| OLD | NEW |
|---|
| (Empty) |
| 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 #include <stdlib.h> | |
| 6 | |
| 7 #include "base/logging.h" | |
| 8 #include "base/strings/string_util.h" | |
| 9 #include "net/base/net_util.h" | |
| 10 #include "net/tools/dump_cache/url_to_filename_encoder.h" | |
| 11 | |
| 12 using std::string; | |
| 13 | |
| 14 namespace { | |
| 15 | |
| 16 // Returns 1 if buf is prefixed by "num_digits" of hex digits | |
| 17 // Teturns 0 otherwise. | |
| 18 // The function checks for '\0' for string termination. | |
| 19 int HexDigitsPrefix(const char* buf, int num_digits) { | |
| 20 for (int i = 0; i < num_digits; i++) { | |
| 21 if (!IsHexDigit(buf[i])) | |
| 22 return 0; // This also detects end of string as '\0' is not xdigit. | |
| 23 } | |
| 24 return 1; | |
| 25 } | |
| 26 | |
| 27 #ifdef WIN32 | |
| 28 #define strtoull _strtoui64 | |
| 29 #endif | |
| 30 | |
| 31 // A simple parser for long long values. Returns the parsed value if a | |
| 32 // valid integer is found; else returns deflt | |
| 33 // UInt64 and Int64 cannot handle decimal numbers with leading 0s. | |
| 34 uint64 ParseLeadingHex64Value(const char *str, uint64 deflt) { | |
| 35 char *error = NULL; | |
| 36 const uint64 value = strtoull(str, &error, 16); | |
| 37 return (error == str) ? deflt : value; | |
| 38 } | |
| 39 | |
| 40 } | |
| 41 | |
| 42 namespace net { | |
| 43 | |
| 44 // The escape character choice is made here -- all code and tests in this | |
| 45 // directory are based off of this constant. However, our testdata | |
| 46 // has tons of dependencies on this, so it cannot be changed without | |
| 47 // re-running those tests and fixing them. | |
| 48 const char UrlToFilenameEncoder::kEscapeChar = ','; | |
| 49 const char UrlToFilenameEncoder::kTruncationChar = '-'; | |
| 50 const size_t UrlToFilenameEncoder::kMaximumSubdirectoryLength = 128; | |
| 51 | |
| 52 void UrlToFilenameEncoder::AppendSegment(string* segment, string* dest) { | |
| 53 CHECK(!segment->empty()); | |
| 54 if ((*segment == ".") || (*segment == "..")) { | |
| 55 dest->append(1, kEscapeChar); | |
| 56 dest->append(*segment); | |
| 57 segment->clear(); | |
| 58 } else { | |
| 59 size_t segment_size = segment->size(); | |
| 60 if (segment_size > kMaximumSubdirectoryLength) { | |
| 61 // We need to inject ",-" at the end of the segment to signify that | |
| 62 // we are inserting an artificial '/'. This means we have to chop | |
| 63 // off at least two characters to make room. | |
| 64 segment_size = kMaximumSubdirectoryLength - 2; | |
| 65 | |
| 66 // But we don't want to break up an escape sequence that happens to lie at | |
| 67 // the end. Escape sequences are at most 2 characters. | |
| 68 if ((*segment)[segment_size - 1] == kEscapeChar) { | |
| 69 segment_size -= 1; | |
| 70 } else if ((*segment)[segment_size - 2] == kEscapeChar) { | |
| 71 segment_size -= 2; | |
| 72 } | |
| 73 dest->append(segment->data(), segment_size); | |
| 74 dest->append(1, kEscapeChar); | |
| 75 dest->append(1, kTruncationChar); | |
| 76 segment->erase(0, segment_size); | |
| 77 | |
| 78 // At this point, if we had segment_size=3, and segment="abcd", | |
| 79 // then after this erase, we will have written "abc,-" and set segment="d" | |
| 80 } else { | |
| 81 dest->append(*segment); | |
| 82 segment->clear(); | |
| 83 } | |
| 84 } | |
| 85 } | |
| 86 | |
| 87 void UrlToFilenameEncoder::EncodeSegment(const string& filename_prefix, | |
| 88 const string& escaped_ending, | |
| 89 char dir_separator, | |
| 90 string* encoded_filename) { | |
| 91 string filename_ending = UrlUtilities::Unescape(escaped_ending); | |
| 92 | |
| 93 char encoded[3]; | |
| 94 int encoded_len; | |
| 95 string segment; | |
| 96 | |
| 97 // TODO(jmarantz): This code would be a bit simpler if we disallowed | |
| 98 // Instaweb allowing filename_prefix to not end in "/". We could | |
