡<b class="foo">Hello</b>, <textarea>O'World</textarea>!`, + }, + { + "range in textarea", + "{{range .A}}{{.}}{{end}}", + "<a><b>", + }, + { + "No tag injection", + `{{"10$"}}<{{"script src,evil.org/pwnd.js"}}...`, + `10$<script src,evil.org/pwnd.js...`, + }, + { + "No comment injection", + `<{{"!--"}}`, + `<!--`, + }, + { + "No RCDATA end tag injection", + `<{{"/textarea "}}...`, + `</textarea ...`, + }, + { + "optional attrs", + `"}}"{{end}}` + + // Double quotes inside if/else. + ` src=` + + `{{if .T}}"?{{""}}"` + + `{{else}}"images/cleardot.gif"{{end}}` + + // Missing space before title, but it is not a + // part of the src attribute. + `{{if .T}}title="{{""}}"{{end}}` + + // Quotes outside if/else. + ` alt="` + + `{{if .T}}{{"<alt>"}}` + + `{{else}}{{if .F}}{{"<title>"}}{{end}}` + + `{{end}}"` + + `>`, + `<img class="iconClass" id="<iconId>" src="?%3ciconPath%3e"title="<title>" alt="<alt>">`, + }, + { + "conditional valueless attr name", + `<input{{if .T}} checked{{end}} name=n>`, + `<input checked name=n>`, + }, + { + "conditional dynamic valueless attr name 1", + `<input{{if .T}} {{"checked"}}{{end}} name=n>`, + `<input checked name=n>`, + }, + { + "conditional dynamic valueless attr name 2", + `<input {{if .T}}{{"checked"}} {{end}}name=n>`, + `<input checked name=n>`, + }, + { + "dynamic attribute name", + `<img on{{"load"}}="alert({{"loaded"}})">`, + // Treated as JS since quotes are inserted. + `<img onload="alert("loaded")">`, + }, + { + "bad dynamic attribute name 1", + // Allow checked, selected, disabled, but not JS or + // CSS attributes. + `<input {{"onchange"}}="{{"doEvil()"}}">`, + `<input ZgotmplZ="doEvil()">`, + }, + { + "bad dynamic attribute name 2", + `<div {{"sTyle"}}="{{"color: expression(alert(1337))"}}">`, + `<div ZgotmplZ="color: expression(alert(1337))">`, + }, + { + "bad dynamic attribute name 3", + // Allow title or alt, but not a URL. + `<img {{"src"}}="{{"javascript:doEvil()"}}">`, + `<img ZgotmplZ="javascript:doEvil()">`, + }, + { + "bad dynamic attribute name 4", + // Structure preservation requires values to associate + // with a consistent attribute. + `<input checked {{""}}="Whose value am I?">`, + `<input checked ZgotmplZ="Whose value am I?">`, + }, + { + "dynamic element name", + `<h{{3}}><table><t{{"head"}}>...</h{{3}}>`, + `<h3><table><thead>...</h3>`, + }, + { + "bad dynamic element name", + // Dynamic element names are typically used to switch + // between (thead, tfoot, tbody), (ul, ol), (th, td), + // and other replaceable sets. + // We do not currently easily support (ul, ol). + // If we do change to support that, this test should + // catch failures to filter out special tag names which + // would violate the structure preservation property -- + // if any special tag name could be substituted, then + // the content could be raw text/RCDATA for some inputs + // and regular HTML content for others. + `<{{"script"}}>{{"doEvil()"}}</{{"script"}}>`, + `<script>doEvil()</script>`, + }, + } + + for _, test := range tests { + tmpl := New(test.name) + tmpl = Must(tmpl.Parse(test.input)) + // Check for bug 6459: Tree field was not set in Parse. + if tmpl.Tree != tmpl.text.Tree { + t.Errorf("%s: tree not set properly", test.name) + continue + } + b := new(bytes.Buffer) + if err := tmpl.Execute(b, data); err != nil { + t.Errorf("%s: template execution failed: %s", test.name, err) + continue + } + if w, g := test.output, b.String(); w != g { + t.Errorf("%s: escaped output: want\n\t%q\ngot\n\t%q", test.name, w, g) + continue + } + b.Reset() + if err := tmpl.Execute(b, pdata); err != nil { + t.Errorf("%s: template execution failed for pointer: %s", test.name, err) + continue + } + if w, g := test.output, b.String(); w != g { + t.Errorf("%s: escaped output for pointer: want\n\t%q\ngot\n\t%q", test.name, w, g) + continue + } + if tmpl.Tree != tmpl.text.Tree { + t.Errorf("%s: tree mismatch", test.name) + continue + } + } +} + +func TestEscapeSet(t *testing.T) { + type dataItem struct { + Children []*dataItem + X string + } + + data := dataItem{ + Children: []*dataItem{ + {X: "foo"}, + {X: "<bar>"}, + { + Children: []*dataItem{ + {X: "baz"}, + }, + }, + }, + } + + tests := []struct { + inputs map[string]string + want string + }{ + // The trivial set. + { + map[string]string{ + "main": ``, + }, + ``, + }, + // A template called in the start context. + { + map[string]string{ + "main": `Hello, {{template "helper"}}!`, + // Not a valid top level HTML template. + // "<b" is not a full tag. + "helper": `{{"<World>"}}`, + }, + `Hello, <World>!`, + }, + // A template called in a context other than the start. + { + map[string]string{ + "main": `<a onclick='a = {{template "helper"}};'>`, + // Not a valid top level HTML template. + // "<b" is not a full tag. + "helper": `{{"<a>"}}<b`, + }, + `<a onclick='a = "\u003ca\u003e"<b;'>`, + }, + // A recursive template that ends in its start context. + { + map[string]string{ + "main": `{{range .Children}}{{template "main" .}}{{else}}{{.X}} {{end}}`, + }, + `foo <bar> baz `, + }, + // A recursive helper template that ends in its start context. + { + map[string]string{ + "main": `{{template "helper" .}}`, + "helper": `{{if .Children}}<ul>{{range .Children}}<li>{{template "main" .}}</li>{{end}}</ul>{{else}}{{.X}}{{end}}`, + }, + `<ul><li>foo</li><li><bar></li><li><ul><li>baz</li></ul></li></ul>`, + }, + // Co-recursive templates that end in its start context. + { + map[string]string{ + "main": `<blockquote>{{range .Children}}{{template "helper" .}}{{end}}</blockquote>`, + "helper": `{{if .Children}}{{template "main" .}}{{else}}{{.X}}<br>{{end}}`, + }, + `<blockquote>foo<br><bar><br><blockquote>baz<br></blockquote></blockquote>`, + }, + // A template that is called in two different contexts. + { + map[string]string{ + "main": `<button onclick="title='{{template "helper"}}'; ...">{{template "helper"}}</button>`, + "helper": `{{11}} of {{"<100>"}}`, + }, + `<button onclick="title='11 of \x3c100\x3e'; ...">11 of <100></button>`, + }, + // A non-recursive template that ends in a different context. + // helper starts in jsCtxRegexp and ends in jsCtxDivOp. + { + map[string]string{ + "main": `<script>var x={{template "helper"}}/{{"42"}};</script>`, + "helper": "{{126}}", + }, + `<script>var x= 126 /"42";</script>`, + }, + // A recursive template that ends in a similar context. + { + map[string]string{ + "main": `<script>var x=[{{template "countdown" 4}}];</script>`, + "countdown": `{{.}}{{if .}},{{template "countdown" . | pred}}{{end}}`, + }, + `<script>var x=[ 4 , 3 , 2 , 1 , 0 ];</script>`, + }, + // A recursive template that ends in a different context. + /* + { + map[string]string{ + "main": `<a href="/foo{{template "helper" .}}">`, + "helper": `{{if .Children}}{{range .Children}}{{template "helper" .}}{{end}}{{else}}?x={{.X}}{{end}}`, + }, + `<a href="/foo?x=foo?x=%3cbar%3e?x=baz">`, + }, + */ + } + + // pred is a template function that returns the predecessor of a + // natural number for testing recursive templates. + fns := FuncMap{"pred": func(a ...interface{}) (interface{}, error) { + if len(a) == 1 { + if i, _ := a[0].(int); i > 0 { + return i - 1, nil + } + } + return nil, fmt.Errorf("undefined pred(%v)", a) + }} + + for _, test := range tests { + source := "" + for name, body := range test.inputs { + source += fmt.Sprintf("{{define %q}}%s{{end}} ", name, body) + } + tmpl, err := New("root").Funcs(fns).Parse(source) + if err != nil { + t.Errorf("error parsing %q: %v", source, err) + continue + } + var b bytes.Buffer + + if err := tmpl.ExecuteTemplate(&b, "main", data); err != nil { + t.Errorf("%q executing %v", err.Error(), tmpl.Lookup("main")) + continue + } + if got := b.String(); test.want != got { + t.Errorf("want\n\t%q\ngot\n\t%q", test.want, got) + } + } + +} + +func TestErrors(t *testing.T) { + tests := []struct { + input string + err string + }{ + // Non-error cases. + { + "{{if .Cond}}<a>{{else}}<b>{{end}}", + "", + }, + { + "{{if .Cond}}<a>{{end}}", + "", + }, + { + "{{if .Cond}}{{else}}<b>{{end}}", + "", + }, + { + "{{with .Cond}}<div>{{end}}", + "", + }, + { + "{{range .Items}}<a>{{end}}", + "", + }, + { + "<a href='/foo?{{range .Items}}&{{.K}}={{.V}}{{end}}'>", + "", + }, + // Error cases. + { + "{{if .Cond}}<a{{end}}", + "z:1:5: {{if}} branches", + }, + { + "{{if .Cond}}\n{{else}}\n<a{{end}}", + "z:1:5: {{if}} branches", + }, + { + // Missing quote in the else branch. + `{{if .Cond}}<a href="foo">{{else}}<a href="bar>{{end}}`, + "z:1:5: {{if}} branches", + }, + { + // Different kind of attribute: href implies a URL. + "<a {{if .Cond}}href='{{else}}title='{{end}}{{.X}}'>", + "z:1:8: {{if}} branches", + }, + { + "\n{{with .X}}<a{{end}}", + "z:2:7: {{with}} branches", + }, + { + "\n{{with .X}}<a>{{else}}<a{{end}}", + "z:2:7: {{with}} branches", + }, + { + "{{range .Items}}<a{{end}}", + `z:1: on range loop re-entry: "<" in attribute name: "<a"`, + }, + { + "\n{{range .Items}} x='<a{{end}}", + "z:2:8: on range loop re-entry: {{range}} branches", + }, + { + "<a b=1 c={{.H}}", + "z: ends in a non-text context: {stateAttr delimSpaceOrTagEnd", + }, + { + "<script>foo();", + "z: ends in a non-text context: {stateJS", + }, + { + `<a href="{{if .F}}/foo?a={{else}}/bar/{{end}}{{.H}}">`, + "z:1:47: {{.H}} appears in an ambiguous URL context", + }, + { + `<a onclick="alert('Hello \`, + `unfinished escape sequence in JS string: "Hello \\"`, + }, + { + `<a onclick='alert("Hello\, World\`, + `unfinished escape sequence in JS string: "Hello\\, World\\"`, + }, + { + `<a onclick='alert(/x+\`, + `unfinished escape sequence in JS string: "x+\\"`, + }, + { + `<a onclick="/foo[\]/`, + `unfinished JS regexp charset: "foo[\\]/"`, + }, + { + // It is ambiguous whether 1.5 should be 1\.5 or 1.5. + // Either `var x = 1/- 1.5 /i.test(x)` + // where `i.test(x)` is a method call of reference i, + // or `/-1\.5/i.test(x)` which is a method call on a + // case insensitive regular expression. + `<script>{{if false}}var x = 