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/history_backend.h"
6
7 #include <list>
8 #include <map>
9 #include <set>
10 #include <vector>
11
12 #include "base/command_line.h"
13 #include "base/compiler_specific.h"
14 #include "base/file_util.h"
15 #include "base/memory/scoped_ptr.h"
16 #include "base/memory/scoped_vector.h"
17 #include "base/message_loop.h"
18 #include "base/metrics/histogram.h"
19 #include "base/string_util.h"
20 #include "base/time.h"
21 #include "chrome/browser/autocomplete/history_url_provider.h"
22 #include "chrome/browser/bookmarks/bookmark_service.h"
23 #include "chrome/browser/history/download_create_info.h"
24 #include "chrome/browser/history/history_notifications.h"
25 #include "chrome/browser/history/history_publisher.h"
26 #include "chrome/browser/history/in_memory_history_backend.h"
27 #include "chrome/browser/history/page_usage_data.h"
28 #include "chrome/browser/history/top_sites.h"
29 #include "chrome/common/chrome_constants.h"
30 #include "chrome/common/url_constants.h"
31 #include "content/common/notification_type.h"
32 #include "googleurl/src/gurl.h"
33 #include "grit/chromium_strings.h"
34 #include "grit/generated_resources.h"
35 #include "net/base/registry_controlled_domain.h"
36
37 using base::Time;
38 using base::TimeDelta;
39 using base::TimeTicks;
40
41 /* The HistoryBackend consists of a number of components:
42
43 HistoryDatabase (stores past 3 months of history)
44 URLDatabase (stores a list of URLs)
45 DownloadDatabase (stores a list of downloads)
46 VisitDatabase (stores a list of visits for the URLs)
47 VisitSegmentDatabase (stores groups of URLs for the most visited view).
48
49 ArchivedDatabase (stores history older than 3 months)
50 URLDatabase (stores a list of URLs)
51 DownloadDatabase (stores a list of downloads)
52 VisitDatabase (stores a list of visits for the URLs)
53
54 (this does not store visit segments as they expire after 3 mos.)
55
56 TextDatabaseManager (manages multiple text database for different times)
57 TextDatabase (represents a single month of full-text index).
58 ...more TextDatabase objects...
59
60 ExpireHistoryBackend (manages moving things from HistoryDatabase to
61 the ArchivedDatabase and deleting)
62 */
63
64 namespace history {
65
66 // How long we keep segment data for in days. Currently 3 months.
67 // This value needs to be greater or equal to
68 // MostVisitedModel::kMostVisitedScope but we don't want to introduce a direct
69 // dependency between MostVisitedModel and the history backend.
70 static const int kSegmentDataRetention = 90;
71
72 // The number of milliseconds we'll wait to do a commit, so that things are
73 // batched together.
74 static const int kCommitIntervalMs = 10000;
75
76 // The amount of time before we re-fetch the favicon.
77 static const int kFaviconRefetchDays = 7;
78
79 // GetSessionTabs returns all open tabs, or tabs closed kSessionCloseTimeWindow
80 // seconds ago.
81 static const int kSessionCloseTimeWindowSecs = 10;
82
83 // The maximum number of items we'll allow in the redirect list before
84 // deleting some.
85 static const int kMaxRedirectCount = 32;
86
87 // The number of days old a history entry can be before it is considered "old"
88 // and is archived.
89 static const int kArchiveDaysThreshold = 90;
90
91 // Converts from PageUsageData to MostVisitedURL. |redirects| is a
92 // list of redirects for this URL. Empty list means no redirects.
MakeMostVisitedURL(const PageUsageData & page_data,const RedirectList & redirects)93 MostVisitedURL MakeMostVisitedURL(const PageUsageData& page_data,
94 const RedirectList& redirects) {
95 MostVisitedURL mv;
96 mv.url = page_data.GetURL();
97 mv.title = page_data.GetTitle();
98 if (redirects.empty()) {
99 // Redirects must contain at least the target url.
100 mv.redirects.push_back(mv.url);
101 } else {
102 mv.redirects = redirects;
103 if (mv.redirects[mv.redirects.size() - 1] != mv.url) {
104 // The last url must be the target url.
105 mv.redirects.push_back(mv.url);
106 }
107 }
108 return mv;
109 }
110
111 // This task is run on a timer so that commits happen at regular intervals
112 // so they are batched together. The important thing about this class is that
113 // it supports canceling of the task so the reference to the backend will be
114 // freed. The problem is that when history is shutting down, there is likely
115 // to be one of these commits still pending and holding a reference.
116 //
117 // The backend can call Cancel to have this task release the reference. The
118 // task will still run (if we ever get to processing the event before
119 // shutdown), but it will not do anything.
120 //
121 // Note that this is a refcounted object and is not a task in itself. It should
122 // be assigned to a RunnableMethod.
123 //
124 // TODO(brettw): bug 1165182: This should be replaced with a
125 // ScopedRunnableMethodFactory which will handle everything automatically (like
126 // we do in ExpireHistoryBackend).
127 class CommitLaterTask : public base::RefCounted<CommitLaterTask> {
128 public:
CommitLaterTask(HistoryBackend * history_backend)129 explicit CommitLaterTask(HistoryBackend* history_backend)
130 : history_backend_(history_backend) {
131 }
132
133 // The backend will call this function if it is being destroyed so that we
134 // release our reference.
Cancel()135 void Cancel() {
136 history_backend_ = NULL;
137 }
138
RunCommit()139 void RunCommit() {
140 if (history_backend_.get())
141 history_backend_->Commit();
142 }
143
144 private:
145 friend class base::RefCounted<CommitLaterTask>;
146
~CommitLaterTask()147 ~CommitLaterTask() {}
148
149 scoped_refptr<HistoryBackend> history_backend_;
150 };
151
152 // Handles querying first the main database, then the full text database if that
153 // fails. It will optionally keep track of all URLs seen so duplicates can be
154 // eliminated. This is used by the querying sub-functions.
155 //
156 // TODO(brettw): This class may be able to be simplified or eliminated. After
157 // this was written, QueryResults can efficiently look up by URL, so the need
158 // for this extra set of previously queried URLs is less important.
159 class HistoryBackend::URLQuerier {
160 public:
URLQuerier(URLDatabase * main_db,URLDatabase * archived_db,bool track_unique)161 URLQuerier(URLDatabase* main_db, URLDatabase* archived_db, bool track_unique)
162 : main_db_(main_db),
163 archived_db_(archived_db),
164 track_unique_(track_unique) {
165 }
166
167 // When we're tracking unique URLs, returns true if this URL has been
168 // previously queried. Only call when tracking unique URLs.
HasURL(const GURL & url)169 bool HasURL(const GURL& url) {
170 DCHECK(track_unique_);
171 return unique_urls_.find(url) != unique_urls_.end();
172 }
173
GetRowForURL(const GURL & url,URLRow * row)174 bool GetRowForURL(const GURL& url, URLRow* row) {
175 if (!main_db_->GetRowForURL(url, row)) {
176 if (!archived_db_ || !archived_db_->GetRowForURL(url, row)) {
177 // This row is neither in the main nor the archived DB.
178 return false;
179 }
180 }
181
182 if (track_unique_)
183 unique_urls_.insert(url);
184 return true;
185 }
186
187 private:
188 URLDatabase* main_db_; // Guaranteed non-NULL.
189 URLDatabase* archived_db_; // Possibly NULL.
190
191 bool track_unique_;
192
193 // When track_unique_ is set, this is updated with every URL seen so far.
194 std::set<GURL> unique_urls_;
195
196 DISALLOW_COPY_AND_ASSIGN(URLQuerier);
197 };
198
199 // HistoryBackend --------------------------------------------------------------
200
HistoryBackend(const FilePath & history_dir,Delegate * delegate,BookmarkService * bookmark_service)201 HistoryBackend::HistoryBackend(const FilePath& history_dir,
202 Delegate* delegate,
203 BookmarkService* bookmark_service)
204 : delegate_(delegate),
205 history_dir_(history_dir),
206 ALLOW_THIS_IN_INITIALIZER_LIST(expirer_(this, bookmark_service)),
207 recent_redirects_(kMaxRedirectCount),
208 backend_destroy_message_loop_(NULL),
209 backend_destroy_task_(NULL),
210 segment_queried_(false),
211 bookmark_service_(bookmark_service) {
212 }
213
~HistoryBackend()214 HistoryBackend::~HistoryBackend() {
215 DCHECK(!scheduled_commit_) << "Deleting without cleanup";
216 ReleaseDBTasks();
217
218 // First close the databases before optionally running the "destroy" task.
219 if (db_.get()) {
220 // Commit the long-running transaction.
221 db_->CommitTransaction();
222 db_.reset();
223 }
224 if (thumbnail_db_.get()) {
225 thumbnail_db_->CommitTransaction();
226 thumbnail_db_.reset();
227 }
228 if (archived_db_.get()) {
229 archived_db_->CommitTransaction();
230 archived_db_.reset();
231 }
232 if (text_database_.get()) {
233 text_database_->CommitTransaction();
234 text_database_.reset();
235 }
236
237 if (backend_destroy_task_) {
238 // Notify an interested party (typically a unit test) that we're done.
239 DCHECK(backend_destroy_message_loop_);
240 backend_destroy_message_loop_->PostTask(FROM_HERE, backend_destroy_task_);
241 }
242 }
243
Init(const std::string & languages,bool force_fail)244 void HistoryBackend::Init(const std::string& languages, bool force_fail) {
245 if (!force_fail)
246 InitImpl(languages);
247 delegate_->DBLoaded();
248 }
249
SetOnBackendDestroyTask(MessageLoop * message_loop,Task * task)250 void HistoryBackend::SetOnBackendDestroyTask(MessageLoop* message_loop,
251 Task* task) {
252 if (backend_destroy_task_) {
253 DLOG(WARNING) << "Setting more than one destroy task, overriding";
254 delete backend_destroy_task_;
255 }
256 backend_destroy_message_loop_ = message_loop;
257 backend_destroy_task_ = task;
258 }
259
Closing()260 void HistoryBackend::Closing() {
261 // Any scheduled commit will have a reference to us, we must make it
262 // release that reference before we can be destroyed.
263 CancelScheduledCommit();
264
265 // Release our reference to the delegate, this reference will be keeping the
266 // history service alive.
267 delegate_.reset();
268 }
269
NotifyRenderProcessHostDestruction(const void * host)270 void HistoryBackend::NotifyRenderProcessHostDestruction(const void* host) {
271 tracker_.NotifyRenderProcessHostDestruction(host);
272 }
273
GetThumbnailFileName() const274 FilePath HistoryBackend::GetThumbnailFileName() const {
275 return history_dir_.Append(chrome::kThumbnailsFilename);
276 }
277
GetFaviconsFileName() const278 FilePath HistoryBackend::GetFaviconsFileName() const {
279 return history_dir_.Append(chrome::kFaviconsFilename);
280 }
281
GetArchivedFileName() const282 FilePath HistoryBackend::GetArchivedFileName() const {
283 return history_dir_.Append(chrome::kArchivedHistoryFilename);
284 }
285
GetLastSegmentID(VisitID from_visit)286 SegmentID HistoryBackend::GetLastSegmentID(VisitID from_visit) {
287 // Set is used to detect referrer loops. Should not happen, but can
288 // if the database is corrupt.
289 std::set<VisitID> visit_set;
290 VisitID visit_id = from_visit;
291 while (visit_id) {
292 VisitRow row;
293 if (!db_->GetRowForVisit(visit_id, &row))
294 return 0;
295 if (row.segment_id)
296 return row.segment_id; // Found a visit in this change with a segment.
297
298 // Check the referrer of this visit, if any.
