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  // jsWhitespace contains all of the JS whitespace characters, as defined
    17  // by the \s character class.
    18  // See https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Regular_expressions/Character_classes.
    19  const jsWhitespace = "\f\n\r\t\v\u0020\u00a0\u1680\u2000\u2001\u2002\u2003\u2004\u2005\u2006\u2007\u2008\u2009\u200a\u2028\u2029\u202f\u205f\u3000\ufeff"
    20  
    21  // nextJSCtx returns the context that determines whether a slash after the
    22  // given run of tokens starts a regular expression instead of a division
    23  // operator: / or /=.
    24  //
    25  // This assumes that the token run does not include any string tokens, comment
    26  // tokens, regular expression literal tokens, or division operators.
    27  //
    28  // This fails on some valid but nonsensical JavaScript programs like
    29  // "x = ++/foo/i" which is quite different than "x++/foo/i", but is not known to
    30  // fail on any known useful programs. It is based on the draft
    31  // JavaScript 2.0 lexical grammar and requires one token of lookbehind:
    32  // https://www.mozilla.org/js/language/js20-2000-07/rationale/syntax.html
    33  func nextJSCtx(s []byte, preceding jsCtx) jsCtx {
    34  	// Trim all JS whitespace characters
    35  	s = bytes.TrimRight(s, jsWhitespace)
    36  	if len(s) == 0 {
    37  		return preceding
    38  	}
    39  
    40  	// All cases below are in the single-byte UTF-8 group.
    41  	switch c, n := s[len(s)-1], len(s); c {
    42  	case '+', '-':
    43  		// ++ and -- are not regexp preceders, but + and - are whether
    44  		// they are used as infix or prefix operators.
    45  		start := n - 1
    46  		// Count the number of adjacent dashes or pluses.
    47  		for start > 0 && s[start-1] == c {
    48  			start--
    49  		}
    50  		if (n-start)&1 == 1 {
    51  			// Reached for trailing minus signs since "---" is the
    52  			// same as "-- -".
    53  			return jsCtxRegexp
    54  		}
    55  		return jsCtxDivOp
    56  	case '.':
    57  		// Handle "42."
    58  		if n != 1 && '0' <= s[n-2] && s[n-2] <= '9' {
    59  			return jsCtxDivOp
    60  		}
    61  		return jsCtxRegexp
    62  	// Suffixes for all punctuators from section 7.7 of the language spec
    63  	// that only end binary operators not handled above.
    64  	case ',', '<', '>', '=', '*', '%', '&', '|', '^', '?':
    65  		return jsCtxRegexp
    66  	// Suffixes for all punctuators from section 7.7 of the language spec
    67  	// that are prefix operators not handled above.
    68  	case '!', '~':
    69  		return jsCtxRegexp
    70  	// Matches all the punctuators from section 7.7 of the language spec
    71  	// that are open brackets not handled above.
    72  	case '(', '[':
    73  		return jsCtxRegexp
    74  	// Matches all the punctuators from section 7.7 of the language spec
    75  	// that precede expression starts.
    76  	case ':', ';', '{':
    77  		return jsCtxRegexp
    78  	// CAVEAT: the close punctuators ('}', ']', ')') precede div ops and
    79  	// are handled in the default except for '}' which can precede a
    80  	// division op as in
    81  	//    ({ valueOf: function () { return 42 } } / 2
    82  	// which is valid, but, in practice, developers don't divide object
    83  	// literals, so our heuristic works well for code like
    84  	//    function () { ... }  /foo/.test(x) && sideEffect();
    85  	// The ')' punctuator can precede a regular expression as in
    86  	//     if (b) /foo/.test(x) && ...
    87  	// but this is much less likely than
    88  	//     (a + b) / c
    89  	case '}':
    90  		return jsCtxRegexp
    91  	default:
    92  		// Look for an IdentifierName and see if it is a keyword that
    93  		// can precede a regular expression.
    94  		j := n
    95  		for j > 0 && isJSIdentPart(rune(s[j-1])) {
    96  			j--
    97  		}
    98  		if regexpPrecederKeywords[string(s[j:])] {
    99  			return jsCtxRegexp
   100  		}
   101  	}
   102  	// Otherwise is a punctuator not listed above, or
   103  	// a string which precedes a div op, or an identifier
   104  	// which precedes a div op.
   105  	return jsCtxDivOp
   106  }
   107  
   108  // regexpPrecederKeywords is a set of reserved JS keywords that can precede a
   109  // regular expression in JS source.