1{{end}}/-{{"1.5"}}/i.test(x)</script>`, + `'/' could start a division or regexp: "/-"`, + }, + { + `{{template "foo"}}`, + "z:1:11: no such template \"foo\"", + }, + { + `<div{{template "y"}}>` + + // Illegal starting in stateTag but not in stateText. + `{{define "y"}} foo<b{{end}}`, + `"<" in attribute name: " foo<b"`, + }, + { + `<script>reverseList = [{{template "t"}}]</script>` + + // Missing " after recursive call. + `{{define "t"}}{{if .Tail}}{{template "t" .Tail}}{{end}}{{.Head}}",{{end}}`, + `: cannot compute output context for template t$htmltemplate_stateJS_elementScript`, + }, + { + `<input type=button value=onclick=>`, + `html/template:z: "=" in unquoted attr: "onclick="`, + }, + { + `<input type=button value= onclick=>`, + `html/template:z: "=" in unquoted attr: "onclick="`, + }, + { + `<input type=button value= 1+1=2>`, + `html/template:z: "=" in unquoted attr: "1+1=2"`, + }, + { + "<a class=`foo>", + "html/template:z: \"`\" in unquoted attr: \"`foo\"", + }, + { + `<a style=font:'Arial'>`, + `html/template:z: "'" in unquoted attr: "font:'Arial'"`, + }, + { + `<a=foo>`, + `: expected space, attr name, or end of tag, but got "=foo>"`, + }, + } + + for _, test := range tests { + buf := new(bytes.Buffer) + tmpl, err := New("z").Parse(test.input) + if err != nil { + t.Errorf("input=%q: unexpected parse error %s\n", test.input, err) + continue + } + err = tmpl.Execute(buf, nil) + var got string + if err != nil { + got = err.Error() + } + if test.err == "" { + if got != "" { + t.Errorf("input=%q: unexpected error %q", test.input, got) + } + continue + } + if strings.Index(got, test.err) == -1 { + t.Errorf("input=%q: error\n\t%q\ndoes not contain expected string\n\t%q", test.input, got, test.err) + continue + } + // Check that we get the same error if we call Execute again. + if err := tmpl.Execute(buf, nil); err == nil || err.Error() != got { + t.Errorf("input=%q: unexpected error on second call %q", test.input, err) + + } + } +} + +func TestEscapeText(t *testing.T) { + tests := []struct { + input string + output context + }{ + { + ``, + context{}, + }, + { + `Hello, World!`, + context{}, + }, + { + // An orphaned "<" is OK. + `I <3 Ponies!`, + context{}, + }, + { + `<a`, + context{state: stateTag}, + }, + { + `<a `, + context{state: stateTag}, + }, + { + `<a>`, + context{state: stateText}, + }, + { + `<a href`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a on`, + context{state: stateAttrName, attr: attrScript}, + }, + { + `<a href `, + context{state: stateAfterName, attr: attrURL}, + }, + { + `<a style = `, + context{state: stateBeforeValue, attr: attrStyle}, + }, + { + `<a href=`, + context{state: stateBeforeValue, attr: attrURL}, + }, + { + `<a href=x`, + context{state: stateURL, delim: delimSpaceOrTagEnd, urlPart: urlPartPreQuery}, + }, + { + `<a href=x `, + context{state: stateTag}, + }, + { + `<a href=>`, + context{state: stateText}, + }, + { + `<a href=x>`, + context{state: stateText}, + }, + { + `<a href ='`, + context{state: stateURL, delim: delimSingleQuote}, + }, + { + `<a href=''`, + context{state: stateTag}, + }, + { + `<a href= "`, + context{state: stateURL, delim: delimDoubleQuote}, + }, + { + `<a href=""`, + context{state: stateTag}, + }, + { + `<a title="`, + context{state: stateAttr, delim: delimDoubleQuote}, + }, + { + `<a HREF='http:`, + context{state: stateURL, delim: delimSingleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a Href='/`, + context{state: stateURL, delim: delimSingleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href='"`, + context{state: stateURL, delim: delimSingleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href="'`, + context{state: stateURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href=''`, + context{state: stateURL, delim: delimSingleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href=""`, + context{state: stateURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href=""`, + context{state: stateURL, delim: delimDoubleQuote, urlPart: urlPartPreQuery}, + }, + { + `<a href="`, + context{state: stateURL, delim: delimSpaceOrTagEnd, urlPart: urlPartPreQuery}, + }, + { + `<img alt="1">`, + context{state: stateText}, + }, + { + `<img alt="1>"`, + context{state: stateTag}, + }, + { + `<img alt="1>">`, + context{state: stateText}, + }, + { + `<input checked type="checkbox"`, + context{state: stateTag}, + }, + { + `<a onclick="`, + context{state: stateJS, delim: delimDoubleQuote}, + }, + { + `<a onclick="//foo`, + context{state: stateJSLineCmt, delim: delimDoubleQuote}, + }, + { + "<a onclick='//\n", + context{state: stateJS, delim: delimSingleQuote}, + }, + { + "<a onclick='//\r\n", + context{state: stateJS, delim: delimSingleQuote}, + }, + { + "<a onclick='//\u2028", + context{state: stateJS, delim: delimSingleQuote}, + }, + { + `<a onclick="/*`, + context{state: stateJSBlockCmt, delim: delimDoubleQuote}, + }, + { + `<a onclick="/*/`, + context{state: stateJSBlockCmt, delim: delimDoubleQuote}, + }, + { + `<a onclick="/**/`, + context{state: stateJS, delim: delimDoubleQuote}, + }, + { + `<a onkeypress=""`, + context{state: stateJSDqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick='"foo"`, + context{state: stateJS, delim: delimSingleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick='foo'`, + context{state: stateJS, delim: delimSpaceOrTagEnd, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick='foo`, + context{state: stateJSSqStr, delim: delimSpaceOrTagEnd}, + }, + { + `<a onclick=""foo'`, + context{state: stateJSDqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick="'foo"`, + context{state: stateJSSqStr, delim: delimDoubleQuote}, + }, + { + `<A ONCLICK="'`, + context{state: stateJSSqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick="/`, + context{state: stateJSRegexp, delim: delimDoubleQuote}, + }, + { + `<a onclick="'foo'`, + context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick="'foo\'`, + context{state: stateJSSqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick="'foo\'`, + context{state: stateJSSqStr, delim: delimDoubleQuote}, + }, + { + `<a onclick="/foo/`, + context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<script>/foo/ /=`, + context{state: stateJS, element: elementScript}, + }, + { + `<a onclick="1 /foo`, + context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick="1 /*c*/ /foo`, + context{state: stateJS, delim: delimDoubleQuote, jsCtx: jsCtxDivOp}, + }, + { + `<a onclick="/foo[/]`, + context{state: stateJSRegexp, delim: delimDoubleQuote}, + }, + { + `<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: "`, + 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("/`, + 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}, + }, + { + `<!-- foo`, + context{state: stateHTMLCmt}, + }, + { + `<!-->`, + context{state: stateHTMLCmt}, + }, + { + `<!--->`, + context{state: stateHTMLCmt}, + }, + { + `<!-- foo -->`, + context{state: stateText}, + }, + { + `<script`, + context{state: stateTag, element: elementScript}, + }, + { + `<script `, + context{state: stateTag, element: elementScript}, + }, + { + `<script src="foo.js" `, + context{state: stateTag, element: elementScript}, + }, + { + `<script src='foo.js' `, + context{state: stateTag, element: elementScript}, + }, + { + `<script type=text/javascript `, + context{state: stateTag, element: elementScript}, + }, + { + `<script>foo`, + context{state: stateJS, jsCtx: jsCtxDivOp, element: elementScript}, + }, + { + `<script>foo</script>`, + context{state: stateText}, + }, + { + `<script>foo</script><!--`, + context{state: stateHTMLCmt}, + }, + { + `<script>document.write("<p>foo</p>");`, + context{state: stateJS, element: elementScript}, + }, + { + `<script>document.write("<p>foo<\/script>");`, + context{state: stateJS, element: elementScript}, + }, + { + `<script>document.write("<script>alert(1)</script>");`, + context{state: stateText}, + }, + { + `<Script>`, + context{state: stateJS, element: elementScript}, + }, + { + `<SCRIPT>foo`, + context{state: stateJS, jsCtx: jsCtxDivOp, element: elementScript}, + }, + { + `<textarea>value`, + context{state: stateRCDATA, element: elementTextarea}, + }, + { + `<textarea>value</TEXTAREA>`, + context{state: stateText}, + }, + { + `<textarea name=html><b`, + context{state: stateRCDATA, element: elementTextarea}, + }, + { + `<title>value`, + context{state: stateRCDATA, element: elementTitle}, + }, + { + `<style>value`, + context{state: stateCSS, element: elementStyle}, + }, + { + `<a xlink:href`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a xmlns`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a xmlns:foo`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a xmlnsxyz`, + context{state: stateAttrName}, + }, + { + `<a data-url`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a data-iconUri`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a data-urlItem`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a g:`, + context{state: stateAttrName}, + }, + { + `<a g:url`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a g:iconUri`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a g:urlItem`, + context{state: stateAttrName, attr: attrURL}, + }, + { + `<a g:value`, + context{state: stateAttrName}, + }, + { + `<a svg:style='`, + context{state: stateCSS, delim: delimSingleQuote}, + }, + { + `<svg:font-face`, + context{state: stateTag}, + }, + { + `<svg:a svg:onclick="`, + context{state: stateJS, delim: delimDoubleQuote}, + }, + } + + for _, test := range tests { + b, e := []byte(test.input), newEscaper(nil) + c := e.escapeText(context{}, &parse.TextNode{NodeType: parse.NodeText, Text: b}) + if !test.output.eq(c) { + t.Errorf("input %q: want context\n\t%v\ngot\n\t%v", test.input, test.output, c) + continue + } + if test.input != string(b) { + t.Errorf("input %q: text node was modified: want %q got %q", test.input, test.input, b) + continue + } + } +} + +func TestEnsurePipelineContains(t *testing.T) { + tests := []struct { + input, output string + ids []string + }{ + { + "{{.X}}", + ".X", + []string{}, + }, + { + "{{.X | html}}", + ".X | html", + []string{}, + }, + { + "{{.X}}", + ".X | html", + []string{"html"}, + }, + { + "{{.X | html}}", + ".X | html | urlquery", + []string{"urlquery"}, + }, + { + "{{.X | html | urlquery}}", + ".X | html | urlquery", + []string{"urlquery"}, + }, + { + "{{.X | html | urlquery}}", + ".X | html | urlquery", + []string{"html", "urlquery"}, + }, + { + "{{.X | html | urlquery}}", + ".X | html | urlquery", + []string{"html"}, + }, + { + "{{.X | urlquery}}", + ".X | html | urlquery", + []string{"html", "urlquery"}, + }, + { + "{{.X | html | print}}", + ".X | urlquery | html | print", + []string{"urlquery", "html"}, + }, + { + "{{($).X | html | print}}", + "($).X | urlquery | html | print", + []string{"urlquery", "html"}, + }, + } + for i, test := range tests { + tmpl := template.Must(template.New("test").Parse(test.input)) + action, ok := (tmpl.Tree.Root.Nodes[0].