Home | History | Annotate | Download | only in http
      1 // Copyright (c) 2011 The Chromium Authors. All rights reserved.
      2 // Use of this source code is governed by a BSD-style license that can be
      3 // found in the LICENSE file.
      4 
      5 // The rules for parsing content-types were borrowed from Firefox:
      6 // http://lxr.mozilla.org/mozilla/source/netwerk/base/src/nsURLHelper.cpp#834
      7 
      8 #include "net/http/http_util.h"
      9 
     10 #include <algorithm>
     11 
     12 #include "base/basictypes.h"
     13 #include "base/logging.h"
     14 #include "base/string_number_conversions.h"
     15 #include "base/string_piece.h"
     16 #include "base/string_util.h"
     17 
     18 using std::string;
     19 
     20 namespace net {
     21 
     22 //-----------------------------------------------------------------------------
     23 
     24 // Return the index of the closing quote of the string, if any.
     25 static size_t FindStringEnd(const string& line, size_t start, char delim) {
     26   DCHECK(start < line.length() && line[start] == delim &&
     27          (delim == '"' || delim == '\''));
     28 
     29   const char set[] = { delim, '\\', '\0' };
     30   for (;;) {
     31     // start points to either the start quote or the last
     32     // escaped char (the char following a '\\')
     33 
     34     size_t end = line.find_first_of(set, start + 1);
     35     if (end == string::npos)
     36       return line.length();
     37 
     38     if (line[end] == '\\') {
     39       // Hit a backslash-escaped char.  Need to skip over it.
     40       start = end + 1;
     41       if (start == line.length())
     42         return start;
     43 
     44       // Go back to looking for the next escape or the string end
     45       continue;
     46     }
     47 
     48     return end;
     49   }
     50 
     51   NOTREACHED();
     52   return line.length();
     53 }
     54 
     55 //-----------------------------------------------------------------------------
     56 
     57 // static
     58 size_t HttpUtil::FindDelimiter(const string& line, size_t search_start,
     59                                char delimiter) {
     60   do {
     61     // search_start points to the spot from which we should start looking
     62     // for the delimiter.
     63     const char delim_str[] = { delimiter, '"', '\'', '\0' };
     64     size_t cur_delim_pos = line.find_first_of(delim_str, search_start);
     65     if (cur_delim_pos == string::npos)
     66       return line.length();
     67 
     68     char ch = line[cur_delim_pos];
     69     if (ch == delimiter) {
     70       // Found delimiter
     71       return cur_delim_pos;
     72     }
     73 
     74     // We hit the start of a quoted string.  Look for its end.
     75     search_start = FindStringEnd(line, cur_delim_pos, ch);
     76     if (search_start == line.length())
     77       return search_start;
     78 
     79     ++search_start;
     80 
     81     // search_start now points to the first char after the end of the
     82     // string, so just go back to the top of the loop and look for
     83     // |delimiter| again.
     84   } while (true);
     85 
     86   NOTREACHED();
     87   return line.length();
     88 }
     89 
     90 // static
     91 void HttpUtil::ParseContentType(const string& content_type_str,
     92                                 string* mime_type, string* charset,
     93                                 bool *had_charset) {
     94   // Trim leading and trailing whitespace from type.  We include '(' in
     95   // the trailing trim set to catch media-type comments, which are not at all
     96   // standard, but may occur in rare cases.
     97   size_t type_val = content_type_str.find_first_not_of(HTTP_LWS);
     98   type_val = std::min(type_val, content_type_str.length());
     99   size_t type_end = content_type_str.find_first_of(HTTP_LWS ";(", type_val);
    100   if (string::npos == type_end)
    101     type_end = content_type_str.length();
    102 
    103   size_t charset_val = 0;
    104   size_t charset_end = 0;
    105 
    106   // Iterate over parameters
    107   bool type_has_charset = false;
    108   size_t param_start = content_type_str.find_first_of(';', type_end);
    109   if (param_start != string::npos) {
    110     // We have parameters.  Iterate over them.