   110  var regexpPrecederKeywords = map[string]bool{
   111  	"break":      true,
   112  	"case":       true,
   113  	"continue":   true,
   114  	"delete":     true,
   115  	"do":         true,
   116  	"else":       true,
   117  	"finally":    true,
   118  	"in":         true,
   119  	"instanceof": true,
   120  	"return":     true,
   121  	"throw":      true,
   122  	"try":        true,
   123  	"typeof":     true,
   124  	"void":       true,
   125  }
   126  
   127  var jsonMarshalType = reflect.TypeOf((*json.Marshaler)(nil)).Elem()
   128  
   129  // indirectToJSONMarshaler returns the value, after dereferencing as many times
   130  // as necessary to reach the base type (or nil) or an implementation of json.Marshal.
   131  func indirectToJSONMarshaler(a any) any {
   132  	// text/template now supports passing untyped nil as a func call
   133  	// argument, so we must support it. Otherwise we'd panic below, as one
   134  	// cannot call the Type or Interface methods on an invalid
   135  	// reflect.Value. See golang.org/issue/18716.
   136  	if a == nil {
   137  		return nil
   138  	}
   139  
   140  	v := reflect.ValueOf(a)
   141  	for !v.Type().Implements(jsonMarshalType) && v.Kind() == reflect.Pointer && !v.IsNil() {
   142  		v = v.Elem()
   143  	}
   144  	return v.Interface()
   145  }
   146  
   147  // jsValEscaper escapes its inputs to a JS Expression (section 11.14) that has
   148  // neither side-effects nor free variables outside (NaN, Infinity).
   149  func jsValEscaper(args ...any) string {
   150  	var a any
   151  	if len(args) == 1 {
   152  		a = indirectToJSONMarshaler(args[0])
   153  		switch t := a.(type) {
   154  		case JS:
   155  			return string(t)
   156  		case JSStr:
   157  			// TODO: normalize quotes.
   158  			return `"` + string(t) + `"`
   159  		case json.Marshaler:
   160  			// Do not treat as a Stringer.
   161  		case fmt.Stringer:
   162  			a = t.String()
   163  		}
   164  	} else {
   165  		for i, arg := range args {
   166  			args[i] = indirectToJSONMarshaler(arg)
   167  		}
   168  		a = fmt.Sprint(args...)
   169  	}
   170  	// TODO: detect cycles before calling Marshal which loops infinitely on
   171  	// cyclic data. This may be an unacceptable DoS risk.
   172  	b, err := json.Marshal(a)
   173  	if err != nil {
   174  		// Put a space before comment so that if it is flush against
   175  		// a division operator it is not turned into a line comment:
   176  		//     x/{{y}}
   177  		// turning into
   178  		//     x//* error marshaling y:
   179  		//          second line of error message */null
   180  		return fmt.Sprintf(" /* %s */null ", strings.ReplaceAll(err.Error(), "*/", "* /"))
   181  	}
   182  
   183  	// TODO: maybe post-process output to prevent it from containing
   184  	// "<!--", "-->", "<![CDATA[", "]]>", or "</script"
   185  	// in case custom marshalers produce output containing those.
   186  	// Note: Do not use \x escaping to save bytes because it is not JSON compatible and this escaper
   187  	// supports ld+json content-type.
   188  	if len(b) == 0 {
   189  		// In, `x=y/{{.}}*z` a json.Marshaler that produces "" should
   190  		// not cause the output `x=y/*z`.
   191  		return " null "
   192  	}
   193  	first, _ := utf8.DecodeRune(b)
   194  	last, _ := utf8.DecodeLastRune(b)
   195  	var buf strings.Builder
   196  	// Prevent IdentifierNames and NumericLiterals from running into
   197  	// keywords: in, instanceof, typeof, void
   198  	pad := isJSIdentPart(first) || isJSIdentPart(last)
   199  	if pad {
   200  		buf.WriteByte(' ')
   201  	}
   202  	written := 0
   203  	// Make sure that json.Marshal escapes codepoints U+2028 & U+2029
   204  	// so it falls within the subset of JSON which is valid JS.
   205  	for i := 0; i < len(b); {
   206  		rune, n := utf8.DecodeRune(b[i:])
   207  		repl := ""
   208  		if rune == 0x2028 {
   209  			repl = `\u2028`
   210  		} else if rune == 0x2029 {
   211  			repl = `\u2029`
   212  		}
   213  		if repl != "" {
   214  			buf.Write(b[written:i])
   215  			buf.WriteString(repl)
   216  			written = i + n
   217  		}
   218  		i += n
   219  	}
   220  	if buf.Len() != 0 {
   221  		buf.Write(b[written:])
   222  		if pad {
   223  			buf.WriteByte(' ')
   224  		}
   225  		return buf.String()
   226  	}
   227  	return string(b)
   228  }
   229  
   230  // jsStrEscaper produces a string that can be included between quotes in
   231  // JavaScript source, in JavaScript embedded in an HTML5 <script> element,
   232  // or in an HTML5 event handler attribute such as onclick.
