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
18using std::string;
19
20namespace net {
21
22//-----------------------------------------------------------------------------
23
24// Return the index of the closing quote of the string, if any.
25static 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
58size_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
91void 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
190bool 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
211bool 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
282bool 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
305std::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
330bool 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
353bool HttpUtil::IsLWS(char c) {
354  return strchr(HTTP_LWS, c) != NULL;
355}
356
357void 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
369bool 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
376std::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
410std::string HttpUtil::Unquote(const std::string& str) {
411  return Unquote(str.begin(), str.end());
412}
413
414// static
415std::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
439int HttpUtil::LocateStartOfStatusLine(const char* buf, int buf_len) {
440  const int slop = 4;
441  const int http_len = 4;
442#ifdef ANDROID
443  const int icy_len = 3;
444#endif
445
446  if (buf_len >= http_len) {
447    int i_max = std::min(buf_len - http_len, slop);
448    for (int i = 0; i <= i_max; ++i) {
449      if (LowerCaseEqualsASCII(buf + i, buf + i + http_len, "http"))
450        return i;
451#ifdef ANDROID
452      if (LowerCaseEqualsASCII(buf + i, buf + i + icy_len, "icy"))
453        return i;
454#endif
455    }
456  }
457  return -1;  // Not found
458}
459
460int HttpUtil::LocateEndOfHeaders(const char* buf, int buf_len, int i) {
461  bool was_lf = false;
462  char last_c = '\0';
463  for (; i < buf_len; ++i) {
464    char c = buf[i];
465    if (c == '\n') {
466      if (was_lf)
467        return i + 1;
468      was_lf = true;
469    } else if (c != '\r' || last_c != '\n') {
470      was_lf = false;
471    }
472    last_c = c;
473  }
474  return -1;
475}
476
477// In order for a line to be continuable, it must specify a
478// non-blank header-name. Line continuations are specifically for
479// header values -- do not allow headers names to span lines.
480static bool IsLineSegmentContinuable(const char* begin, const char* end) {
481  if (begin == end)
482    return false;
483
484  const char* colon = std::find(begin, end, ':');
485  if (colon == end)
486    return false;
487
488  const char* name_begin = begin;
489  const char* name_end = colon;
490
491  // Name can't be empty.
492  if (name_begin == name_end)
493    return false;
494
495  // Can't start with LWS (this would imply the segment is a continuation)
496  if (HttpUtil::IsLWS(*name_begin))
497    return false;
498
499  return true;
500}
501
502// Helper used by AssembleRawHeaders, to find the end of the status line.
503static const char* FindStatusLineEnd(const char* begin, const char* end) {
504  size_t i = base::StringPiece(begin, end - begin).find_first_of("\r\n");
505  if (i == base::StringPiece::npos)
506    return end;
507  return begin + i;
508}
509
510// Helper used by AssembleRawHeaders, to skip past leading LWS.
511static const char* FindFirstNonLWS(const char* begin, const char* end) {
512  for (const char* cur = begin; cur != end; ++cur) {
513    if (!HttpUtil::IsLWS(*cur))
514      return cur;
515  }
516  return end;  // Not found.
517}
518
519std::string HttpUtil::AssembleRawHeaders(const char* input_begin,
520                                         int input_len) {
521  std::string raw_headers;
522  raw_headers.reserve(input_len);
523
524  const char* input_end = input_begin + input_len;
525
526  // Skip any leading slop, since the consumers of this output
527  // (HttpResponseHeaders) don't deal with it.
528  int status_begin_offset = LocateStartOfStatusLine(input_begin, input_len);
529  if (status_begin_offset != -1)
530    input_begin += status_begin_offset;
531
532  // Copy the status line.
533  const char* status_line_end = FindStatusLineEnd(input_begin, input_end);
534  raw_headers.append(input_begin, status_line_end);
535
536  // After the status line, every subsequent line is a header line segment.
537  // Should a segment start with LWS, it is a continuation of the previous
538  // line's field-value.
539
540  // TODO(ericroman): is this too permissive? (delimits on [\r\n]+)
541  CStringTokenizer lines(status_line_end, input_end, "\r\n");
542
543  // This variable is true when the previous line was continuable.
544  bool prev_line_continuable = false;
545
546  while (lines.GetNext()) {
547    const char* line_begin = lines.token_begin();
548    const char* line_end = lines.token_end();
549
550    if (prev_line_continuable && IsLWS(*line_begin)) {
551      // Join continuation; reduce the leading LWS to a single SP.
