1 // Copyright (c) 2012 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 <algorithm>
8 #include <functional>
9 #include <list>
10 #include <map>
11 #include <set>
12 #include <vector>
13
14 #include "base/basictypes.h"
15 #include "base/bind.h"
16 #include "base/compiler_specific.h"
17 #include "base/files/file_enumerator.h"
18 #include "base/memory/scoped_ptr.h"
19 #include "base/memory/scoped_vector.h"
20 #include "base/message_loop/message_loop.h"
21 #include "base/metrics/histogram.h"
22 #include "base/rand_util.h"
23 #include "base/strings/string_util.h"
24 #include "base/strings/utf_string_conversions.h"
25 #include "base/time/time.h"
26 #include "chrome/browser/chrome_notification_types.h"
27 #include "chrome/browser/history/download_row.h"
28 #include "chrome/browser/history/history_db_task.h"
29 #include "chrome/browser/history/history_db_task.h"
30 #include "chrome/browser/history/history_notifications.h"
31 #include "chrome/browser/history/in_memory_history_backend.h"
32 #include "chrome/browser/history/in_memory_history_backend.h"
33 #include "chrome/browser/history/top_sites.h"
34 #include "chrome/browser/history/typed_url_syncable_service.h"
35 #include "chrome/browser/history/typed_url_syncable_service.h"
36 #include "chrome/browser/history/visit_filter.h"
37 #include "chrome/common/chrome_constants.h"
38 #include "chrome/common/importer/imported_favicon_usage.h"
39 #include "chrome/common/url_constants.h"
40 #include "components/favicon_base/select_favicon_frames.h"
41 #include "components/history/core/browser/history_client.h"
42 #include "components/history/core/browser/keyword_search_term.h"
43 #include "components/history/core/browser/page_usage_data.h"
44 #include "net/base/registry_controlled_domains/registry_controlled_domain.h"
45 #include "sql/error_delegate_util.h"
46 #include "third_party/skia/include/core/SkBitmap.h"
47 #include "ui/gfx/codec/png_codec.h"
48 #include "url/gurl.h"
49
50 #if defined(OS_ANDROID)
51 #include "chrome/browser/history/android/android_provider_backend.h"
52 #endif
53
54 using base::Time;
55 using base::TimeDelta;
56 using base::TimeTicks;
57
58 /* The HistoryBackend consists of two components:
59
60 HistoryDatabase (stores past 3 months of history)
61 URLDatabase (stores a list of URLs)
62 DownloadDatabase (stores a list of downloads)
63 VisitDatabase (stores a list of visits for the URLs)
64 VisitSegmentDatabase (stores groups of URLs for the most visited view).
65
66 ExpireHistoryBackend (manages deleting things older than 3 months)
67 */
68
69 namespace history {
70
71 namespace {
RunUnlessCanceled(const base::Closure & closure,const base::CancelableTaskTracker::IsCanceledCallback & is_canceled)72 void RunUnlessCanceled(
73 const base::Closure& closure,
74 const base::CancelableTaskTracker::IsCanceledCallback& is_canceled) {
75 if (!is_canceled.Run())
76 closure.Run();
77 }
78 } // namespace
79
80 #if defined(OS_ANDROID)
81 // How long we keep segment data for in days. Currently 3 months.
82 // This value needs to be greater or equal to
83 // MostVisitedModel::kMostVisitedScope but we don't want to introduce a direct
84 // dependency between MostVisitedModel and the history backend.
85 const int kSegmentDataRetention = 90;
86 #endif
87
88 // How long we'll wait to do a commit, so that things are batched together.
89 const int kCommitIntervalSeconds = 10;
90
91 // The amount of time before we re-fetch the favicon.
92 const int kFaviconRefetchDays = 7;
93
94 // The maximum number of items we'll allow in the redirect list before
95 // deleting some.
96 const int kMaxRedirectCount = 32;
97
98 // The number of days old a history entry can be before it is considered "old"
99 // and is deleted.
100 const int kExpireDaysThreshold = 90;
101
102 #if defined(OS_ANDROID)
103 // The maximum number of top sites to track when recording top page visit stats.
104 const size_t kPageVisitStatsMaxTopSites = 50;
105 #endif
106
107 // Converts from PageUsageData to MostVisitedURL. |redirects| is a
108 // list of redirects for this URL. Empty list means no redirects.
MakeMostVisitedURL(const PageUsageData & page_data,const RedirectList & redirects)109 MostVisitedURL MakeMostVisitedURL(const PageUsageData& page_data,
110 const RedirectList& redirects) {
111 MostVisitedURL mv;
112 mv.url = page_data.GetURL();
113 mv.title = page_data.GetTitle();
114 if (redirects.empty()) {
115 // Redirects must contain at least the target url.
116 mv.redirects.push_back(mv.url);
117 } else {
118 mv.redirects = redirects;
119 if (mv.redirects[mv.redirects.size() - 1] != mv.url) {
120 // The last url must be the target url.
121 mv.redirects.push_back(mv.url);
122 }
123 }
124 return mv;
125 }
126
127 // This task is run on a timer so that commits happen at regular intervals
128 // so they are batched together. The important thing about this class is that
129 // it supports canceling of the task so the reference to the backend will be
130 // freed. The problem is that when history is shutting down, there is likely
131 // to be one of these commits still pending and holding a reference.
132 //
133 // The backend can call Cancel to have this task release the reference. The
134 // task will still run (if we ever get to processing the event before
135 // shutdown), but it will not do anything.
136 //
137 // Note that this is a refcounted object and is not a task in itself. It should
138 // be assigned to a RunnableMethod.
139 //
140 // TODO(brettw): bug 1165182: This should be replaced with a
141 // base::WeakPtrFactory which will handle everything automatically (like we do
142 // in ExpireHistoryBackend).
143 class CommitLaterTask : public base::RefCounted<CommitLaterTask> {
144 public:
CommitLaterTask(HistoryBackend * history_backend)145 explicit CommitLaterTask(HistoryBackend* history_backend)
146 : history_backend_(history_backend) {
147 }
148
149 // The backend will call this function if it is being destroyed so that we
150 // release our reference.