   233  func jsStrEscaper(args ...any) string {
   234  	s, t := stringify(args...)
   235  	if t == contentTypeJSStr {
   236  		return replace(s, jsStrNormReplacementTable)
   237  	}
   238  	return replace(s, jsStrReplacementTable)
   239  }
   240  
   241  // jsRegexpEscaper behaves like jsStrEscaper but escapes regular expression
   242  // specials so the result is treated literally when included in a regular
   243  // expression literal. /foo{{.X}}bar/ matches the string "foo" followed by
   244  // the literal text of {{.X}} followed by the string "bar".
   245  func jsRegexpEscaper(args ...any) string {
   246  	s, _ := stringify(args...)
   247  	s = replace(s, jsRegexpReplacementTable)
   248  	if s == "" {
   249  		// /{{.X}}/ should not produce a line comment when .X == "".
   250  		return "(?:)"
   251  	}
   252  	return s
   253  }
   254  
   255  // replace replaces each rune r of s with replacementTable[r], provided that
   256  // r < len(replacementTable). If replacementTable[r] is the empty string then
   257  // no replacement is made.
   258  // It also replaces runes U+2028 and U+2029 with the raw strings `\u2028` and
   259  // `\u2029`.
   260  func replace(s string, replacementTable []string) string {
   261  	var b strings.Builder
   262  	r, w, written := rune(0), 0, 0
   263  	for i := 0; i < len(s); i += w {
   264  		// See comment in htmlEscaper.
   265  		r, w = utf8.DecodeRuneInString(s[i:])
   266  		var repl string
   267  		switch {
   268  		case int(r) < len(lowUnicodeReplacementTable):
   269  			repl = lowUnicodeReplacementTable[r]
   270  		case int(r) < len(replacementTable) && replacementTable[r] != "":
   271  			repl = replacementTable[r]
   272  		case r == '\u2028':
   273  			repl = `\u2028`
   274  		case r == '\u2029':
   275  			repl = `\u2029`
   276  		default:
   277  			continue
   278  		}
   279  		if written == 0 {
   280  			b.Grow(len(s))
   281  		}
   282  		b.WriteString(s[written:i])
   283  		b.WriteString(repl)
   284  		written = i + w
   285  	}
   286  	if written == 0 {
   287  		return s
   288  	}
   289  	b.WriteString(s[written:])
   290  	return b.String()
   291  }
   292  
   293  var lowUnicodeReplacementTable = []string{
   294  	0: `\u0000`, 1: `\u0001`, 2: `\u0002`, 3: `\u0003`, 4: `\u0004`, 5: `\u0005`, 6: `\u0006`,
   295  	'\a': `\u0007`,
   296  	'\b': `\u0008`,
   297  	'\t': `\t`,
   298  	'\n': `\n`,
   299  	'\v': `\u000b`, // "\v" == "v" on IE 6.
   300  	'\f': `\f`,
   301  	'\r': `\r`,
   302  	0xe:  `\u000e`, 0xf: `\u000f`, 0x10: `\u0010`, 0x11: `\u0011`, 0x12: `\u0012`, 0x13: `\u0013`,
   303  	0x14: `\u0014`, 0x15: `\u0015`, 0x16: `\u0016`, 0x17: `\u0017`, 0x18: `\u0018`, 0x19: `\u0019`,
   304  	0x1a: `\u001a`, 0x1b: `\u001b`, 0x1c: `\u001c`, 0x1d: `\u001d`, 0x1e: `\u001e`, 0x1f: `\u001f`,
   305  }
   306  
   307  var jsStrReplacementTable = []string{
   308  	0:    `\u0000`,
   309  	'\t': `\t`,
   310  	'\n': `\n`,
   311  	'\v': `\u000b`, // "\v" == "v" on IE 6.
   312  	'\f': `\f`,
   313  	'\r': `\r`,
   314  	// Encode HTML specials as hex so the output can be embedded
   315  	// in HTML attributes without further encoding.
