• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // Copyright (c) 2009 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 <limits>
6 #include <set>
7 #include <string>
8 
9 #include "chrome/browser/history/text_database.h"
10 
11 #include "app/sql/statement.h"
12 #include "app/sql/transaction.h"
13 #include "base/file_util.h"
14 #include "base/logging.h"
15 #include "base/metrics/histogram.h"
16 #include "base/string_number_conversions.h"
17 #include "base/string_util.h"
18 #include "base/utf_string_conversions.h"
19 #include "chrome/browser/diagnostics/sqlite_diagnostics.h"
20 
21 // There are two tables in each database, one full-text search (FTS) table which
22 // indexes the contents and title of the pages. The other is a regular SQLITE
23 // table which contains non-indexed information about the page. All columns of
24 // a FTS table are indexed using the text search algorithm, which isn't what we
25 // want for things like times. If this were in the FTS table, there would be
26 // different words in the index for each time number.
27 //
28 // "pages" FTS table:
29 //   url    URL of the page so searches will match the URL.
30 //   title  Title of the page.
31 //   body   Body of the page.
32 //
33 // "info" regular table:
34 //   time     Time the corresponding FTS entry was visited.
35 //
36 // We do joins across these two tables by using their internal rowids, which we
37 // keep in sync between the two tables. The internal rowid is the only part of
38 // an FTS table that is indexed like a normal table, and the index over it is
39 // free since sqlite always indexes the internal rowid.
40 
41 namespace history {
42 
43 namespace {
44 
45 // Version 1 uses FTS2 for index files.
46 // Version 2 uses FTS3.
47 static const int kCurrentVersionNumber = 2;
48 static const int kCompatibleVersionNumber = 2;
49 
50 // Snippet computation relies on the index of the columns in the original
51 // create statement. These are the 0-based indices (as strings) of the
52 // corresponding columns.
53 const char kTitleColumnIndex[] = "1";
54 const char kBodyColumnIndex[] = "2";
55 
56 // The string prepended to the database identifier to generate the filename.
57 const FilePath::CharType kFilePrefix[] = FILE_PATH_LITERAL("History Index ");
58 
59 }  // namespace
60 
Match()61 TextDatabase::Match::Match() {}
62 
~Match()63 TextDatabase::Match::~Match() {}
64 
TextDatabase(const FilePath & path,DBIdent id,bool allow_create)65 TextDatabase::TextDatabase(const FilePath& path,
66                            DBIdent id,
67                            bool allow_create)
68     : path_(path),
69       ident_(id),
70       allow_create_(allow_create) {
71   // Compute the file name.
72   file_name_ = path_.Append(IDToFileName(ident_));
73 }
74 
~TextDatabase()75 TextDatabase::~TextDatabase() {
76 }
77 
78 // static
file_base()79 const FilePath::CharType* TextDatabase::file_base() {
80   return kFilePrefix;
81 }
82 
83 // static
IDToFileName(DBIdent id)84 FilePath TextDatabase::IDToFileName(DBIdent id) {
85   // Identifiers are intended to be a combination of the year and month, for
86   // example, 200801 for January 2008. We convert this to
87   // "History Index 2008-01". However, we don't make assumptions about this
88   // scheme: the caller should assign IDs as it feels fit with the knowledge
89   // that they will apppear on disk in this form.
90   FilePath::StringType filename(file_base());
91   base::StringAppendF(&filename, FILE_PATH_LITERAL("%d-%02d"),
92                       id / 100, id % 100);
93   return FilePath(filename);
94 }
95 
96 // static
FileNameToID(const FilePath & file_path)97 TextDatabase::DBIdent TextDatabase::FileNameToID(const FilePath& file_path) {
98   FilePath::StringType file_name = file_path.BaseName().value();
99 
100   // We don't actually check the prefix here. Since the file system could
101   // be case insensitive in ways we can't predict (NTFS), checking could
102   // potentially be the wrong thing to do. Instead, we just look for a suffix.
103   static const size_t kIDStringLength = 7;  // Room for "xxxx-xx".
104   if (file_name.length() < kIDStringLength)
105     return 0;
106   const FilePath::StringType suffix(
107       &file_name[file_name.length() - kIDStringLength]);
108 
109   if (suffix.length() != kIDStringLength ||
110       suffix[4] != FILE_PATH_LITERAL('-')) {
111     return 0;
112   }
113 
114   int year, month;
115   base::StringToInt(suffix.begin(), suffix.begin() + 4, &year);
116   base::StringToInt(suffix.begin() + 5, suffix.begin() + 7, &month);
117 
118   return year * 100 + month;
119 }
120 
Init()121 bool TextDatabase::Init() {
122   // Make sure, if we're not allowed to create the file, that it exists.
123   if (!allow_create_) {
124     if (!file_util::PathExists(file_name_))
125       return false;
126   }
127 
128   // Set the exceptional sqlite error handler.
129   db_.set_error_delegate(GetErrorHandlerForTextDb());
130 
131   // Set the database page size to something a little larger to give us
132   // better performance (we're typically seek rather than bandwidth limited).