299 visit_id = row.referring_visit;
300
301 if (visit_set.find(visit_id) != visit_set.end()) {
302 NOTREACHED() << "Loop in referer chain, giving up";
303 break;
304 }
305 visit_set.insert(visit_id);
306 }
307 return 0;
308 }
309
UpdateSegments(const GURL & url,VisitID from_visit,VisitID visit_id,PageTransition::Type transition_type,const Time ts)310 SegmentID HistoryBackend::UpdateSegments(const GURL& url,
311 VisitID from_visit,
312 VisitID visit_id,
313 PageTransition::Type transition_type,
314 const Time ts) {
315 if (!db_.get())
316 return 0;
317
318 // We only consider main frames.
319 if (!PageTransition::IsMainFrame(transition_type))
320 return 0;
321
322 SegmentID segment_id = 0;
323 PageTransition::Type t = PageTransition::StripQualifier(transition_type);
324
325 // Are we at the beginning of a new segment?
326 if (t == PageTransition::TYPED || t == PageTransition::AUTO_BOOKMARK) {
327 // If so, create or get the segment.
328 std::string segment_name = db_->ComputeSegmentName(url);
329 URLID url_id = db_->GetRowForURL(url, NULL);
330 if (!url_id)
331 return 0;
332
333 if (!(segment_id = db_->GetSegmentNamed(segment_name))) {
334 if (!(segment_id = db_->CreateSegment(url_id, segment_name))) {
335 NOTREACHED();
336 return 0;
337 }
338 } else {
339 // Note: if we update an existing segment, we update the url used to
340 // represent that segment in order to minimize stale most visited
341 // images.
342 db_->UpdateSegmentRepresentationURL(segment_id, url_id);
343 }
344 } else {
345 // Note: it is possible there is no segment ID set for this visit chain.
346 // This can happen if the initial navigation wasn't AUTO_BOOKMARK or
347 // TYPED. (For example GENERATED). In this case this visit doesn't count
348 // toward any segment.
349 if (!(segment_id = GetLastSegmentID(from_visit)))
350 return 0;
351 }
352
353 // Set the segment in the visit.
354 if (!db_->SetSegmentID(visit_id, segment_id)) {
355 NOTREACHED();
356 return 0;
357 }
358
359 // Finally, increase the counter for that segment / day.
360 if (!db_->IncreaseSegmentVisitCount(segment_id, ts, 1)) {
361 NOTREACHED();
362 return 0;
363 }
364 return segment_id;
365 }
366
AddPage(scoped_refptr<HistoryAddPageArgs> request)367 void HistoryBackend::AddPage(scoped_refptr<HistoryAddPageArgs> request) {
368 if (!db_.get())
369 return;
370
371 // Will be filled with the URL ID and the visit ID of the last addition.
372 std::pair<URLID, VisitID> last_ids(0, tracker_.GetLastVisit(
373 request->id_scope, request->page_id, request->referrer));
374
375 VisitID from_visit_id = last_ids.second;
376
377 // If a redirect chain is given, we expect the last item in that chain to be
378 // the final URL.
379 DCHECK(request->redirects.empty() ||
380 request->redirects.back() == request->url);
381
382 // Avoid duplicating times in the database, at least as long as pages are
383 // added in order. However, we don't want to disallow pages from recording
384 // times earlier than our last_recorded_time_, because someone might set
385 // their machine's clock back.
386 if (last_requested_time_ == request->time) {
387 last_recorded_time_ = last_recorded_time_ + TimeDelta::FromMicroseconds(1);
388 } else {
389 last_requested_time_ = request->time;
390 last_recorded_time_ = last_requested_time_;
391 }
392
393 // If the user is adding older history, we need to make sure our times
394 // are correct.
395 if (request->time < first_recorded_time_)
396 first_recorded_time_ = request->time;
397
398 PageTransition::Type transition =
399 PageTransition::StripQualifier(request->transition);
400 bool is_keyword_generated = (transition == PageTransition::KEYWORD_GENERATED);
401
402 if (request->redirects.size() <= 1) {
403 // The single entry is both a chain start and end.
404 PageTransition::Type t = request->transition |
405 PageTransition::CHAIN_START | PageTransition::CHAIN_END;
406
407 // No redirect case (one element means just the page itself).
408 last_ids = AddPageVisit(request->url, last_recorded_time_,
409 last_ids.second, t, request->visit_source);
410
411 // Update the segment for this visit. KEYWORD_GENERATED visits should not
412 // result in changing most visited, so we don't update segments (most
413 // visited db).
414 if (!is_keyword_generated) {
415 UpdateSegments(request->url, from_visit_id, last_ids.second, t,
416 last_recorded_time_);
417 }
418 } else {
419 // Redirect case. Add the redirect chain.
420
421 PageTransition::Type redirect_info = PageTransition::CHAIN_START;
422
423 if (request->redirects[0].SchemeIs(chrome::kAboutScheme)) {
424 // When the redirect source + referrer is "about" we skip it. This
425 // happens when a page opens a new frame/window to about:blank and then
426 // script sets the URL to somewhere else (used to hide the referrer). It
427 // would be nice to keep all these redirects properly but we don't ever
428 // see the initial about:blank load, so we don't know where the
429 // subsequent client redirect came from.
430 //
431 // In this case, we just don't bother hooking up the source of the
432 // redirects, so we remove it.
433 request->redirects.erase(request->redirects.begin());
434 } else if (request->transition & PageTransition::CLIENT_REDIRECT) {
435 redirect_info = PageTransition::CLIENT_REDIRECT;
436 // The first entry in the redirect chain initiated a client redirect.
437 // We don't add this to the database since the referrer is already
438 // there, so we skip over it but change the transition type of the first
439 // transition to client redirect.
440 //
441 // The referrer is invalid when restoring a session that features an
442 // https tab that redirects to a different host or to http. In this
443 // case we don't need to reconnect the new redirect with the existing
444 // chain.
445 if (request->referrer.is_valid()) {
446 DCHECK(request->referrer == request->redirects[0]);
447 request->redirects.erase(request->redirects.begin());
448
449 // If the navigation entry for this visit has replaced that for the
450 // first visit, remove the CHAIN_END marker from the first visit. This
451 // can be called a lot, for example, the page cycler, and most of the
452 // time we won't have changed anything.
453 VisitRow visit_row;
454 if (request->did_replace_entry &&
455 db_->GetRowForVisit(last_ids.second, &visit_row) &&
456 visit_row.transition | PageTransition::CHAIN_END) {
457 visit_row.transition &= ~PageTransition::CHAIN_END;
458 db_->UpdateVisitRow(visit_row);
459 }
460 }
461 }
462
463 for (size_t redirect_index = 0; redirect_index < request->redirects.size();
464 redirect_index++) {
465 PageTransition::Type t = transition | redirect_info;
466
467 // If this is the last transition, add a CHAIN_END marker
468 if (redirect_index == (request->redirects.size() - 1))
469 t = t | PageTransition::CHAIN_END;
470
471 // Record all redirect visits with the same timestamp. We don't display
472 // them anyway, and if we ever decide to, we can reconstruct their order
473 // from the redirect chain.
474 last_ids = AddPageVisit(request->redirects[redirect_index],
475 last_recorded_time_, last_ids.second,
476 t, request->visit_source);
477 if (t & PageTransition::CHAIN_START) {
478 // Update the segment for this visit.
479 UpdateSegments(request->redirects[redirect_index],
480 from_visit_id, last_ids.second, t, last_recorded_time_);
481 }
482
483 // Subsequent transitions in the redirect list must all be sever
484 // redirects.
485 redirect_info = PageTransition::SERVER_REDIRECT;
486 }
487
488 // Last, save this redirect chain for later so we can set titles & favicons
489 // on the redirected pages properly. It is indexed by the destination page.
490 recent_redirects_.Put(request->url, request->redirects);
491 }
492
493 // TODO(brettw) bug 1140015: Add an "add page" notification so the history
494 // views can keep in sync.
495
496 // Add the last visit to the tracker so we can get outgoing transitions.
497 // TODO(evanm): Due to http://b/1194536 we lose the referrers of a subframe
498 // navigation anyway, so last_visit_id is always zero for them. But adding
499 // them here confuses main frame history, so we skip them for now.
500 if (transition != PageTransition::AUTO_SUBFRAME &&
501 transition != PageTransition::MANUAL_SUBFRAME && !is_keyword_generated) {
502 tracker_.AddVisit(request->id_scope, request->page_id, request->url,
503 last_ids.second);
504 }
505
506 if (text_database_.get()) {
507 text_database_->AddPageURL(request->url, last_ids.first, last_ids.second,
508 last_recorded_time_);
509 }
510
511 ScheduleCommit();
512 }
513
InitImpl(const std::string & languages)514 void HistoryBackend::InitImpl(const std::string& languages) {
515 DCHECK(!db_.get()) << "Initializing HistoryBackend twice";
516 // In the rare case where the db fails to initialize a dialog may get shown
517 // the blocks the caller, yet allows other messages through. For this reason
518 // we only set db_ to the created database if creation is successful. That
519 // way other methods won't do anything as db_ is still NULL.
520
521 TimeTicks beginning_time = TimeTicks::Now();
522
523 // Compute the file names. Note that the index file can be removed when the
524 // text db manager is finished being hooked up.
525 FilePath history_name = history_dir_.Append(chrome::kHistoryFilename);
526 FilePath thumbnail_name = GetThumbnailFileName();
527 FilePath archived_name = GetArchivedFileName();
528 FilePath tmp_bookmarks_file = history_dir_.Append(
529 chrome::kHistoryBookmarksFileName);
530
531 // History database.
532 db_.reset(new HistoryDatabase());
533 sql::InitStatus status = db_->Init(history_name, tmp_bookmarks_file);
534 switch (status) {
535 case sql::INIT_OK:
536 break;
537 case sql::INIT_FAILURE:
538 // A NULL db_ will cause all calls on this object to notice this error
539 // and to not continue.
540 delegate_->NotifyProfileError(status);
541 db_.reset();
542 return;
543 default:
544 NOTREACHED();
545 }
546
547 // Fill the in-memory database and send it back to the history service on the
548 // main thread.
549 InMemoryHistoryBackend* mem_backend = new InMemoryHistoryBackend;
550 if (mem_backend->Init(history_name, history_dir_, db_.get(), languages))
551 delegate_->SetInMemoryBackend(mem_backend); // Takes ownership of pointer.
552 else
553 delete mem_backend; // Error case, run without the in-memory DB.
554 db_->BeginExclusiveMode(); // Must be after the mem backend read the data.
555
556 // Create the history publisher which needs to be passed on to the text and
557 // thumbnail databases for publishing history.
558 history_publisher_.reset(new HistoryPublisher());
559 if (!history_publisher_->Init()) {
560 // The init may fail when there are no indexers wanting our history.
561 // Hence no need to log the failure.
562 history_publisher_.reset();
563 }
564
565 // Full-text database. This has to be first so we can pass it to the
566 // HistoryDatabase for migration.
567 text_database_.reset(new TextDatabaseManager(history_dir_,
568 db_.get(), db_.get()));
569 if (!text_database_->Init(history_publisher_.get())) {
570 LOG(WARNING) << "Text database initialization failed, running without it.";
571 text_database_.reset();
572 }
573 if (db_->needs_version_17_migration()) {
574 // See needs_version_17_migration() decl for more. In this case, we want
575 // to erase all the text database files. This must be done after the text
576 // database manager has been initialized, since it knows about all the
577 // files it manages.
578 text_database_->DeleteAll();
579 }
580
581 // Thumbnail database.
582 thumbnail_db_.reset(new ThumbnailDatabase());
583 if (!db_->GetNeedsThumbnailMigration()) {
584 // No convertion needed - use new filename right away.
585 thumbnail_name = GetFaviconsFileName();
586 }
587 if (thumbnail_db_->Init(thumbnail_name,
588 history_publisher_.get(),
589 db_.get()) != sql::INIT_OK) {
590 // Unlike the main database, we don't error out when the database is too
591 // new because this error is much less severe. Generally, this shouldn't
592 // happen since the thumbnail and main datbase versions should be in sync.
593 // We'll just continue without thumbnails & favicons in this case or any
594 // other error.