Cancel()151 void Cancel() {
152 history_backend_ = NULL;
153 }
154
RunCommit()155 void RunCommit() {
156 if (history_backend_.get())
157 history_backend_->Commit();
158 }
159
160 private:
161 friend class base::RefCounted<CommitLaterTask>;
162
~CommitLaterTask()163 ~CommitLaterTask() {}
164
165 scoped_refptr<HistoryBackend> history_backend_;
166 };
167
168
QueuedHistoryDBTask(scoped_ptr<HistoryDBTask> task,scoped_refptr<base::SingleThreadTaskRunner> origin_loop,const base::CancelableTaskTracker::IsCanceledCallback & is_canceled)169 QueuedHistoryDBTask::QueuedHistoryDBTask(
170 scoped_ptr<HistoryDBTask> task,
171 scoped_refptr<base::SingleThreadTaskRunner> origin_loop,
172 const base::CancelableTaskTracker::IsCanceledCallback& is_canceled)
173 : task_(task.Pass()), origin_loop_(origin_loop), is_canceled_(is_canceled) {
174 DCHECK(task_);
175 DCHECK(origin_loop_.get());
176 DCHECK(!is_canceled_.is_null());
177 }
178
~QueuedHistoryDBTask()179 QueuedHistoryDBTask::~QueuedHistoryDBTask() {
180 // Ensure that |task_| is destroyed on its origin thread.
181 origin_loop_->PostTask(
182 FROM_HERE,
183 base::Bind(&base::DeletePointer<HistoryDBTask>,
184 base::Unretained(task_.release())));
185 }
186
is_canceled()187 bool QueuedHistoryDBTask::is_canceled() {
188 return is_canceled_.Run();
189 }
190
Run(HistoryBackend * backend,HistoryDatabase * db)191 bool QueuedHistoryDBTask::Run(HistoryBackend* backend,
192 HistoryDatabase* db) {
193 return task_->RunOnDBThread(backend, db);
194 }
195
DoneRun()196 void QueuedHistoryDBTask::DoneRun() {
197 origin_loop_->PostTask(
198 FROM_HERE,
199 base::Bind(&RunUnlessCanceled,
200 base::Bind(&HistoryDBTask::DoneRunOnMainThread,
201 base::Unretained(task_.get())),
202 is_canceled_));
203 }
204
205 // HistoryBackend --------------------------------------------------------------
206
HistoryBackend(const base::FilePath & history_dir,Delegate * delegate,HistoryClient * history_client)207 HistoryBackend::HistoryBackend(const base::FilePath& history_dir,
208 Delegate* delegate,
209 HistoryClient* history_client)
210 : delegate_(delegate),
211 history_dir_(history_dir),
212 scheduled_kill_db_(false),
213 expirer_(this, history_client),
214 recent_redirects_(kMaxRedirectCount),
215 backend_destroy_message_loop_(NULL),
216 segment_queried_(false),
217 history_client_(history_client) {
218 }
219
~HistoryBackend()220 HistoryBackend::~HistoryBackend() {
221 DCHECK(!scheduled_commit_.get()) << "Deleting without cleanup";
222 STLDeleteContainerPointers(queued_history_db_tasks_.begin(),
223 queued_history_db_tasks_.end());
224 queued_history_db_tasks_.clear();
225
226 #if defined(OS_ANDROID)
227 // Release AndroidProviderBackend before other objects.
228 android_provider_backend_.reset();
229 #endif
230
231 // First close the databases before optionally running the "destroy" task.
232 CloseAllDatabases();
233
234 if (!backend_destroy_task_.is_null()) {
235 // Notify an interested party (typically a unit test) that we're done.
236 DCHECK(backend_destroy_message_loop_);
237 backend_destroy_message_loop_->PostTask(FROM_HERE, backend_destroy_task_);
238 }
239
240 #if defined(OS_ANDROID)
241 sql::Connection::Delete(GetAndroidCacheFileName());
242 #endif
243 }
244
Init(const std::string & languages,bool force_fail)245 void HistoryBackend::Init(const std::string& languages, bool force_fail) {
246 if (!force_fail)
247 InitImpl(languages);
248 delegate_->DBLoaded();
249 typed_url_syncable_service_.reset(new TypedUrlSyncableService(this));
250 memory_pressure_listener_.reset(new base::MemoryPressureListener(
251 base::Bind(&HistoryBackend::OnMemoryPressure, base::Unretained(this))));
252 #if defined(OS_ANDROID)
253 PopulateMostVisitedURLMap();
254 #endif
255 }
256
SetOnBackendDestroyTask(base::MessageLoop * message_loop,const base::Closure & task)257 void HistoryBackend::SetOnBackendDestroyTask(base::MessageLoop* message_loop,
258 const base::Closure& task) {
259 if (!backend_destroy_task_.is_null())
260 DLOG(WARNING) << "Setting more than one destroy task, overriding";
261 backend_destroy_message_loop_ = message_loop;
262 backend_destroy_task_ = task;
263 }
264
Closing()265 void HistoryBackend::Closing() {
266 // Any scheduled commit will have a reference to us, we must make it
267 // release that reference before we can be destroyed.
268 CancelScheduledCommit();
269
270 // Release our reference to the delegate, this reference will be keeping the
271 // history service alive.
272 delegate_.reset();
273 }
274
ClearCachedDataForContextID(ContextID context_id)275 void HistoryBackend::ClearCachedDataForContextID(ContextID context_id) {
276 tracker_.ClearCachedDataForContextID(context_id);
277 }
278
GetThumbnailFileName() const279 base::FilePath HistoryBackend::GetThumbnailFileName() const {
280 return history_dir_.Append(chrome::kThumbnailsFilename);
281 }
282
GetFaviconsFileName() const283 base::FilePath HistoryBackend::GetFaviconsFileName() const {
284 return history_dir_.Append(chrome::kFaviconsFilename);
285 }
286
GetArchivedFileName() const287 base::FilePath HistoryBackend::GetArchivedFileName() const {
288 return history_dir_.Append(chrome::kArchivedHistoryFilename);
289 }
290
291 #if defined(OS_ANDROID)
GetAndroidCacheFileName() const292 base::FilePath HistoryBackend::GetAndroidCacheFileName() const {
293 return history_dir_.Append(chrome::kAndroidCacheFilename);
294 }
295 #endif
296
GetLastSegmentID(VisitID from_visit)297 SegmentID HistoryBackend::GetLastSegmentID(VisitID from_visit) {
298 // Set is used to detect referrer loops. Should not happen, but can
299 // if the database is corrupt.
