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 "chrome/browser/history/snippet.h" 6 7 #include <algorithm> 8 9 #include "base/logging.h" 10 #include "base/memory/scoped_ptr.h" 11 #include "base/string_split.h" 12 #include "base/string_util.h" 13 #include "base/utf_string_conversions.h" 14 #include "unicode/brkiter.h" 15 #include "unicode/utext.h" 16 #include "unicode/utf8.h" 17 18 namespace { 19 20 bool PairFirstLessThan(const Snippet::MatchPosition& a, 21 const Snippet::MatchPosition& b) { 22 return a.first < b.first; 23 } 24 25 // Combines all pairs after offset in match_positions that are contained 26 // or touch the pair at offset. 27 void CoalescePositionsFrom(size_t offset, 28 Snippet::MatchPositions* match_positions) { 29 DCHECK(offset < match_positions->size()); 30 Snippet::MatchPosition& pair((*match_positions)[offset]); 31 ++offset; 32 while (offset < match_positions->size() && 33 pair.second >= (*match_positions)[offset].first) { 34 pair.second = std::max(pair.second, (*match_positions)[offset].second); 35 match_positions->erase(match_positions->begin() + offset); 36 } 37 } 38 39 // Makes sure there is a pair in match_positions that contains the specified 40 // range. This keeps the pairs ordered in match_positions by first, and makes 41 // sure none of the pairs in match_positions touch each other. 42 void AddMatch(size_t start, 43 size_t end, 44 Snippet::MatchPositions* match_positions) { 45 DCHECK(start < end); 46 DCHECK(match_positions); 47 Snippet::MatchPosition pair(start, end); 48 if (match_positions->empty()) { 49 match_positions->push_back(pair); 50 return; 51 } 52 // There's at least one match. Find the position of the new match, 53 // potentially extending pairs around it. 54 Snippet::MatchPositions::iterator i = 55 std::lower_bound(match_positions->begin(), match_positions->end(), 56 pair, &PairFirstLessThan); 57 if (i != match_positions->end() && i->first == start) { 58 // Match not at the end and there is already a pair with the same 59 // start. 60 if (end > i->second) { 61 // New pair extends beyond existing pair. Extend existing pair and 62 // coalesce matches after it. 63 i->second = end; 64 CoalescePositionsFrom(i - match_positions->begin(), match_positions); 65 } // else case, new pair completely contained in existing pair, nothing 66 // to do. 67 } else if (i == match_positions->begin()) { 68 // Match at the beginning and the first pair doesn't have the same 69 // start. Insert new pair and coalesce matches after it. 70 match_positions->insert(i, pair); 71 CoalescePositionsFrom(0, match_positions); 72 } else { 73 // Not at the beginning (but may be at the end). 74 --i; 75 if (start <= i->second && end > i->second) { 76 // Previous element contains match. Extend it and coalesce. 77 i->second = end; 78 CoalescePositionsFrom(i - match_positions->begin(), match_positions); 79 } else if (end > i->second) { 80 // Region doesn't touch previous element. See if region touches current 81 // element. 82 ++i; 83 if (i == match_positions->end() || end < i->first) { 84 match_positions->insert(i, pair); 85 } else { 86 i->first = start; 87 i->second = end; 88 CoalescePositionsFrom(i - match_positions->begin(), match_positions); 89 } 90 } 91 } 92 } 93 94 // Converts an index in a utf8 string into the index in the corresponding utf16 95 // string and returns the utf16 index. This is intended to be called in a loop 96 // iterating through a utf8 string. 97 // 98 // utf8_string: the utf8 string. 99 // utf8_length: length of the utf8 string. 100 // offset: the utf8 offset to convert. 101 // utf8_pos: current offset in the utf8 string. This is modified and on return 102 // matches offset. 103 // wide_pos: current index in the wide string. This is the same as the return 104 // value. 