595 LOG(WARNING) << "Could not initialize the thumbnail database.";
596 thumbnail_db_.reset();
597 }
598
599 if (db_->GetNeedsThumbnailMigration()) {
600 VLOG(1) << "Starting TopSites migration";
601 delegate_->StartTopSitesMigration();
602 }
603
604 // Archived database.
605 if (db_->needs_version_17_migration()) {
606 // See needs_version_17_migration() decl for more. In this case, we want
607 // to delete the archived database and need to do so before we try to
608 // open the file. We can ignore any error (maybe the file doesn't exist).
609 file_util::Delete(archived_name, false);
610 }
611 archived_db_.reset(new ArchivedDatabase());
612 if (!archived_db_->Init(archived_name)) {
613 LOG(WARNING) << "Could not initialize the archived database.";
614 archived_db_.reset();
615 }
616
617 // Tell the expiration module about all the nice databases we made. This must
618 // happen before db_->Init() is called since the callback ForceArchiveHistory
619 // may need to expire stuff.
620 //
621 // *sigh*, this can all be cleaned up when that migration code is removed.
622 // The main DB initialization should intuitively be first (not that it
623 // actually matters) and the expirer should be set last.
624 expirer_.SetDatabases(db_.get(), archived_db_.get(),
625 thumbnail_db_.get(), text_database_.get());
626
627 // Open the long-running transaction.
628 db_->BeginTransaction();
629 if (thumbnail_db_.get())
630 thumbnail_db_->BeginTransaction();
631 if (archived_db_.get())
632 archived_db_->BeginTransaction();
633 if (text_database_.get())
634 text_database_->BeginTransaction();
635
636 // Get the first item in our database.
637 db_->GetStartDate(&first_recorded_time_);
638
639 // Start expiring old stuff.
640 expirer_.StartArchivingOldStuff(TimeDelta::FromDays(kArchiveDaysThreshold));
641
642 HISTOGRAM_TIMES("History.InitTime",
643 TimeTicks::Now() - beginning_time);
644 }
645
AddPageVisit(const GURL & url,Time time,VisitID referring_visit,PageTransition::Type transition,VisitSource visit_source)646 std::pair<URLID, VisitID> HistoryBackend::AddPageVisit(
647 const GURL& url,
648 Time time,
649 VisitID referring_visit,
650 PageTransition::Type transition,
651 VisitSource visit_source) {
652 // Top-level frame navigations are visible, everything else is hidden
653 bool new_hidden = !PageTransition::IsMainFrame(transition);
654
655 // NOTE: This code must stay in sync with
656 // ExpireHistoryBackend::ExpireURLsForVisits().
657 // TODO(pkasting): http://b/1148304 We shouldn't be marking so many URLs as
658 // typed, which would eliminate the need for this code.
659 int typed_increment = 0;
660 PageTransition::Type transition_type =
661 PageTransition::StripQualifier(transition);
662 if ((transition_type == PageTransition::TYPED &&
663 !PageTransition::IsRedirect(transition)) ||
664 transition_type == PageTransition::KEYWORD_GENERATED)
665 typed_increment = 1;
666
667 // See if this URL is already in the DB.
668 URLRow url_info(url);
669 URLID url_id = db_->GetRowForURL(url, &url_info);
670 if (url_id) {
671 // Update of an existing row.
672 if (PageTransition::StripQualifier(transition) != PageTransition::RELOAD)
673 url_info.set_visit_count(url_info.visit_count() + 1);
674 if (typed_increment)
675 url_info.set_typed_count(url_info.typed_count() + typed_increment);
676 url_info.set_last_visit(time);
677
678 // Only allow un-hiding of pages, never hiding.
679 if (!new_hidden)
680 url_info.set_hidden(false);
681
682 db_->UpdateURLRow(url_id, url_info);
683 } else {
684 // Addition of a new row.
685 url_info.set_visit_count(1);
686 url_info.set_typed_count(typed_increment);
687 url_info.set_last_visit(time);
688 url_info.set_hidden(new_hidden);
689
690 url_id = db_->AddURL(url_info);
691 if (!url_id) {
692 NOTREACHED() << "Adding URL failed.";
693 return std::make_pair(0, 0);
694 }
695 url_info.id_ = url_id;
696
697 // We don't actually add the URL to the full text index at this point. It
698 // might be nice to do this so that even if we get no title or body, the
699 // user can search for URL components and get the page.
700 //
701 // However, in most cases, we'll get at least a title and usually contents,
702 // and this add will be redundant, slowing everything down. As a result,
703 // we ignore this edge case.
704 }
705
706 // Add the visit with the time to the database.
707 VisitRow visit_info(url_id, time, referring_visit, transition, 0);
708 VisitID visit_id = db_->AddVisit(&visit_info, visit_source);
709
710 if (visit_info.visit_time < first_recorded_time_)
711 first_recorded_time_ = visit_info.visit_time;
712
713 // Broadcast a notification of the visit.
714 if (visit_id) {
715 URLVisitedDetails* details = new URLVisitedDetails;
716 details->transition = transition;
717 details->row = url_info;
718 // TODO(meelapshah) Disabled due to potential PageCycler regression.
719 // Re-enable this.
720 // GetMostRecentRedirectsTo(url, &details->redirects);
721 BroadcastNotifications(NotificationType::HISTORY_URL_VISITED, details);
722 } else {
723 VLOG(0) << "Failed to build visit insert statement: "
724 << "url_id = " << url_id;
725 }
726
727 return std::make_pair(url_id, visit_id);
728 }
729
AddPagesWithDetails(const std::vector<URLRow> & urls,VisitSource visit_source)730 void HistoryBackend::AddPagesWithDetails(const std::vector<URLRow>& urls,
731 VisitSource visit_source) {
732 if (!db_.get())
733 return;
734
735 scoped_ptr<URLsModifiedDetails> modified(new URLsModifiedDetails);
736 for (std::vector<URLRow>::const_iterator i = urls.begin();
737 i != urls.end(); ++i) {
738 DCHECK(!i->last_visit().is_null());
739
740 // We will add to either the archived database or the main one depending on
741 // the date of the added visit.
742 URLDatabase* url_database;
743 VisitDatabase* visit_database;
744 if (i->last_visit() < expirer_.GetCurrentArchiveTime()) {
745 if (!archived_db_.get())
746 return; // No archived database to save it to, just forget this.
747 url_database = archived_db_.get();
748 visit_database = archived_db_.get();
749 } else {
750 url_database = db_.get();
751 visit_database = db_.get();
752 }
753
754 URLRow existing_url;
755 URLID url_id = url_database->GetRowForURL(i->url(), &existing_url);
756 if (!url_id) {
757 // Add the page if it doesn't exist.
758 url_id = url_database->AddURL(*i);
759 if (!url_id) {
760 NOTREACHED() << "Could not add row to DB";
761 return;
762 }
763
764 if (i->typed_count() > 0)
765 modified->changed_urls.push_back(*i);
766 }
767
768 // Add the page to the full text index. This function is also used for
769 // importing. Even though we don't have page contents, we can at least
770 // add the title and URL to the index so they can be searched. We don't
771 // bother to delete any already-existing FTS entries for the URL, since
772 // this is normally called on import.
773 //
774 // If you ever import *after* first run (selecting import from the menu),
775 // then these additional entries will "shadow" the originals when querying
776 // for the most recent match only, and the user won't get snippets. This is
777 // a very minor issue, and fixing it will make import slower, so we don't
778 // bother.
779 bool has_indexed = false;
780 if (text_database_.get()) {
781 // We do not have to make it update the visit database, below, we will
782 // create the visit entry with the indexed flag set.
783 has_indexed = text_database_->AddPageData(i->url(), url_id, 0,
784 i->last_visit(),
785 i->title(), string16());
786 }
787
788 // Make up a visit to correspond to that page.
789 VisitRow visit_info(url_id, i->last_visit(), 0,
790 PageTransition::LINK | PageTransition::CHAIN_START |
791 PageTransition::CHAIN_END, 0);
792 visit_info.is_indexed = has_indexed;
793 if (!visit_database->AddVisit(&visit_info, visit_source)) {
794 NOTREACHED() << "Adding visit failed.";
795 return;
796 }
797
798 if (visit_info.visit_time < first_recorded_time_)
799 first_recorded_time_ = visit_info.visit_time;
800 }
801
802 // Broadcast a notification for typed URLs that have been modified. This
803 // will be picked up by the in-memory URL database on the main thread.
804 //
805 // TODO(brettw) bug 1140015: Add an "add page" notification so the history
806 // views can keep in sync.
807 BroadcastNotifications(NotificationType::HISTORY_TYPED_URLS_MODIFIED,
808 modified.release());
809
810 ScheduleCommit();
811 }
812
SetPageTitle(const GURL & url,const string16 & title)813 void HistoryBackend::SetPageTitle(const GURL& url,
814 const string16& title) {
815 if (!db_.get())
816 return;
817
818 // Search for recent redirects which should get the same title. We make a
819 // dummy list containing the exact URL visited if there are no redirects so
820 // the processing below can be the same.
821 history::RedirectList dummy_list;
822 history::RedirectList* redirects;
823 RedirectCache::iterator iter = recent_redirects_.Get(url);
824 if (iter != recent_redirects_.end()) {
825 redirects = &iter->second;
826
827 // This redirect chain should have the destination URL as the last item.
828 DCHECK(!redirects->empty());
829 DCHECK(redirects->back() == url);
830 } else {
831 // No redirect chain stored, make up one containing the URL we want so we
832 // can use the same logic below.
833 dummy_list.push_back(url);
834 redirects = &dummy_list;
835 }
836
837 bool typed_url_changed = false;
838 std::vector<URLRow> changed_urls;
839 for (size_t i = 0; i < redirects->size(); i++) {
840 URLRow row;
841 URLID row_id = db_->GetRowForURL(redirects->at(i), &row);
842 if (row_id && row.title() != title) {
843 row.set_title(title);
844 db_->UpdateURLRow(row_id, row);
845 changed_urls.push_back(row);
846 if (row.typed_count() > 0)
847 typed_url_changed = true;
848 }
849 }
850
851 // Broadcast notifications for typed URLs that have changed. This will
852 // update the in-memory database.
853 //
854 // TODO(brettw) bug 1140020: Broadcast for all changes (not just typed),
855 // in which case some logic can be removed.
856 if (typed_url_changed) {
857 URLsModifiedDetails* modified =
858 new URLsModifiedDetails;
859 for (size_t i = 0; i < changed_urls.size(); i++) {
860 if (changed_urls[i].typed_count() > 0)
861 modified->changed_urls.push_back(changed_urls[i]);
862 }
863 BroadcastNotifications(NotificationType::HISTORY_TYPED_URLS_MODIFIED,
864 modified);
865 }
866
867 // Update the full text index.
868 if (text_database_.get())
869 text_database_->AddPageTitle(url, title);
870
871 // Only bother committing if things changed.
872 if (!changed_urls.empty())
873 ScheduleCommit();
874 }
875
AddPageNoVisitForBookmark(const GURL & url)876 void HistoryBackend::AddPageNoVisitForBookmark(const GURL& url) {
877 if (!db_.get())
878 return;
879
880 URLRow url_info(url);
881 URLID url_id = db_->GetRowForURL(url, &url_info);
882 if (url_id) {
883 // URL is already known, nothing to do.
884 return;
885 }
886 url_info.set_last_visit(Time::Now());
887 // Mark the page hidden. If the user types it in, it'll unhide.
888 url_info.set_hidden(true);
889
890 db_->AddURL(url_info);
891 }
892
IterateURLs(HistoryService::URLEnumerator * iterator)893 void HistoryBackend::IterateURLs(HistoryService::URLEnumerator* iterator) {
894 if (db_.get()) {
895 HistoryDatabase::URLEnumerator e;
896 if (db_->InitURLEnumeratorForEverything(&e)) {
897 URLRow info;
898 while (e.GetNextURL(&info)) {
899 iterator->OnURL(info.url());
900 }
901 iterator->OnComplete(true); // Success.