    111     size_t cur_param_start = param_start + 1;
    112     do {
    113       size_t cur_param_end =
    114           FindDelimiter(content_type_str, cur_param_start, ';');
    115 
    116       size_t param_name_start = content_type_str.find_first_not_of(
    117           HTTP_LWS, cur_param_start);
    118       param_name_start = std::min(param_name_start, cur_param_end);
    119 
    120       static const char charset_str[] = "charset=";
    121       size_t charset_end_offset = std::min(
    122           param_name_start + sizeof(charset_str) - 1, cur_param_end);
    123       if (LowerCaseEqualsASCII(
    124               content_type_str.begin() + param_name_start,
    125               content_type_str.begin() + charset_end_offset, charset_str)) {
    126         charset_val = param_name_start + sizeof(charset_str) - 1;
    127         charset_end = cur_param_end;
    128         type_has_charset = true;
    129       }
    130 
    131       cur_param_start = cur_param_end + 1;
    132     } while (cur_param_start < content_type_str.length());
    133   }
    134 
    135   if (type_has_charset) {
    136     // Trim leading and trailing whitespace from charset_val.  We include
    137     // '(' in the trailing trim set to catch media-type comments, which are
    138     // not at all standard, but may occur in rare cases.
    139     charset_val = content_type_str.find_first_not_of(HTTP_LWS, charset_val);
    140     charset_val = std::min(charset_val, charset_end);
    141     char first_char = content_type_str[charset_val];
    142     if (first_char == '"' || first_char == '\'') {
    143       charset_end = FindStringEnd(content_type_str, charset_val, first_char);
    144       ++charset_val;
    145       DCHECK(charset_end >= charset_val);
    146     } else {
    147       charset_end = std::min(content_type_str.find_first_of(HTTP_LWS ";(",
    148                                                             charset_val),
    149                              charset_end);
    150     }
    151   }
    152 
    153   // if the server sent "*/*", it is meaningless, so do not store it.
    154   // also, if type_val is the same as mime_type, then just update the
    155   // charset.  however, if charset is empty and mime_type hasn't
    156   // changed, then don't wipe-out an existing charset.  We
    157   // also want to reject a mime-type if it does not include a slash.
    158   // some servers give junk after the charset parameter, which may
    159   // include a comma, so this check makes us a bit more tolerant.
    160   if (content_type_str.length() != 0 &&
    161       content_type_str != "*/*" &&
    162       content_type_str.find_first_of('/') != string::npos) {
    163     // Common case here is that mime_type is empty
    164     bool eq = !mime_type->empty() &&
    165               LowerCaseEqualsASCII(content_type_str.begin() + type_val,
    166                                    content_type_str.begin() + type_end,
    167                                    mime_type->data());
    168     if (!eq) {
    169       mime_type->assign(content_type_str.begin() + type_val,
    170                         content_type_str.begin() + type_end);
    171       StringToLowerASCII(mime_type);
    172     }
    173     if ((!eq && *had_charset) || type_has_charset) {
    174       *had_charset = true;
    175       charset->assign(content_type_str.begin() + charset_val,
    176                       content_type_str.begin() + charset_end);
    177       StringToLowerASCII(charset);
    178     }
    179   }
    180 }
    181 
    182 // static
    183 // Parse the Range header according to RFC 2616 14.35.1
    184 // ranges-specifier = byte-ranges-specifier
    185 // byte-ranges-specifier = bytes-unit "=" byte-range-set
    186 // byte-range-set  = 1#( byte-range-spec | suffix-byte-range-spec )
    187 // byte-range-spec = first-byte-pos "-" [last-byte-pos]
    188 // first-byte-pos  = 1*DIGIT
    189 // last-byte-pos   = 1*DIGIT
    190 bool HttpUtil::ParseRanges(const std::string& headers,
    191                            std::vector<HttpByteRange>* ranges) {
    192   std::string ranges_specifier;
    193   HttpUtil::HeadersIterator it(headers.begin(), headers.end(), "\r\n");
    194 
    195   while (it.GetNext()) {
    196     // Look for "Range" header.
    197     if (!LowerCaseEqualsASCII(it.name(), "range"))
    198       continue;
    199     ranges_specifier = it.values();
    200     // We just care about the first "Range" header, so break here.
    201     break;
    202   }
    203 
    204   if (ranges_specifier.empty())
    205     return false;
    206 
    207   return ParseRangeHeader(ranges_specifier, ranges);
    208 }
    209 
    210 // static
    211 bool HttpUtil::ParseRangeHeader(const std::string& ranges_specifier,
    212                                 std::vector<HttpByteRange>* ranges) {
    213   size_t equal_char_offset = ranges_specifier.find('=');
    214   if (equal_char_offset == std::string::npos)
    215     return false;
    216 
    217   // Try to extract bytes-unit part.