300 std::set<VisitID> visit_set;
301 VisitID visit_id = from_visit;
302 while (visit_id) {
303 VisitRow row;
304 if (!db_->GetRowForVisit(visit_id, &row))
305 return 0;
306 if (row.segment_id)
307 return row.segment_id; // Found a visit in this change with a segment.
308
309 // Check the referrer of this visit, if any.
310 visit_id = row.referring_visit;
311
312 if (visit_set.find(visit_id) != visit_set.end()) {
313 NOTREACHED() << "Loop in referer chain, giving up";
314 break;
315 }
316 visit_set.insert(visit_id);
317 }
318 return 0;
319 }
320
UpdateSegments(const GURL & url,VisitID from_visit,VisitID visit_id,ui::PageTransition transition_type,const Time ts)321 SegmentID HistoryBackend::UpdateSegments(
322 const GURL& url,
323 VisitID from_visit,
324 VisitID visit_id,
325 ui::PageTransition transition_type,
326 const Time ts) {
327 if (!db_)
328 return 0;
329
330 // We only consider main frames.
331 if (!ui::PageTransitionIsMainFrame(transition_type))
332 return 0;
333
334 SegmentID segment_id = 0;
335 ui::PageTransition t =
336 ui::PageTransitionStripQualifier(transition_type);
337
338 // Are we at the beginning of a new segment?
339 // Note that navigating to an existing entry (with back/forward) reuses the
340 // same transition type. We are not adding it as a new segment in that case
341 // because if this was the target of a redirect, we might end up with
342 // 2 entries for the same final URL. Ex: User types google.net, gets
343 // redirected to google.com. A segment is created for google.net. On
344 // google.com users navigates through a link, then press back. That last
345 // navigation is for the entry google.com transition typed. We end up adding
346 // a segment for that one as well. So we end up with google.net and google.com
347 // in the segment table, showing as 2 entries in the NTP.
348 // Note also that we should still be updating the visit count for that segment
349 // which we are not doing now. It should be addressed when
350 // http://crbug.com/96860 is fixed.
351 if ((t == ui::PAGE_TRANSITION_TYPED ||
352 t == ui::PAGE_TRANSITION_AUTO_BOOKMARK) &&
353 (transition_type & ui::PAGE_TRANSITION_FORWARD_BACK) == 0) {
354 // If so, create or get the segment.
355 std::string segment_name = db_->ComputeSegmentName(url);
356 URLID url_id = db_->GetRowForURL(url, NULL);
357 if (!url_id)
358 return 0;
359
360 segment_id = db_->GetSegmentNamed(segment_name);
361 if (!segment_id) {
362 segment_id = db_->CreateSegment(url_id, segment_name);
363 if (!segment_id) {
364 NOTREACHED();
365 return 0;
366 }
367 } else {
368 // Note: if we update an existing segment, we update the url used to
369 // represent that segment in order to minimize stale most visited
370 // images.
371 db_->UpdateSegmentRepresentationURL(segment_id, url_id);
372 }
373 } else {
374 // Note: it is possible there is no segment ID set for this visit chain.
375 // This can happen if the initial navigation wasn't AUTO_BOOKMARK or
376 // TYPED. (For example GENERATED). In this case this visit doesn't count
377 // toward any segment.
378 segment_id = GetLastSegmentID(from_visit);
379 if (!segment_id)
380 return 0;
381 }
382
383 // Set the segment in the visit.
384 if (!db_->SetSegmentID(visit_id, segment_id)) {
385 NOTREACHED();
386 return 0;
387 }
388
389 // Finally, increase the counter for that segment / day.
390 if (!db_->IncreaseSegmentVisitCount(segment_id, ts, 1)) {
391 NOTREACHED();
392 return 0;
393 }
394 return segment_id;
395 }
396
UpdateWithPageEndTime(ContextID context_id,int32 page_id,const GURL & url,Time end_ts)397 void HistoryBackend::UpdateWithPageEndTime(ContextID context_id,
398 int32 page_id,
399 const GURL& url,
400 Time end_ts) {
401 // Will be filled with the URL ID and the visit ID of the last addition.
402 VisitID visit_id = tracker_.GetLastVisit(context_id, page_id, url);
403 UpdateVisitDuration(visit_id, end_ts);
404 }
405
UpdateVisitDuration(VisitID visit_id,const Time end_ts)406 void HistoryBackend::UpdateVisitDuration(VisitID visit_id, const Time end_ts) {
407 if (!db_)
408 return;
409
410 // Get the starting visit_time for visit_id.
411 VisitRow visit_row;
412 if (db_->GetRowForVisit(visit_id, &visit_row)) {
413 // We should never have a negative duration time even when time is skewed.
414 visit_row.visit_duration = end_ts > visit_row.visit_time ?
415 end_ts - visit_row.visit_time : TimeDelta::FromMicroseconds(0);
416 db_->UpdateVisitRow(visit_row);
417 }
418 }
419
AddPage(const HistoryAddPageArgs & request)420 void HistoryBackend::AddPage(const HistoryAddPageArgs& request) {
421 if (!db_)
422 return;
423
424 // Will be filled with the URL ID and the visit ID of the last addition.
425 std::pair<URLID, VisitID> last_ids(0, tracker_.GetLastVisit(
426 request.context_id, request.page_id, request.referrer));
427
428 VisitID from_visit_id = last_ids.second;
429
430 // If a redirect chain is given, we expect the last item in that chain to be
431 // the final URL.
432 DCHECK(request.redirects.empty() ||
433 request.redirects.back() == request.url);
434
435 // If the user is adding older history, we need to make sure our times
436 // are correct.
437 if (request.time < first_recorded_time_)
438 first_recorded_time_ = request.time;
439
440 ui::PageTransition request_transition = request.transition;
441 ui::PageTransition stripped_transition =
442 ui::PageTransitionStripQualifier(request_transition);
443 bool is_keyword_generated =
444 (stripped_transition == ui::PAGE_TRANSITION_KEYWORD_GENERATED);
445
446 // If the user is navigating to a not-previously-typed intranet hostname,
447 // change the transition to TYPED so that the omnibox will learn that this is
448 // a known host.
