Source file src/pkg/html/template/js.go

     1	// Copyright 2011 The Go Authors. All rights reserved.
     2	// Use of this source code is governed by a BSD-style
     3	// license that can be found in the LICENSE file.
     4	
     5	package template
     6	
     7	import (
     8		"bytes"
     9		"encoding/json"
    10		"fmt"
    11		"reflect"
    12		"strings"
    13		"unicode/utf8"
    14	)
    15	
    16	// nextJSCtx returns the context that determines whether a slash after the
    17	// given run of tokens tokens starts a regular expression instead of a division
    18	// operator: / or /=.
    19	//
    20	// This assumes that the token run does not include any string tokens, comment
    21	// tokens, regular expression literal tokens, or division operators.
    22	//
    23	// This fails on some valid but nonsensical JavaScript programs like
    24	// "x = ++/foo/i" which is quite different than "x++/foo/i", but is not known to
    25	// fail on any known useful programs. It is based on the draft
    26	// JavaScript 2.0 lexical grammar and requires one token of lookbehind:
    27	// http://www.mozilla.org/js/language/js20-2000-07/rationale/syntax.html
    28	func nextJSCtx(s []byte, preceding jsCtx) jsCtx {
    29		s = bytes.TrimRight(s, "\t\n\f\r \u2028\u2029")
    30		if len(s) == 0 {
    31			return preceding
    32		}
    33	
    34		// All cases below are in the single-byte UTF-8 group.
    35		switch c, n := s[len(s)-1], len(s); c {
    36		case '+', '-':
    37			// ++ and -- are not regexp preceders, but + and - are whether
    38			// they are used as infix or prefix operators.
    39			start := n - 1
    40			// Count the number of adjacent dashes or pluses.
    41			for start > 0 && s[start-1] == c {
    42				start--
    43			}
    44			if (n-start)&1 == 1 {
    45				// Reached for trailing minus signs since "---" is the
    46				// same as "-- -".
    47				return jsCtxRegexp
    48			}
    49			return jsCtxDivOp
    50		case '.':
    51			// Handle "42."
    52			if n != 1 && '0' <= s[n-2] && s[n-2] <= '9' {
    53				return jsCtxDivOp
    54			}
    55			return jsCtxRegexp
    56		// Suffixes for all punctuators from section 7.7 of the language spec
    57		// that only end binary operators not handled above.
    58		case ',', '<', '>', '=', '*', '%', '&', '|', '^', '?':
    59			return jsCtxRegexp
    60		// Suffixes for all punctuators from section 7.7 of the language spec
    61		// that are prefix operators not handled above.
    62		case '!', '~':
    63			return jsCtxRegexp
    64		// Matches all the punctuators from section 7.7 of the language spec
    65		// that are open brackets not handled above.
    66		case '(', '[':
    67			return jsCtxRegexp
    68		// Matches all the punctuators from section 7.7 of the language spec
    69		// that precede expression starts.
    70		case ':', ';', '{':
    71			return jsCtxRegexp
    72		// CAVEAT: the close punctuators ('}', ']', ')') precede div ops and
    73		// are handled in the default except for '}' which can precede a
    74		// division op as in
    75		//    ({ valueOf: function () { return 42 } } / 2
    76		// which is valid, but, in practice, developers don't divide object
    77		// literals, so our heuristic works well for code like
    78		//    function () { ... }  /foo/.test(x) && sideEffect();
    79		// The ')' punctuator can precede a regular expression as in
    80		//     if (b) /foo/.test(x) && ...
    81		// but this is much less likely than
    82		//     (a + b) / c
    83		case '}':
    84			return jsCtxRegexp
    85		default:
    86			// Look for an IdentifierName and see if it is a keyword that
    87			// can precede a regular expression.
    88			j := n
    89			for j > 0 && isJSIdentPart(rune(s[j-1])) {
    90				j--
    91			}
    92			if regexpPrecederKeywords[string(s[j:])] {
    93				return jsCtxRegexp
    94			}
    95		}
    96		// Otherwise is a punctuator not listed above, or
    97		// a string which precedes a div op, or an identifier
    98		// which precedes a div op.
    99		return jsCtxDivOp
   100	}
   101	
   102	// regexPrecederKeywords is a set of reserved JS keywords that can precede a
   103	// regular expression in JS source.
   104	var regexpPrecederKeywords = map[string]bool{
   105		"break":      true,
   106		"case":       true,
   107		"continue":   true,
   108		"delete":     true,
   109		"do":         true,
   110		"else":       true,
   111		"finally":    true,
   112		"in":         true,
   113		"instanceof": true,
   114		"return":     true,
   115		"throw":      true,
   116		"try":        true,
   117		"typeof":     true,
   118		"void":       true,
   119	}
   120	
   121	var jsonMarshalType = reflect.TypeOf((*json.Marshaler)(nil)).Elem()
   122	
   123	// indirectToJSONMarshaler returns the value, after dereferencing as many times
   124	// as necessary to reach the base type (or nil) or an implementation of json.Marshal.