    218   std::string::const_iterator bytes_unit_begin = ranges_specifier.begin();
    219   std::string::const_iterator bytes_unit_end = bytes_unit_begin +
    220                                                equal_char_offset;
    221   std::string::const_iterator byte_range_set_begin = bytes_unit_end + 1;
    222   std::string::const_iterator byte_range_set_end = ranges_specifier.end();
    223 
    224   TrimLWS(&bytes_unit_begin, &bytes_unit_end);
    225   // "bytes" unit identifier is not found.
    226   if (!LowerCaseEqualsASCII(bytes_unit_begin, bytes_unit_end, "bytes"))
    227     return false;
    228 
    229   ValuesIterator byte_range_set_iterator(byte_range_set_begin,
    230                                          byte_range_set_end, ',');
    231   while (byte_range_set_iterator.GetNext()) {
    232     size_t minus_char_offset = byte_range_set_iterator.value().find('-');
    233     // If '-' character is not found, reports failure.
    234     if (minus_char_offset == std::string::npos)
    235       return false;
    236 
    237     std::string::const_iterator first_byte_pos_begin =
    238         byte_range_set_iterator.value_begin();
    239     std::string::const_iterator first_byte_pos_end =
    240         first_byte_pos_begin +  minus_char_offset;
    241     TrimLWS(&first_byte_pos_begin, &first_byte_pos_end);
    242     std::string first_byte_pos(first_byte_pos_begin, first_byte_pos_end);
    243 
    244     HttpByteRange range;
    245     // Try to obtain first-byte-pos.
    246     if (!first_byte_pos.empty()) {
    247       int64 first_byte_position = -1;
    248       if (!base::StringToInt64(first_byte_pos, &first_byte_position))
    249         return false;
    250       range.set_first_byte_position(first_byte_position);
    251     }
    252 
    253     std::string::const_iterator last_byte_pos_begin =
    254         byte_range_set_iterator.value_begin() + minus_char_offset + 1;
    255     std::string::const_iterator last_byte_pos_end =
    256         byte_range_set_iterator.value_end();
    257     TrimLWS(&last_byte_pos_begin, &last_byte_pos_end);
    258     std::string last_byte_pos(last_byte_pos_begin, last_byte_pos_end);
    259 
    260     // We have last-byte-pos or suffix-byte-range-spec in this case.
    261     if (!last_byte_pos.empty()) {
    262       int64 last_byte_position;
    263       if (!base::StringToInt64(last_byte_pos, &last_byte_position))
    264         return false;
    265       if (range.HasFirstBytePosition())
    266         range.set_last_byte_position(last_byte_position);
    267       else
    268         range.set_suffix_length(last_byte_position);
    269     } else if (!range.HasFirstBytePosition()) {
    270       return false;
    271     }
    272 
    273     // Do a final check on the HttpByteRange object.
    274     if (!range.IsValid())
    275       return false;
    276     ranges->push_back(range);
    277   }
    278   return !ranges->empty();
    279 }
    280 
    281 // static
    282 bool HttpUtil::HasHeader(const std::string& headers, const char* name) {
    283   size_t name_len = strlen(name);
    284   string::const_iterator it =
    285       std::search(headers.begin(),
    286                   headers.end(),
    287                   name,
    288                   name + name_len,
    289                   base::CaseInsensitiveCompareASCII<char>());
    290   if (it == headers.end())
    291     return false;
    292 
    293   // ensure match is prefixed by newline
    294   if (it != headers.begin() && it[-1] != '\n')
    295     return false;
    296 
    297   // ensure match is suffixed by colon
    298   if (it + name_len >= headers.end() || it[name_len] != ':')
    299     return false;
    300 
    301   return true;
    302 }
    303 
    304 // static
    305 std::string HttpUtil::StripHeaders(const std::string& headers,
    306                                    const char* const headers_to_remove[],
    307                                    size_t headers_to_remove_len) {
    308   std::string stripped_headers;
    309   net::HttpUtil::HeadersIterator it(headers.begin(), headers.end(), "\r\n");
    310 
    311   while (it.GetNext()) {
    312     bool should_remove = false;
    313     for (size_t i = 0; i < headers_to_remove_len; ++i) {
    314       if (LowerCaseEqualsASCII(it.name_begin(), it.name_end(),
    315                                headers_to_remove[i])) {
    316         should_remove = true;
    317         break;
    318       }
    319     }
    320     if (!should_remove) {
    321       // Assume that name and values are on the same line.