105 size_t AdvanceAndReturnUTF16Pos(const char* utf8_string, 106 int32_t utf8_length, 107 int32_t offset, 108 int32_t* utf8_pos, 109 size_t* utf16_pos) { 110 DCHECK(offset >= *utf8_pos && offset <= utf8_length); 111 112 UChar32 wide_char; 113 while (*utf8_pos < offset) { 114 U8_NEXT(utf8_string, *utf8_pos, utf8_length, wide_char); 115 *utf16_pos += (wide_char <= 0xFFFF) ? 1 : 2; 116 } 117 return *utf16_pos; 118 } 119 120 // Given a character break iterator over a UTF-8 string, set the iterator 121 // position to |*utf8_pos| and move by |count| characters. |count| can 122 // be either positive or negative. 123 void MoveByNGraphemes(icu::BreakIterator* bi, int count, size_t* utf8_pos) { 124 // Ignore the return value. A side effect of the current position 125 // being set at or following |*utf8_pos| is exploited here. 126 // It's simpler than calling following(n) and then previous(). 127 // isBoundary() is not very fast, but should be good enough for the 128 // snippet generation. If not, revisit the way we scan in ComputeSnippet. 129 bi->isBoundary(*utf8_pos); 130 bi->next(count); 131 *utf8_pos = static_cast<size_t>(bi->current()); 132 } 133 134 // The amount of context to include for a given hit. Note that it's counted 135 // in terms of graphemes rather than bytes. 136 const int kSnippetContext = 50; 137 138 // Returns true if next match falls within a snippet window 139 // from the previous match. The window size is counted in terms 140 // of graphemes rather than bytes in UTF-8. 141 bool IsNextMatchWithinSnippetWindow(icu::BreakIterator* bi, 142 size_t previous_match_end, 143 size_t next_match_start) { 144 // If it's within a window in terms of bytes, it's certain 145 // that it's within a window in terms of graphemes as well. 146 if (next_match_start < previous_match_end + kSnippetContext) 147 return true; 148 bi->isBoundary(previous_match_end); 149 // An alternative to this is to call |bi->next()| at most 150 // kSnippetContext times, compare |bi->current()| with |next_match_start| 151 // after each call and return early if possible. There are other 152 // heuristics to speed things up if necessary, but it's not likely that 153 // we need to bother. 154 bi->next(kSnippetContext); 155 int64 current = bi->current(); 156 return (next_match_start < static_cast<uint64>(current) || 157 current == icu::BreakIterator::DONE); 158 } 159 160 } // namespace 161 162 // static 163 void Snippet::ExtractMatchPositions(const std::string& offsets_str, 164 const std::string& column_num, 165 MatchPositions* match_positions) { 166 DCHECK(match_positions); 167 if (offsets_str.empty()) 168 return; 169 std::vector<std::string> offsets; 170 base::SplitString(offsets_str, ' ', &offsets); 171 // SQLite offsets are sets of four integers: 172 // column, query term, match offset, match length 173 // Matches within a string are marked by (start, end) pairs. 174 for (size_t i = 0; i < offsets.size() - 3; i += 4) { 175 if (offsets[i] != column_num) 176 continue; 177 const size_t start = atoi(offsets[i + 2].c_str()); 178 const size_t end = start + atoi(offsets[i + 3].c_str()); 179 // Switch to DCHECK after debugging http://crbug.com/15261. 180 CHECK(end >= start); 181 AddMatch(start, end, match_positions); 182 } 183 } 184 185 // static 186 void Snippet::ConvertMatchPositionsToWide( 187 const std::string& utf8_string, 188 Snippet::MatchPositions* match_positions) { 189 DCHECK(match_positions); 190 int32_t utf8_pos = 0; 191 size_t utf16_pos = 0; 192 const char* utf8_cstring = utf8_string.c_str(); 193 const int32_t utf8_length = static_cast<int32_t>(utf8_string.size()); 194 for (Snippet::MatchPositions::iterator i = match_positions->begin(); 195 i != match_positions->end(); ++i) { 196 i->first = AdvanceAndReturnUTF16Pos(utf8_cstring, utf8_length, 197 i->first, &utf8_pos, &utf16_pos); 198 i->second = AdvanceAndReturnUTF16Pos(utf8_cstring, utf8_length, 199 i->second, &utf8_pos, &utf16_pos); 200 } 201 } 202 203 Snippet::Snippet() { 204 } 205 206 Snippet::~Snippet() { 207 } 208 209 void Snippet::ComputeSnippet(const MatchPositions& match_positions, 210 const std::string& document) { 211 // The length of snippets we try to produce. 