mediatype.go

     1  // Copyright 2010 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 mime
     6  
     7  import (
     8  	"errors"
     9  	"fmt"
    10  	"maps"
    11  	"slices"
    12  	"strings"
    13  	"unicode"
    14  )
    15  
    16  // FormatMediaType serializes mediatype t and the parameters
    17  // param as a media type conforming to RFC 2045 and RFC 2616.
    18  // The type and parameter names are written in lower-case.
    19  // When any of the arguments result in a standard violation then
    20  // FormatMediaType returns the empty string.
    21  func FormatMediaType(t string, param map[string]string) string {
    22  	var b strings.Builder
    23  	if major, sub, ok := strings.Cut(t, "/"); !ok {
    24  		if !isToken(t) {
    25  			return ""
    26  		}
    27  		b.WriteString(strings.ToLower(t))
    28  	} else {
    29  		if !isToken(major) || !isToken(sub) {
    30  			return ""
    31  		}
    32  		b.WriteString(strings.ToLower(major))
    33  		b.WriteByte('/')
    34  		b.WriteString(strings.ToLower(sub))
    35  	}
    36  
    37  	for _, attribute := range slices.Sorted(maps.Keys(param)) {
    38  		value := param[attribute]
    39  		b.WriteByte(';')
    40  		b.WriteByte(' ')
    41  		if !isToken(attribute) {
    42  			return ""
    43  		}
    44  		b.WriteString(strings.ToLower(attribute))
    45  
    46  		needEnc := needsEncoding(value)
    47  		if needEnc {
    48  			// RFC 2231 section 4
    49  			b.WriteByte('*')
    50  		}
    51  		b.WriteByte('=')
    52  
    53  		if needEnc {
    54  			b.WriteString("utf-8''")
    55  
    56  			offset := 0
    57  			for index := 0; index < len(value); index++ {
    58  				ch := value[index]
    59  				// {RFC 2231 section 7}
    60  				// attribute-char := <any (US-ASCII) CHAR except SPACE, CTLs, "*", "'", "%", or tspecials>
    61  				if ch <= ' ' || ch >= 0x7F ||
    62  					ch == '*' || ch == '\'' || ch == '%' ||
    63  					isTSpecial(ch) {
    64  
    65  					b.WriteString(value[offset:index])
    66  					offset = index + 1
    67  
    68  					b.WriteByte('%')
    69  					b.WriteByte(upperhex[ch>>4])
    70  					b.WriteByte(upperhex[ch&0x0F])
    71  				}
    72  			}
    73  			b.WriteString(value[offset:])
    74  			continue
    75  		}
    76  
    77  		if isToken(value) {
    78  			b.WriteString(value)
    79  			continue
    80  		}
    81  
    82  		b.WriteByte('"')
    83  		offset := 0
    84  		for index := 0; index < len(value); index++ {
    85  			character := value[index]
    86  			if character == '"' || character == '\\' {
    87  				b.WriteString(value[offset:index])
    88  				offset = index
    89  				b.WriteByte('\\')
    90  			}
    91  		}
    92  		b.WriteString(value[offset:])
    93  		b.WriteByte('"')
    94  	}
    95  	return b.String()
    96  }
    97  
    98  func checkMediaTypeDisposition(s string) error {
    99  	typ, rest := consumeToken(s)
   100  	if typ == "" {
   101  		return errNoMediaType
   102  	}
   103  	if rest == "" {
   104  		return nil
   105  	}
   106  	var ok bool
   107  	if rest, ok = strings.CutPrefix(rest, "/"); !ok {
   108  		return errNoSlashAfterFirstToken
   109  	}
   110  	subtype, rest := consumeToken(rest)
   111  	if subtype == "" {
   112  		return errNoTokenAfterSlash
   113  	}
   114  	if rest != "" {
   115  		return errUnexpectedContentAfterMediaSubtype
   116  	}
   117  	return nil
   118  }
   119  
   120  var (
   121  	errNoMediaType                        = errors.New("mime: no media type")
   122  	errNoSlashAfterFirstToken             = errors.New("mime: expected slash after first token")
   123  	errNoTokenAfterSlash                  = errors.New("mime: expected token after slash")
   124  	errUnexpectedContentAfterMediaSubtype = errors.New("mime: unexpected content after media subtype")
   125  )
   126  
   127  // ErrInvalidMediaParameter is returned by [ParseMediaType] if
   128  // the media type value was found but there was an error parsing
   129  // the optional parameters
   130  var ErrInvalidMediaParameter = errors.New("mime: invalid media parameter")
   131  
   132  // ParseMediaType parses a media type value and any optional
   133  // parameters, per RFC 1521.  Media types are the values in
   134  // Content-Type and Content-Disposition headers (RFC 2183).