    322       stripped_headers.append(it.name_begin(), it.values_end());
    323       stripped_headers.append("\r\n");
    324     }
    325   }
    326   return stripped_headers;
    327 }
    328 
    329 // static
    330 bool HttpUtil::IsNonCoalescingHeader(string::const_iterator name_begin,
    331                                      string::const_iterator name_end) {
    332   // NOTE: "set-cookie2" headers do not support expires attributes, so we don't
    333   // have to list them here.
    334   const char* kNonCoalescingHeaders[] = {
    335     "date",
    336     "expires",
    337     "last-modified",
    338     "location",  // See bug 1050541 for details
    339     "retry-after",
    340     "set-cookie",
    341     // The format of auth-challenges mixes both space separated tokens and
    342     // comma separated properties, so coalescing on comma won't work.
    343     "www-authenticate",
    344     "proxy-authenticate"
    345   };
    346   for (size_t i = 0; i < arraysize(kNonCoalescingHeaders); ++i) {
    347     if (LowerCaseEqualsASCII(name_begin, name_end, kNonCoalescingHeaders[i]))
    348       return true;
    349   }
    350   return false;
    351 }
    352 
    353 bool HttpUtil::IsLWS(char c) {
    354   return strchr(HTTP_LWS, c) != NULL;
    355 }
    356 
    357 void HttpUtil::TrimLWS(string::const_iterator* begin,
    358                        string::const_iterator* end) {
    359   // leading whitespace
    360   while (*begin < *end && IsLWS((*begin)[0]))
    361     ++(*begin);
    362 
    363   // trailing whitespace
    364   while (*begin < *end && IsLWS((*end)[-1]))
    365     --(*end);
    366 }
    367 
    368 // static
    369 bool HttpUtil::IsQuote(char c) {
    370   // Single quote mark isn't actually part of quoted-text production,
    371   // but apparently some servers rely on this.
    372   return c == '"' || c == '\'';
    373 }
    374 
    375 // static
    376 std::string HttpUtil::Unquote(std::string::const_iterator begin,
    377                               std::string::const_iterator end) {
    378   // Empty string
    379   if (begin == end)
    380     return std::string();
    381 
    382   // Nothing to unquote.
    383   if (!IsQuote(*begin))
    384     return std::string(begin, end);
    385 
    386   // No terminal quote mark.
    387   if (end - begin < 2 || *begin != *(end - 1))
    388     return std::string(begin, end);
    389 
    390   // Strip quotemarks
    391   ++begin;
    392   --end;
    393 
    394   // Unescape quoted-pair (defined in RFC 2616 section 2.2)
    395   std::string unescaped;
    396   bool prev_escape = false;
    397   for (; begin != end; ++begin) {
    398     char c = *begin;
    399     if (c == '\\' && !prev_escape) {
    400       prev_escape = true;
    401       continue;
    402     }
    403     prev_escape = false;
    404     unescaped.push_back(c);
    405   }
    406   return unescaped;
    407 }
    408 
    409 // static
    410 std::string HttpUtil::Unquote(const std::string& str) {
    411   return Unquote(str.begin(), str.end());
    412 }
    413 
    414 // static
    415 std::string HttpUtil::Quote(const std::string& str) {
    416   std::string escaped;
    417   escaped.reserve(2 + str.size());
    418 
    419   std::string::const_iterator begin = str.begin();
    420   std::string::const_iterator end = str.end();
    421 
    422   // Esape any backslashes or quotemarks within the string, and
    423   // then surround with quotes.
    424   escaped.push_back('"');
    425   for (; begin != end; ++begin) {
    426     char c = *begin;
    427     if (c == '"' || c == '\\')
    428       escaped.push_back('\\');
    429     escaped.push_back(c);
    430   }
    431   escaped.push_back('"');
    432   return escaped;
    433 }
    434 
    435 // Find the "http" substring in a status line. This allows for
    436 // some slop at the start. If the "http" string could not be found
    437 // then returns -1.