(*parse.ActionNode)) + if !ok { + t.Errorf("#%d: First node is not an action: %s", i, test.input) + continue + } + pipe := action.Pipe + ensurePipelineContains(pipe, test.ids) + got := pipe.String() + if got != test.output { + t.Errorf("#%d: %s, %v: want\n\t%s\ngot\n\t%s", i, test.input, test.ids, test.output, got) + } + } +} + +func TestEscapeErrorsNotIgnorable(t *testing.T) { + var b bytes.Buffer + tmpl, _ := New("dangerous").Parse("<a") + err := tmpl.Execute(&b, nil) + if err == nil { + t.Errorf("Expected error") + } else if b.Len() != 0 { + t.Errorf("Emitted output despite escaping failure") + } +} + +func TestEscapeSetErrorsNotIgnorable(t *testing.T) { + var b bytes.Buffer + tmpl, err := New("root").Parse(`{{define "t"}}<a{{end}}`) + if err != nil { + t.Errorf("failed to parse set: %q", err) + } + err = tmpl.ExecuteTemplate(&b, "t", nil) + if err == nil { + t.Errorf("Expected error") + } else if b.Len() != 0 { + t.Errorf("Emitted output despite escaping failure") + } +} + +func TestRedundantFuncs(t *testing.T) { + inputs := []interface{}{ + "\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f" + + "\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !"#$%&'()*+,-./` + + `0123456789:;<=>?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ[\]^_` + + "`abcdefghijklmno" + + "pqrstuvwxyz{|}~\x7f" + + "\u00A0\u0100\u2028\u2029\ufeff\ufdec\ufffd\uffff\U0001D11E" + + "&%22\\", + CSS(`a[href =~ "//example.com"]#foo`), + HTML(`Hello, <b>World</b> &tc!`), + HTMLAttr(` dir="ltr"`), + JS(`c && alert("Hello, World!");`), + JSStr(`Hello, World & O'Reilly\x21`), + URL(`greeting=H%69&addressee=(World)`), + } + + for n0, m := range redundantFuncs { + f0 := funcMap[n0].(func(...interface{}) string) + for n1 := range m { + f1 := funcMap[n1].(func(...interface{}) string) + for _, input := range inputs { + want := f0(input) + if got := f1(want); want != got { + t.Errorf("%s %s with %T %q: want\n\t%q,\ngot\n\t%q", n0, n1, input, input, want, got) + } + } + } + } +} + +func TestIndirectPrint(t *testing.T) { + a := 3 + ap := &a + b := "hello" + bp := &b + bpp := &bp + tmpl := Must(New("t").Parse(`{{.}}`)) + var buf bytes.Buffer + err := tmpl.Execute(&buf, ap) + if err != nil { + t.Errorf("Unexpected error: %s", err) + } else if buf.String() != "3" { + t.Errorf(`Expected "3"; got %q`, buf.String()) + } + buf.Reset() + err = tmpl.Execute(&buf, bpp) + if err != nil { + t.Errorf("Unexpected error: %s", err) + } else if buf.String() != "hello" { + t.Errorf(`Expected "hello"; got %q`, buf.String()) + } +} + +// This is a test for issue 3272. +func TestEmptyTemplate(t *testing.T) { + page := Must(New("page").ParseFiles(os.DevNull)) + if err := page.ExecuteTemplate(os.Stdout, "page", "nothing"); err == nil { + t.Fatal("expected error") + } +} + +type Issue7379 int + +func (Issue7379) SomeMethod(x int) string { + return fmt.Sprintf("<%d>", x) +} + +// This is a test for issue 7379: type assertion error caused panic, and then +// the code to handle the panic breaks escaping. It's hard to see the second +// problem once the first is fixed, but its fix is trivial so we let that go. See +// the discussion for issue 7379. +func TestPipeToMethodIsEscaped(t *testing.T) { + tmpl := Must(New("x").Parse("<html>{{0 | .SomeMethod}}</html>\n")) + tryExec := func() string { + defer func() { + panicValue := recover() + if panicValue != nil { + t.Errorf("panicked: %v\n", panicValue) + } + }() + var b bytes.Buffer + tmpl.Execute(&b, Issue7379(0)) + return b.String() + } + for i := 0; i < 3; i++ { + str := tryExec() + const expect = "<html><0></html>\n" + if str != expect { + t.Errorf("expected %q got %q", expect, str) + } + } +} + +func BenchmarkEscapedExecute(b *testing.B) { + tmpl := Must(New("t").Parse(`<a onclick="alert('{{.}}')">{{.}}</a>`)) + var buf bytes.Buffer + b.ResetTimer() + for i := 0; i < b.N; i++ { + tmpl.Execute(&buf, "foo & 'bar' & baz") + buf.Reset() + } +} diff --git a/template/html.go b/template/html.go new file mode 100644 index 00000000..9c069efd --- /dev/null +++ b/template/html.go @@ -0,0 +1,257 @@ +// 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 template + +import ( + "bytes" + "fmt" + "strings" + "unicode/utf8" +) + +// htmlNospaceEscaper escapes for inclusion in unquoted attribute values. +func htmlNospaceEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTML { + return htmlReplacer(stripTags(s), htmlNospaceNormReplacementTable, false) + } + return htmlReplacer(s, htmlNospaceReplacementTable, false) +} + +// attrEscaper escapes for inclusion in quoted attribute values. +func attrEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTML { + return htmlReplacer(stripTags(s), htmlNormReplacementTable, true) + } + return htmlReplacer(s, htmlReplacementTable, true) +} + +// rcdataEscaper escapes for inclusion in an RCDATA element body. +func rcdataEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTML { + return htmlReplacer(s, htmlNormReplacementTable, true) + } + return htmlReplacer(s, htmlReplacementTable, true) +} + +// htmlEscaper escapes for inclusion in HTML text. +func htmlEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTML { + return s + } + return htmlReplacer(s, htmlReplacementTable, true) +} + +// htmlReplacementTable contains the runes that need to be escaped +// inside a quoted attribute value or in a text node. +var htmlReplacementTable = []string{ + // http://www.w3.org/TR/html5/syntax.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/syntax.html#before-attribute-value-state + 0: "\uFFFD", + '"': """, + '&': "&", + '\'': "'", + '+': "+", + '<': "<", + '>': ">", +} + +// htmlNormReplacementTable is like htmlReplacementTable but without '&' to +// avoid over-encoding existing entities. +var htmlNormReplacementTable = []string{ + 0: "\uFFFD", + '"': """, + '\'': "'", + '+': "+", + '<': "<", + '>': ">", +} + +// htmlNospaceReplacementTable contains the runes that need to be escaped +// inside an unquoted attribute value. +// 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 htmlNospaceReplacementTable = []string{ + 0: "�", + '\t': " ", + '\n': " ", + '\v': " ", + '\f': "", + '\r': " ", + ' ': " ", + '"': """, + '&': "&", + '\'': "'", + '+': "+", + '<': "<", + '=': "=", + '>': ">", + // A parse error in the attribute value (unquoted) and + // before attribute value states. + // Treated as a quoting character by IE. + '`': "`", +} + +// htmlNospaceNormReplacementTable is like htmlNospaceReplacementTable but +// without '&' to avoid over-encoding existing entities. +var htmlNospaceNormReplacementTable = []string{ + 0: "�", + '\t': " ", + '\n': " ", + '\v': " ", + '\f': "", + '\r': " ", + ' ': " ", + '"': """, + '\'': "'", + '+': "+", + '<': "<", + '=': "=", + '>': ">", + // A parse error in the attribute value (unquoted) and + // before attribute value states. + // Treated as a quoting character by IE. + '`': "`", +} + +// htmlReplacer returns s with runes replaced according to replacementTable +// and when badRunes is true, certain bad runes are allowed through unescaped. +func htmlReplacer(s string, replacementTable []string, badRunes bool) string { + written, b := 0, new(bytes.Buffer) + for i, r := range s { + if int(r) < len(replacementTable) { + if repl := replacementTable[r]; len(repl) != 0 { + b.WriteString(s[written:i]) + b.WriteString(repl) + // Valid as long as replacementTable doesn't + // include anything above 0x7f. + written = i + utf8.RuneLen(r) + } + } else if badRunes { + // No-op. + // IE does not allow these ranges in unquoted attrs. + } else if 0xfdd0 <= r && r <= 0xfdef || 0xfff0 <= r && r <= 0xffff { + fmt.Fprintf(b, "%s&#x%x;", s[written:i], r) + written = i + utf8.RuneLen(r) + } + } + if written == 0 { + return s + } + b.WriteString(s[written:]) + return b.String() +} + +// stripTags takes a snippet of HTML and returns only the text content. +// For example, `<b>¡Hi!</b> <script>...