902 return;
903 }
904 }
905 iterator->OnComplete(false); // Failure.
906 }
907
GetAllTypedURLs(std::vector<history::URLRow> * urls)908 bool HistoryBackend::GetAllTypedURLs(std::vector<history::URLRow>* urls) {
909 if (db_.get())
910 return db_->GetAllTypedUrls(urls);
911 return false;
912 }
913
GetVisitsForURL(URLID id,VisitVector * visits)914 bool HistoryBackend::GetVisitsForURL(URLID id, VisitVector* visits) {
915 if (db_.get())
916 return db_->GetVisitsForURL(id, visits);
917 return false;
918 }
919
UpdateURL(URLID id,const history::URLRow & url)920 bool HistoryBackend::UpdateURL(URLID id, const history::URLRow& url) {
921 if (db_.get())
922 return db_->UpdateURLRow(id, url);
923 return false;
924 }
925
AddVisits(const GURL & url,const std::vector<base::Time> & visits,VisitSource visit_source)926 bool HistoryBackend::AddVisits(const GURL& url,
927 const std::vector<base::Time>& visits,
928 VisitSource visit_source) {
929 if (db_.get()) {
930 for (std::vector<base::Time>::const_iterator visit = visits.begin();
931 visit != visits.end(); ++visit) {
932 if (!AddPageVisit(url, *visit, 0, 0, visit_source).first) {
933 return false;
934 }
935 }
936 ScheduleCommit();
937 return true;
938 }
939 return false;
940 }
941
RemoveVisits(const VisitVector & visits)942 bool HistoryBackend::RemoveVisits(const VisitVector& visits) {
943 if (db_.get()) {
944 std::map<URLID, int> url_visits_removed;
945 for (VisitVector::const_iterator visit = visits.begin();
946 visit != visits.end(); ++visit) {
947 db_->DeleteVisit(*visit);
948 std::map<URLID, int>::iterator visit_count =
949 url_visits_removed.find(visit->url_id);
950 if (visit_count == url_visits_removed.end()) {
951 url_visits_removed[visit->url_id] = 1;
952 } else {
953 ++visit_count->second;
954 }
955 }
956 for (std::map<URLID, int>::iterator count = url_visits_removed.begin();
957 count != url_visits_removed.end(); ++count) {
958 history::URLRow url_row;
959 if (!db_->GetURLRow(count->first, &url_row)) {
960 return false;
961 }
962 DCHECK(count->second <= url_row.visit_count());
963 url_row.set_visit_count(url_row.visit_count() - count->second);
964 if (!db_->UpdateURLRow(url_row.id(), url_row)) {
965 return false;
966 }
967 }
968 ScheduleCommit();
969 return true;
970 }
971 return false;
972 }
973
GetURL(const GURL & url,history::URLRow * url_row)974 bool HistoryBackend::GetURL(const GURL& url, history::URLRow* url_row) {
975 if (db_.get())
976 return db_->GetRowForURL(url, url_row) != 0;
977 return false;
978 }
979
QueryURL(scoped_refptr<QueryURLRequest> request,const GURL & url,bool want_visits)980 void HistoryBackend::QueryURL(scoped_refptr<QueryURLRequest> request,
981 const GURL& url,
982 bool want_visits) {
983 if (request->canceled())
984 return;
985
986 bool success = false;
987 URLRow* row = &request->value.a;
988 VisitVector* visits = &request->value.b;
989 if (db_.get()) {
990 if (db_->GetRowForURL(url, row)) {
991 // Have a row.
992 success = true;
993
994 // Optionally query the visits.
995 if (want_visits)
996 db_->GetVisitsForURL(row->id(), visits);
997 }
998 }
999 request->ForwardResult(QueryURLRequest::TupleType(request->handle(), success,
1000 row, visits));
1001 }
1002
1003 // Segment usage ---------------------------------------------------------------
1004
DeleteOldSegmentData()1005 void HistoryBackend::DeleteOldSegmentData() {
1006 if (db_.get())
1007 db_->DeleteSegmentData(Time::Now() -
1008 TimeDelta::FromDays(kSegmentDataRetention));
1009 }
1010
SetSegmentPresentationIndex(SegmentID segment_id,int index)1011 void HistoryBackend::SetSegmentPresentationIndex(SegmentID segment_id,
1012 int index) {
1013 if (db_.get())
1014 db_->SetSegmentPresentationIndex(segment_id, index);
1015 }
1016
QuerySegmentUsage(scoped_refptr<QuerySegmentUsageRequest> request,const Time from_time,int max_result_count)1017 void HistoryBackend::QuerySegmentUsage(
1018 scoped_refptr<QuerySegmentUsageRequest> request,
1019 const Time from_time,
1020 int max_result_count) {
1021 if (request->canceled())
1022 return;
1023
1024 if (db_.get()) {
1025 db_->QuerySegmentUsage(from_time, max_result_count, &request->value.get());
1026
1027 // If this is the first time we query segments, invoke
1028 // DeleteOldSegmentData asynchronously. We do this to cleanup old
1029 // entries.
1030 if (!segment_queried_) {
1031 segment_queried_ = true;
1032 MessageLoop::current()->PostTask(FROM_HERE,
1033 NewRunnableMethod(this, &HistoryBackend::DeleteOldSegmentData));
1034 }
1035 }
1036 request->ForwardResult(
1037 QuerySegmentUsageRequest::TupleType(request->handle(),
1038 &request->value.get()));
1039 }
1040
1041 // Keyword visits --------------------------------------------------------------
1042
SetKeywordSearchTermsForURL(const GURL & url,TemplateURLID keyword_id,const string16 & term)1043 void HistoryBackend::SetKeywordSearchTermsForURL(const GURL& url,
1044 TemplateURLID keyword_id,
1045 const string16& term) {
1046 if (!db_.get())
1047 return;
1048
1049 // Get the ID for this URL.
1050 URLRow url_row;
1051 if (!db_->GetRowForURL(url, &url_row)) {
1052 // There is a small possibility the url was deleted before the keyword
1053 // was added. Ignore the request.
1054 return;
1055 }
1056
1057 db_->SetKeywordSearchTermsForURL(url_row.id(), keyword_id, term);
1058
1059 // details is deleted by BroadcastNotifications.
1060 KeywordSearchTermDetails* details = new KeywordSearchTermDetails;
1061 details->url = url;
1062 details->keyword_id = keyword_id;
1063 details->term = term;
1064 BroadcastNotifications(NotificationType::HISTORY_KEYWORD_SEARCH_TERM_UPDATED,
1065 details);
1066 ScheduleCommit();
1067 }
1068
DeleteAllSearchTermsForKeyword(TemplateURLID keyword_id)1069 void HistoryBackend::DeleteAllSearchTermsForKeyword(
1070 TemplateURLID keyword_id) {
1071 if (!db_.get())
1072 return;
1073
1074 db_->DeleteAllSearchTermsForKeyword(keyword_id);
1075 // TODO(sky): bug 1168470. Need to move from archive dbs too.
1076 ScheduleCommit();
1077 }
1078
GetMostRecentKeywordSearchTerms(scoped_refptr<GetMostRecentKeywordSearchTermsRequest> request,TemplateURLID keyword_id,const string16 & prefix,int max_count)1079 void HistoryBackend::GetMostRecentKeywordSearchTerms(
1080 scoped_refptr<GetMostRecentKeywordSearchTermsRequest> request,
1081 TemplateURLID keyword_id,
1082 const string16& prefix,
1083 int max_count) {
1084 if (request->canceled())
1085 return;
1086
1087 if (db_.get()) {
1088 db_->GetMostRecentKeywordSearchTerms(keyword_id, prefix, max_count,
1089 &(request->value));
1090 }
1091 request->ForwardResult(
1092 GetMostRecentKeywordSearchTermsRequest::TupleType(request->handle(),
1093 &request->value));
1094 }
1095
1096 // Downloads -------------------------------------------------------------------
1097
1098 // Get all the download entries from the database.
QueryDownloads(scoped_refptr<DownloadQueryRequest> request)1099 void HistoryBackend::QueryDownloads(
1100 scoped_refptr<DownloadQueryRequest> request) {
1101 if (request->canceled())
1102 return;
1103 if (db_.get())
1104 db_->QueryDownloads(&request->value);
1105 request->ForwardResult(DownloadQueryRequest::TupleType(&request->value));
1106 }
1107
1108 // Clean up entries that has been corrupted (because of the crash, for example).
CleanUpInProgressEntries()1109 void HistoryBackend::CleanUpInProgressEntries() {
1110 if (db_.get()) {
1111 // If some "in progress" entries were not updated when Chrome exited, they
1112 // need to be cleaned up.
1113 db_->CleanUpInProgressEntries();
1114 }
1115 }
1116
1117 // Update a particular download entry.
UpdateDownload(int64 received_bytes,int32 state,int64 db_handle)1118 void HistoryBackend::UpdateDownload(int64 received_bytes,
1119 int32 state,
1120 int64 db_handle) {
1121 if (db_.get())
1122 db_->UpdateDownload(received_bytes, state, db_handle);
1123 }
1124
1125 // Update the path of a particular download entry.
UpdateDownloadPath(const FilePath & path,int64 db_handle)1126 void HistoryBackend::UpdateDownloadPath(const FilePath& path,
1127 int64 db_handle) {
1128 if (db_.get())
1129 db_->UpdateDownloadPath(path, db_handle);
1130 }
1131
1132 // Create a new download entry and pass back the db_handle to it.
CreateDownload(scoped_refptr<DownloadCreateRequest> request,const DownloadCreateInfo & create_info)1133 void HistoryBackend::CreateDownload(
1134 scoped_refptr<DownloadCreateRequest> request,
1135 const DownloadCreateInfo& create_info) {
1136 int64 db_handle = 0;
1137 if (!request->canceled()) {
1138 if (db_.get())
1139 db_handle = db_->CreateDownload(create_info);
1140 request->ForwardResult(DownloadCreateRequest::TupleType(create_info,
1141 db_handle));
1142 }
1143 }
1144
RemoveDownload(int64 db_handle)1145 void HistoryBackend::RemoveDownload(int64 db_handle) {
1146 if (db_.get())
1147 db_->RemoveDownload(db_handle);
1148 }
1149
RemoveDownloadsBetween(const Time remove_begin,const Time remove_end)1150 void HistoryBackend::RemoveDownloadsBetween(const Time remove_begin,
1151 const Time remove_end) {
1152 if (db_.get())
1153 db_->RemoveDownloadsBetween(remove_begin, remove_end);
1154 }
1155
QueryHistory(scoped_refptr<QueryHistoryRequest> request,const string16 & text_query,const QueryOptions & options)1156 void HistoryBackend::QueryHistory(scoped_refptr<QueryHistoryRequest> request,
1157 const string16& text_query,
1158 const QueryOptions& options) {
1159 if (request->canceled())
1160 return;
1161
1162 TimeTicks beginning_time = TimeTicks::Now();
1163
1164 if (db_.get()) {
1165 if (text_query.empty()) {
1166 // Basic history query for the main database.
1167 QueryHistoryBasic(db_.get(), db_.get(), options, &request->value);
1168
1169 // Now query the archived database. This is a bit tricky because we don't
1170 // want to query it if the queried time range isn't going to find anything
1171 // in it.
1172 // TODO(brettw) bug 1171036: do blimpie querying for the archived database
1173 // as well.
1174 // if (archived_db_.get() &&
1175 // expirer_.GetCurrentArchiveTime() - TimeDelta::FromDays(7)) {
1176 } else {
1177 // Full text history query.
1178 QueryHistoryFTS(text_query, options, &request->value);
1179 }
1180 }
1181
1182 request->ForwardResult(QueryHistoryRequest::TupleType(request->handle(),
1183 &request->value));
1184
1185 UMA_HISTOGRAM_TIMES("History.QueryHistory",
1186 TimeTicks::Now() - beginning_time);
1187 }
1188
1189 // Basic time-based querying of history.