    438 // static
    439 int HttpUtil::LocateStartOfStatusLine(const char* buf, int buf_len) {
    440   const int slop = 4;
    441   const int http_len = 4;
    442 
    443   if (buf_len >= http_len) {
    444     int i_max = std::min(buf_len - http_len, slop);
    445     for (int i = 0; i <= i_max; ++i) {
    446       if (LowerCaseEqualsASCII(buf + i, buf + i + http_len, "http"))
    447         return i;
    448     }
    449   }
    450   return -1;  // Not found
    451 }
    452 
    453 int HttpUtil::LocateEndOfHeaders(const char* buf, int buf_len, int i) {
    454   bool was_lf = false;
    455   char last_c = '\0';
    456   for (; i < buf_len; ++i) {
    457     char c = buf[i];
    458     if (c == '\n') {
    459       if (was_lf)
    460         return i + 1;
    461       was_lf = true;
    462     } else if (c != '\r' || last_c != '\n') {
    463       was_lf = false;
    464     }
    465     last_c = c;
    466   }
    467   return -1;
    468 }
    469 
    470 // In order for a line to be continuable, it must specify a
    471 // non-blank header-name. Line continuations are specifically for
    472 // header values -- do not allow headers names to span lines.
    473 static bool IsLineSegmentContinuable(const char* begin, const char* end) {
    474   if (begin == end)
    475     return false;
    476 
    477   const char* colon = std::find(begin, end, ':');
    478   if (colon == end)
    479     return false;
    480 
    481   const char* name_begin = begin;
    482   const char* name_end = colon;
    483 
    484   // Name can't be empty.
    485   if (name_begin == name_end)
    486     return false;
    487 
    488   // Can't start with LWS (this would imply the segment is a continuation)
    489   if (HttpUtil::IsLWS(*name_begin))
    490     return false;
    491 
    492   return true;
    493 }
    494 
    495 // Helper used by AssembleRawHeaders, to find the end of the status line.
    496 static const char* FindStatusLineEnd(const char* begin, const char* end) {
    497   size_t i = base::StringPiece(begin, end - begin).find_first_of("\r\n");
    498   if (i == base::StringPiece::npos)
    499     return end;
    500   return begin + i;
    501 }
    502 
    503 // Helper used by AssembleRawHeaders, to skip past leading LWS.
    504 static const char* FindFirstNonLWS(const char* begin, const char* end) {
    505   for (const char* cur = begin; cur != end; ++cur) {
    506     if (!HttpUtil::IsLWS(*cur))
    507       return cur;
    508   }
    509   return end;  // Not found.
    510 }
    511 
    512 std::string HttpUtil::AssembleRawHeaders(const char* input_begin,
    513                                          int input_len) {
    514   std::string raw_headers;
    515   raw_headers.reserve(input_len);
    516 
    517   const char* input_end = input_begin + input_len;
    518 
    519   // Skip any leading slop, since the consumers of this output
    520   // (HttpResponseHeaders) don't deal with it.
    521   int status_begin_offset = LocateStartOfStatusLine(input_begin, input_len);
    522   if (status_begin_offset != -1)
    523     input_begin += status_begin_offset;
    524 
    525   // Copy the status line.
    526   const char* status_line_end = FindStatusLineEnd(input_begin, input_end);
    527   raw_headers.append(input_begin, status_line_end);
    528 
    529   // After the status line, every subsequent line is a header line segment.
    530   // Should a segment start with LWS, it is a continuation of the previous
    531   // line's field-value.
    532 
    533   // TODO(ericroman): is this too permissive? (delimits on [\r\n]+)
    534   CStringTokenizer lines(status_line_end, input_end, "\r\n");
    535 
    536   // This variable is true when the previous line was continuable.
    537   bool prev_line_continuable = false;
    538 
    539   while (lines.GetNext()) {
    540     const char* line_begin = lines.token_begin();
    541     const char* line_end = lines.token_end();
    542 
    543     if (prev_line_continuable && IsLWS(*line_begin)) {
    544       // Join continuation; reduce the leading LWS to a single SP.
    545       raw_headers.push_back(' ');
    546       raw_headers.append(FindFirstNonLWS(line_begin, line_end), line_end);
    547     } else {
    548       // Terminate the previous line.
    549       raw_headers.push_back('\0');
    550 
    551       // Copy the raw data to output.
    552       raw_headers.append(line_begin, line_end);
    553 
    554       // Check if the current line can be continued.
    555       prev_line_continuable = IsLineSegmentContinuable(line_begin, line_end);
    556     }
    557   }
    558 
    559   raw_headers.append("\0\0", 2);
    560   return raw_headers;
    561 }
    562 
    563 // TODO(jungshik): 1. If the list is 'fr-CA,fr-FR,en,de', we have to add
    564 // 'fr' after 'fr-CA' with the same q-value as 'fr-CA' because
    565 // web servers, in general, do not fall back to 'fr' and may end up picking
    566 // 'en' which has a lower preference than 'fr-CA' and 'fr-FR'.