| 99 // then change the is routine to just take one input string. | |
| 100 size_t start_of_segment = filename_prefix.find_last_of(dir_separator); | |
| 101 if (start_of_segment == string::npos) { | |
| 102 segment = filename_prefix; | |
| 103 } else { | |
| 104 segment = filename_prefix.substr(start_of_segment + 1); | |
| 105 *encoded_filename = filename_prefix.substr(0, start_of_segment + 1); | |
| 106 } | |
| 107 | |
| 108 size_t index = 0; | |
| 109 // Special case the first / to avoid adding a leading kEscapeChar. | |
| 110 if (!filename_ending.empty() && (filename_ending[0] == dir_separator)) { | |
| 111 encoded_filename->append(segment); | |
| 112 segment.clear(); | |
| 113 encoded_filename->append(1, dir_separator); | |
| 114 ++index; | |
| 115 } | |
| 116 | |
| 117 for (; index < filename_ending.length(); ++index) { | |
| 118 unsigned char ch = static_cast<unsigned char>(filename_ending[index]); | |
| 119 | |
| 120 // Note: instead of outputing an empty segment, we let the second slash | |
| 121 // be escaped below. | |
| 122 if ((ch == dir_separator) && !segment.empty()) { | |
| 123 AppendSegment(&segment, encoded_filename); | |
| 124 encoded_filename->append(1, dir_separator); | |
| 125 segment.clear(); | |
| 126 } else { | |
| 127 // After removing unsafe chars the only safe ones are _.=+- and alphanums. | |
| 128 if ((ch == '_') || (ch == '.') || (ch == '=') || (ch == '+') || | |
| 129 (ch == '-') || (('0' <= ch) && (ch <= '9')) || | |
| 130 (('A' <= ch) && (ch <= 'Z')) || (('a' <= ch) && (ch <= 'z'))) { | |
| 131 encoded[0] = ch; | |
| 132 encoded_len = 1; | |
| 133 } else { | |
| 134 encoded[0] = kEscapeChar; | |
| 135 encoded[1] = ch / 16; | |
| 136 encoded[1] += (encoded[1] >= 10) ? 'A' - 10 : '0'; | |
| 137 encoded[2] = ch % 16; | |
| 138 encoded[2] += (encoded[2] >= 10) ? 'A' - 10 : '0'; | |
| 139 encoded_len = 3; | |
| 140 } | |
| 141 segment.append(encoded, encoded_len); | |
| 142 | |
| 143 // If segment is too big, we must chop it into chunks. | |
| 144 if (segment.size() > kMaximumSubdirectoryLength) { | |
| 145 AppendSegment(&segment, encoded_filename); | |
| 146 encoded_filename->append(1, dir_separator); | |
| 147 } | |
| 148 } | |
| 149 } | |
| 150 | |
| 151 // Append "," to the leaf filename so the leaf can also be a branch., e.g. | |
| 152 // allow http://a/b/c and http://a/b/c/d to co-exist as files "/a/b/c," and | |
| 153 // /a/b/c/d". So we will rename the "d" here to "d,". If doing that pushed | |
| 154 // us over the 128 char limit, then we will need to append "/" and the | |
| 155 // remaining chars. | |
| 156 segment += kEscapeChar; | |
| 157 AppendSegment(&segment, encoded_filename); | |
| 158 if (!segment.empty()) { | |
| 159 // The last overflow segment is special, because we appended in | |
| 160 // kEscapeChar above. We won't need to check it again for size | |
| 161 // or further escaping. | |
| 162 encoded_filename->append(1, dir_separator); | |
| 163 encoded_filename->append(segment); | |
| 164 } | |
| 165 } | |
| 166 | |
| 167 // Note: this decoder is not the exact inverse of the EncodeSegment above, | |
| 168 // because it does not take into account a prefix. | |
| 169 bool UrlToFilenameEncoder::Decode(const string& encoded_filename, | |
| 170 char dir_separator, | |
| 171 string* decoded_url) { | |
| 172 enum State { | |
| 173 kStart, | |
| 174 kEscape, | |
| 175 kFirstDigit, | |
| 176 kTruncate, | |
| 177 kEscapeDot | |
| 178 }; | |
| 179 State state = kStart; | |
| 180 char hex_buffer[3]; | |
| 181 hex_buffer[2] = '\0'; | |
| 182 for (size_t i = 0; i < encoded_filename.size(); ++i) { | |
| 183 char ch = encoded_filename[i]; | |
| 184 switch (state) { | |
| 185 case kStart: | |
| 186 if (ch == kEscapeChar) { | |
| 187 state = kEscape; | |
| 188 } else if (ch == dir_separator) { | |
| 189 decoded_url->append(1, '/'); // URLs only use '/' not '\\' | |
| 190 } else { | |