QueryHistoryBasic(URLDatabase * url_db,VisitDatabase * visit_db,const QueryOptions & options,QueryResults * result)1190 void HistoryBackend::QueryHistoryBasic(URLDatabase* url_db,
1191 VisitDatabase* visit_db,
1192 const QueryOptions& options,
1193 QueryResults* result) {
1194 // First get all visits.
1195 VisitVector visits;
1196 visit_db->GetVisibleVisitsInRange(options.begin_time, options.end_time,
1197 options.max_count, &visits);
1198 DCHECK(options.max_count == 0 ||
1199 static_cast<int>(visits.size()) <= options.max_count);
1200
1201 // Now add them and the URL rows to the results.
1202 URLResult url_result;
1203 for (size_t i = 0; i < visits.size(); i++) {
1204 const VisitRow visit = visits[i];
1205
1206 // Add a result row for this visit, get the URL info from the DB.
1207 if (!url_db->GetURLRow(visit.url_id, &url_result)) {
1208 VLOG(0) << "Failed to get id " << visit.url_id
1209 << " from history.urls.";
1210 continue; // DB out of sync and URL doesn't exist, try to recover.
1211 }
1212
1213 if (!url_result.url().is_valid()) {
1214 VLOG(0) << "Got invalid URL from history.urls with id "
1215 << visit.url_id << ": "
1216 << url_result.url().possibly_invalid_spec();
1217 continue; // Don't report invalid URLs in case of corruption.
1218 }
1219
1220 // The archived database may be out of sync with respect to starring,
1221 // titles, last visit date, etc. Therefore, we query the main DB if the
1222 // current URL database is not the main one.
1223 if (url_db == db_.get()) {
1224 // Currently querying the archived DB, update with the main database to
1225 // catch any interesting stuff. This will update it if it exists in the
1226 // main DB, and do nothing otherwise.
1227 db_->GetRowForURL(url_result.url(), &url_result);
1228 }
1229
1230 url_result.set_visit_time(visit.visit_time);
1231
1232 // We don't set any of the query-specific parts of the URLResult, since
1233 // snippets and stuff don't apply to basic querying.
1234 result->AppendURLBySwapping(&url_result);
1235 }
1236
1237 if (options.begin_time <= first_recorded_time_)
1238 result->set_reached_beginning(true);
1239 }
1240
QueryHistoryFTS(const string16 & text_query,const QueryOptions & options,QueryResults * result)1241 void HistoryBackend::QueryHistoryFTS(const string16& text_query,
1242 const QueryOptions& options,
1243 QueryResults* result) {
1244 if (!text_database_.get())
1245 return;
1246
1247 // Full text query, first get all the FTS results in the time range.
1248 std::vector<TextDatabase::Match> fts_matches;
1249 Time first_time_searched;
1250 text_database_->GetTextMatches(text_query, options,
1251 &fts_matches, &first_time_searched);
1252
1253 URLQuerier querier(db_.get(), archived_db_.get(), true);
1254
1255 // Now get the row and visit information for each one.
1256 URLResult url_result; // Declare outside loop to prevent re-construction.
1257 for (size_t i = 0; i < fts_matches.size(); i++) {
1258 if (options.max_count != 0 &&
1259 static_cast<int>(result->size()) >= options.max_count)
1260 break; // Got too many items.
1261
1262 // Get the URL, querying the main and archived databases as necessary. If
1263 // this is not found, the history and full text search databases are out
1264 // of sync and we give up with this result.
1265 if (!querier.GetRowForURL(fts_matches[i].url, &url_result))
1266 continue;
1267
1268 if (!url_result.url().is_valid())
1269 continue; // Don't report invalid URLs in case of corruption.
1270
1271 // Copy over the FTS stuff that the URLDatabase doesn't know about.
1272 // We do this with swap() to avoid copying, since we know we don't
1273 // need the original any more. Note that we override the title with the
1274 // one from FTS, since that will match the title_match_positions (the
1275 // FTS title and the history DB title may differ).
1276 url_result.set_title(fts_matches[i].title);
1277 url_result.title_match_positions_.swap(
1278 fts_matches[i].title_match_positions);
1279 url_result.snippet_.Swap(&fts_matches[i].snippet);
1280
1281 // The visit time also comes from the full text search database. Since it
1282 // has the time, we can avoid an extra query of the visits table.
1283 url_result.set_visit_time(fts_matches[i].time);
1284
1285 // Add it to the vector, this will clear our |url_row| object as a
1286 // result of the swap.
1287 result->AppendURLBySwapping(&url_result);
1288 }
1289
1290 if (options.begin_time <= first_recorded_time_)
1291 result->set_reached_beginning(true);
1292 }
1293
1294 // Frontend to GetMostRecentRedirectsFrom from the history thread.
QueryRedirectsFrom(scoped_refptr<QueryRedirectsRequest> request,const GURL & url)1295 void HistoryBackend::QueryRedirectsFrom(
1296 scoped_refptr<QueryRedirectsRequest> request,
1297 const GURL& url) {
1298 if (request->canceled())
1299 return;
1300 bool success = GetMostRecentRedirectsFrom(url, &request->value);
1301 request->ForwardResult(QueryRedirectsRequest::TupleType(
1302 request->handle(), url, success, &request->value));
1303 }
1304
QueryRedirectsTo(scoped_refptr<QueryRedirectsRequest> request,const GURL & url)1305 void HistoryBackend::QueryRedirectsTo(
1306 scoped_refptr<QueryRedirectsRequest> request,
1307 const GURL& url) {
1308 if (request->canceled())
1309 return;
1310 bool success = GetMostRecentRedirectsTo(url, &request->value);
1311 request->ForwardResult(QueryRedirectsRequest::TupleType(
1312 request->handle(), url, success, &request->value));
1313 }
1314
GetVisitCountToHost(scoped_refptr<GetVisitCountToHostRequest> request,const GURL & url)1315 void HistoryBackend::GetVisitCountToHost(
1316 scoped_refptr<GetVisitCountToHostRequest> request,
1317 const GURL& url) {
1318 if (request->canceled())
1319 return;
1320 int count = 0;
1321 Time first_visit;
1322 const bool success = (db_.get() && db_->GetVisitCountToHost(url, &count,
1323 &first_visit));
1324 request->ForwardResult(GetVisitCountToHostRequest::TupleType(
1325 request->handle(), success, count, first_visit));
1326 }
1327
QueryTopURLsAndRedirects(scoped_refptr<QueryTopURLsAndRedirectsRequest> request,int result_count)1328 void HistoryBackend::QueryTopURLsAndRedirects(
1329 scoped_refptr<QueryTopURLsAndRedirectsRequest> request,
1330 int result_count) {
1331 if (request->canceled())
1332 return;
1333
1334 if (!db_.get()) {
1335 request->ForwardResult(QueryTopURLsAndRedirectsRequest::TupleType(
1336 request->handle(), false, NULL, NULL));
1337 return;
1338 }
1339
1340 std::vector<GURL>* top_urls = &request->value.a;
1341 history::RedirectMap* redirects = &request->value.b;
1342
1343 ScopedVector<PageUsageData> data;
1344 db_->QuerySegmentUsage(base::Time::Now() - base::TimeDelta::FromDays(90),
1345 result_count, &data.get());
1346
1347 for (size_t i = 0; i < data.size(); ++i) {
1348 top_urls->push_back(data[i]->GetURL());
1349 RefCountedVector<GURL>* list = new RefCountedVector<GURL>;
1350 GetMostRecentRedirectsFrom(top_urls->back(), &list->data);
1351 (*redirects)[top_urls->back()] = list;
1352 }
1353
1354 request->ForwardResult(QueryTopURLsAndRedirectsRequest::TupleType(
1355 request->handle(), true, top_urls, redirects));
1356 }
1357
1358 // Will replace QueryTopURLsAndRedirectsRequest.
QueryMostVisitedURLs(scoped_refptr<QueryMostVisitedURLsRequest> request,int result_count,int days_back)1359 void HistoryBackend::QueryMostVisitedURLs(
1360 scoped_refptr<QueryMostVisitedURLsRequest> request,
1361 int result_count,
1362 int days_back) {
1363 if (request->canceled())
1364 return;
1365
1366 if (!db_.get()) {
1367 // No History Database - return an empty list.
1368 request->ForwardResult(QueryMostVisitedURLsRequest::TupleType(
1369 request->handle(), MostVisitedURLList()));
1370 return;
1371 }
1372
1373 MostVisitedURLList* result = &request->value;
1374 QueryMostVisitedURLsImpl(result_count, days_back, result);
1375 request->ForwardResult(QueryMostVisitedURLsRequest::TupleType(
1376 request->handle(), *result));
1377 }
1378
QueryMostVisitedURLsImpl(int result_count,int days_back,MostVisitedURLList * result)1379 void HistoryBackend::QueryMostVisitedURLsImpl(int result_count,
1380 int days_back,
1381 MostVisitedURLList* result) {
1382 if (!db_.get())
1383 return;
1384
1385 ScopedVector<PageUsageData> data;
1386 db_->QuerySegmentUsage(base::Time::Now() -
1387 base::TimeDelta::FromDays(days_back),
1388 result_count, &data.get());
1389
1390 for (size_t i = 0; i < data.size(); ++i) {
1391 PageUsageData* current_data = data[i];
1392 RedirectList redirects;
1393 GetMostRecentRedirectsFrom(current_data->GetURL(), &redirects);
1394 MostVisitedURL url = MakeMostVisitedURL(*current_data, redirects);
1395 result->push_back(url);
1396 }
1397 }
1398
GetRedirectsFromSpecificVisit(VisitID cur_visit,history::RedirectList * redirects)1399 void HistoryBackend::GetRedirectsFromSpecificVisit(
1400 VisitID cur_visit, history::RedirectList* redirects) {
1401 // Follow any redirects from the given visit and add them to the list.
1402 // It *should* be impossible to get a circular chain here, but we check
1403 // just in case to avoid infinite loops.
1404 GURL cur_url;
1405 std::set<VisitID> visit_set;
1406 visit_set.insert(cur_visit);
1407 while (db_->GetRedirectFromVisit(cur_visit, &cur_visit, &cur_url)) {
1408 if (visit_set.find(cur_visit) != visit_set.end()) {
1409 NOTREACHED() << "Loop in visit chain, giving up";
1410 return;
1411 }
1412 visit_set.insert(cur_visit);
1413 redirects->push_back(cur_url);
1414 }
1415 }
1416
GetRedirectsToSpecificVisit(VisitID cur_visit,history::RedirectList * redirects)1417 void HistoryBackend::GetRedirectsToSpecificVisit(
1418 VisitID cur_visit,
1419 history::RedirectList* redirects) {
1420 // Follow redirects going to cur_visit. These are added to |redirects| in
1421 // the order they are found. If a redirect chain looks like A -> B -> C and
1422 // |cur_visit| = C, redirects will be {B, A} in that order.
1423 if (!db_.get())
1424 return;
1425
1426 GURL cur_url;
1427 std::set<VisitID> visit_set;
1428 visit_set.insert(cur_visit);
1429 while (db_->GetRedirectToVisit(cur_visit, &cur_visit, &cur_url)) {
1430 if (visit_set.find(cur_visit) != visit_set.end()) {
1431 NOTREACHED() << "Loop in visit chain, giving up";
1432 return;
1433 }
1434 visit_set.insert(cur_visit);
1435 redirects->push_back(cur_url);
1436 }
1437 }
1438
GetMostRecentRedirectsFrom(const GURL & from_url,history::RedirectList * redirects)1439 bool HistoryBackend::GetMostRecentRedirectsFrom(
1440 const GURL& from_url,
1441 history::RedirectList* redirects) {
1442 redirects->clear();
1443 if (!db_.get())
1444 return false;
1445
1446 URLID from_url_id = db_->GetRowForURL(from_url, NULL);
1447 VisitID cur_visit = db_->GetMostRecentVisitForURL(from_url_id, NULL);
1448 if (!cur_visit)
1449 return false; // No visits for URL.