    567 // 2. This function assumes that the input is a comma separated list
    568 // without any whitespace. As long as it comes from the preference and
    569 // a user does not manually edit the preference file, it's the case. Still,
    570 // we may have to make it more robust.
    571 std::string HttpUtil::GenerateAcceptLanguageHeader(
    572     const std::string& raw_language_list) {
    573   // We use integers for qvalue and qvalue decrement that are 10 times
    574   // larger than actual values to avoid a problem with comparing
    575   // two floating point numbers.
    576   const unsigned int kQvalueDecrement10 = 2;
    577   unsigned int qvalue10 = 10;
    578   StringTokenizer t(raw_language_list, ",");
    579   std::string lang_list_with_q;
    580   while (t.GetNext()) {
    581     std::string language = t.token();
    582     if (qvalue10 == 10) {
    583       // q=1.0 is implicit.
    584       lang_list_with_q = language;
    585     } else {
    586       DCHECK_LT(qvalue10, 10U);
    587       base::StringAppendF(&lang_list_with_q, ",%s;q=0.%d", language.c_str(),
    588                           qvalue10);
    589     }
    590     // It does not make sense to have 'q=0'.
    591     if (qvalue10 > kQvalueDecrement10)
    592       qvalue10 -= kQvalueDecrement10;
    593   }
    594   return lang_list_with_q;
    595 }
    596 
    597 std::string HttpUtil::GenerateAcceptCharsetHeader(const std::string& charset) {
    598   std::string charset_with_q = charset;
    599   if (LowerCaseEqualsASCII(charset, "utf-8")) {
    600     charset_with_q += ",*;q=0.5";
    601   } else {
    602     charset_with_q += ",utf-8;q=0.7,*;q=0.3";
    603   }
    604   return charset_with_q;
    605 }
    606 
    607 void HttpUtil::AppendHeaderIfMissing(const char* header_name,
    608                                      const std::string& header_value,
    609                                      std::string* headers) {
    610   if (header_value.empty())
    611     return;
    612   if (net::HttpUtil::HasHeader(*headers, header_name))
    613     return;
    614   *headers += std::string(header_name) + ": " + header_value + "\r\n";
    615 }
    616 
    617 // BNF from section 4.2 of RFC 2616:
    618 //
    619 //   message-header = field-name ":" [ field-value ]
    620 //   field-name     = token
    621 //   field-value    = *( field-content | LWS )
    622 //   field-content  = <the OCTETs making up the field-value
    623 //                     and consisting of either *TEXT or combinations
    624 //                     of token, separators, and quoted-string>
    625 //
    626 
    627 HttpUtil::HeadersIterator::HeadersIterator(string::const_iterator headers_begin,
    628                                            string::const_iterator headers_end,
    629                                            const std::string& line_delimiter)
    630     : lines_(headers_begin, headers_end, line_delimiter) {
    631 }
    632 
    633 HttpUtil::HeadersIterator::~HeadersIterator() {
    634 }
    635 
    636 bool HttpUtil::HeadersIterator::GetNext() {
    637   while (lines_.GetNext()) {
    638     name_begin_ = lines_.token_begin();
    639     values_end_ = lines_.token_end();
    640 
    641     string::const_iterator colon = find(name_begin_, values_end_, ':');
    642     if (colon == values_end_)
    643       continue;  // skip malformed header
    644 
    645     name_end_ = colon;
    646 
    647     // If the name starts with LWS, it is an invalid line.
    648     // Leading LWS implies a line continuation, and these should have
    649     // already been joined by AssembleRawHeaders().
    650     if (name_begin_ == name_end_ || IsLWS(*name_begin_))
    651       continue;
    652 
    653     TrimLWS(&name_begin_, &name_end_);
    654     if (name_begin_ == name_end_)
    655       continue;  // skip malformed header
    656 
    657     values_begin_ = colon + 1;
    658     TrimLWS(&values_begin_, &values_end_);
    659 
    660     // if we got a header name, then we are done.