212 // We can generate longer snippets but stop once we cross kSnippetMaxLength. 213 const size_t kSnippetMaxLength = 200; 214 const string16 kEllipsis = ASCIIToUTF16(" ... "); 215 216 UText* document_utext = NULL; 217 UErrorCode status = U_ZERO_ERROR; 218 document_utext = utext_openUTF8(document_utext, document.data(), 219 document.size(), &status); 220 // Locale does not matter because there's no per-locale customization 221 // for character iterator. 222 scoped_ptr<icu::BreakIterator> bi(icu::BreakIterator::createCharacterInstance( 223 icu::Locale::getDefault(), status)); 224 bi->setText(document_utext, status); 225 DCHECK(U_SUCCESS(status)); 226 227 // We build the snippet by iterating through the matches and then grabbing 228 // context around each match. If matches are near enough each other (within 229 // kSnippetContext), we skip the "..." between them. 230 string16 snippet; 231 size_t start = 0; 232 for (size_t i = 0; i < match_positions.size(); ++i) { 233 // Some shorter names for the current match. 234 const size_t match_start = match_positions[i].first; 235 const size_t match_end = match_positions[i].second; 236 237 // Switch to DCHECK after debugging http://crbug.com/15261. 238 CHECK(match_end > match_start); 239 CHECK(match_end <= document.size()); 240 241 // Add the context, if any, to show before the match. 242 size_t context_start = match_start; 243 MoveByNGraphemes(bi.get(), -kSnippetContext, &context_start); 244 start = std::max(start, context_start); 245 if (start < match_start) { 246 if (start > 0) 247 snippet += kEllipsis; 248 // Switch to DCHECK after debugging http://crbug.com/15261. 249 CHECK(start < document.size()); 250 snippet += UTF8ToUTF16(document.substr(start, match_start - start)); 251 } 252 253 // Add the match. 254 const size_t first = snippet.size(); 255 snippet += UTF8ToUTF16(document.substr(match_start, 256 match_end - match_start)); 257 matches_.push_back(std::make_pair(first, snippet.size())); 258 259 // Compute the context, if any, to show after the match. 260 size_t end; 261 // Check if the next match falls within our snippet window. 262 if (i + 1 < match_positions.size() && 263 IsNextMatchWithinSnippetWindow(bi.get(), match_end, 264 match_positions[i + 1].first)) { 265 // Yes, it's within the window. Make the end context extend just up 266 // to the next match. 267 end = match_positions[i + 1].first; 268 // Switch to DCHECK after debugging http://crbug.com/15261. 269 CHECK(end >= match_end); 270 CHECK(end <= document.size()); 271 snippet += UTF8ToUTF16(document.substr(match_end, end - match_end)); 272 } else { 273 // No, there's either no next match or the next match is too far away. 274 end = match_end; 275 MoveByNGraphemes(bi.get(), kSnippetContext, &end); 276 // Switch to DCHECK after debugging http://crbug.com/15261. 277 CHECK(end >= match_end); 278 CHECK(end <= document.size()); 279 snippet += UTF8ToUTF16(document.substr(match_end, end - match_end)); 280 if (end < document.size()) 281 snippet += kEllipsis; 282 } 283 start = end; 284 285 // Stop here if we have enough snippet computed. 286 if (snippet.size() >= kSnippetMaxLength) 287 break; 288 } 289 290 utext_close(document_utext); 291 swap(text_, snippet); 292 } 293 294 void Snippet::Swap(Snippet* other) { 295 text_.swap(other->text_); 296 matches_.swap(other->matches_); 297 } 298