1450
1451 GetRedirectsFromSpecificVisit(cur_visit, redirects);
1452 return true;
1453 }
1454
GetMostRecentRedirectsTo(const GURL & to_url,history::RedirectList * redirects)1455 bool HistoryBackend::GetMostRecentRedirectsTo(
1456 const GURL& to_url,
1457 history::RedirectList* redirects) {
1458 redirects->clear();
1459 if (!db_.get())
1460 return false;
1461
1462 URLID to_url_id = db_->GetRowForURL(to_url, NULL);
1463 VisitID cur_visit = db_->GetMostRecentVisitForURL(to_url_id, NULL);
1464 if (!cur_visit)
1465 return false; // No visits for URL.
1466
1467 GetRedirectsToSpecificVisit(cur_visit, redirects);
1468 return true;
1469 }
1470
ScheduleAutocomplete(HistoryURLProvider * provider,HistoryURLProviderParams * params)1471 void HistoryBackend::ScheduleAutocomplete(HistoryURLProvider* provider,
1472 HistoryURLProviderParams* params) {
1473 // ExecuteWithDB should handle the NULL database case.
1474 provider->ExecuteWithDB(this, db_.get(), params);
1475 }
1476
SetPageContents(const GURL & url,const string16 & contents)1477 void HistoryBackend::SetPageContents(const GURL& url,
1478 const string16& contents) {
1479 // This is histogrammed in the text database manager.
1480 if (!text_database_.get())
1481 return;
1482 text_database_->AddPageContents(url, contents);
1483 }
1484
SetPageThumbnail(const GURL & url,const SkBitmap & thumbnail,const ThumbnailScore & score)1485 void HistoryBackend::SetPageThumbnail(
1486 const GURL& url,
1487 const SkBitmap& thumbnail,
1488 const ThumbnailScore& score) {
1489 if (!db_.get() || !thumbnail_db_.get())
1490 return;
1491
1492 URLRow url_row;
1493 URLID url_id = db_->GetRowForURL(url, &url_row);
1494 if (url_id) {
1495 thumbnail_db_->SetPageThumbnail(url, url_id, thumbnail, score,
1496 url_row.last_visit());
1497 }
1498
1499 ScheduleCommit();
1500 }
1501
GetPageThumbnail(scoped_refptr<GetPageThumbnailRequest> request,const GURL & page_url)1502 void HistoryBackend::GetPageThumbnail(
1503 scoped_refptr<GetPageThumbnailRequest> request,
1504 const GURL& page_url) {
1505 if (request->canceled())
1506 return;
1507
1508 scoped_refptr<RefCountedBytes> data;
1509 GetPageThumbnailDirectly(page_url, &data);
1510
1511 request->ForwardResult(GetPageThumbnailRequest::TupleType(
1512 request->handle(), data));
1513 }
1514
GetPageThumbnailDirectly(const GURL & page_url,scoped_refptr<RefCountedBytes> * data)1515 void HistoryBackend::GetPageThumbnailDirectly(
1516 const GURL& page_url,
1517 scoped_refptr<RefCountedBytes>* data) {
1518 if (thumbnail_db_.get()) {
1519 *data = new RefCountedBytes;
1520
1521 // Time the result.
1522 TimeTicks beginning_time = TimeTicks::Now();
1523
1524 history::RedirectList redirects;
1525 URLID url_id;
1526 bool success = false;
1527
1528 // If there are some redirects, try to get a thumbnail from the last
1529 // redirect destination.
1530 if (GetMostRecentRedirectsFrom(page_url, &redirects) &&
1531 !redirects.empty()) {
1532 if ((url_id = db_->GetRowForURL(redirects.back(), NULL)))
1533 success = thumbnail_db_->GetPageThumbnail(url_id, &(*data)->data);
1534 }
1535
1536 // If we don't have a thumbnail from redirects, try the URL directly.
1537 if (!success) {
1538 if ((url_id = db_->GetRowForURL(page_url, NULL)))
1539 success = thumbnail_db_->GetPageThumbnail(url_id, &(*data)->data);
1540 }
1541
1542 // In this rare case, we start to mine the older redirect sessions
1543 // from the visit table to try to find a thumbnail.
1544 if (!success) {
1545 success = GetThumbnailFromOlderRedirect(page_url, &(*data)->data);
1546 }
1547
1548 if (!success)
1549 *data = NULL; // This will tell the callback there was an error.
1550
1551 UMA_HISTOGRAM_TIMES("History.GetPageThumbnail",
1552 TimeTicks::Now() - beginning_time);
1553 }
1554 }
1555
MigrateThumbnailsDatabase()1556 void HistoryBackend::MigrateThumbnailsDatabase() {
1557 // If there is no History DB, we can't record that the migration was done.
1558 // It will be recorded on the next run.
1559 if (db_.get()) {
1560 // If there is no thumbnail DB, we can still record a successful migration.
1561 if (thumbnail_db_.get()) {
1562 thumbnail_db_->RenameAndDropThumbnails(GetThumbnailFileName(),
1563 GetFaviconsFileName());
1564 }
1565 db_->ThumbnailMigrationDone();
1566 }
1567 }
1568
GetThumbnailFromOlderRedirect(const GURL & page_url,std::vector<unsigned char> * data)1569 bool HistoryBackend::GetThumbnailFromOlderRedirect(
1570 const GURL& page_url,
1571 std::vector<unsigned char>* data) {
1572 // Look at a few previous visit sessions.
1573 VisitVector older_sessions;
1574 URLID page_url_id = db_->GetRowForURL(page_url, NULL);
1575 static const int kVisitsToSearchForThumbnail = 4;
1576 db_->GetMostRecentVisitsForURL(
1577 page_url_id, kVisitsToSearchForThumbnail, &older_sessions);
1578
1579 // Iterate across all those previous visits, and see if any of the
1580 // final destinations of those redirect chains have a good thumbnail
1581 // for us.
1582 bool success = false;
1583 for (VisitVector::const_iterator it = older_sessions.begin();
1584 !success && it != older_sessions.end(); ++it) {
1585 history::RedirectList redirects;
1586 if (it->visit_id) {
1587 GetRedirectsFromSpecificVisit(it->visit_id, &redirects);
1588
1589 if (!redirects.empty()) {
1590 URLID url_id;
1591 if ((url_id = db_->GetRowForURL(redirects.back(), NULL)))
1592 success = thumbnail_db_->GetPageThumbnail(url_id, data);
1593 }
1594 }
1595 }
1596
1597 return success;
1598 }
1599
GetFavicon(scoped_refptr<GetFaviconRequest> request,const GURL & icon_url,int icon_types)1600 void HistoryBackend::GetFavicon(scoped_refptr<GetFaviconRequest> request,
1601 const GURL& icon_url,
1602 int icon_types) {
1603 UpdateFaviconMappingAndFetchImpl(NULL, icon_url, request, icon_types);
1604 }
1605
UpdateFaviconMappingAndFetch(scoped_refptr<GetFaviconRequest> request,const GURL & page_url,const GURL & icon_url,IconType icon_type)1606 void HistoryBackend::UpdateFaviconMappingAndFetch(
1607 scoped_refptr<GetFaviconRequest> request,
1608 const GURL& page_url,
1609 const GURL& icon_url,
1610 IconType icon_type) {
1611 UpdateFaviconMappingAndFetchImpl(&page_url, icon_url, request, icon_type);
1612 }
1613
SetFaviconOutOfDateForPage(const GURL & page_url)1614 void HistoryBackend::SetFaviconOutOfDateForPage(const GURL& page_url) {
1615 std::vector<IconMapping> icon_mappings;
1616
1617 if (!thumbnail_db_.get() ||
1618 !thumbnail_db_->GetIconMappingsForPageURL(page_url,
1619 &icon_mappings))
1620 return;
1621
1622 for (std::vector<IconMapping>::iterator m = icon_mappings.begin();
1623 m != icon_mappings.end(); ++m) {
1624 thumbnail_db_->SetFaviconLastUpdateTime(m->icon_id, Time());
1625 }
1626 ScheduleCommit();
1627 }
1628
SetImportedFavicons(const std::vector<ImportedFaviconUsage> & favicon_usage)1629 void HistoryBackend::SetImportedFavicons(
1630 const std::vector<ImportedFaviconUsage>& favicon_usage) {
1631 if (!db_.get() || !thumbnail_db_.get())
1632 return;
1633
1634 Time now = Time::Now();
1635
1636 // Track all URLs that had their favicons set or updated.
1637 std::set<GURL> favicons_changed;
1638
1639 for (size_t i = 0; i < favicon_usage.size(); i++) {
1640 FaviconID favicon_id = thumbnail_db_->GetFaviconIDForFaviconURL(
1641 favicon_usage[i].favicon_url, history::FAVICON, NULL);
1642 if (!favicon_id) {
1643 // This favicon doesn't exist yet, so we create it using the given data.
1644 favicon_id = thumbnail_db_->AddFavicon(favicon_usage[i].favicon_url,
1645 history::FAVICON);
1646 if (!favicon_id)
1647 continue; // Unable to add the favicon.
1648 thumbnail_db_->SetFavicon(favicon_id,
1649 new RefCountedBytes(favicon_usage[i].png_data), now);
1650 }
1651
1652 // Save the mapping from all the URLs to the favicon.
1653 BookmarkService* bookmark_service = GetBookmarkService();
1654 for (std::set<GURL>::const_iterator url = favicon_usage[i].urls.begin();
1655 url != favicon_usage[i].urls.end(); ++url) {
1656 URLRow url_row;
1657 if (!db_->GetRowForURL(*url, &url_row)) {
1658 // If the URL is present as a bookmark, add the url in history to
1659 // save the favicon mapping. This will match with what history db does
1660 // for regular bookmarked URLs with favicons - when history db is
1661 // cleaned, we keep an entry in the db with 0 visits as long as that
1662 // url is bookmarked.
1663 if (bookmark_service && bookmark_service_->IsBookmarked(*url)) {
1664 URLRow url_info(*url);
1665 url_info.set_visit_count(0);
1666 url_info.set_typed_count(0);
1667 url_info.set_last_visit(base::Time());
1668 url_info.set_hidden(false);
1669 db_->AddURL(url_info);
1670 thumbnail_db_->AddIconMapping(*url, favicon_id);
1671 favicons_changed.insert(*url);
1672 }
1673 } else {
1674 if (!thumbnail_db_->GetIconMappingForPageURL(*url, FAVICON, NULL)) {
1675 // URL is present in history, update the favicon *only* if it is not
1676 // set already.
1677 thumbnail_db_->AddIconMapping(*url, favicon_id);
1678 favicons_changed.insert(*url);
1679 }
1680 }
1681 }
1682 }
1683
1684 if (!favicons_changed.empty()) {
1685 // Send the notification about the changed favicon URLs.
1686 FaviconChangeDetails* changed_details = new FaviconChangeDetails;
1687 changed_details->urls.swap(favicons_changed);
1688 BroadcastNotifications(NotificationType::FAVICON_CHANGED, changed_details);
1689 }
1690 }
1691
UpdateFaviconMappingAndFetchImpl(const GURL * page_url,const GURL & icon_url,scoped_refptr<GetFaviconRequest> request,int icon_types)1692 void HistoryBackend::UpdateFaviconMappingAndFetchImpl(
1693 const GURL* page_url,
1694 const GURL& icon_url,
1695 scoped_refptr<GetFaviconRequest> request,
1696 int icon_types) {
1697 // Check only a single type was given when the page_url was specified.