133   // This only has an effect before any tables have been created, otherwise
134   // this is a NOP. Must be a power of 2 and a max of 8192.
135   db_.set_page_size(4096);
136 
137   // The default cache size is 2000 which give >8MB of data. Since we will often
138   // have 2-3 of these objects, each with their own 8MB, this adds up very fast.
139   // We therefore reduce the size so when there are multiple objects, we're not
140   // too big.
141   db_.set_cache_size(512);
142 
143   // Run the database in exclusive mode. Nobody else should be accessing the
144   // database while we're running, and this will give somewhat improved perf.
145   db_.set_exclusive_locking();
146 
147   // Attach the database to our index file.
148   if (!db_.Open(file_name_))
149     return false;
150 
151   // Meta table tracking version information.
152   if (!meta_table_.Init(&db_, kCurrentVersionNumber, kCompatibleVersionNumber))
153     return false;
154   if (meta_table_.GetCompatibleVersionNumber() > kCurrentVersionNumber) {
155     // This version is too new. We don't bother notifying the user on this
156     // error, and just fail to use the file. Normally if they have version skew,
157     // they will get it for the main history file and it won't be necessary
158     // here. If that's not the case, since this is only indexed data, it's
159     // probably better to just not give FTS results than strange errors when
160     // everything else is working OK.
161     LOG(WARNING) << "Text database is too new.";
162     return false;
163   }
164 
165   return CreateTables();
166 }
167 
BeginTransaction()168 void TextDatabase::BeginTransaction() {
169   db_.BeginTransaction();
170 }
171 
CommitTransaction()172 void TextDatabase::CommitTransaction() {
173   db_.CommitTransaction();
174 }
175 
CreateTables()176 bool TextDatabase::CreateTables() {
177   // FTS table of page contents.
178   if (!db_.DoesTableExist("pages")) {
179     if (!db_.Execute("CREATE VIRTUAL TABLE pages USING fts3("
180                      "TOKENIZE icu,"
181                      "url LONGVARCHAR,"
182                      "title LONGVARCHAR,"
183                      "body LONGVARCHAR)"))
184       return false;
185   }
186 
187   // Non-FTS table containing URLs and times so we can efficiently find them
188   // using a regular index (all FTS columns are special and are treated as
189   // full-text-search, which is not what we want when retrieving this data).
190   if (!db_.DoesTableExist("info")) {
191     // Note that there is no point in creating an index over time. Since
192     // we must always query the entire FTS table (it can not efficiently do
193     // subsets), we will always end up doing that first, and joining the info
194     // table off of that.
195     if (!db_.Execute("CREATE TABLE info(time INTEGER NOT NULL)"))
196       return false;
197   }
198 
199   // Create the index. This will fail when the index already exists, so we just
200   // ignore the error.
201   db_.Execute("CREATE INDEX info_time ON info(time)");
202   return true;
203 }
204 
AddPageData(base::Time time,const std::string & url,const std::string & title,const std::string & contents)205 bool TextDatabase::AddPageData(base::Time time,
206                                const std::string& url,
207                                const std::string& title,
208                                const std::string& contents) {
209   sql::Transaction committer(&db_);
210   if (!committer.Begin())
211     return false;
212 
213   // Add to the pages table.
214   sql::Statement add_to_pages(db_.GetCachedStatement(SQL_FROM_HERE,
215       "INSERT INTO pages (url, title, body) VALUES (?,?,?)"));
216   if (!add_to_pages) {
217     NOTREACHED() << db_.GetErrorMessage();
218     return false;
219   }
220   add_to_pages.BindString(0, url);
221   add_to_pages.BindString(1, title);
222   add_to_pages.BindString(2, contents);
223   if (!add_to_pages.Run()) {
224     NOTREACHED() << db_.GetErrorMessage();
225     return false;
226   }
227 
228   int64 rowid = db_.GetLastInsertRowId();
229 
230   // Add to the info table with the same rowid.
231   sql::Statement add_to_info(db_.GetCachedStatement(SQL_FROM_HERE,
232       "INSERT INTO info (rowid, time) VALUES (?,?)"));
233   if (!add_to_info) {
234     NOTREACHED() << db_.GetErrorMessage();
235     return false;
236   }
237   add_to_info.BindInt64(0, rowid);
238   add_to_info.BindInt64(1, time.ToInternalValue());
239   if (!add_to_info.Run()) {
240     NOTREACHED() << db_.GetErrorMessage();
241     return false;
242   }
243 
244   return committer.Commit();
245 }
246 
DeletePageData(base::Time time,const std::string & url)247 void TextDatabase::DeletePageData(base::Time time, const std::string& url) {
248   // First get all rows that match. Selecing on time (which has an index) allows
249   // us to avoid brute-force searches on the full-text-index table (there will
250   // generally be only one match per time).
251   sql::Statement select_ids(db_.GetCachedStatement(SQL_FROM_HERE,
252       "SELECT info.rowid "
253       "FROM info JOIN pages ON info.rowid = pages.rowid "
254       "WHERE info.time=? AND pages.url=?"));
255   if (!select_ids)
256     return;
257   select_ids.BindInt64(0, time.ToInternalValue());
258   select_ids.BindString(1, url);
259   std::set<int64> rows_to_delete;
260   while (select_ids.Step())
261     rows_to_delete.insert(select_ids.ColumnInt64(0));
262 
263   // Delete from the pages table.