449 bool has_redirects = request.redirects.size() > 1;
450 if (ui::PageTransitionIsMainFrame(request_transition) &&
451 (stripped_transition != ui::PAGE_TRANSITION_TYPED) &&
452 !is_keyword_generated) {
453 const GURL& origin_url(has_redirects ?
454 request.redirects[0] : request.url);
455 if (origin_url.SchemeIs(url::kHttpScheme) ||
456 origin_url.SchemeIs(url::kHttpsScheme) ||
457 origin_url.SchemeIs(url::kFtpScheme)) {
458 std::string host(origin_url.host());
459 size_t registry_length =
460 net::registry_controlled_domains::GetRegistryLength(
461 host,
462 net::registry_controlled_domains::EXCLUDE_UNKNOWN_REGISTRIES,
463 net::registry_controlled_domains::EXCLUDE_PRIVATE_REGISTRIES);
464 if (registry_length == 0 && !db_->IsTypedHost(host)) {
465 stripped_transition = ui::PAGE_TRANSITION_TYPED;
466 request_transition =
467 ui::PageTransitionFromInt(
468 stripped_transition |
469 ui::PageTransitionGetQualifier(request_transition));
470 }
471 }
472 }
473
474 if (!has_redirects) {
475 // The single entry is both a chain start and end.
476 ui::PageTransition t = ui::PageTransitionFromInt(
477 request_transition |
478 ui::PAGE_TRANSITION_CHAIN_START |
479 ui::PAGE_TRANSITION_CHAIN_END);
480
481 // No redirect case (one element means just the page itself).
482 last_ids = AddPageVisit(request.url, request.time,
483 last_ids.second, t, request.visit_source);
484
485 // Update the segment for this visit. KEYWORD_GENERATED visits should not
486 // result in changing most visited, so we don't update segments (most
487 // visited db).
488 if (!is_keyword_generated) {
489 UpdateSegments(request.url, from_visit_id, last_ids.second, t,
490 request.time);
491
492 // Update the referrer's duration.
493 UpdateVisitDuration(from_visit_id, request.time);
494 }
495 } else {
496 // Redirect case. Add the redirect chain.
497
498 ui::PageTransition redirect_info =
499 ui::PAGE_TRANSITION_CHAIN_START;
500
501 RedirectList redirects = request.redirects;
502 if (redirects[0].SchemeIs(url::kAboutScheme)) {
503 // When the redirect source + referrer is "about" we skip it. This
504 // happens when a page opens a new frame/window to about:blank and then
505 // script sets the URL to somewhere else (used to hide the referrer). It
506 // would be nice to keep all these redirects properly but we don't ever
507 // see the initial about:blank load, so we don't know where the
508 // subsequent client redirect came from.
509 //
510 // In this case, we just don't bother hooking up the source of the
511 // redirects, so we remove it.
512 redirects.erase(redirects.begin());
513 } else if (request_transition & ui::PAGE_TRANSITION_CLIENT_REDIRECT) {
514 redirect_info = ui::PAGE_TRANSITION_CLIENT_REDIRECT;
515 // The first entry in the redirect chain initiated a client redirect.
516 // We don't add this to the database since the referrer is already
517 // there, so we skip over it but change the transition type of the first
518 // transition to client redirect.
519 //
520 // The referrer is invalid when restoring a session that features an
521 // https tab that redirects to a different host or to http. In this
522 // case we don't need to reconnect the new redirect with the existing
523 // chain.
524 if (request.referrer.is_valid()) {
525 DCHECK(request.referrer == redirects[0]);
526 redirects.erase(redirects.begin());
527
528 // If the navigation entry for this visit has replaced that for the
529 // first visit, remove the CHAIN_END marker from the first visit. This
530 // can be called a lot, for example, the page cycler, and most of the
531 // time we won't have changed anything.
532 VisitRow visit_row;
533 if (request.did_replace_entry &&
534 db_->GetRowForVisit(last_ids.second, &visit_row) &&
535 visit_row.transition & ui::PAGE_TRANSITION_CHAIN_END) {
536 visit_row.transition = ui::PageTransitionFromInt(
537 visit_row.transition & ~ui::PAGE_TRANSITION_CHAIN_END);
538 db_->UpdateVisitRow(visit_row);
539 }
540 }
541 }
542
543 for (size_t redirect_index = 0; redirect_index < redirects.size();
544 redirect_index++) {
545 ui::PageTransition t =
546 ui::PageTransitionFromInt(stripped_transition | redirect_info);
547
548 // If this is the last transition, add a CHAIN_END marker
549 if (redirect_index == (redirects.size() - 1)) {
550 t = ui::PageTransitionFromInt(
551 t | ui::PAGE_TRANSITION_CHAIN_END);
552 }
553
554 // Record all redirect visits with the same timestamp. We don't display
555 // them anyway, and if we ever decide to, we can reconstruct their order
556 // from the redirect chain.
557 last_ids = AddPageVisit(redirects[redirect_index],
558 request.time, last_ids.second,
559 t, request.visit_source);
560 if (t & ui::PAGE_TRANSITION_CHAIN_START) {
561 // Update the segment for this visit.
562 UpdateSegments(redirects[redirect_index],
563 from_visit_id, last_ids.second, t, request.time);
564
565 // Update the visit_details for this visit.
566 UpdateVisitDuration(from_visit_id, request.time);
567 }
568
569 // Subsequent transitions in the redirect list must all be server
570 // redirects.
571 redirect_info = ui::PAGE_TRANSITION_SERVER_REDIRECT;
572 }
573
574 // Last, save this redirect chain for later so we can set titles & favicons
575 // on the redirected pages properly.
576 recent_redirects_.Put(request.url, redirects);
577 }
578
579 // TODO(brettw) bug 1140015: Add an "add page" notification so the history
580 // views can keep in sync.
581
582 // Add the last visit to the tracker so we can get outgoing transitions.