552      raw_headers.push_back(' ');
553      raw_headers.append(FindFirstNonLWS(line_begin, line_end), line_end);
554    } else {
555      // Terminate the previous line.
556      raw_headers.push_back('\0');
557
558      // Copy the raw data to output.
559      raw_headers.append(line_begin, line_end);
560
561      // Check if the current line can be continued.
562      prev_line_continuable = IsLineSegmentContinuable(line_begin, line_end);
563    }
564  }
565
566  raw_headers.append("\0\0", 2);
567  return raw_headers;
568}
569
570// TODO(jungshik): 1. If the list is 'fr-CA,fr-FR,en,de', we have to add
571// 'fr' after 'fr-CA' with the same q-value as 'fr-CA' because
572// web servers, in general, do not fall back to 'fr' and may end up picking
573// 'en' which has a lower preference than 'fr-CA' and 'fr-FR'.
574// 2. This function assumes that the input is a comma separated list
575// without any whitespace. As long as it comes from the preference and
576// a user does not manually edit the preference file, it's the case. Still,
577// we may have to make it more robust.
578std::string HttpUtil::GenerateAcceptLanguageHeader(
579    const std::string& raw_language_list) {
580  // We use integers for qvalue and qvalue decrement that are 10 times
581  // larger than actual values to avoid a problem with comparing
582  // two floating point numbers.
583  const unsigned int kQvalueDecrement10 = 2;
584  unsigned int qvalue10 = 10;
585  StringTokenizer t(raw_language_list, ",");
586  std::string lang_list_with_q;
587  while (t.GetNext()) {
588    std::string language = t.token();
589    if (qvalue10 == 10) {
590      // q=1.0 is implicit.
591      lang_list_with_q = language;
592    } else {
593      DCHECK_LT(qvalue10, 10U);
594      base::StringAppendF(&lang_list_with_q, ",%s;q=0.%d", language.c_str(),
595                          qvalue10);
596    }
597    // It does not make sense to have 'q=0'.
598    if (qvalue10 > kQvalueDecrement10)
599      qvalue10 -= kQvalueDecrement10;
600  }
601  return lang_list_with_q;
602}
603
604std::string HttpUtil::GenerateAcceptCharsetHeader(const std::string& charset) {
605  std::string charset_with_q = charset;
606  if (LowerCaseEqualsASCII(charset, "utf-8")) {
607    charset_with_q += ",*;q=0.5";
608  } else {
609    charset_with_q += ",utf-8;q=0.7,*;q=0.3";
610  }
611  return charset_with_q;
612}
613
614void HttpUtil::AppendHeaderIfMissing(const char* header_name,
615                                     const std::string& header_value,
616                                     std::string* headers) {
617  if (header_value.empty())
618    return;
619  if (net::HttpUtil::HasHeader(*headers, header_name))
620    return;
621  *headers += std::string(header_name) + ": " + header_value + "\r\n";
622}
623
624// BNF from section 4.2 of RFC 2616:
625//
626//   message-header = field-name ":" [ field-value ]
627//   field-name     = token
628//   field-value    = *( field-content | LWS )
629//   field-content  = <the OCTETs making up the field-value
630//                     and consisting of either *TEXT or combinations
631//                     of token, separators, and quoted-string>
632//
633
634HttpUtil::HeadersIterator::HeadersIterator(string::const_iterator headers_begin,
635                                           string::const_iterator headers_end,
636                                           const std::string& line_delimiter)
637    : lines_(headers_begin, headers_end, line_delimiter) {
638}
639
640HttpUtil::HeadersIterator::~HeadersIterator() {
641}
642
643bool HttpUtil::HeadersIterator::GetNext() {
644  while (lines_.GetNext()) {
645    name_begin_ = lines_.token_begin();
646    values_end_ = lines_.token_end();
647
648    string::const_iterator colon = find(name_begin_, values_end_, ':');
649    if (colon == values_end_)
650      continue;  // skip malformed header
651
652    name_end_ = colon;
653
654    // If the name starts with LWS, it is an invalid line.
655    // Leading LWS implies a line continuation, and these should have
656    // already been joined by AssembleRawHeaders().