</script>` -> `¡Hi! `. +func stripTags(html string) string { + var b bytes.Buffer + 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 { + 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. + j := i1 + if d.state != c.state { + for j1 := j - 1; j1 >= i; j1-- { + if s[j1] == '<' { + j = j1 + break + } + } + } + b.Write(s[i:j]) + } else { + allText = false + } + c, i = d, i1 + continue + } + i1 := i + bytes.IndexAny(s[i:], delimEnds[c.delim]) + if i1 < i { + break + } + if c.delim != delimSpaceOrTagEnd { + // Consume any quote. + i1++ + } + c, i = context{state: stateTag, element: c.element}, i1 + } + if allText { + return html + } else if c.state == stateText || c.state == stateRCDATA { + b.Write(s[i:]) + } + return b.String() +} + +// htmlNameFilter accepts valid parts of an HTML attribute or tag name or +// a known-safe HTML attribute. +func htmlNameFilter(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeHTMLAttr { + return s + } + if len(s) == 0 { + // Avoid violation of structure preservation. + // <input checked {{.K}}={{.V}}>. + // Without this, if .K is empty then .V is the value of + // checked, but otherwise .V is the value of the attribute + // named .K. + return filterFailsafe + } + s = strings.ToLower(s) + if t := attrType(s); t != contentTypePlain { + // TODO: Split attr and element name part filters so we can whitelist + // attributes. + return filterFailsafe + } + for _, r := range s { + switch { + case '0' <= r && r <= '9': + case 'a' <= r && r <= 'z': + default: + return filterFailsafe + } + } + return s +} + +// commentEscaper returns the empty string regardless of input. +// Comment content does not correspond to any parsed structure or +// human-readable content, so the simplest and most secure policy is to drop +// content interpolated into comments. +// This approach is equally valid whether or not static comment content is +// removed from the template. +func commentEscaper(args ...interface{}) string { + return "" +} diff --git a/template/html_test.go b/template/html_test.go new file mode 100644 index 00000000..b9b97038 --- /dev/null +++ b/template/html_test.go @@ -0,0 +1,94 @@ +// 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 template + +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\ufdec\U0001D11E") + + want := ("�\x01\x02\x03\x04\x05\x06\x07" + + "\x08 \x0E\x0F" + + "\x10\x11\x12\x13\x14\x15\x16\x17" + + "\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !"#$%&'()*+,-./` + + `0123456789:;<=>?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ[\]^_` + + ``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 TestStripTags(t *testing.T) { + tests := []struct { + input, want string + }{ + {"", ""}, + {"Hello, World!", "Hello, World!"}, + {"foo&bar", "foo&bar"}, + {`Hello <a href="www.example.com/">World</a>!`, "Hello World!"}, + {"Foo <textarea>Bar</textarea> Baz", "Foo Bar Baz"}, + {"Foo <!-- Bar --> Baz", "Foo Baz"}, + {"<", "<"}, + {"foo < bar", "foo < bar"}, + {`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 { + if got := stripTags(test.input); got != test.want { + t.Errorf("%q: want %q, got %q", test.input, test.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.") + } +} + +func BenchmarkStripTags(b *testing.B) { + for i := 0; i < b.N; i++ { + stripTags("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>") + } +} + +func BenchmarkStripTagsNoSpecials(b *testing.B) { + for i := 0; i < b.N; i++ { + stripTags("The quick, brown fox jumps over the lazy dog.") + } +} diff --git a/template/js.go b/template/js.go new file mode 100644 index 00000000..999a61ed --- /dev/null +++ b/template/js.go @@ -0,0 +1,362 @@ +// 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 template + +import ( + "bytes" + "encoding/json" + "fmt" + "reflect" + "strings" + "unicode/utf8" +) + +// nextJSCtx returns the context that determines whether a slash after the +// given run of tokens starts a regular expression instead of a division +// operator: / or /=. +// +// This assumes that the token run does not include any string tokens, comment +// tokens, regular expression literal tokens, or division operators. +// +// This fails on some valid but nonsensical JavaScript programs like +// "x = ++/foo/i" which is quite different than "x++/foo/i", but is not known to +// fail on any known useful programs. It is based on the draft +// JavaScript 2.0 lexical grammar and requires one token of lookbehind: +// http://www.mozilla.org/js/language/js20-2000-07/rationale/syntax.html +func nextJSCtx(s []byte, preceding jsCtx) jsCtx { + s = bytes.TrimRight(s, "\t\n\f\r \u2028\u2029") + if len(s) == 0 { + return preceding + } + + // All cases below are in the single-byte UTF-8 group. + switch c, n := s[len(s)-1], len(s); c { + case '+', '-': + // ++ and -- are not regexp preceders, but + and - are whether + // they are used as infix or prefix operators. + start := n - 1 + // Count the number of adjacent dashes or pluses. + for start > 0 && s[start-1] == c { + start-- + } + if (n-start)&1 == 1 { + // Reached for trailing minus signs since "---" is the + // same as "-- -". + return jsCtxRegexp + } + return jsCtxDivOp + case '.': + // Handle "42." + if n != 1 && '0' <= s[n-2] && s[n-2] <= '9' { + return jsCtxDivOp + } + return jsCtxRegexp + // Suffixes for all punctuators from section 7.7 of the language spec + // that only end binary operators not handled above. + case ',', '<', '>', '=', '*', '%', '&', '|', '^', '?': + return jsCtxRegexp + // Suffixes for all punctuators from section 7.7 of the language spec + // that are prefix operators not handled above. + case '!', '~': + return jsCtxRegexp + // Matches all the punctuators from section 7.7 of the language spec + // that are open brackets not handled above. + case '(', '[': + return jsCtxRegexp + // Matches all the punctuators from section 7.7 of the language spec + // that precede expression starts. + case ':', ';', '{': + return jsCtxRegexp + // CAVEAT: the close punctuators ('}', ']', ')') precede div ops and + // are handled in the default except for '}' which can precede a + // division op as in + // ({ valueOf: function () { return 42 } } / 2 + // which is valid, but, in practice, developers don't divide object + // literals, so our heuristic works well for code like + // function () { ... } /foo/.test(x) && sideEffect(); + // The ')' punctuator can precede a regular expression as in + // if (b) /foo/.test(x) && ... + // but this is much less likely than + // (a + b) / c + case '}': + return jsCtxRegexp + default: + // Look for an IdentifierName and see if it is a keyword that + // can precede a regular expression. + j := n + for j > 0 && isJSIdentPart(rune(s[j-1])) { + j-- + } + if regexpPrecederKeywords[string(s[j:])] { + return jsCtxRegexp + } + } + // Otherwise is a punctuator not listed above, or + // a string which precedes a div op, or an identifier + // which precedes a div op. + return jsCtxDivOp +} + +// regexpPrecederKeywords is a set of reserved JS keywords that can precede a +// regular expression in JS source. +var regexpPrecederKeywords = map[string]bool{ + "break": true, + "case": true, + "continue": true, + "delete": true, + "do": true, + "else": true, + "finally": true, + "in": true, + "instanceof": true, + "return": true, + "throw": true, + "try": true, + "typeof": true, + "void": true, +} + +var jsonMarshalType = reflect.TypeOf((*json.Marshaler)(nil)).Elem() + +// indirectToJSONMarshaler returns the value, after dereferencing as many times +// as necessary to reach the base type (or nil) or an implementation of json.Marshal. +func indirectToJSONMarshaler(a interface{}) interface{} { + v := reflect.ValueOf(a) + for !v.Type().Implements(jsonMarshalType) && v.Kind() == reflect.Ptr && !v.IsNil() { + v = v.Elem() + } + return v.Interface() +} + +// jsValEscaper escapes its inputs to a JS Expression (section 11.14) that has +// neither side-effects nor free variables outside (NaN, Infinity). +func jsValEscaper(args ...interface{}) string { + var a interface{} + if len(args) == 1 { + a = indirectToJSONMarshaler(args[0]) + switch t := a.(type) { + case JS: + return string(t) + case JSStr: + // TODO: normalize quotes. + return `"` + string(t) + `"` + case json.Marshaler: + // Do not treat as a Stringer. + case fmt.Stringer: + a = t.String() + } + } else { + for i, arg := range args { + args[i] = indirectToJSONMarshaler(arg) + } + a = fmt.Sprint(args...) + } + // TODO: detect cycles before calling Marshal which loops infinitely on + // cyclic data. This may be an unacceptable DoS risk. + + b, err := json.Marshal(a) + if err != nil { + // Put a space before comment so that if it is flush against + // a division operator it is not turned into a line comment: + // x/{{y}} + // turning into + // x//* error marshalling y: + // second line of error message */null + return fmt.Sprintf(" /* %s */null ", strings.Replace(err.Error(), "*/", "* /", -1)) + } + + // TODO: maybe post-process output to prevent it from containing + // "<!--", "-->", "<![CDATA[", "]]>", or "</script" + // in case custom marshallers produce output containing those. + + // TODO: Maybe abbreviate \u00ab to \xab to produce more compact output. + if len(b) == 0 { + // In, `x=y/{{.}}*z` a json.Marshaler that produces "" should + // not cause the output `x=y/*z`. + return " null " + } + first, _ := utf8.DecodeRune(b) + last, _ := utf8.DecodeLastRune(b) + var buf bytes.Buffer + // Prevent IdentifierNames and NumericLiterals from running into + // keywords: in, instanceof, typeof, void + pad := isJSIdentPart(first) || isJSIdentPart(last) + if pad { + buf.WriteByte(' ') + } + written := 0 + // Make sure that json.Marshal escapes codepoints U+2028 & U+2029 + // so it falls within the subset of JSON which is valid JS. + for i := 0; i < len(b); { + rune, n := utf8.DecodeRune(b[i:]) + repl := "" + if rune == 0x2028 { + repl = `\u2028` + } else if rune == 0x2029 { + repl = `\u2029` + } + if repl != "" { + buf.Write(b[written:i]) + buf.WriteString(repl) + written = i + n + } + i += n + } + if buf.Len() != 0 { + buf.Write(b[written:]) + if pad { + buf.WriteByte(' ') + } + b = buf.Bytes() + } + return string(b) +} + +// jsStrEscaper produces a string that can be included between quotes in +// JavaScript source, in JavaScript embedded in an HTML5 <script> element, +// or in an HTML5 event handler attribute such as onclick. +func jsStrEscaper(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeJSStr { + return replace(s, jsStrNormReplacementTable) + } + return replace(s, jsStrReplacementTable) +} + +// jsRegexpEscaper behaves like jsStrEscaper but escapes regular expression +// specials so the result is treated literally when included in a regular +// expression literal. /foo{{.X}}bar/ matches the string "foo" followed by +// the literal text of {{.X}} followed by the string "bar". +func jsRegexpEscaper(args ...interface{}) string { + s, _ := stringify(args...) + s = replace(s, jsRegexpReplacementTable) + if s == "" { + // /{{.X}}/ should not produce a line comment when .X == "". + return "(?:)" + } + return s +} + +// replace replaces each rune r of s with