1698 DCHECK(!page_url || (page_url && (icon_types == FAVICON ||
1699 icon_types == TOUCH_ICON || icon_types == TOUCH_PRECOMPOSED_ICON)));
1700
1701 if (request->canceled())
1702 return;
1703
1704 FaviconData favicon;
1705
1706 if (thumbnail_db_.get()) {
1707 const FaviconID favicon_id =
1708 thumbnail_db_->GetFaviconIDForFaviconURL(
1709 icon_url, icon_types, &favicon.icon_type);
1710 if (favicon_id) {
1711 scoped_refptr<RefCountedBytes> data = new RefCountedBytes();
1712 favicon.known_icon = true;
1713 Time last_updated;
1714 if (thumbnail_db_->GetFavicon(favicon_id, &last_updated, &data->data,
1715 NULL)) {
1716 favicon.expired = (Time::Now() - last_updated) >
1717 TimeDelta::FromDays(kFaviconRefetchDays);
1718 favicon.image_data = data;
1719 }
1720
1721 if (page_url)
1722 SetFaviconMapping(*page_url, favicon_id, favicon.icon_type);
1723 }
1724 // else case, haven't cached entry yet. Caller is responsible for
1725 // downloading the favicon and invoking SetFavicon.
1726 }
1727 request->ForwardResult(GetFaviconRequest::TupleType(
1728 request->handle(), favicon));
1729 }
1730
GetFaviconForURL(scoped_refptr<GetFaviconRequest> request,const GURL & page_url,int icon_types)1731 void HistoryBackend::GetFaviconForURL(
1732 scoped_refptr<GetFaviconRequest> request,
1733 const GURL& page_url,
1734 int icon_types) {
1735 if (request->canceled())
1736 return;
1737
1738 FaviconData favicon;
1739
1740 if (db_.get() && thumbnail_db_.get()) {
1741 // Time the query.
1742 TimeTicks beginning_time = TimeTicks::Now();
1743
1744 std::vector<IconMapping> icon_mappings;
1745 Time last_updated;
1746 scoped_refptr<RefCountedBytes> data = new RefCountedBytes();
1747 if (thumbnail_db_->GetIconMappingsForPageURL(page_url, &icon_mappings) &&
1748 (icon_mappings.front().icon_type & icon_types) &&
1749 thumbnail_db_->GetFavicon(icon_mappings.front().icon_id, &last_updated,
1750 &data->data, &favicon.icon_url)) {
1751 favicon.known_icon = true;
1752 favicon.expired = (Time::Now() - last_updated) >
1753 TimeDelta::FromDays(kFaviconRefetchDays);
1754 favicon.icon_type = icon_mappings.front().icon_type;
1755 favicon.image_data = data;
1756 }
1757
1758 UMA_HISTOGRAM_TIMES("History.GetFavIconForURL", // historical name
1759 TimeTicks::Now() - beginning_time);
1760 }
1761
1762 request->ForwardResult(
1763 GetFaviconRequest::TupleType(request->handle(), favicon));
1764 }
1765
SetFavicon(const GURL & page_url,const GURL & icon_url,scoped_refptr<RefCountedMemory> data,IconType icon_type)1766 void HistoryBackend::SetFavicon(
1767 const GURL& page_url,
1768 const GURL& icon_url,
1769 scoped_refptr<RefCountedMemory> data,
1770 IconType icon_type) {
1771 DCHECK(data.get());
1772 if (!thumbnail_db_.get() || !db_.get())
1773 return;
1774
1775 FaviconID id = thumbnail_db_->GetFaviconIDForFaviconURL(
1776 icon_url, icon_type, NULL);
1777 if (!id)
1778 id = thumbnail_db_->AddFavicon(icon_url, icon_type);
1779
1780 // Set the image data.
1781 thumbnail_db_->SetFavicon(id, data, Time::Now());
1782
1783 SetFaviconMapping(page_url, id, icon_type);
1784 }
1785
SetFaviconMapping(const GURL & page_url,FaviconID id,IconType icon_type)1786 void HistoryBackend::SetFaviconMapping(const GURL& page_url,
1787 FaviconID id,
1788 IconType icon_type) {
1789 if (!thumbnail_db_.get())
1790 return;
1791
1792 // Find all the pages whose favicons we should set, we want to set it for
1793 // all the pages in the redirect chain if it redirected.
1794 history::RedirectList dummy_list;
1795 history::RedirectList* redirects;
1796 RedirectCache::iterator iter = recent_redirects_.Get(page_url);
1797 if (iter != recent_redirects_.end()) {
1798 redirects = &iter->second;
1799
1800 // This redirect chain should have the destination URL as the last item.
1801 DCHECK(!redirects->empty());
1802 DCHECK(redirects->back() == page_url);
1803 } else {
1804 // No redirect chain stored, make up one containing the URL we want to we
1805 // can use the same logic below.
1806 dummy_list.push_back(page_url);
1807 redirects = &dummy_list;
1808 }
1809
1810 std::set<GURL> favicons_changed;
1811
1812 // Save page <-> favicon association.
1813 for (history::RedirectList::const_iterator i(redirects->begin());
1814 i != redirects->end(); ++i) {
1815 FaviconID replaced_id;
1816 if (AddOrUpdateIconMapping(*i, id, icon_type, &replaced_id)) {
1817 // The page's favicon ID changed. This means that the one we just
1818 // changed from could have been orphaned, and we need to re-check it.
1819 // This is not super fast, but this case will get triggered rarely,
1820 // since normally a page will always map to the same favicon ID. It
1821 // will mostly happen for favicons we import.
1822 if (replaced_id && !thumbnail_db_->HasMappingFor(replaced_id))
1823 thumbnail_db_->DeleteFavicon(replaced_id);
1824
1825 favicons_changed.insert(*i);
1826 }
1827 }
1828
1829 // Send the notification about the changed favicons.
1830 FaviconChangeDetails* changed_details = new FaviconChangeDetails;
1831 changed_details->urls.swap(favicons_changed);
1832 BroadcastNotifications(NotificationType::FAVICON_CHANGED, changed_details);
1833
1834 ScheduleCommit();
1835 }
1836
AddOrUpdateIconMapping(const GURL & page_url,FaviconID id,IconType icon_type,FaviconID * replaced_icon)1837 bool HistoryBackend::AddOrUpdateIconMapping(const GURL& page_url,
1838 FaviconID id,
1839 IconType icon_type,
1840 FaviconID* replaced_icon) {
1841 *replaced_icon = 0;
1842 std::vector<IconMapping> icon_mappings;
1843 if (!thumbnail_db_->GetIconMappingsForPageURL(page_url, &icon_mappings)) {
1844 // There is no mapping add it directly.
1845 thumbnail_db_->AddIconMapping(page_url, id);
1846 return true;
1847 }
1848 // Iterate all matched icon mappings,
1849 // a. If the given icon id and matched icon id are same, return.
1850 // b. If the given icon type and matched icon type are same, but icon id
1851 // are not, update the IconMapping.
1852 // c. If the given icon_type and matched icon type are not same, but
1853 // either of them is ICON_TOUCH or ICON_PRECOMPOSED_TOUCH, update the
1854 // IconMapping.
1855 // d. Otherwise add a icon mapping.
1856 for (std::vector<IconMapping>::iterator m = icon_mappings.begin();
1857 m != icon_mappings.end(); ++m) {
1858 if (m->icon_id == id)
1859 // The mapping is already there.
1860 return false;
1861
1862 if ((icon_type == TOUCH_ICON && m->icon_type == TOUCH_PRECOMPOSED_ICON) ||
1863 (icon_type == TOUCH_PRECOMPOSED_ICON && m->icon_type == TOUCH_ICON) ||
1864 (icon_type == m->icon_type)) {
1865 thumbnail_db_->UpdateIconMapping(m->mapping_id, id);
1866 *replaced_icon = m->icon_id;
1867 return true;
1868 }
1869 }
1870 thumbnail_db_->AddIconMapping(page_url, id);
1871 return true;
1872 }
1873
Commit()1874 void HistoryBackend::Commit() {
1875 if (!db_.get())
1876 return;
1877
1878 // Note that a commit may not actually have been scheduled if a caller
1879 // explicitly calls this instead of using ScheduleCommit. Likewise, we
1880 // may reset the flag written by a pending commit. But this is OK! It
1881 // will merely cause extra commits (which is kind of the idea). We
1882 // could optimize more for this case (we may get two extra commits in
1883 // some cases) but it hasn't been important yet.
1884 CancelScheduledCommit();
1885
1886 db_->CommitTransaction();
1887 DCHECK(db_->transaction_nesting() == 0) << "Somebody left a transaction open";
1888 db_->BeginTransaction();
1889
1890 if (thumbnail_db_.get()) {
1891 thumbnail_db_->CommitTransaction();
1892 DCHECK(thumbnail_db_->transaction_nesting() == 0) <<
1893 "Somebody left a transaction open";
1894 thumbnail_db_->BeginTransaction();
1895 }
1896
1897 if (archived_db_.get()) {
1898 archived_db_->CommitTransaction();
1899 archived_db_->BeginTransaction();
1900 }
1901
1902 if (text_database_.get()) {
1903 text_database_->CommitTransaction();
1904 text_database_->BeginTransaction();
1905 }
1906 }
1907
ScheduleCommit()1908 void HistoryBackend::ScheduleCommit() {
1909 if (scheduled_commit_.get())
1910 return;
1911 scheduled_commit_ = new CommitLaterTask(this);
1912 MessageLoop::current()->PostDelayedTask(FROM_HERE,
1913 NewRunnableMethod(scheduled_commit_.get(),
1914 &CommitLaterTask::RunCommit),
1915 kCommitIntervalMs);
1916 }
1917
CancelScheduledCommit()1918 void HistoryBackend::CancelScheduledCommit() {
1919 if (scheduled_commit_) {
1920 scheduled_commit_->Cancel();
1921 scheduled_commit_ = NULL;
1922 }
1923 }
1924
ProcessDBTaskImpl()1925 void HistoryBackend::ProcessDBTaskImpl() {
1926 if (!db_.get()) {
1927 // db went away, release all the refs.
1928 ReleaseDBTasks();
1929 return;
1930 }
1931
1932 // Remove any canceled tasks.
1933 while (!db_task_requests_.empty() && db_task_requests_.front()->canceled()) {
1934 db_task_requests_.front()->Release();
1935 db_task_requests_.pop_front();
1936 }
1937 if (db_task_requests_.empty())
1938 return;
1939
1940 // Run the first task.
1941 HistoryDBTaskRequest* request = db_task_requests_.front();
1942 db_task_requests_.pop_front();
1943 if (request->value->RunOnDBThread(this, db_.get())) {
1944 // The task is done. Notify the callback.
1945 request->ForwardResult(HistoryDBTaskRequest::TupleType());
1946 // We AddRef'd the request before adding, need to release it now.
1947 request->Release();
1948 } else {
1949 // Tasks wants to run some more. Schedule it at the end of current tasks.
1950 db_task_requests_.push_back(request);
1951 // And process it after an invoke later.
1952 MessageLoop::current()->PostTask(FROM_HERE, NewRunnableMethod(
1953 this, &HistoryBackend::ProcessDBTaskImpl));
1954 }
1955 }
1956
ReleaseDBTasks()1957 void HistoryBackend::ReleaseDBTasks() {
1958 for (std::list<HistoryDBTaskRequest*>::iterator i =
1959 db_task_requests_.begin(); i != db_task_requests_.end(); ++i) {
1960 (*i)->Release();
1961 }
1962 db_task_requests_.clear();
1963 }
1964
1965 ////////////////////////////////////////////////////////////////////////////////
1966 //
1967 // Generic operations
1968 //
1969 ////////////////////////////////////////////////////////////////////////////////
1970
DeleteURLs(const std::vector<GURL> & urls)1971 void HistoryBackend::DeleteURLs(const std::vector<GURL>& urls) {
1972 for (std::vector<GURL>::const_iterator url = urls.begin(); url != urls.end();
1973 ++url) {
1974 expirer_.DeleteURL(*url);
1975 }
1976
1977 db_->GetStartDate(&first_recorded_time_);
1978 // Force a commit, if the user is deleting something for privacy reasons, we
1979 // want to get it on disk ASAP.