   135  // On success, ParseMediaType returns the media type converted
   136  // to lowercase and trimmed of white space and a non-nil map.
   137  // If there is an error parsing the optional parameter,
   138  // the media type will be returned along with the error
   139  // [ErrInvalidMediaParameter].
   140  // The returned map, params, maps from the lowercase
   141  // attribute to the attribute value with its case preserved.
   142  func ParseMediaType(v string) (mediatype string, params map[string]string, err error) {
   143  	base, _, _ := strings.Cut(v, ";")
   144  	mediatype = strings.TrimSpace(strings.ToLower(base))
   145  
   146  	err = checkMediaTypeDisposition(mediatype)
   147  	if err != nil {
   148  		return "", nil, err
   149  	}
   150  
   151  	params = make(map[string]string)
   152  
   153  	// Map of base parameter name -> parameter name -> value
   154  	// for parameters containing a '*' character.
   155  	// Lazily initialized.
   156  	var continuation map[string]map[string]string
   157  
   158  	v = v[len(base):]
   159  	for len(v) > 0 {
   160  		v = strings.TrimLeftFunc(v, unicode.IsSpace)
   161  		if len(v) == 0 {
   162  			break
   163  		}
   164  		key, value, rest := consumeMediaParam(v)
   165  		if key == "" {
   166  			if strings.TrimSpace(rest) == ";" {
   167  				// Ignore trailing semicolons.
   168  				// Not an error.
   169  				break
   170  			}
   171  			// Parse error.
   172  			return mediatype, nil, ErrInvalidMediaParameter
   173  		}
   174  
   175  		pmap := params
   176  		if baseName, _, ok := strings.Cut(key, "*"); ok {
   177  			if continuation == nil {
   178  				continuation = make(map[string]map[string]string)
   179  			}
   180  			if pmap, ok = continuation[baseName]; !ok {
   181  				continuation[baseName] = make(map[string]string)
   182  				pmap = continuation[baseName]
   183  			}
   184  		}
   185  		if v, exists := pmap[key]; exists && v != value {
   186  			// Duplicate parameter names are incorrect, but we allow them if they are equal.
   187  			return "", nil, errDuplicateParamName
   188  		}
   189  		pmap[key] = value
   190  		v = rest
   191  	}
   192  
   193  	// Stitch together any continuations or things with stars
   194  	// (i.e. RFC 2231 things with stars: "foo*0" or "foo*")
   195  	var buf strings.Builder
   196  	for key, pieceMap := range continuation {
   197  		singlePartKey := key + "*"
   198  		if v, ok := pieceMap[singlePartKey]; ok {
   199  			if decv, ok := decode2231Enc(v); ok {
   200  				params[key] = decv
   201  			}
   202  			continue
   203  		}
   204  
   205  		buf.Reset()
   206  		valid := false
   207  		for n := 0; ; n++ {
   208  			simplePart := fmt.Sprintf("%s*%d", key, n)
   209  			if v, ok := pieceMap[simplePart]; ok {
   210  				valid = true
   211  				buf.WriteString(v)
   212  				continue
   213  			}
   214  			encodedPart := simplePart + "*"
   215  			v, ok := pieceMap[encodedPart]
   216  			if !ok {
   217  				break
   218  			}
   219  			valid = true
   220  			if n == 0 {
   221  				if decv, ok := decode2231Enc(v); ok {
   222  					buf.WriteString(decv)
   223  				}
   224  			} else {
   225  				decv, _ := percentHexUnescape(v)
   226  				buf.WriteString(decv)
   227  			}
   228  		}
   229  		if valid {
   230  			params[key] = buf.String()
   231  		}
   232  	}
   233  
   234  	return
   235  }
   236  
   237  var errDuplicateParamName = errors.New("mime: duplicate parameter name")
   238  
   239  func decode2231Enc(v string) (string, bool) {
   240  	charset, v, ok := strings.Cut(v, "'")
   241  	if !ok {
   242  		return "", false
   243  	}
   244  	// TODO: ignoring the language part for now. If anybody needs it, we'll
   245  	// need to decide how to expose it in the API. But I'm not sure
   246  	// anybody uses it in practice.
   247  	_, extOtherVals, ok := strings.Cut(v, "'")
   248  	if !ok {
   249  		return "", false
   250  	}
   251  	charset = strings.ToLower(charset)
   252  	switch charset {
   253  	case "us-ascii", "utf-8":
   254  	default:
   255  		// Empty or unsupported encoding.
   256  		return "", false
   257  	}
   258  	return percentHexUnescape(extOtherVals)
   259  }
   260  
   261  // consumeToken consumes a token from the beginning of provided
   262  // string, per RFC 2045 section 5.1 (referenced from 2183), and return
   263  // the token consumed and the rest of the string. Returns ("", v) on
   264  // failure to consume at least one character.