   316  	'"':  `\u0022`,
   317  	'`':  `\u0060`,
   318  	'&':  `\u0026`,
   319  	'\'': `\u0027`,
   320  	'+':  `\u002b`,
   321  	'/':  `\/`,
   322  	'<':  `\u003c`,
   323  	'>':  `\u003e`,
   324  	'\\': `\\`,
   325  }
   326  
   327  // jsStrNormReplacementTable is like jsStrReplacementTable but does not
   328  // overencode existing escapes since this table has no entry for `\`.
   329  var jsStrNormReplacementTable = []string{
   330  	0:    `\u0000`,
   331  	'\t': `\t`,
   332  	'\n': `\n`,
   333  	'\v': `\u000b`, // "\v" == "v" on IE 6.
   334  	'\f': `\f`,
   335  	'\r': `\r`,
   336  	// Encode HTML specials as hex so the output can be embedded
   337  	// in HTML attributes without further encoding.
   338  	'"':  `\u0022`,
   339  	'&':  `\u0026`,
   340  	'\'': `\u0027`,
   341  	'`':  `\u0060`,
   342  	'+':  `\u002b`,
   343  	'/':  `\/`,
   344  	'<':  `\u003c`,
   345  	'>':  `\u003e`,
   346  }
   347  var jsRegexpReplacementTable = []string{
   348  	0:    `\u0000`,
   349  	'\t': `\t`,
   350  	'\n': `\n`,
   351  	'\v': `\u000b`, // "\v" == "v" on IE 6.
   352  	'\f': `\f`,
   353  	'\r': `\r`,
   354  	// Encode HTML specials as hex so the output can be embedded
   355  	// in HTML attributes without further encoding.
   356  	'"':  `\u0022`,
   357  	'$':  `\$`,
   358  	'&':  `\u0026`,
   359  	'\'': `\u0027`,
   360  	'(':  `\(`,
   361  	')':  `\)`,
   362  	'*':  `\*`,
   363  	'+':  `\u002b`,
   364  	'-':  `\-`,
   365  	'.':  `\.`,
   366  	'/':  `\/`,
   367  	'<':  `\u003c`,
   368  	'>':  `\u003e`,
   369  	'?':  `\?`,
   370  	'[':  `\[`,
   371  	'\\': `\\`,
   372  	']':  `\]`,
   373  	'^':  `\^`,
   374  	'{':  `\{`,
   375  	'|':  `\|`,
   376  	'}':  `\}`,
   377  }
   378  
   379  // isJSIdentPart reports whether the given rune is a JS identifier part.
   380  // It does not handle all the non-Latin letters, joiners, and combining marks,
   381  // but it does handle every codepoint that can occur in a numeric literal or
   382  // a keyword.
   383  func isJSIdentPart(r rune) bool {
   384  	switch {
   385  	case r == '$':
   386  		return true
   387  	case '0' <= r && r <= '9':
   388  		return true
   389  	case 'A' <= r && r <= 'Z':
   390  		return true
   391  	case r == '_':
   392  		return true
   393  	case 'a' <= r && r <= 'z':
   394  		return true
   395  	}
   396  	return false
   397  }
   398  
   399  // isJSType reports whether the given MIME type should be considered JavaScript.
   400  //
   401  // It is used to determine whether a script tag with a type attribute is a javascript container.
   402  func isJSType(mimeType string) bool {
   403  	// per
   404  	//   https://www.w3.org/TR/html5/scripting-1.html#attr-script-type
   405  	//   https://tools.ietf.org/html/rfc7231#section-3.1.1
   406  	//   https://tools.ietf.org/html/rfc4329#section-3
   407  	//   https://www.ietf.org/rfc/rfc4627.txt
   408  	// discard parameters
   409  	mimeType, _, _ = strings.Cut(mimeType, ";")
   410  	mimeType = strings.ToLower(mimeType)
   411  	mimeType = strings.TrimSpace(mimeType)
   412  	switch mimeType {
   413  	case
   414  		"application/ecmascript",
   415  		"application/javascript",
   416  		"application/json",
   417  		"application/ld+json",
   418  		"application/x-ecmascript",
   419  		"application/x-javascript",
   420  		"module",
   421  		"text/ecmascript",
   422  		"text/javascript",
   423  		"text/javascript1.0",
   424  		"text/javascript1.1",
   425  		"text/javascript1.2",
   426  		"text/javascript1.3",
   427  		"text/javascript1.4",
   428  		"text/javascript1.5",
   429  		"text/jscript",
   430  		"text/livescript",
   431  		"text/x-ecmascript",
   432  		"text/x-javascript":
   433  		return true
   434  	default:
   435  		return false
   436  	}
   437  }
   438
View as plain text