   125	func indirectToJSONMarshaler(a interface{}) interface{} {
   126		v := reflect.ValueOf(a)
   127		for !v.Type().Implements(jsonMarshalType) && v.Kind() == reflect.Ptr && !v.IsNil() {
   128			v = v.Elem()
   129		}
   130		return v.Interface()
   131	}
   132	
   133	// jsValEscaper escapes its inputs to a JS Expression (section 11.14) that has
   134	// neither side-effects nor free variables outside (NaN, Infinity).
   135	func jsValEscaper(args ...interface{}) string {
   136		var a interface{}
   137		if len(args) == 1 {
   138			a = indirectToJSONMarshaler(args[0])
   139			switch t := a.(type) {
   140			case JS:
   141				return string(t)
   142			case JSStr:
   143				// TODO: normalize quotes.
   144				return `"` + string(t) + `"`
   145			case json.Marshaler:
   146				// Do not treat as a Stringer.
   147			case fmt.Stringer:
   148				a = t.String()
   149			}
   150		} else {
   151			for i, arg := range args {
   152				args[i] = indirectToJSONMarshaler(arg)
   153			}
   154			a = fmt.Sprint(args...)
   155		}
   156		// TODO: detect cycles before calling Marshal which loops infinitely on
   157		// cyclic data. This may be an unacceptable DoS risk.
   158	
   159		b, err := json.Marshal(a)
   160		if err != nil {
   161			// Put a space before comment so that if it is flush against
   162			// a division operator it is not turned into a line comment:
   163			//     x/{{y}}
   164			// turning into
   165			//     x//* error marshalling y:
   166			//          second line of error message */null
   167			return fmt.Sprintf(" /* %s */null ", strings.Replace(err.Error(), "*/", "* /", -1))
   168		}
   169	
   170		// TODO: maybe post-process output to prevent it from containing
   171		// "<!--", "-->", "<![CDATA[", "]]>", or "</script"
   172		// in case custom marshallers produce output containing those.
   173	
   174		// TODO: Maybe abbreviate \u00ab to \xab to produce more compact output.
   175		if len(b) == 0 {
   176			// In, `x=y/{{.}}*z` a json.Marshaler that produces "" should
   177			// not cause the output `x=y/*z`.
   178			return " null "
   179		}
   180		first, _ := utf8.DecodeRune(b)
   181		last, _ := utf8.DecodeLastRune(b)
   182		var buf bytes.Buffer
   183		// Prevent IdentifierNames and NumericLiterals from running into
   184		// keywords: in, instanceof, typeof, void
   185		pad := isJSIdentPart(first) || isJSIdentPart(last)
   186		if pad {
   187			buf.WriteByte(' ')
   188		}
   189		written := 0
   190		// Make sure that json.Marshal escapes codepoints U+2028 & U+2029
   191		// so it falls within the subset of JSON which is valid JS.
   192		for i := 0; i < len(b); {
   193			rune, n := utf8.DecodeRune(b[i:])
   194			repl := ""
   195			if rune == 0x2028 {
   196				repl = `\u2028`
   197			} else if rune == 0x2029 {
   198				repl = `\u2029`
   199			}
   200			if repl != "" {
   201				buf.Write(b[written:i])
   202				buf.WriteString(repl)
   203				written = i + n
   204			}
   205			i += n
   206		}
   207		if buf.Len() != 0 {
   208			buf.Write(b[written:])
   209			if pad {
   210				buf.WriteByte(' ')
   211			}
   212			b = buf.Bytes()
   213		}
   214		return string(b)
   215	}
   216	
   217	// jsStrEscaper produces a string that can be included between quotes in
   218	// JavaScript source, in JavaScript embedded in an HTML5 <script> element,
   219	// or in an HTML5 event handler attribute such as onclick.
   220	func jsStrEscaper(args ...interface{}) string {
   221		s, t := stringify(args...)
   222		if t == contentTypeJSStr {
   223			return replace(s, jsStrNormReplacementTable)
   224		}
   225		return replace(s, jsStrReplacementTable)
   226	}
   227	
   228	// jsRegexpEscaper behaves like jsStrEscaper but escapes regular expression
   229	// specials so the result is treated literally when included in a regular
   230	// expression literal. /foo{{.X}}bar/ matches the string "foo" followed by
   231	// the literal text of {{.X}} followed by the string "bar".