657    if (name_begin_ == name_end_ || IsLWS(*name_begin_))
658      continue;
659
660    TrimLWS(&name_begin_, &name_end_);
661    if (name_begin_ == name_end_)
662      continue;  // skip malformed header
663
664    values_begin_ = colon + 1;
665    TrimLWS(&values_begin_, &values_end_);
666
667    // if we got a header name, then we are done.
668    return true;
669  }
670  return false;
671}
672
673bool HttpUtil::HeadersIterator::AdvanceTo(const char* name) {
674  DCHECK(name != NULL);
675  DCHECK_EQ(0, StringToLowerASCII<std::string>(name).compare(name))
676      << "the header name must be in all lower case";
677
678  while (GetNext()) {
679    if (LowerCaseEqualsASCII(name_begin_, name_end_, name)) {
680      return true;
681    }
682  }
683
684  return false;
685}
686
687HttpUtil::ValuesIterator::ValuesIterator(
688    string::const_iterator values_begin,
689    string::const_iterator values_end,
690    char delimiter)
691    : values_(values_begin, values_end, string(1, delimiter)) {
692  values_.set_quote_chars("\'\"");
693}
694
695HttpUtil::ValuesIterator::~ValuesIterator() {
696}
697
698bool HttpUtil::ValuesIterator::GetNext() {
699  while (values_.GetNext()) {
700    value_begin_ = values_.token_begin();
701    value_end_ = values_.token_end();
702    TrimLWS(&value_begin_, &value_end_);
703
704    // bypass empty values.
705    if (value_begin_ != value_end_)
706      return true;
707  }
708  return false;
709}
710
711HttpUtil::NameValuePairsIterator::NameValuePairsIterator(
712    string::const_iterator begin,
713    string::const_iterator end,
714    char delimiter)
715    : props_(begin, end, delimiter),
716      valid_(true),
717      begin_(begin),
718      end_(end),
719      name_begin_(end),
720      name_end_(end),
721      value_begin_(end),
722      value_end_(end),
723      value_is_quoted_(false) {
724}
725
726HttpUtil::NameValuePairsIterator::~NameValuePairsIterator() {}
727
728// We expect properties to be formatted as one of:
729//   name="value"
730//   name='value'
731//   name='\'value\''
732//   name=value
733//   name = value
734//   name=
735// Due to buggy implementations found in some embedded devices, we also
736// accept values with missing close quotemark (http://crbug.com/39836):
737//   name="value
738bool HttpUtil::NameValuePairsIterator::GetNext() {
739  if (!props_.GetNext())
740    return false;
741
742  // Set the value as everything. Next we will split out the name.
743  value_begin_ = props_.value_begin();
744  value_end_ = props_.value_end();
745  name_begin_ = name_end_ = value_end_;
746
747  // Scan for the equals sign.
748  std::string::const_iterator equals = std::find(value_begin_, value_end_, '=');
749  if (equals == value_end_ || equals == value_begin_)
750    return valid_ = false;  // Malformed, no equals sign
751
752  // Verify that the equals sign we found wasn't inside of quote marks.
753  for (std::string::const_iterator it = value_begin_; it != equals; ++it) {
754    if (HttpUtil::IsQuote(*it))
755      return valid_ = false;  // Malformed, quote appears before equals sign
756  }
757
758  name_begin_ = value_begin_;
759  name_end_ = equals;
760  value_begin_ = equals + 1;
761
762  TrimLWS(&name_begin_, &name_end_);
763  TrimLWS(&value_begin_, &value_end_);
764  value_is_quoted_ = false;
765  unquoted_value_.clear();
766
767  if (value_begin_ == value_end_)
768    return valid_ = false;  // Malformed, value is empty
769
770  if (HttpUtil::IsQuote(*value_begin_)) {
771    // Trim surrounding quotemarks off the value
772    if (*value_begin_ != *(value_end_ - 1) || value_begin_ + 1 == value_end_) {
773      // NOTE: This is not as graceful as it sounds:
774      // * quoted-pairs will no longer be unquoted
775      //   (["\"hello] should give ["hello]).
776      // * Does not detect when the final quote is escaped
777      //   (["value\"] should give [value"])
778      ++value_begin_;  // Gracefully recover from mismatching quotes.
779    } else {
780      value_is_quoted_ = true;
781      // Do not store iterators into this. See declaration of unquoted_value_.
782      unquoted_value_ = HttpUtil::Unquote(value_begin_, value_end_);
783    }
784  }
785
786  return true;
787}
788
789}  // namespace net
790