replacementTable[r], provided that +// r < len(replacementTable). If replacementTable[r] is the empty string then +// no replacement is made. +// It also replaces runes U+2028 and U+2029 with the raw strings `\u2028` and +// `\u2029`. +func replace(s string, replacementTable []string) string { + var b bytes.Buffer + written := 0 + for i, r := range s { + var repl string + switch { + case int(r) < len(replacementTable) && replacementTable[r] != "": + repl = replacementTable[r] + case r == '\u2028': + repl = `\u2028` + case r == '\u2029': + repl = `\u2029` + default: + continue + } + b.WriteString(s[written:i]) + b.WriteString(repl) + written = i + utf8.RuneLen(r) + } + if written == 0 { + return s + } + b.WriteString(s[written:]) + return b.String() +} + +var jsStrReplacementTable = []string{ + 0: `\0`, + '\t': `\t`, + '\n': `\n`, + '\v': `\x0b`, // "\v" == "v" on IE 6. + '\f': `\f`, + '\r': `\r`, + // Encode HTML specials as hex so the output can be embedded + // in HTML attributes without further encoding. + '"': `\x22`, + '&': `\x26`, + '\'': `\x27`, + '+': `\x2b`, + '/': `\/`, + '<': `\x3c`, + '>': `\x3e`, + '\\': `\\`, +} + +// jsStrNormReplacementTable is like jsStrReplacementTable but does not +// overencode existing escapes since this table has no entry for `\`. +var jsStrNormReplacementTable = []string{ + 0: `\0`, + '\t': `\t`, + '\n': `\n`, + '\v': `\x0b`, // "\v" == "v" on IE 6. + '\f': `\f`, + '\r': `\r`, + // Encode HTML specials as hex so the output can be embedded + // in HTML attributes without further encoding. + '"': `\x22`, + '&': `\x26`, + '\'': `\x27`, + '+': `\x2b`, + '/': `\/`, + '<': `\x3c`, + '>': `\x3e`, +} + +var jsRegexpReplacementTable = []string{ + 0: `\0`, + '\t': `\t`, + '\n': `\n`, + '\v': `\x0b`, // "\v" == "v" on IE 6. + '\f': `\f`, + '\r': `\r`, + // Encode HTML specials as hex so the output can be embedded + // in HTML attributes without further encoding. + '"': `\x22`, + '$': `\$`, + '&': `\x26`, + '\'': `\x27`, + '(': `\(`, + ')': `\)`, + '*': `\*`, + '+': `\x2b`, + '-': `\-`, + '.': `\.`, + '/': `\/`, + '<': `\x3c`, + '>': `\x3e`, + '?': `\?`, + '[': `\[`, + '\\': `\\`, + ']': `\]`, + '^': `\^`, + '{': `\{`, + '|': `\|`, + '}': `\}`, +} + +// isJSIdentPart reports 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. +func isJSIdentPart(r rune) bool { + switch { + case r == '$': + return true + case '0' <= r && r <= '9': + return true + case 'A' <= r && r <= 'Z': + return true + case r == '_': + return true + case 'a' <= r && r <= 'z': + return true + } + return false +} diff --git a/template/js_test.go b/template/js_test.go new file mode 100644 index 00000000..7af7997d --- /dev/null +++ b/template/js_test.go @@ -0,0 +1,401 @@ +// 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 template + +import ( + "bytes" + "math" + "strings" + "testing" +) + +func TestNextJsCtx(t *testing.T) { + tests := []struct { + jsCtx jsCtx + s string + }{ + // Statement terminators precede regexps. + {jsCtxRegexp, ";"}, + // This is not airtight. + // ({ valueOf: function () { return 1 } } / 2) + // is valid JavaScript but in practice, devs do not do this. + // A block followed by a statement starting with a RegExp is + // much more common: + // while (x) {...} /foo/.test(x) || panic() + {jsCtxRegexp, "}"}, + // But member, call, grouping, and array expression terminators + // precede div ops. + {jsCtxDivOp, ")"}, + {jsCtxDivOp, "]"}, + // At the start of a primary expression, array, or expression + // statement, expect a regexp. + {jsCtxRegexp, "("}, + {jsCtxRegexp, "["}, + {jsCtxRegexp, "{"}, + // Assignment operators precede regexps as do all exclusively + // prefix and binary operators. + {jsCtxRegexp, "="}, + {jsCtxRegexp, "+="}, + {jsCtxRegexp, "*="}, + {jsCtxRegexp, "*"}, + {jsCtxRegexp, "!"}, + // Whether the + or - is infix or prefix, it cannot precede a + // div op. + {jsCtxRegexp, "+"}, + {jsCtxRegexp, "-"}, + // An incr/decr op precedes a div operator. + // This is not airtight. In (g = ++/h/i) a regexp follows a + // pre-increment operator, but in practice devs do not try to + // increment or decrement regular expressions. + // (g++/h/i) where ++ is a postfix operator on g is much more + // common. + {jsCtxDivOp, "--"}, + {jsCtxDivOp, "++"}, + {jsCtxDivOp, "x--"}, + // When we have many dashes or pluses, then they are grouped + // left to right. + {jsCtxRegexp, "x---"}, // A postfix -- then a -. + // return followed by a slash returns the regexp literal or the + // slash starts a regexp literal in an expression statement that + // is dead code. + {jsCtxRegexp, "return"}, + {jsCtxRegexp, "return "}, + {jsCtxRegexp, "return\t"}, + {jsCtxRegexp, "return\n"}, + {jsCtxRegexp, "return\u2028"}, + // Identifiers can be divided and cannot validly be preceded by + // a regular expressions. Semicolon insertion cannot happen + // between an identifier and a regular expression on a new line + // because the one token lookahead for semicolon insertion has + // to conclude that it could be a div binary op and treat it as + // such. + {jsCtxDivOp, "x"}, + {jsCtxDivOp, "x "}, + {jsCtxDivOp, "x\t"}, + {jsCtxDivOp, "x\n"}, + {jsCtxDivOp, "x\u2028"}, + {jsCtxDivOp, "preturn"}, + // Numbers precede div ops. + {jsCtxDivOp, "0"}, + // Dots that are part of a number are div preceders. + {jsCtxDivOp, "0."}, + } + + for _, test := range tests { + if nextJSCtx([]byte(test.s), jsCtxRegexp) != test.jsCtx { + t.Errorf("want %s got %q", test.jsCtx, test.s) + } + if nextJSCtx([]byte(test.s), jsCtxDivOp) != test.jsCtx { + t.Errorf("want %s got %q", test.jsCtx, test.s) + } + } + + if nextJSCtx([]byte(" "), jsCtxRegexp) != jsCtxRegexp { + t.Error("Blank tokens") + } + + if nextJSCtx([]byte(" "), jsCtxDivOp) != jsCtxDivOp { + t.Error("Blank tokens") + } +} + +func TestJSValEscaper(t *testing.T) { + tests := []struct { + x interface{} + js string + }{ + {int(42), " 42 "}, + {uint(42), " 42 "}, + {int16(42), " 42 "}, + {uint16(42), " 42 "}, + {int32(-42), " -42 "}, + {uint32(42), " 42 "}, + {int16(-42), " -42 "}, + {uint16(42), " 42 "}, + {int64(-42), " -42 "}, + {uint64(42), " 42 "}, + {uint64(1) << 53, " 9007199254740992 "}, + // ulp(1 << 53) > 1 so this loses precision in JS + // but it is still a representable integer literal. + {uint64(1)<<53 + 1, " 9007199254740993 "}, + {float32(1.0), " 1 "}, + {float32(-1.0), " -1 "}, + {float32(0.5), " 0.5 "}, + {float32(-0.5), " -0.5 "}, + {float32(1.0) / float32(256), " 0.00390625 "}, + {float32(0), " 0 "}, + {math.Copysign(0, -1), " -0 "}, + {float64(1.0), " 1 "}, + {float64(-1.0), " -1 "}, + {float64(0.5), " 0.5 "}, + {float64(-0.5), " -0.5 "}, + {float64(0), " 0 "}, + {math.Copysign(0, -1), " -0 "}, + {"", `""`}, + {"foo", `"foo"`}, + // Newlines. + {"\r\n\u2028\u2029", `"\r\n\u2028\u2029"`}, + // "\v" == "v" on IE 6 so use "\x0b" instead. + {"\t\x0b", `"\t\u000b"`}, + {struct{ X, Y int }{1, 2}, `{"X":1,"Y":2}`}, + {[]interface{}{}, "[]"}, + {[]interface{}{42, "foo", nil}, `[42,"foo",null]`}, + {[]string{"<!--", "</script>", "-->"}, `["\u003c!--","\u003c/script\u003e","--\u003e"]`}, + {"<!--", `"\u003c!--"`}, + {"-->", `"--\u003e"`}, + {"<![CDATA[", `"\u003c![CDATA["`}, + {"]]>", `"]]\u003e"`}, + {"</script", `"\u003c/script"`}, + {"\U0001D11E", "\"\U0001D11E\""}, // or "\uD834\uDD1E" + } + + for _, test := range tests { + if js := jsValEscaper(test.x); js != test.js { + t.Errorf("%+v: want\n\t%q\ngot\n\t%q", test.x, test.js, js) + } + // Make sure that escaping corner cases are not broken + // by nesting. + a := []interface{}{test.x} + want := "[" + strings.TrimSpace(test.js) + "]" + if js := jsValEscaper(a); js != want { + t.Errorf("%+v: want\n\t%q\ngot\n\t%q", a, want, js) + } + } +} + +func TestJSStrEscaper(t *testing.T) { + tests := []struct { + x interface{} + esc string + }{ + {"", ``}, + {"foo", `foo`}, + {"\u0000", `\0`}, + {"\t", `\t`}, + {"\n", `\n`}, + {"\r", `\r`}, + {"\u2028", `\u2028`}, + {"\u2029", `\u2029`}, + {"\\", `\\`}, + {"\\n", `\\n`}, + {"foo\r\nbar", `foo\r\nbar`}, + // Preserve attribute boundaries. + {`"`, `\x22`}, + {`'`, `\x27`}, + // Allow embedding in HTML without further escaping. + {`&`, `\x26amp;`}, + // Prevent breaking out of text node and element boundaries. + {"</script>", `\x3c\/script\x3e`}, + {"<![CDATA[", `\x3c![CDATA[`}, + {"]]>", `]]\x3e`}, + // http://dev.w3.org/html5/markup/aria/syntax.html#escaping-text-span + // "The text in style, script, title, and textarea elements + // must not have an escaping text span start that is not + // followed by an escaping text span end." + // Furthermore, spoofing an escaping text span end could lead + // to different interpretation of a </script> sequence otherwise + // masked by the escaping text span, and spoofing a start could + // allow regular text content to be interpreted as script + // allowing script execution via a combination of a JS string + // injection followed by an HTML text injection. + {"<!--", `\x3c!--`}, + {"-->", `--\x3e`}, + // From http://code.google.com/p/doctype/wiki/ArticleUtf7 + {"+ADw-script+AD4-alert(1)+ADw-/script+AD4-", + `\x2bADw-script\x2bAD4-alert(1)\x2bADw-\/script\x2bAD4-`, + }, + // Invalid UTF-8 sequence + {"foo\xA0bar", "foo\xA0bar"}, + // Invalid unicode scalar value. + {"foo\xed\xa0\x80bar", "foo\xed\xa0\x80bar"}, + } + + for _, test := range tests { + esc := jsStrEscaper(test.x) + if esc != test.esc { + t.Errorf("%q: want %q got %q", test.x, test.esc, esc) + } + } +} + +func TestJSRegexpEscaper(t *testing.T) { + tests := []struct { + x interface{} + esc string + }{ + {"", `(?:)`}, + {"foo", `foo`}, + {"\u0000", `\0`}, + {"\t", `\t`}, + {"\n", `\n`}, + {"\r", `\r`}, + {"\u2028", `\u2028`}, + {"\u2029", `\u2029`}, + {"\\", `\\`}, + {"\\n", `\\n`}, + {"foo\r\nbar", `foo\r\nbar`}, + // Preserve attribute boundaries. + {`"`, `\x22`}, + {`'`, `\x27`}, + // Allow embedding in HTML without further escaping. + {`&`, `\x26amp;`}, + // Prevent breaking out of text node and element boundaries. + {"</script>", `\x3c\/script\x3e`}, + {"<![CDATA[", `\x3c!