583 // TODO(evanm): Due to http://b/1194536 we lose the referrers of a subframe
584 // navigation anyway, so last_visit_id is always zero for them. But adding
585 // them here confuses main frame history, so we skip them for now.
586 if (stripped_transition != ui::PAGE_TRANSITION_AUTO_SUBFRAME &&
587 stripped_transition != ui::PAGE_TRANSITION_MANUAL_SUBFRAME &&
588 !is_keyword_generated) {
589 tracker_.AddVisit(request.context_id, request.page_id, request.url,
590 last_ids.second);
591 }
592
593 ScheduleCommit();
594 }
595
InitImpl(const std::string & languages)596 void HistoryBackend::InitImpl(const std::string& languages) {
597 DCHECK(!db_) << "Initializing HistoryBackend twice";
598 // In the rare case where the db fails to initialize a dialog may get shown
599 // the blocks the caller, yet allows other messages through. For this reason
600 // we only set db_ to the created database if creation is successful. That
601 // way other methods won't do anything as db_ is still NULL.
602
603 TimeTicks beginning_time = TimeTicks::Now();
604
605 // Compute the file names.
606 base::FilePath history_name = history_dir_.Append(chrome::kHistoryFilename);
607 base::FilePath thumbnail_name = GetFaviconsFileName();
608 base::FilePath archived_name = GetArchivedFileName();
609
610 // Delete the old index database files which are no longer used.
611 DeleteFTSIndexDatabases();
612
613 // History database.
614 db_.reset(new HistoryDatabase());
615
616 // Unretained to avoid a ref loop with db_.
617 db_->set_error_callback(
618 base::Bind(&HistoryBackend::DatabaseErrorCallback,
619 base::Unretained(this)));
620
621 sql::InitStatus status = db_->Init(history_name);
622 switch (status) {
623 case sql::INIT_OK:
624 break;
625 case sql::INIT_FAILURE: {
626 // A NULL db_ will cause all calls on this object to notice this error
627 // and to not continue. If the error callback scheduled killing the
628 // database, the task it posted has not executed yet. Try killing the
629 // database now before we close it.
630 bool kill_db = scheduled_kill_db_;
631 if (kill_db)
632 KillHistoryDatabase();
633 UMA_HISTOGRAM_BOOLEAN("History.AttemptedToFixProfileError", kill_db);
634 delegate_->NotifyProfileError(status);
635 db_.reset();
636 return;
637 }
638 default:
639 NOTREACHED();
640 }
641
642 // Fill the in-memory database and send it back to the history service on the
643 // main thread.
644 {
645 scoped_ptr<InMemoryHistoryBackend> mem_backend(new InMemoryHistoryBackend);
646 if (mem_backend->Init(history_name))
647 delegate_->SetInMemoryBackend(mem_backend.Pass());
648 }
649 db_->BeginExclusiveMode(); // Must be after the mem backend read the data.
650
651 // Thumbnail database.
652 // TODO(shess): "thumbnail database" these days only stores
653 // favicons. Thumbnails are stored in "top sites". Consider
654 // renaming "thumbnail" references to "favicons" or something of the
655 // sort.
656 thumbnail_db_.reset(new ThumbnailDatabase(history_client_));
657 if (thumbnail_db_->Init(thumbnail_name) != sql::INIT_OK) {
658 // Unlike the main database, we don't error out when the database is too
659 // new because this error is much less severe. Generally, this shouldn't
660 // happen since the thumbnail and main database versions should be in sync.
661 // We'll just continue without thumbnails & favicons in this case or any
662 // other error.
663 LOG(WARNING) << "Could not initialize the thumbnail database.";
664 thumbnail_db_.reset();
665 }
666
667 // Nuke any files corresponding to the legacy Archived History Database, which
668 // previously retained expired (> 3 months old) history entries, but, in the
669 // end, was not used for much, and consequently has been removed as of M37.
670 // TODO(engedy): Remove this code after the end of 2014.
671 sql::Connection::Delete(archived_name);
672
673 // Generate the history and thumbnail database metrics only after performing
674 // any migration work.
675 if (base::RandInt(1, 100) == 50) {
676 // Only do this computation sometimes since it can be expensive.
677 db_->ComputeDatabaseMetrics(history_name);
678 if (thumbnail_db_)
679 thumbnail_db_->ComputeDatabaseMetrics();
680 }
681
682 expirer_.SetDatabases(db_.get(), thumbnail_db_.get());
683
684 // Open the long-running transaction.
685 db_->BeginTransaction();
686 if (thumbnail_db_)
687 thumbnail_db_->BeginTransaction();
688
689 // Get the first item in our database.
690 db_->GetStartDate(&first_recorded_time_);
691
692 // Start expiring old stuff.
693 expirer_.StartExpiringOldStuff(TimeDelta::FromDays(kExpireDaysThreshold));
694
695 #if defined(OS_ANDROID)
696 if (thumbnail_db_) {
697 android_provider_backend_.reset(
698 new AndroidProviderBackend(GetAndroidCacheFileName(),
699 db_.get(),
700 thumbnail_db_.get(),
701 history_client_,
702 delegate_.get()));
703 }
704 #endif
705
706 LOCAL_HISTOGRAM_TIMES("History.InitTime", TimeTicks::Now() - beginning_time);
707 }
708
OnMemoryPressure(base::MemoryPressureListener::MemoryPressureLevel memory_pressure_level)709 void HistoryBackend::OnMemoryPressure(
710 base::MemoryPressureListener::MemoryPressureLevel memory_pressure_level) {
711 bool trim_aggressively = memory_pressure_level ==
712 base::MemoryPressureListener::MEMORY_PRESSURE_CRITICAL;
713 if (db_)
714 db_->TrimMemory(trim_aggressively);
715 if (thumbnail_db_)
716 thumbnail_db_->TrimMemory(trim_aggressively);
717 }
718
CloseAllDatabases()719 void HistoryBackend::CloseAllDatabases() {
720 if (db_) {
721 // Commit the long-running transaction.
722 db_->CommitTransaction();
723 db_.reset();
724 // Forget the first recorded time since the database is closed.