   265  func consumeToken(v string) (token, rest string) {
   266  	for i := range len(v) {
   267  		if !isTokenChar(v[i]) {
   268  			return v[:i], v[i:]
   269  		}
   270  	}
   271  	return v, ""
   272  }
   273  
   274  // consumeValue consumes a "value" per RFC 2045, where a value is
   275  // either a 'token' or a 'quoted-string'.  On success, consumeValue
   276  // returns the value consumed (and de-quoted/escaped, if a
   277  // quoted-string) and the rest of the string. On failure, returns
   278  // ("", v).
   279  func consumeValue(v string) (value, rest string) {
   280  	if v == "" {
   281  		return
   282  	}
   283  	if v[0] != '"' {
   284  		return consumeToken(v)
   285  	}
   286  
   287  	// parse a quoted-string
   288  	buffer := new(strings.Builder)
   289  	for i := 1; i < len(v); i++ {
   290  		r := v[i]
   291  		if r == '"' {
   292  			return buffer.String(), v[i+1:]
   293  		}
   294  		// When MSIE sends a full file path (in "intranet mode"), it does not
   295  		// escape backslashes: "C:\dev\go\foo.txt", not "C:\\dev\\go\\foo.txt".
   296  		//
   297  		// No known MIME generators emit unnecessary backslash escapes
   298  		// for simple token characters like numbers and letters.
   299  		//
   300  		// If we see an unnecessary backslash escape, assume it is from MSIE
   301  		// and intended as a literal backslash. This makes Go servers deal better
   302  		// with MSIE without affecting the way they handle conforming MIME
   303  		// generators.
   304  		if r == '\\' && i+1 < len(v) && isTSpecial(v[i+1]) {
   305  			buffer.WriteByte(v[i+1])
   306  			i++
   307  			continue
   308  		}
   309  		if r == '\r' || r == '\n' {
   310  			return "", v
   311  		}
   312  		buffer.WriteByte(v[i])
   313  	}
   314  	// Did not find end quote.
   315  	return "", v
   316  }
   317  
   318  func consumeMediaParam(v string) (param, value, rest string) {
   319  	rest = strings.TrimLeftFunc(v, unicode.IsSpace)
   320  	var ok bool
   321  	if rest, ok = strings.CutPrefix(rest, ";"); !ok {
   322  		return "", "", v
   323  	}
   324  
   325  	rest = strings.TrimLeftFunc(rest, unicode.IsSpace)
   326  	param, rest = consumeToken(rest)
   327  	param = strings.ToLower(param)
   328  	if param == "" {
   329  		return "", "", v
   330  	}
   331  
   332  	rest = strings.TrimLeftFunc(rest, unicode.IsSpace)
   333  	if rest, ok = strings.CutPrefix(rest, "="); !ok {
   334  		return "", "", v
   335  	}
   336  	rest = strings.TrimLeftFunc(rest, unicode.IsSpace)
   337  	value, rest2 := consumeValue(rest)
   338  	if value == "" && rest2 == rest {
   339  		return "", "", v
   340  	}
   341  	rest = rest2
   342  	return param, value, rest
   343  }
   344  
   345  func percentHexUnescape(s string) (string, bool) {
   346  	// Count %, check that they're well-formed.
   347  	percents := 0
   348  	for i := 0; i < len(s); {
   349  		if s[i] != '%' {
   350  			i++
   351  			continue
   352  		}
   353  		percents++
   354  		if i+2 >= len(s) || !ishex(s[i+1]) || !ishex(s[i+2]) {
   355  			return "", false
   356  		}
   357  		i += 3
   358  	}
   359  	if percents == 0 {
   360  		return s, true
   361  	}
   362  
   363  	t := make([]byte, len(s)-2*percents)
   364  	j := 0
   365  	for i := 0; i < len(s); {
   366  		switch s[i] {
   367  		case '%':
   368  			t[j] = unhex(s[i+1])<<4 | unhex(s[i+2])
   369  			j++
   370  			i += 3
   371  		default:
   372  			t[j] = s[i]
   373  			j++
   374  			i++
   375  		}
   376  	}
   377  	return string(t), true
   378  }
   379  
   380  func ishex(c byte) bool {
   381  	switch {
   382  	case '0' <= c && c <= '9':
   383  		return true
   384  	case 'a' <= c && c <= 'f':
   385  		return true
   386  	case 'A' <= c && c <= 'F':
   387  		return true
   388  	}
   389  	return false
   390  }
   391  
   392  func unhex(c byte) byte {
   393  	switch {
   394  	case '0' <= c && c <= '9':
   395  		return c - '0'
   396  	case 'a' <= c && c <= 'f':
   397  		return c - 'a' + 10
   398  	case 'A' <= c && c <= 'F':
   399  		return c - 'A' + 10
   400  	}
   401  	return 0
   402  }
   403
View as plain text