\[CDATA\[`}, + {"]]>", `\]\]\x3e`}, + // Escaping text spans. + {"<!--", `\x3c!\-\-`}, + {"-->", `\-\-\x3e`}, + {"*", `\*`}, + {"+", `\x2b`}, + {"?", `\?`}, + {"[](){}", `\[\]\(\)\{\}`}, + {"$foo|x.y", `\$foo\|x\.y`}, + {"x^y", `x\^y`}, + } + + for _, test := range tests { + esc := jsRegexpEscaper(test.x) + if esc != test.esc { + t.Errorf("%q: want %q got %q", test.x, test.esc, esc) + } + } +} + +func TestEscapersOnLower7AndSelectHighCodepoints(t *testing.T) { + input := ("\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f" + + "\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !"#$%&'()*+,-./` + + `0123456789:;<=>?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ[\]^_` + + "`abcdefghijklmno" + + "pqrstuvwxyz{|}~\x7f" + + "\u00A0\u0100\u2028\u2029\ufeff\U0001D11E") + + tests := []struct { + name string + escaper func(...interface{}) string + escaped string + }{ + { + "jsStrEscaper", + jsStrEscaper, + "\\0\x01\x02\x03\x04\x05\x06\x07" + + "\x08\\t\\n\\x0b\\f\\r\x0E\x0F" + + "\x10\x11\x12\x13\x14\x15\x16\x17" + + "\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !\x22#$%\x26\x27()*\x2b,-.\/` + + `0123456789:;\x3c=\x3e?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ[\\]^_` + + "`abcdefghijklmno" + + "pqrstuvwxyz{|}~\x7f" + + "\u00A0\u0100\\u2028\\u2029\ufeff\U0001D11E", + }, + { + "jsRegexpEscaper", + jsRegexpEscaper, + "\\0\x01\x02\x03\x04\x05\x06\x07" + + "\x08\\t\\n\\x0b\\f\\r\x0E\x0F" + + "\x10\x11\x12\x13\x14\x15\x16\x17" + + "\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" + + ` !\x22#\$%\x26\x27\(\)\*\x2b,\-\.\/` + + `0123456789:;\x3c=\x3e\?` + + `@ABCDEFGHIJKLMNO` + + `PQRSTUVWXYZ\[\\\]\^_` + + "`abcdefghijklmno" + + `pqrstuvwxyz\{\|\}~` + "\u007f" + + "\u00A0\u0100\\u2028\\u2029\ufeff\U0001D11E", + }, + } + + for _, test := range tests { + if s := test.escaper(input); s != test.escaped { + t.Errorf("%s once: want\n\t%q\ngot\n\t%q", test.name, test.escaped, s) + continue + } + + // Escape it rune by rune to make sure that any + // fast-path checking does not break escaping. + var buf bytes.Buffer + for _, c := range input { + buf.WriteString(test.escaper(string(c))) + } + + if s := buf.String(); s != test.escaped { + t.Errorf("%s rune-wise: want\n\t%q\ngot\n\t%q", test.name, test.escaped, s) + continue + } + } +} + +func BenchmarkJSValEscaperWithNum(b *testing.B) { + for i := 0; i < b.N; i++ { + jsValEscaper(3.141592654) + } +} + +func BenchmarkJSValEscaperWithStr(b *testing.B) { + for i := 0; i < b.N; i++ { + jsValEscaper("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>") + } +} + +func BenchmarkJSValEscaperWithStrNoSpecials(b *testing.B) { + for i := 0; i < b.N; i++ { + jsValEscaper("The quick, brown fox jumps over the lazy dog") + } +} + +func BenchmarkJSValEscaperWithObj(b *testing.B) { + o := struct { + S string + N int + }{ + "The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>\u2028", + 42, + } + for i := 0; i < b.N; i++ { + jsValEscaper(o) + } +} + +func BenchmarkJSValEscaperWithObjNoSpecials(b *testing.B) { + o := struct { + S string + N int + }{ + "The quick, brown fox jumps over the lazy dog", + 42, + } + for i := 0; i < b.N; i++ { + jsValEscaper(o) + } +} + +func BenchmarkJSStrEscaperNoSpecials(b *testing.B) { + for i := 0; i < b.N; i++ { + jsStrEscaper("The quick, brown fox jumps over the lazy dog.") + } +} + +func BenchmarkJSStrEscaper(b *testing.B) { + for i := 0; i < b.N; i++ { + jsStrEscaper("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>") + } +} + +func BenchmarkJSRegexpEscaperNoSpecials(b *testing.B) { + for i := 0; i < b.N; i++ { + jsRegexpEscaper("The quick, brown fox jumps over the lazy dog") + } +} + +func BenchmarkJSRegexpEscaper(b *testing.B) { + for i := 0; i < b.N; i++ { + jsRegexpEscaper("The <i>quick</i>,\r\n<span style='color:brown'>brown</span> fox jumps\u2028over the <canine class=\"lazy\">dog</canine>") + } +} diff --git a/template/template.go b/template/template.go new file mode 100644 index 00000000..65b5bff5 --- /dev/null +++ b/template/template.go @@ -0,0 +1,390 @@ +// 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 template + +import ( + "fmt" + "io" + "io/ioutil" + "path/filepath" + "sync" + "text/template" + "text/template/parse" +) + +// Template is a specialized Template from "text/template" that produces a safe +// HTML document fragment. +type Template struct { + // Sticky error if escaping fails. + escapeErr error + // We could embed the text/template field, but it's safer not to because + // we need to keep our version of the name space and the underlying + // template's in sync. + text *template.Template + // The underlying template's parse tree, updated to be HTML-safe. + Tree *parse.Tree + *nameSpace // common to all associated templates +} + +// escapeOK is a sentinel value used to indicate valid escaping. +var escapeOK = fmt.Errorf("template escaped correctly") + +// nameSpace is the data structure shared by all templates in an association. +type nameSpace struct { + mu sync.Mutex + set map[string]*Template +} + +// Templates returns a slice of the templates associated with t, including t +// itself. +func (t *Template) Templates() []*Template { + ns := t.nameSpace + ns.mu.Lock() + defer ns.mu.Unlock() + // Return a slice so we don't expose the map. + m := make([]*Template, 0, len(ns.set)) + for _, v := range ns.set { + m = append(m, v) + } + return m +} + +// escape escapes all associated templates. +func (t *Template) escape() error { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + if t.escapeErr == nil { + if err := escapeTemplate(t, t.text.Root, t.Name()); err != nil { + return err + } + } else if t.escapeErr != escapeOK { + return t.escapeErr + } + return nil +} + +// Execute applies a parsed template to the specified data object, +// writing the output to wr. +// If an error occurs executing the template or writing its output, +// execution stops, but partial results may already have been written to +// the output writer. +// A template may be executed safely in parallel. +func (t *Template) Execute(wr io.Writer, data interface{}) error { + if err := t.escape(); err != nil { + return err + } + return t.text.Execute(wr, data) +} + +// ExecuteTemplate applies the template associated with t that has the given +// name to the specified data object and writes the output to wr. +// If an error occurs executing the template or writing its output, +// execution stops, but partial results may already have been written to +// the output writer. +// A template may be executed safely in parallel. +func (t *Template) ExecuteTemplate(wr io.Writer, name string, data interface{}) error { + tmpl, err := t.lookupAndEscapeTemplate(name) + if err != nil { + return err + } + return tmpl.text.Execute(wr, data) +} + +// lookupAndEscapeTemplate guarantees that the template with the given name +// is escaped, or returns an error if it cannot be. It returns the named +// template. +func (t *Template) lookupAndEscapeTemplate(name string) (tmpl *Template, err error) { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + tmpl = t.set[name] + if tmpl == nil { + return nil, fmt.Errorf("html/template: %q is undefined", name) + } + if tmpl.escapeErr != nil && tmpl.escapeErr != escapeOK { + return nil, tmpl.escapeErr + } + if tmpl.text.Tree == nil || tmpl.text.Root == nil { + return nil, fmt.Errorf("html/template: %q is an incomplete template", name) + } + if t.text.Lookup(name) == nil { + panic("html/template internal error: template escaping out of sync") + } + if tmpl.escapeErr == nil { + err = escapeTemplate(tmpl, tmpl.text.Root, name) + } + return tmpl, err +} + +// Parse parses a string into a template. Nested template definitions +// will be associated with the top-level template t. Parse may be +// called multiple times to parse definitions of templates to associate +// with t. It is an error if a resulting template is non-empty (contains +// content other than template definitions) and would replace a +// non-empty template with the same name. (In multiple calls to Parse +// with the same receiver template, only one call can contain text +// other than space, comments, and template definitions.) +func (t *Template) Parse(src string) (*Template, error) { + t.nameSpace.mu.Lock() + t.escapeErr = nil + t.nameSpace.mu.Unlock() + ret, err := t.text.Parse(src) + if err != nil { + return nil, err + } + // In general, all the named templates might have changed underfoot. + // Regardless, some new ones may have been defined. + // The template.Template set has been updated; update ours. + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + for _, v := range ret.Templates() { + name := v.Name() + tmpl := t.set[name] + if tmpl == nil { + tmpl = t.new(name) + } + // Restore our record of this text/template to its unescaped original state. + tmpl.escapeErr = nil + tmpl.text = v + tmpl.Tree = v.Tree + } + return t, nil +} + +// AddParseTree creates a new template with the name and parse tree +// and associates it with t. +// +// It returns an error if t has already been executed. +func (t *Template) AddParseTree(name string, tree *parse.Tree) (*Template, error) { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + if t.escapeErr != nil { + return nil, fmt.Errorf("html/template: cannot AddParseTree to %q after it has executed", t.Name()) + } + text, err := t.text.AddParseTree(name, tree) + if err != nil { + return nil, err + } + ret := &Template{ + nil, + text, + text.Tree, + t.nameSpace, + } + t.set[name] = ret + return ret, nil +} + +// Clone returns a duplicate of the template, including all associated +// templates. The actual representation is not copied, but the name space of +// associated templates is, so further calls to Parse in the copy will add +// templates to the copy but not to the original. Clone can be used to prepare +// common templates and use them with variant definitions for other templates +// by adding the variants after the clone is made. +// +// It