725 first_recorded_time_ = base::Time();
726 }
727 if (thumbnail_db_) {
728 thumbnail_db_->CommitTransaction();
729 thumbnail_db_.reset();
730 }
731 }
732
AddPageVisit(const GURL & url,Time time,VisitID referring_visit,ui::PageTransition transition,VisitSource visit_source)733 std::pair<URLID, VisitID> HistoryBackend::AddPageVisit(
734 const GURL& url,
735 Time time,
736 VisitID referring_visit,
737 ui::PageTransition transition,
738 VisitSource visit_source) {
739 // Top-level frame navigations are visible, everything else is hidden
740 bool new_hidden = !ui::PageTransitionIsMainFrame(transition);
741
742 // NOTE: This code must stay in sync with
743 // ExpireHistoryBackend::ExpireURLsForVisits().
744 // TODO(pkasting): http://b/1148304 We shouldn't be marking so many URLs as
745 // typed, which would eliminate the need for this code.
746 int typed_increment = 0;
747 ui::PageTransition transition_type =
748 ui::PageTransitionStripQualifier(transition);
749 if ((transition_type == ui::PAGE_TRANSITION_TYPED &&
750 !ui::PageTransitionIsRedirect(transition)) ||
751 transition_type == ui::PAGE_TRANSITION_KEYWORD_GENERATED)
752 typed_increment = 1;
753
754 #if defined(OS_ANDROID)
755 // Only count the page visit if it came from user browsing and only count it
756 // once when cycling through a redirect chain.
757 if (visit_source == SOURCE_BROWSED &&
758 (transition & ui::PAGE_TRANSITION_CHAIN_END) != 0) {
759 RecordTopPageVisitStats(url);
760 }
761 #endif
762
763 // See if this URL is already in the DB.
764 URLRow url_info(url);
765 URLID url_id = db_->GetRowForURL(url, &url_info);
766 if (url_id) {
767 // Update of an existing row.
768 if (ui::PageTransitionStripQualifier(transition) !=
769 ui::PAGE_TRANSITION_RELOAD)
770 url_info.set_visit_count(url_info.visit_count() + 1);
771 if (typed_increment)
772 url_info.set_typed_count(url_info.typed_count() + typed_increment);
773 if (url_info.last_visit() < time)
774 url_info.set_last_visit(time);
775
776 // Only allow un-hiding of pages, never hiding.
777 if (!new_hidden)
778 url_info.set_hidden(false);
779
780 db_->UpdateURLRow(url_id, url_info);
781 } else {
782 // Addition of a new row.
783 url_info.set_visit_count(1);
784 url_info.set_typed_count(typed_increment);
785 url_info.set_last_visit(time);
786 url_info.set_hidden(new_hidden);
787
788 url_id = db_->AddURL(url_info);
789 if (!url_id) {
790 NOTREACHED() << "Adding URL failed.";
791 return std::make_pair(0, 0);
792 }
793 url_info.id_ = url_id;
794 }
795
796 // Add the visit with the time to the database.
797 VisitRow visit_info(url_id, time, referring_visit, transition, 0);
798 VisitID visit_id = db_->AddVisit(&visit_info, visit_source);
799 NotifyVisitObservers(visit_info);
800
801 if (visit_info.visit_time < first_recorded_time_)
802 first_recorded_time_ = visit_info.visit_time;
803
804 // Broadcast a notification of the visit.
805 if (visit_id) {
806 if (typed_url_syncable_service_.get())
807 typed_url_syncable_service_->OnUrlVisited(transition, &url_info);
808
809 scoped_ptr<URLVisitedDetails> details(new URLVisitedDetails);
810 details->transition = transition;
811 details->row = url_info;
812 details->visit_time = time;
813 // TODO(meelapshah) Disabled due to potential PageCycler regression.
814 // Re-enable this.
815 // QueryRedirectsTo(url, &details->redirects);
816 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URL_VISITED,
817 details.PassAs<HistoryDetails>());
818 } else {
819 VLOG(0) << "Failed to build visit insert statement: "
820 << "url_id = " << url_id;
821 }
822
823 return std::make_pair(url_id, visit_id);
824 }
825
AddPagesWithDetails(const URLRows & urls,VisitSource visit_source)826 void HistoryBackend::AddPagesWithDetails(const URLRows& urls,
827 VisitSource visit_source) {
828 if (!db_)
829 return;
830
831 scoped_ptr<URLsModifiedDetails> modified(new URLsModifiedDetails);
832 for (URLRows::const_iterator i = urls.begin(); i != urls.end(); ++i) {
833 DCHECK(!i->last_visit().is_null());
834
835 // As of M37, we no longer maintain an archived database, ignore old visits.
836 if (IsExpiredVisitTime(i->last_visit()))
837 continue;
838
839 URLRow existing_url;
840 URLID url_id = db_->GetRowForURL(i->url(), &existing_url);
841 if (!url_id) {
842 // Add the page if it doesn't exist.
843 url_id = db_->AddURL(*i);
844 if (!url_id) {
845 NOTREACHED() << "Could not add row to DB";
846 return;
847 }
848
849 modified->changed_urls.push_back(*i);
850 modified->changed_urls.back().set_id(url_id); // i->id_ is likely 0.
851 }
852
853 // Sync code manages the visits itself.
854 if (visit_source != SOURCE_SYNCED) {
855 // Make up a visit to correspond to the last visit to the page.
856 VisitRow visit_info(url_id, i->last_visit(), 0,
857 ui::PageTransitionFromInt(
858 ui::PAGE_TRANSITION_LINK |
859 ui::PAGE_TRANSITION_CHAIN_START |
860 ui::PAGE_TRANSITION_CHAIN_END), 0);
861 if (!db_->AddVisit(&visit_info, visit_source)) {
862 NOTREACHED() << "Adding visit failed.";
863 return;
864 }
865 NotifyVisitObservers(visit_info);
866
867 if (visit_info.visit_time < first_recorded_time_)
868 first_recorded_time_ = visit_info.visit_time;
869 }
870 }
871
872 if (typed_url_syncable_service_.get())
873 typed_url_syncable_service_->OnUrlsModified(&modified->changed_urls);
874
875 // Broadcast a notification for typed URLs that have been modified. This
876 // will be picked up by the in-memory URL database on the main thread.
877 //
878 // TODO(brettw) bug 1140015: Add an "add page" notification so the history
879 // views can keep in sync.