1980 Commit();
1981 }
1982
DeleteURL(const GURL & url)1983 void HistoryBackend::DeleteURL(const GURL& url) {
1984 expirer_.DeleteURL(url);
1985
1986 db_->GetStartDate(&first_recorded_time_);
1987 // Force a commit, if the user is deleting something for privacy reasons, we
1988 // want to get it on disk ASAP.
1989 Commit();
1990 }
1991
ExpireHistoryBetween(scoped_refptr<ExpireHistoryRequest> request,const std::set<GURL> & restrict_urls,Time begin_time,Time end_time)1992 void HistoryBackend::ExpireHistoryBetween(
1993 scoped_refptr<ExpireHistoryRequest> request,
1994 const std::set<GURL>& restrict_urls,
1995 Time begin_time,
1996 Time end_time) {
1997 if (request->canceled())
1998 return;
1999
2000 if (db_.get()) {
2001 if (begin_time.is_null() && end_time.is_null() && restrict_urls.empty()) {
2002 // Special case deleting all history so it can be faster and to reduce the
2003 // possibility of an information leak.
2004 DeleteAllHistory();
2005 } else {
2006 // Clearing parts of history, have the expirer do the depend
2007 expirer_.ExpireHistoryBetween(restrict_urls, begin_time, end_time);
2008
2009 // Force a commit, if the user is deleting something for privacy reasons,
2010 // we want to get it on disk ASAP.
2011 Commit();
2012 }
2013 }
2014
2015 if (begin_time <= first_recorded_time_)
2016 db_->GetStartDate(&first_recorded_time_);
2017
2018 request->ForwardResult(ExpireHistoryRequest::TupleType());
2019
2020 if (history_publisher_.get() && restrict_urls.empty())
2021 history_publisher_->DeleteUserHistoryBetween(begin_time, end_time);
2022 }
2023
URLsNoLongerBookmarked(const std::set<GURL> & urls)2024 void HistoryBackend::URLsNoLongerBookmarked(const std::set<GURL>& urls) {
2025 if (!db_.get())
2026 return;
2027
2028 for (std::set<GURL>::const_iterator i = urls.begin(); i != urls.end(); ++i) {
2029 URLRow url_row;
2030 if (!db_->GetRowForURL(*i, &url_row))
2031 continue; // The URL isn't in the db; nothing to do.
2032
2033 VisitVector visits;
2034 db_->GetVisitsForURL(url_row.id(), &visits);
2035
2036 if (visits.empty())
2037 expirer_.DeleteURL(*i); // There are no more visits; nuke the URL.
2038 }
2039 }
2040
ProcessDBTask(scoped_refptr<HistoryDBTaskRequest> request)2041 void HistoryBackend::ProcessDBTask(
2042 scoped_refptr<HistoryDBTaskRequest> request) {
2043 DCHECK(request.get());
2044 if (request->canceled())
2045 return;
2046
2047 bool task_scheduled = !db_task_requests_.empty();
2048 // Make sure we up the refcount of the request. ProcessDBTaskImpl will
2049 // release when done with the task.
2050 request->AddRef();
2051 db_task_requests_.push_back(request.get());
2052 if (!task_scheduled) {
2053 // No other tasks are scheduled. Process request now.
2054 ProcessDBTaskImpl();
2055 }
2056 }
2057
BroadcastNotifications(NotificationType type,HistoryDetails * details_deleted)2058 void HistoryBackend::BroadcastNotifications(
2059 NotificationType type,
2060 HistoryDetails* details_deleted) {
2061 DCHECK(delegate_.get());
2062 delegate_->BroadcastNotifications(type, details_deleted);
2063 }
2064
2065 // Deleting --------------------------------------------------------------------
2066
DeleteAllHistory()2067 void HistoryBackend::DeleteAllHistory() {
2068 // Our approach to deleting all history is:
2069 // 1. Copy the bookmarks and their dependencies to new tables with temporary
2070 // names.
2071 // 2. Delete the original tables. Since tables can not share pages, we know
2072 // that any data we don't want to keep is now in an unused page.
2073 // 3. Renaming the temporary tables to match the original.
2074 // 4. Vacuuming the database to delete the unused pages.
2075 //
2076 // Since we are likely to have very few bookmarks and their dependencies
2077 // compared to all history, this is also much faster than just deleting from
2078 // the original tables directly.
2079
2080 // Get the bookmarked URLs.
2081 std::vector<GURL> starred_urls;
2082 BookmarkService* bookmark_service = GetBookmarkService();
2083 if (bookmark_service)
2084 bookmark_service_->GetBookmarks(&starred_urls);
2085
2086 std::vector<URLRow> kept_urls;
2087 for (size_t i = 0; i < starred_urls.size(); i++) {
2088 URLRow row;
2089 if (!db_->GetRowForURL(starred_urls[i], &row))
2090 continue;
2091
2092 // Clear the last visit time so when we write these rows they are "clean."
2093 row.set_last_visit(Time());
2094 row.set_visit_count(0);
2095 row.set_typed_count(0);
2096 kept_urls.push_back(row);
2097 }
2098
2099 // Clear thumbnail and favicon history. The favicons for the given URLs will
2100 // be kept.
2101 if (!ClearAllThumbnailHistory(&kept_urls)) {
2102 LOG(ERROR) << "Thumbnail history could not be cleared";
2103 // We continue in this error case. If the user wants to delete their
2104 // history, we should delete as much as we can.
2105 }
2106
2107 // ClearAllMainHistory will change the IDs of the URLs in kept_urls. Therfore,
2108 // we clear the list afterwards to make sure nobody uses this invalid data.
2109 if (!ClearAllMainHistory(kept_urls))
2110 LOG(ERROR) << "Main history could not be cleared";
2111 kept_urls.clear();
2112
2113 // Delete FTS files & archived history.
2114 if (text_database_.get()) {
2115 // We assume that the text database has one transaction on them that we need
2116 // to close & restart (the long-running history transaction).
2117 text_database_->CommitTransaction();
2118 text_database_->DeleteAll();
2119 text_database_->BeginTransaction();
2120 }
2121
2122 if (archived_db_.get()) {
2123 // Close the database and delete the file.
2124 archived_db_.reset();
2125 FilePath archived_file_name = GetArchivedFileName();
2126 file_util::Delete(archived_file_name, false);
2127
2128 // Now re-initialize the database (which may fail).
2129 archived_db_.reset(new ArchivedDatabase());
2130 if (!archived_db_->Init(archived_file_name)) {
2131 LOG(WARNING) << "Could not initialize the archived database.";
2132 archived_db_.reset();
2133 } else {
2134 // Open our long-running transaction on this database.
2135 archived_db_->BeginTransaction();
2136 }
2137 }
2138
2139 db_->GetStartDate(&first_recorded_time_);
2140
2141 // Send out the notfication that history is cleared. The in-memory datdabase
2142 // will pick this up and clear itself.
2143 URLsDeletedDetails* details = new URLsDeletedDetails;
2144 details->all_history = true;
2145 BroadcastNotifications(NotificationType::HISTORY_URLS_DELETED, details);
2146 }
2147
ClearAllThumbnailHistory(std::vector<URLRow> * kept_urls)2148 bool HistoryBackend::ClearAllThumbnailHistory(
2149 std::vector<URLRow>* kept_urls) {
2150 if (!thumbnail_db_.get()) {
2151 // When we have no reference to the thumbnail database, maybe there was an
2152 // error opening it. In this case, we just try to blow it away to try to
2153 // fix the error if it exists. This may fail, in which case either the
2154 // file doesn't exist or there's no more we can do.
2155 file_util::Delete(GetThumbnailFileName(), false);
2156 return true;
2157 }
2158
2159 // Create the duplicate favicon table, this is where the favicons we want
2160 // to keep will be stored.
2161 if (!thumbnail_db_->InitTemporaryFaviconsTable())
2162 return false;
2163
2164 if (!thumbnail_db_->InitTemporaryIconMappingTable())
2165 return false;
2166
2167 // This maps existing favicon IDs to the ones in the temporary table.
2168 typedef std::map<FaviconID, FaviconID> FaviconMap;
2169 FaviconMap copied_favicons;
2170
2171 // Copy all unique favicons to the temporary table, and update all the
2172 // URLs to have the new IDs.
2173 for (std::vector<URLRow>::iterator i = kept_urls->begin();
2174 i != kept_urls->end(); ++i) {
2175 std::vector<IconMapping> icon_mappings;
2176 if (!thumbnail_db_->GetIconMappingsForPageURL(i->url(), &icon_mappings))
2177 continue;
2178
2179 for (std::vector<IconMapping>::iterator m = icon_mappings.begin();
2180 m != icon_mappings.end(); ++m) {
2181 FaviconID old_id = m->icon_id;
2182 FaviconID new_id;
2183 FaviconMap::const_iterator found = copied_favicons.find(old_id);
2184 if (found == copied_favicons.end()) {
2185 new_id = thumbnail_db_->CopyToTemporaryFaviconTable(old_id);
2186 copied_favicons[old_id] = new_id;
2187 } else {
2188 // We already encountered a URL that used this favicon, use the ID we
2189 // previously got.
2190 new_id = found->second;
2191 }
2192 // Add Icon mapping, and we don't care wheteher it suceeded or not.
2193 thumbnail_db_->AddToTemporaryIconMappingTable(i->url(), new_id);
2194 }
2195 }
2196
2197 // Rename the duplicate favicon and icon_mapping back table and recreate the
2198 // other tables. This will make the database consistent again.
2199 thumbnail_db_->CommitTemporaryFaviconTable();
2200 thumbnail_db_->CommitTemporaryIconMappingTable();
2201
2202 thumbnail_db_->RecreateThumbnailTable();
2203
2204 // Vacuum to remove all the pages associated with the dropped tables. There
2205 // must be no transaction open on the table when we do this. We assume that
2206 // our long-running transaction is open, so we complete it and start it again.
2207 DCHECK(thumbnail_db_->transaction_nesting() == 1);
2208 thumbnail_db_->CommitTransaction();
2209 thumbnail_db_->Vacuum();
2210 thumbnail_db_->BeginTransaction();
2211 return true;
2212 }
2213
ClearAllMainHistory(const std::vector<URLRow> & kept_urls)2214 bool HistoryBackend::ClearAllMainHistory(
2215 const std::vector<URLRow>& kept_urls) {
2216 // Create the duplicate URL table. We will copy the kept URLs into this.
2217 if (!db_->CreateTemporaryURLTable())
2218 return false;
2219
2220 // Insert the URLs into the temporary table, we need to keep a map of changed
2221 // IDs since the ID will be different in the new table.
2222 typedef std::map<URLID, URLID> URLIDMap;
2223 URLIDMap old_to_new; // Maps original ID to new one.
2224 for (std::vector<URLRow>::const_iterator i = kept_urls.begin();
2225 i != kept_urls.end();
2226 ++i) {
2227 URLID new_id = db_->AddTemporaryURL(*i);
2228 old_to_new[i->id()] = new_id;
2229 }
2230
2231 // Replace the original URL table with the temporary one.
2232 if (!db_->CommitTemporaryURLTable())
2233 return false;
2234
2235 // Delete the old tables and recreate them empty.
2236 db_->RecreateAllTablesButURL();
2237
2238 // Vacuum to reclaim the space from the dropped tables. This must be done
2239 // when there is no transaction open, and we assume that our long-running
2240 // transaction is currently open.
2241 db_->CommitTransaction();
2242 db_->Vacuum();
2243 db_->BeginTransaction();
2244 db_->GetStartDate(&first_recorded_time_);
2245
2246 return true;
2247 }
2248
GetBookmarkService()2249 BookmarkService* HistoryBackend::GetBookmarkService() {
2250 if (bookmark_service_)
2251 bookmark_service_->BlockTillLoaded();
2252 return bookmark_service_;
2253 }
2254
2255 } // namespace history
2256