returns an error if t has already been executed. +func (t *Template) Clone() (*Template, error) { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + if t.escapeErr != nil { + return nil, fmt.Errorf("html/template: cannot Clone %q after it has executed", t.Name()) + } + textClone, err := t.text.Clone() + if err != nil { + return nil, err + } + ret := &Template{ + nil, + textClone, + textClone.Tree, + &nameSpace{ + set: make(map[string]*Template), + }, + } + for _, x := range textClone.Templates() { + name := x.Name() + src := t.set[name] + if src == nil || src.escapeErr != nil { + return nil, fmt.Errorf("html/template: cannot Clone %q after it has executed", t.Name()) + } + x.Tree = x.Tree.Copy() + ret.set[name] = &Template{ + nil, + x, + x.Tree, + ret.nameSpace, + } + } + return ret, nil +} + +// New allocates a new HTML template with the given name. +func New(name string) *Template { + tmpl := &Template{ + nil, + template.New(name), + nil, + &nameSpace{ + set: make(map[string]*Template), + }, + } + tmpl.set[name] = tmpl + return tmpl +} + +// New allocates a new HTML template associated with the given one +// and with the same delimiters. The association, which is transitive, +// allows one template to invoke another with a {{template}} action. +func (t *Template) New(name string) *Template { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + return t.new(name) +} + +// new is the implementation of New, without the lock. +func (t *Template) new(name string) *Template { + tmpl := &Template{ + nil, + t.text.New(name), + nil, + t.nameSpace, + } + tmpl.set[name] = tmpl + return tmpl +} + +// Name returns the name of the template. +func (t *Template) Name() string { + return t.text.Name() +} + +// FuncMap is the type of the map defining the mapping from names to +// functions. Each function must have either a single return value, or two +// return values of which the second has type error. In that case, if the +// second (error) argument evaluates to non-nil during execution, execution +// terminates and Execute returns that error. FuncMap has the same base type +// as FuncMap in "text/template", copied here so clients need not import +// "text/template". +type FuncMap map[string]interface{} + +// Funcs adds the elements of the argument map to the template's function map. +// It panics if a value in the map is not a function with appropriate return +// type. However, it is legal to overwrite elements of the map. The return +// value is the template, so calls can be chained. +func (t *Template) Funcs(funcMap FuncMap) *Template { + t.text.Funcs(template.FuncMap(funcMap)) + return t +} + +// Delims sets the action delimiters to the specified strings, to be used in +// subsequent calls to Parse, ParseFiles, or ParseGlob. Nested template +// definitions will inherit the settings. An empty delimiter stands for the +// corresponding default: {{ or }}. +// The return value is the template, so calls can be chained. +func (t *Template) Delims(left, right string) *Template { + t.text.Delims(left, right) + return t +} + +// Lookup returns the template with the given name that is associated with t, +// or nil if there is no such template. +func (t *Template) Lookup(name string) *Template { + t.nameSpace.mu.Lock() + defer t.nameSpace.mu.Unlock() + return t.set[name] +} + +// Must is a helper that wraps a call to a function returning (*Template, error) +// and panics if the error is non-nil. It is intended for use in variable initializations +// such as +// var t = template.Must(template.New("name").Parse("html")) +func Must(t *Template, err error) *Template { + if err != nil { + panic(err) + } + return t +} + +// ParseFiles creates a new Template and parses the template definitions from +// the named files. The returned template's name will have the (base) name and +// (parsed) contents of the first file. There must be at least one file. +// If an error occurs, parsing stops and the returned *Template is nil. +func ParseFiles(filenames ...string) (*Template, error) { + return parseFiles(nil, filenames...) +} + +// ParseFiles parses the named files and associates the resulting templates with +// t. If an error occurs, parsing stops and the returned template is nil; +// otherwise it is t. There must be at least one file. +func (t *Template) ParseFiles(filenames ...string) (*Template, error) { + return parseFiles(t, filenames...) +} + +// parseFiles is the helper for the method and function. If the argument +// template is nil, it is created from the first file. +func parseFiles(t *Template, filenames ...string) (*Template, error) { + if len(filenames) == 0 { + // Not really a problem, but be consistent. + return nil, fmt.Errorf("html/template: no files named in call to ParseFiles") + } + for _, filename := range filenames { + b, err := ioutil.ReadFile(filename) + if err != nil { + return nil, err + } + s := string(b) + // name := filepath.Base(filename) + name := filename + // First template becomes return value if not already defined, + // and we use that one for subsequent New calls to associate + // all the templates together. Also, if this file has the same name + // as t, this file becomes the contents of t, so + // t, err := New(name).Funcs(xxx).ParseFiles(name) + // works. Otherwise we create a new template associated with t. + var tmpl *Template + if t == nil { + t = New(name) + } + if name == t.Name() { + tmpl = t + } else { + tmpl = t.New(name) + } + _, err = tmpl.Parse(s) + if err != nil { + return nil, err + } + } + return t, nil +} + +// ParseGlob creates a new Template and parses the template definitions from the +// files identified by the pattern, which must match at least one file. The +// returned template will have the (base) name and (parsed) contents of the +// first file matched by the pattern. ParseGlob is equivalent to calling +// ParseFiles with the list of files matched by the pattern. +func ParseGlob(pattern string) (*Template, error) { + return parseGlob(nil, pattern) +} + +// ParseGlob parses the template definitions in the files identified by the +// pattern and associates the resulting templates with t. The pattern is +// processed by filepath.Glob and must match at least one file. ParseGlob is +// equivalent to calling t.ParseFiles with the list of files matched by the +// pattern. +func (t *Template) ParseGlob(pattern string) (*Template, error) { + return parseGlob(t, pattern) +} + +// parseGlob is the implementation of the function and method ParseGlob. +func parseGlob(t *Template, pattern string) (*Template, error) { + filenames, err := filepath.Glob(pattern) + if err != nil { + return nil, err + } + if len(filenames) == 0 { + return nil, fmt.Errorf("html/template: pattern matches no files: %#q", pattern) + } + return parseFiles(t, filenames...) +} diff --git a/template/transition.go b/template/transition.go new file mode 100644 index 00000000..b486fcd2 --- /dev/null +++ b/template/transition.go @@ -0,0 +1,550 @@ +// 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 template + +import ( + "bytes" + "strings" +) + +// transitionFunc is the array of context transition functions for text nodes. +// A transition function takes a context and template text input, and returns +// the updated context and the number of bytes consumed from the front of the +// input. +var transitionFunc = [...]func(context, []byte) (context, int){ + stateText: tText, + stateTag: tTag, + stateAttrName: tAttrName, + stateAfterName: tAfterName, + stateBeforeValue: tBeforeValue, + stateHTMLCmt: tHTMLCmt, + stateRCDATA: tSpecialTagEnd, + stateAttr: tAttr, + stateURL: tURL, + stateJS: tJS, + stateJSDqStr: tJSDelimited, + stateJSSqStr: tJSDelimited, + stateJSRegexp: tJSDelimited, + stateJSBlockCmt: tBlockCmt, + stateJSLineCmt: tLineCmt, + stateCSS: tCSS, + stateCSSDqStr: tCSSStr, + stateCSSSqStr: tCSSStr, + stateCSSDqURL: tCSSStr, + stateCSSSqURL: tCSSStr, + stateCSSURL: tCSSStr, + stateCSSBlockCmt: tBlockCmt, + stateCSSLineCmt: tLineCmt, + stateError: tError, +} + +var commentStart = []byte("<!--") +var commentEnd = []byte("-->") + +// tText is the context transition function for the text state. +func tText(c context, s []byte) (context, int) { + k := 0 + for { + i := k + bytes.IndexByte(s[k:], '<') + if i < k || i+1 == len(s) { + return c, len(s) + } else if i+4 <= len(s) && bytes.Equal(commentStart, s[i:i+4]) { + return context{state: stateHTMLCmt}, i + 4 + } + i++ + end := false + if s[i] == '/' { + if i+1 == len(s) { + return c, len(s) + } + 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 + } + k = j + } +} + +var elementContentType = [...]state{ + elementNone: stateText, + elementScript: stateJS, + elementStyle: stateCSS, + elementTextarea: stateRCDATA, + elementTitle: stateRCDATA, +} + +// tTag is the context transition function for the tag state. +func tTag(c context, s []byte) (context, int) { + // Find the attribute name. + i := eatWhiteSpace(s, 0) + if i == len(s) { + return c, len(s) + } + if s[i] == '>' { + return context{ + state: elementContentType[c.element], + element: c.element, + }, i + 1 + } + j, err := eatAttrName(s, i) + if err != nil { + return context{state: stateError, err: err}, len(s) + } + state, attr := stateTag, attrNone + if i == j { + return context{ + state: stateError, + err: errorf(ErrBadHTML, nil, 0, "expected space, attr name, or end of tag, but got %q", s[i:]), + }, len(s) + } + switch attrType(string(s[i:j])) { + case contentTypeURL: + attr = attrURL + case contentTypeCSS: + attr = attrStyle + case contentTypeJS: + attr = attrScript + } + if j == len(s) { + state = stateAttrName + } else { + state = stateAfterName + } + return context{state: state, element: c.element, attr: attr}, j +} + +// tAttrName is the context transition function for stateAttrName. +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) { + c.state = stateAfterName + } + return c, i +} + +// tAfterName is the context transition function for stateAfterName. +func tAfterName(c