880 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URLS_MODIFIED,
881 modified.PassAs<HistoryDetails>());
882
883 ScheduleCommit();
884 }
885
IsExpiredVisitTime(const base::Time & time)886 bool HistoryBackend::IsExpiredVisitTime(const base::Time& time) {
887 return time < expirer_.GetCurrentExpirationTime();
888 }
889
SetPageTitle(const GURL & url,const base::string16 & title)890 void HistoryBackend::SetPageTitle(const GURL& url,
891 const base::string16& title) {
892 if (!db_)
893 return;
894
895 // Search for recent redirects which should get the same title. We make a
896 // dummy list containing the exact URL visited if there are no redirects so
897 // the processing below can be the same.
898 history::RedirectList dummy_list;
899 history::RedirectList* redirects;
900 RedirectCache::iterator iter = recent_redirects_.Get(url);
901 if (iter != recent_redirects_.end()) {
902 redirects = &iter->second;
903
904 // This redirect chain should have the destination URL as the last item.
905 DCHECK(!redirects->empty());
906 DCHECK(redirects->back() == url);
907 } else {
908 // No redirect chain stored, make up one containing the URL we want so we
909 // can use the same logic below.
910 dummy_list.push_back(url);
911 redirects = &dummy_list;
912 }
913
914 scoped_ptr<URLsModifiedDetails> details(new URLsModifiedDetails);
915 for (size_t i = 0; i < redirects->size(); i++) {
916 URLRow row;
917 URLID row_id = db_->GetRowForURL(redirects->at(i), &row);
918 if (row_id && row.title() != title) {
919 row.set_title(title);
920 db_->UpdateURLRow(row_id, row);
921 details->changed_urls.push_back(row);
922 }
923 }
924
925 // Broadcast notifications for any URLs that have changed. This will
926 // update the in-memory database and the InMemoryURLIndex.
927 if (!details->changed_urls.empty()) {
928 if (typed_url_syncable_service_.get())
929 typed_url_syncable_service_->OnUrlsModified(&details->changed_urls);
930 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URLS_MODIFIED,
931 details.PassAs<HistoryDetails>());
932 ScheduleCommit();
933 }
934 }
935
AddPageNoVisitForBookmark(const GURL & url,const base::string16 & title)936 void HistoryBackend::AddPageNoVisitForBookmark(const GURL& url,
937 const base::string16& title) {
938 if (!db_)
939 return;
940
941 URLRow url_info(url);
942 URLID url_id = db_->GetRowForURL(url, &url_info);
943 if (url_id) {
944 // URL is already known, nothing to do.
945 return;
946 }
947
948 if (!title.empty()) {
949 url_info.set_title(title);
950 } else {
951 url_info.set_title(base::UTF8ToUTF16(url.spec()));
952 }
953
954 url_info.set_last_visit(Time::Now());
955 // Mark the page hidden. If the user types it in, it'll unhide.
956 url_info.set_hidden(true);
957
958 db_->AddURL(url_info);
959 }
960
IterateURLs(const scoped_refptr<visitedlink::VisitedLinkDelegate::URLEnumerator> & iterator)961 void HistoryBackend::IterateURLs(
962 const scoped_refptr<visitedlink::VisitedLinkDelegate::URLEnumerator>&
963 iterator) {
964 if (db_) {
965 HistoryDatabase::URLEnumerator e;
966 if (db_->InitURLEnumeratorForEverything(&e)) {
967 URLRow info;
968 while (e.GetNextURL(&info)) {
969 iterator->OnURL(info.url());
970 }
971 iterator->OnComplete(true); // Success.
972 return;
973 }
974 }
975 iterator->OnComplete(false); // Failure.
976 }
977
GetAllTypedURLs(URLRows * urls)978 bool HistoryBackend::GetAllTypedURLs(URLRows* urls) {
979 if (db_)
980 return db_->GetAllTypedUrls(urls);
981 return false;
982 }
983
GetVisitsForURL(URLID id,VisitVector * visits)984 bool HistoryBackend::GetVisitsForURL(URLID id, VisitVector* visits) {
985 if (db_)
986 return db_->GetVisitsForURL(id, visits);
987 return false;
988 }
989
GetMostRecentVisitsForURL(URLID id,int max_visits,VisitVector * visits)990 bool HistoryBackend::GetMostRecentVisitsForURL(URLID id,
991 int max_visits,
992 VisitVector* visits) {
993 if (db_)
994 return db_->GetMostRecentVisitsForURL(id, max_visits, visits);
995 return false;
996 }
997
UpdateURLs(const history::URLRows & urls)998 size_t HistoryBackend::UpdateURLs(const history::URLRows& urls) {
999 if (!db_)
1000 return 0;
1001
1002 scoped_ptr<URLsModifiedDetails> details(new URLsModifiedDetails);
1003 for (history::URLRows::const_iterator it = urls.begin(); it != urls.end();
1004 ++it) {
1005 DCHECK(it->id());
1006 if (db_->UpdateURLRow(it->id(), *it))
1007 details->changed_urls.push_back(*it);
1008 }
1009
1010 // Broadcast notifications for any URLs that have actually been changed. This
1011 // will update the in-memory database and the InMemoryURLIndex.