264   sql::Statement delete_page(db_.GetCachedStatement(SQL_FROM_HERE,
265       "DELETE FROM pages WHERE rowid=?"));
266   if (!delete_page)
267     return;
268   for (std::set<int64>::const_iterator i = rows_to_delete.begin();
269        i != rows_to_delete.end(); ++i) {
270     delete_page.BindInt64(0, *i);
271     if (!delete_page.Run()) {
272       NOTREACHED();
273       return;
274     }
275     delete_page.Reset();
276   }
277 
278   // Delete from the info table.
279   sql::Statement delete_info(db_.GetCachedStatement(SQL_FROM_HERE,
280       "DELETE FROM info WHERE rowid=?"));
281   if (!delete_info)
282     return;
283   for (std::set<int64>::const_iterator i = rows_to_delete.begin();
284        i != rows_to_delete.end(); ++i) {
285     delete_info.BindInt64(0, *i);
286     if (!delete_info.Run()) {
287       NOTREACHED();
288       return;
289     }
290     delete_info.Reset();
291   }
292 }
293 
Optimize()294 void TextDatabase::Optimize() {
295   sql::Statement statement(db_.GetCachedStatement(SQL_FROM_HERE,
296       "SELECT OPTIMIZE(pages) FROM pages LIMIT 1"));
297   if (!statement)
298     return;
299   statement.Run();
300 }
301 
GetTextMatches(const std::string & query,const QueryOptions & options,std::vector<Match> * results,URLSet * found_urls,base::Time * first_time_searched)302 void TextDatabase::GetTextMatches(const std::string& query,
303                                   const QueryOptions& options,
304                                   std::vector<Match>* results,
305                                   URLSet* found_urls,
306                                   base::Time* first_time_searched) {
307   *first_time_searched = options.begin_time;
308 
309   sql::Statement statement(db_.GetCachedStatement(SQL_FROM_HERE,
310     "SELECT url, title, time, offsets(pages), body "
311         "FROM pages LEFT OUTER JOIN info ON pages.rowid = info.rowid "
312         "WHERE pages MATCH ? AND time >= ? AND time < ? "
313         "ORDER BY time DESC "
314         "LIMIT ?"));
315   if (!statement)
316     return;
317 
318   // When their values indicate "unspecified", saturate the numbers to the max
319   // or min to get the correct result.
320   int64 effective_begin_time = options.begin_time.is_null() ?
321       0 : options.begin_time.ToInternalValue();
322   int64 effective_end_time = options.end_time.is_null() ?
323       std::numeric_limits<int64>::max() : options.end_time.ToInternalValue();
324   int effective_max_count = options.max_count ?
325       options.max_count : std::numeric_limits<int>::max();
326 
327   statement.BindString(0, query);
328   statement.BindInt64(1, effective_begin_time);
329   statement.BindInt64(2, effective_end_time);
330   statement.BindInt(3, effective_max_count);
331 
332   while (statement.Step()) {
333     // TODO(brettw) allow canceling the query in the middle.
334     // if (canceled_or_something)
335     //   break;
336 
337     GURL url(statement.ColumnString(0));
338     URLSet::const_iterator found_url = found_urls->find(url);
339     if (found_url != found_urls->end())
340       continue;  // Don't add this duplicate.
341 
342     // Fill the results into the vector (avoid copying the URL with Swap()).
343     results->resize(results->size() + 1);
344     Match& match = results->at(results->size() - 1);
345     match.url.Swap(&url);
346 
347     match.title = statement.ColumnString16(1);
348     match.time = base::Time::FromInternalValue(statement.ColumnInt64(2));
349 
350     // Extract any matches in the title.
351     std::string offsets_str = statement.ColumnString(3);
352     Snippet::ExtractMatchPositions(offsets_str, kTitleColumnIndex,
353                                    &match.title_match_positions);
354     Snippet::ConvertMatchPositionsToWide(statement.ColumnString(1),
355                                          &match.title_match_positions);
356 
357     // Extract the matches in the body.
358     Snippet::MatchPositions match_positions;
359     Snippet::ExtractMatchPositions(offsets_str, kBodyColumnIndex,
360                                    &match_positions);
361 
362     // Compute the snippet based on those matches.
363     std::string body = statement.ColumnString(4);
364     match.snippet.ComputeSnippet(match_positions, body);
365   }
366 
367   // When we have returned all the results possible (or determined that there
368   // are none), then we have searched all the time requested, so we can
369   // set the first_time_searched to that value.
370   if (results->empty() ||
371       options.max_count == 0 ||  // Special case for wanting all the results.
372       static_cast<int>(results->size()) < options.max_count) {
373     *first_time_searched = options.begin_time;
374   } else {
375     // Since we got the results in order, we know the last item is the last
376     // time we considered.
377     *first_time_searched = results->back().time;
378   }
379 
380   statement.Reset();
381 }
382 
383 }  // namespace history
384