| 191 decoded_url->append(1, ch); | |
| 192 } | |
| 193 break; | |
| 194 case kEscape: | |
| 195 if (HexDigitsPrefix(&ch, 1) == 1) { | |
| 196 hex_buffer[0] = ch; | |
| 197 state = kFirstDigit; | |
| 198 } else if (ch == kTruncationChar) { | |
| 199 state = kTruncate; | |
| 200 } else if (ch == '.') { | |
| 201 decoded_url->append(1, '.'); | |
| 202 state = kEscapeDot; // Look for at most one more dot. | |
| 203 } else if (ch == dir_separator) { | |
| 204 // Consider url "//x". This was once encoded to "/,/x,". | |
| 205 // This code is what skips the first Escape. | |
| 206 decoded_url->append(1, '/'); // URLs only use '/' not '\\' | |
| 207 state = kStart; | |
| 208 } else { | |
| 209 return false; | |
| 210 } | |
| 211 break; | |
| 212 case kFirstDigit: | |
| 213 if (HexDigitsPrefix(&ch, 1) == 1) { | |
| 214 hex_buffer[1] = ch; | |
| 215 uint64 hex_value = ParseLeadingHex64Value(hex_buffer, 0); | |
| 216 decoded_url->append(1, static_cast<char>(hex_value)); | |
| 217 state = kStart; | |
| 218 } else { | |
| 219 return false; | |
| 220 } | |
| 221 break; | |
| 222 case kTruncate: | |
| 223 if (ch == dir_separator) { | |
| 224 // Skip this separator, it was only put in to break up long | |
| 225 // path segments, but is not part of the URL. | |
| 226 state = kStart; | |
| 227 } else { | |
| 228 return false; | |
| 229 } | |
| 230 break; | |
| 231 case kEscapeDot: | |
| 232 decoded_url->append(1, ch); | |
| 233 state = kStart; | |
| 234 break; | |
| 235 } | |
| 236 } | |
| 237 | |
| 238 // All legal encoded filenames end in kEscapeChar. | |
| 239 return (state == kEscape); | |
| 240 } | |
| 241 | |
| 242 // Escape the given input |path| and chop any individual components | |
| 243 // of the path which are greater than kMaximumSubdirectoryLength characters | |
| 244 // into two chunks. | |
| 245 // | |
| 246 // This legacy version has several issues with aliasing of different URLs, | |
| 247 // inability to represent both /a/b/c and /a/b/c/d, and inability to decode | |
| 248 // the filenames back into URLs. | |
| 249 // | |
| 250 // But there is a large body of slurped data which depends on this format, | |
| 251 // so leave it as the default for spdy_in_mem_edsm_server. | |
| 252 string UrlToFilenameEncoder::LegacyEscape(const string& path) { | |
| 253 string output; | |
| 254 | |
| 255 // Note: We also chop paths into medium sized 'chunks'. | |
| 256 // This is due to the incompetence of the windows | |
| 257 // filesystem, which still hasn't figured out how | |
| 258 // to deal with long filenames. | |
| 259 int last_slash = 0; | |
| 260 for (size_t index = 0; index < path.length(); index++) { | |
| 261 char ch = path[index]; | |
| 262 if (ch == 0x5C) | |
| 263 last_slash = index; | |
| 264 if ((ch == 0x2D) || // hyphen | |
| 265 (ch == 0x5C) || (ch == 0x5F) || // backslash, underscore | |
| 266 ((0x30 <= ch) && (ch <= 0x39)) || // Digits [0-9] | |
| 267 ((0x41 <= ch) && (ch <= 0x5A)) || // Uppercase [A-Z] | |
| 268 ((0x61 <= ch) && (ch <= 0x7A))) { // Lowercase [a-z] | |
| 269 output.append(&path[index], 1); | |
| 270 } else { | |
| 271 char encoded[3]; | |
| 272 encoded[0] = 'x'; | |
| 273 encoded[1] = ch / 16; | |
| 274 encoded[1] += (encoded[1] >= 10) ? 'A' - 10 : '0'; | |
| 275 encoded[2] = ch % 16; | |
| 276 encoded[2] += (encoded[2] >= 10) ? 'A' - 10 : '0'; | |
| 277 output.append(encoded, 3); | |
| 278 } | |
| 279 if (index - last_slash > kMaximumSubdirectoryLength) { | |
| 280 #ifdef WIN32 | |
| 281 char slash = '\\'; | |
| 282 #else | |
| 283 char slash = '/'; | |
| 284 #endif | |
| 285 output.append(&slash, 1); | |
| 286 last_slash = index; | |
| 287 } | |
| 288 } | |
| 289 return output; | |
| 290 } | |
| 291 | |
| 292 } // namespace net | |
| OLD | NEW |
|---|
Chromium Code Reviews
Issue 992733002:
Remove //net (except for Android test stuff) and sdch (Closed)
Base URL: git@github.com:domokit/mojo.git@master