   232	func jsRegexpEscaper(args ...interface{}) string {
   233		s, _ := stringify(args...)
   234		s = replace(s, jsRegexpReplacementTable)
   235		if s == "" {
   236			// /{{.X}}/ should not produce a line comment when .X == "".
   237			return "(?:)"
   238		}
   239		return s
   240	}
   241	
   242	// replace replaces each rune r of s with replacementTable[r], provided that
   243	// r < len(replacementTable). If replacementTable[r] is the empty string then
   244	// no replacement is made.
   245	// It also replaces runes U+2028 and U+2029 with the raw strings `\u2028` and
   246	// `\u2029`.
   247	func replace(s string, replacementTable []string) string {
   248		var b bytes.Buffer
   249		written := 0
   250		for i, r := range s {
   251			var repl string
   252			switch {
   253			case int(r) < len(replacementTable) && replacementTable[r] != "":
   254				repl = replacementTable[r]
   255			case r == '\u2028':
   256				repl = `\u2028`
   257			case r == '\u2029':
   258				repl = `\u2029`
   259			default:
   260				continue
   261			}
   262			b.WriteString(s[written:i])
   263			b.WriteString(repl)
   264			written = i + utf8.RuneLen(r)
   265		}
   266		if written == 0 {
   267			return s
   268		}
   269		b.WriteString(s[written:])
   270		return b.String()
   271	}
   272	
   273	var jsStrReplacementTable = []string{
   274		0:    `\0`,
   275		'\t': `\t`,
   276		'\n': `\n`,
   277		'\v': `\x0b`, // "\v" == "v" on IE 6.
   278		'\f': `\f`,
   279		'\r': `\r`,
   280		// Encode HTML specials as hex so the output can be embedded
   281		// in HTML attributes without further encoding.
   282		'"':  `\x22`,
   283		'&':  `\x26`,
   284		'\'': `\x27`,
   285		'+':  `\x2b`,
   286		'/':  `\/`,
   287		'<':  `\x3c`,
   288		'>':  `\x3e`,
   289		'\\': `\\`,
   290	}
   291	
   292	// jsStrNormReplacementTable is like jsStrReplacementTable but does not
   293	// overencode existing escapes since this table has no entry for `\`.
   294	var jsStrNormReplacementTable = []string{
   295		0:    `\0`,
   296		'\t': `\t`,
   297		'\n': `\n`,
   298		'\v': `\x0b`, // "\v" == "v" on IE 6.
   299		'\f': `\f`,
   300		'\r': `\r`,
   301		// Encode HTML specials as hex so the output can be embedded
   302		// in HTML attributes without further encoding.
   303		'"':  `\x22`,
   304		'&':  `\x26`,
   305		'\'': `\x27`,
   306		'+':  `\x2b`,
   307		'/':  `\/`,
   308		'<':  `\x3c`,
   309		'>':  `\x3e`,
   310	}
   311	
   312	var jsRegexpReplacementTable = []string{
   313		0:    `\0`,
   314		'\t': `\t`,
   315		'\n': `\n`,
   316		'\v': `\x0b`, // "\v" == "v" on IE 6.
   317		'\f': `\f`,
   318		'\r': `\r`,
   319		// Encode HTML specials as hex so the output can be embedded
   320		// in HTML attributes without further encoding.
   321		'"':  `\x22`,
   322		'$':  `\$`,
   323		'&':  `\x26`,
   324		'\'': `\x27`,
   325		'(':  `\(`,
   326		')':  `\)`,
   327		'*':  `\*`,
   328		'+':  `\x2b`,
   329		'-':  `\-`,
   330		'.':  `\.`,
   331		'/':  `\/`,
   332		'<':  `\x3c`,
   333		'>':  `\x3e`,
   334		'?':  `\?`,
   335		'[':  `\[`,
   336		'\\': `\\`,
   337		']':  `\]`,
   338		'^':  `\^`,
   339		'{':  `\{`,
   340		'|':  `\|`,
   341		'}':  `\}`,
   342	}
   343	
   344	// isJSIdentPart returns whether the given rune is a JS identifier part.
   345	// It does not handle all the non-Latin letters, joiners, and combining marks,
   346	// but it does handle every codepoint that can occur in a numeric literal or
   347	// a keyword.
   348	func isJSIdentPart(r rune) bool {
   349		switch {
   350		case r == '$':
   351			return true
   352		case '0' <= r && r <= '9':
   353			return true
   354		case 'A' <= r && r <= 'Z':
   355			return true
   356		case r == '_':
   357			return true
   358		case 'a' <= r && r <= 'z':
   359			return true
   360		}
   361		return false
   362	}