context, s []byte) (context, int) { + // Look for the start of the value. + i := eatWhiteSpace(s, 0) + if i == len(s) { + return c, len(s) + } else if s[i] != '=' { + // Occurs due to tag ending '>', and valueless attribute. + c.state = stateTag + return c, i + } + c.state = stateBeforeValue + // Consume the "=". + return c, i + 1 +} + +var attrStartStates = [...]state{ + attrNone: stateAttr, + attrScript: stateJS, + attrStyle: stateCSS, + attrURL: stateURL, +} + +// tBeforeValue is the context transition function for stateBeforeValue. +func tBeforeValue(c context, s []byte) (context, int) { + i := eatWhiteSpace(s, 0) + if i == len(s) { + return c, len(s) + } + // Find the attribute delimiter. + delim := delimSpaceOrTagEnd + switch s[i] { + case '\'': + delim, i = delimSingleQuote, i+1 + case '"': + delim, i = delimDoubleQuote, i+1 + } + c.state, c.delim, c.attr = attrStartStates[c.attr], delim, attrNone + return c, i +} + +// tHTMLCmt is the context transition function for stateHTMLCmt. +func tHTMLCmt(c context, s []byte) (context, int) { + if i := bytes.Index(s, commentEnd); i != -1 { + return context{}, i + 3 + } + return c, len(s) +} + +// specialTagEndMarkers maps element types to the character sequence that +// case-insensitively signals the end of the special tag body. +var specialTagEndMarkers = [...]string{ + elementScript: "</script", + elementStyle: "</style", + elementTextarea: "</textarea", + elementTitle: "</title", +} + +// tSpecialTagEnd is the context transition function for raw text and RCDATA +// element states. +func tSpecialTagEnd(c context, s []byte) (context, int) { + if c.element != elementNone { + if i := strings.Index(strings.ToLower(string(s)), specialTagEndMarkers[c.element]); i != -1 { + return context{}, i + } + } + return c, len(s) +} + +// tAttr is the context transition function for the attribute state. +func tAttr(c context, s []byte) (context, int) { + return c, len(s) +} + +// tURL is the context transition function for the URL state. +func tURL(c context, s []byte) (context, int) { + if bytes.IndexAny(s, "#?") >= 0 { + c.urlPart = urlPartQueryOrFrag + } else if len(s) != eatWhiteSpace(s, 0) && c.urlPart == urlPartNone { + // HTML5 uses "Valid URL potentially surrounded by spaces" for + // attrs: http://www.w3.org/TR/html5/index.html#attributes-1 + c.urlPart = urlPartPreQuery + } + return c, len(s) +} + +// tJS is the context transition function for the JS state. +func tJS(c context, s []byte) (context, int) { + i := bytes.IndexAny(s, `"'/`) + if i == -1 { + // Entire input is non string, comment, regexp tokens. + c.jsCtx = nextJSCtx(s, c.jsCtx) + return c, len(s) + } + c.jsCtx = nextJSCtx(s[:i], c.jsCtx) + switch s[i] { + case '"': + c.state, c.jsCtx = stateJSDqStr, jsCtxRegexp + case '\'': + c.state, c.jsCtx = stateJSSqStr, jsCtxRegexp + case '/': + switch { + case i+1 < len(s) && s[i+1] == '/': + c.state, i = stateJSLineCmt, i+1 + case i+1 < len(s) && s[i+1] == '*': + c.state, i = stateJSBlockCmt, i+1 + case c.jsCtx == jsCtxRegexp: + c.state = stateJSRegexp + case c.jsCtx == jsCtxDivOp: + c.jsCtx = jsCtxRegexp + default: + return context{ + state: stateError, + err: errorf(ErrSlashAmbig, nil, 0, "'/' could start a division or regexp: %.32q", s[i:]), + }, len(s) + } + default: + panic("unreachable") + } + return c, i + 1 +} + +// 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:], specials) + if i < k { + break + } + switch s[i] { + case '\\': + i++ + if i == len(s) { + return context{ + state: stateError, + err: errorf(ErrPartialEscape, nil, 0, "unfinished escape sequence in JS string: %q", s), + }, len(s) + } + case '[': + inCharset = true + case ']': + inCharset = false + default: + // end delimiter + if !inCharset { + c.state, c.jsCtx = stateJS, jsCtxDivOp + return c, i + 1 + } + } + k = i + 1 + } + + if inCharset { + // This can be fixed by making context richer if interpolation + // into charsets is desired. + return context{ + state: stateError, + err: errorf(ErrPartialCharset, nil, 0, "unfinished JS regexp charset: %q", s), + }, len(s) + } + + return c, len(s) +} + +var blockCommentEnd = []byte("*/") + +// tBlockCmt is the context transition function for /*comment*/ states. +func tBlockCmt(c context, s []byte) (context, int) { + i := bytes.Index(s, blockCommentEnd) + if i == -1 { + return c, len(s) + } + switch c.state { + case stateJSBlockCmt: + c.state = stateJS + case stateCSSBlockCmt: + c.state = stateCSS + default: + panic(c.state.String()) + } + return c, i + 2 +} + +// tLineCmt is the context transition function for //comment states. +func tLineCmt(c context, s []byte) (context, int) { + 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, lineTerminators) + if i == -1 { + return c, len(s) + } + 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 + // recognized separately by the lexical grammar and becomes part + // of the stream of input elements for the syntactic grammar." + return c, i +} + +// tCSS is the context transition function for the CSS state. +func tCSS(c context, s []byte) (context, int) { + // 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. + + k := 0 + for { + i := k + bytes.IndexAny(s[k:], `("'/`) + if i < k { + return c, len(s) + } + switch s[i] { + case '(': + // Look for url to the left. + p := bytes.TrimRight(s[:i], "\t\n\f\r ") + if endsWithCSSKeyword(p, "url") { + j := len(s) - len(bytes.TrimLeft(s[i+1:], "\t\n\f\r ")) + switch { + case j != len(s) && s[j] == '"': + c.state, j = stateCSSDqURL, j+1 + case j != len(s) && s[j] == '\'': + c.state, j = stateCSSSqURL, j+1 + default: + c.state = stateCSSURL + } + return c, j + } + case '/': + if i+1 < len(s) { + switch s[i+1] { + case '/': + c.state = stateCSSLineCmt + return c, i + 2 + case '*': + c.state = stateCSSBlockCmt + return c, i + 2 + } + } + case '"': + c.state = stateCSSDqStr + return c, i + 1 + case '\'': + c.state = stateCSSSqStr + return c, i + 1 + } + k = i + 1 + } +} + +// tCSSStr is the context transition function for the CSS string and URL states. +func tCSSStr(c context, s []byte) (context, int) { + 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()) + } + + k := 0 + for { + i := k + bytes.IndexAny(s[k:], endAndEsc) + if i < k { + c, nread := tURL(c, decodeCSS(s[k:])) + return c, k + nread + } + if s[i] == '\\' { + i++ + if i == len(s) { + return context{ + state: stateError, + err: errorf(ErrPartialEscape, nil, 0, "unfinished escape sequence in CSS string: %q", s), + }, len(s) + } + } else { + c.state = stateCSS + return c, i + 1 + } + c, _ = tURL(c, decodeCSS(s[:i+1])) + k = i + 1 + } +} + +// tError is the context transition function for the error state. +func tError(c context, s []byte) (context, int) { + return c, len(s) +} + +// eatAttrName returns the largest j such that s[i:j] is an attribute name. +// It returns an error if s[i:] does not look like it begins with an +// attribute name, such as encountering a quote mark without a preceding +// equals sign. +func eatAttrName(s []byte, i int) (int, *Error) { + for j := i; j < len(s); j++ { + switch s[j] { + case ' ', '\t', '\n', '\f', '\r', '=', '>': + return j, nil + case '\'', '"', '<': + // These result in a parse warning in HTML5 and are + // indicative of serious problems if seen in an attr + // name in a template. + return -1, errorf(ErrBadHTML, nil, 0, "%q in attribute name: %.32q", s[j:j+1], s) + default: + // No-op. + } + } + return len(s), nil +} + +var elementNameMap = map[string]element{ + "script": elementScript, + "style": elementStyle, + "textarea": elementTextarea, + "title": elementTitle, +} + +// asciiAlpha reports whether c is an ASCII letter. +func asciiAlpha(c byte) bool { + return 'A' <= c && c <= 'Z' || 'a' <= c && c <= 'z' +} + +// asciiAlphaNum reports whether c is an ASCII letter or digit. +func asciiAlphaNum(c byte) bool { + return asciiAlpha(c) || '0' <= c && c <= '9' +} + +// eatTagName returns the largest j such that s[i:j] is a tag name and the tag type. +func eatTagName(s []byte, i int) (int, element) { + if i == len(s) || !asciiAlpha(s[i]) { + return i, elementNone + } + j := i + 1 + for j < len(s) { + x := s[j] + if asciiAlphaNum(x) { + j++ + continue + } + // Allow "x-y" or "x:y" but not "x-", "-y", or "x--y". + if (x == ':' || x == '-') && j+1 < len(s) && asciiAlphaNum(s[j+1]) { + j += 2 + continue + } + break + } + return j, elementNameMap[strings.ToLower(string(s[i:j]))] +} + +// eatWhiteSpace returns the largest j such that s[i:j] is white space. +func eatWhiteSpace(s []byte, i int) int { + for j := i; j < len(s); j++ { + switch s[j] { + case ' ', '\t', '\n', '\f', '\r': + // No-op. + default: + return j + } + } + return len(s) +} diff --git a/template/url.go b/template/url.go new file mode 100644 index 00000000..2ca76bf3 --- /dev/null +++ b/template/url.go @@ -0,0 +1,105 @@ +// 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 template + +import ( + "bytes" + "fmt" + "strings" +) + +// urlFilter returns its input unless it contains an unsafe protocol in which +// case it defangs the entire URL. +func urlFilter(args ...interface{}) string { + s, t := stringify(args...) + if t == contentTypeURL { + return s + } + if i := strings.IndexRune(s, ':'); i >= 0 && strings.IndexRune(s[:i], '/') < 0 { + protocol := strings.ToLower(s[:i]) + if protocol != "http" && protocol != "https" && protocol != "mailto" { + return "#" + filterFailsafe + } + } + 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 '&'. +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, t := stringify(args...) + if t == contentTypeURL { + norm = true + } + 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 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() +} diff --git a/template/url_test.go b/template/url_test.go new file mode 100644 index 00000000..5182e9d7 --- /dev/null +++ b/template/url_test.go @@ -0,0 +1,112 @@ +// 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 template + +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") + } +}