    661     return true;
    662   }
    663   return false;
    664 }
    665 
    666 bool HttpUtil::HeadersIterator::AdvanceTo(const char* name) {
    667   DCHECK(name != NULL);
    668   DCHECK_EQ(0, StringToLowerASCII<std::string>(name).compare(name))
    669       << "the header name must be in all lower case";
    670 
    671   while (GetNext()) {
    672     if (LowerCaseEqualsASCII(name_begin_, name_end_, name)) {
    673       return true;
    674     }
    675   }
    676 
    677   return false;
    678 }
    679 
    680 HttpUtil::ValuesIterator::ValuesIterator(
    681     string::const_iterator values_begin,
    682     string::const_iterator values_end,
    683     char delimiter)
    684     : values_(values_begin, values_end, string(1, delimiter)) {
    685   values_.set_quote_chars("\'\"");
    686 }
    687 
    688 HttpUtil::ValuesIterator::~ValuesIterator() {
    689 }
    690 
    691 bool HttpUtil::ValuesIterator::GetNext() {
    692   while (values_.GetNext()) {
    693     value_begin_ = values_.token_begin();
    694     value_end_ = values_.token_end();
    695     TrimLWS(&value_begin_, &value_end_);
    696 
    697     // bypass empty values.
    698     if (value_begin_ != value_end_)
    699       return true;
    700   }
    701   return false;
    702 }
    703 
    704 HttpUtil::NameValuePairsIterator::NameValuePairsIterator(
    705     string::const_iterator begin,
    706     string::const_iterator end,
    707     char delimiter)
    708     : props_(begin, end, delimiter),
    709       valid_(true),
    710       begin_(begin),
    711       end_(end),
    712       name_begin_(end),
    713       name_end_(end),
    714       value_begin_(end),
    715       value_end_(end),
    716       value_is_quoted_(false) {
    717 }
    718 
    719 HttpUtil::NameValuePairsIterator::~NameValuePairsIterator() {}
    720 
    721 // We expect properties to be formatted as one of:
    722 //   name="value"
    723 //   name='value'
    724 //   name='\'value\''
    725 //   name=value
    726 //   name = value
    727 //   name=
    728 // Due to buggy implementations found in some embedded devices, we also
    729 // accept values with missing close quotemark (http://crbug.com/39836):
    730 //   name="value
    731 bool HttpUtil::NameValuePairsIterator::GetNext() {
    732   if (!props_.GetNext())
    733     return false;
    734 
    735   // Set the value as everything. Next we will split out the name.
    736   value_begin_ = props_.value_begin();
    737   value_end_ = props_.value_end();
    738   name_begin_ = name_end_ = value_end_;
    739 
    740   // Scan for the equals sign.
    741   std::string::const_iterator equals = std::find(value_begin_, value_end_, '=');
    742   if (equals == value_end_ || equals == value_begin_)
    743     return valid_ = false;  // Malformed, no equals sign
    744 
    745   // Verify that the equals sign we found wasn't inside of quote marks.
    746   for (std::string::const_iterator it = value_begin_; it != equals; ++it) {
    747     if (HttpUtil::IsQuote(*it))
    748       return valid_ = false;  // Malformed, quote appears before equals sign
    749   }
    750 
    751   name_begin_ = value_begin_;
    752   name_end_ = equals;
    753   value_begin_ = equals + 1;
    754 
    755   TrimLWS(&name_begin_, &name_end_);
    756   TrimLWS(&value_begin_, &value_end_);
    757   value_is_quoted_ = false;
    758   unquoted_value_.clear();
    759 
    760   if (value_begin_ == value_end_)
    761     return valid_ = false;  // Malformed, value is empty
    762 
    763   if (HttpUtil::IsQuote(*value_begin_)) {
    764     // Trim surrounding quotemarks off the value
    765     if (*value_begin_ != *(value_end_ - 1) || value_begin_ + 1 == value_end_) {
    766       // NOTE: This is not as graceful as it sounds:
    767       // * quoted-pairs will no longer be unquoted
    768       //   (["\"hello] should give ["hello]).
    769       // * Does not detect when the final quote is escaped
    770       //   (["value\"] should give [value"])
    771       ++value_begin_;  // Gracefully recover from mismatching quotes.
    772     } else {
    773       value_is_quoted_ = true;
    774       // Do not store iterators into this. See declaration of unquoted_value_.
    775       unquoted_value_ = HttpUtil::Unquote(value_begin_, value_end_);
    776     }
    777   }
    778 
    779   return true;
    780 }
    781 
    782 }  // namespace net
    783