1012 size_t num_updated_records = details->changed_urls.size();
1013 if (num_updated_records) {
1014 if (typed_url_syncable_service_)
1015 typed_url_syncable_service_->OnUrlsModified(&details->changed_urls);
1016 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URLS_MODIFIED,
1017 details.PassAs<HistoryDetails>());
1018 ScheduleCommit();
1019 }
1020 return num_updated_records;
1021 }
1022
AddVisits(const GURL & url,const std::vector<VisitInfo> & visits,VisitSource visit_source)1023 bool HistoryBackend::AddVisits(const GURL& url,
1024 const std::vector<VisitInfo>& visits,
1025 VisitSource visit_source) {
1026 if (db_) {
1027 for (std::vector<VisitInfo>::const_iterator visit = visits.begin();
1028 visit != visits.end(); ++visit) {
1029 if (!AddPageVisit(
1030 url, visit->first, 0, visit->second, visit_source).first) {
1031 return false;
1032 }
1033 }
1034 ScheduleCommit();
1035 return true;
1036 }
1037 return false;
1038 }
1039
RemoveVisits(const VisitVector & visits)1040 bool HistoryBackend::RemoveVisits(const VisitVector& visits) {
1041 if (!db_)
1042 return false;
1043
1044 expirer_.ExpireVisits(visits);
1045 ScheduleCommit();
1046 return true;
1047 }
1048
GetVisitsSource(const VisitVector & visits,VisitSourceMap * sources)1049 bool HistoryBackend::GetVisitsSource(const VisitVector& visits,
1050 VisitSourceMap* sources) {
1051 if (!db_)
1052 return false;
1053
1054 db_->GetVisitsSource(visits, sources);
1055 return true;
1056 }
1057
GetURL(const GURL & url,history::URLRow * url_row)1058 bool HistoryBackend::GetURL(const GURL& url, history::URLRow* url_row) {
1059 if (db_)
1060 return db_->GetRowForURL(url, url_row) != 0;
1061 return false;
1062 }
1063
QueryURL(const GURL & url,bool want_visits,QueryURLResult * result)1064 void HistoryBackend::QueryURL(const GURL& url,
1065 bool want_visits,
1066 QueryURLResult* result) {
1067 DCHECK(result);
1068 result->success = db_ && db_->GetRowForURL(url, &result->row);
1069 // Optionally query the visits.
1070 if (result->success && want_visits)
1071 db_->GetVisitsForURL(result->row.id(), &result->visits);
1072 }
1073
GetTypedUrlSyncableService() const1074 TypedUrlSyncableService* HistoryBackend::GetTypedUrlSyncableService() const {
1075 return typed_url_syncable_service_.get();
1076 }
1077
1078 // Keyword visits --------------------------------------------------------------
1079
SetKeywordSearchTermsForURL(const GURL & url,KeywordID keyword_id,const base::string16 & term)1080 void HistoryBackend::SetKeywordSearchTermsForURL(const GURL& url,
1081 KeywordID keyword_id,
1082 const base::string16& term) {
1083 if (!db_)
1084 return;
1085
1086 // Get the ID for this URL.
1087 URLRow row;
1088 if (!db_->GetRowForURL(url, &row)) {
1089 // There is a small possibility the url was deleted before the keyword
1090 // was added. Ignore the request.
1091 return;
1092 }
1093
1094 db_->SetKeywordSearchTermsForURL(row.id(), keyword_id, term);
1095
1096 BroadcastNotifications(
1097 chrome::NOTIFICATION_HISTORY_KEYWORD_SEARCH_TERM_UPDATED,
1098 scoped_ptr<HistoryDetails>(
1099 new KeywordSearchUpdatedDetails(row, keyword_id, term)));
1100 ScheduleCommit();
1101 }
1102
DeleteAllSearchTermsForKeyword(KeywordID keyword_id)1103 void HistoryBackend::DeleteAllSearchTermsForKeyword(KeywordID keyword_id) {
1104 if (!db_)
1105 return;
1106
1107 db_->DeleteAllSearchTermsForKeyword(keyword_id);
1108 ScheduleCommit();
1109 }
1110
DeleteKeywordSearchTermForURL(const GURL & url)1111 void HistoryBackend::DeleteKeywordSearchTermForURL(const GURL& url) {
1112 if (!db_)
1113 return;
1114
1115 URLID url_id = db_->GetRowForURL(url, NULL);
1116 if (!url_id)
1117 return;
1118 db_->DeleteKeywordSearchTermForURL(url_id);
1119
1120 BroadcastNotifications(
1121 chrome::NOTIFICATION_HISTORY_KEYWORD_SEARCH_TERM_DELETED,
1122 scoped_ptr<HistoryDetails>(new KeywordSearchDeletedDetails(url_id)));
1123 ScheduleCommit();
1124 }
1125
DeleteMatchingURLsForKeyword(KeywordID keyword_id,const base::string16 & term)1126 void HistoryBackend::DeleteMatchingURLsForKeyword(KeywordID keyword_id,
1127 const base::string16& term) {
1128 if (!db_)
1129 return;
1130
1131 std::vector<KeywordSearchTermRow> rows;
1132 if (db_->GetKeywordSearchTermRows(term, &rows)) {
1133 std::vector<GURL> items_to_delete;
1134 URLRow row;
1135 for (std::vector<KeywordSearchTermRow>::iterator it = rows.begin();
1136 it != rows.end(); ++it) {
1137 if ((it->keyword_id == keyword_id) && db_->GetURLRow(it->url_id, &row))
1138 items_to_delete.push_back(row.url());
1139 }
1140 DeleteURLs(items_to_delete);
1141 }
1142 }
1143
1144 // Downloads -------------------------------------------------------------------
1145
GetNextDownloadId()1146 uint32 HistoryBackend::GetNextDownloadId() {
1147 return db_ ? db_->GetNextDownloadId() : content::DownloadItem::kInvalidId;
1148 }
1149
1150 // Get all the download entries from the database.
QueryDownloads(std::vector<DownloadRow> * rows)1151 void HistoryBackend::QueryDownloads(std::vector<DownloadRow>* rows) {
1152 if (db_)
1153 db_->QueryDownloads(rows);
1154 }
1155
1156 // Update a particular download entry.
UpdateDownload(const history::DownloadRow & data)1157 void HistoryBackend::UpdateDownload(const history::DownloadRow& data) {
1158 if (!db_)
1159 return;
1160 db_->UpdateDownload(data);
1161 ScheduleCommit();
1162 }
1163
CreateDownload(const history::DownloadRow & history_info)1164 bool HistoryBackend::CreateDownload(const history::DownloadRow& history_info) {
1165 if (!db_)
1166 return false;
1167 bool success = db_->CreateDownload(history_info);
1168 ScheduleCommit();
1169 return success;
1170 }
1171
RemoveDownloads(const std::set<uint32> & ids)1172 void HistoryBackend::RemoveDownloads(const std::set<uint32>& ids) {
1173 if (!db_)
1174 return;
1175 size_t downloads_count_before = db_->CountDownloads();
1176 base::TimeTicks started_removing = base::TimeTicks::Now();
1177 // HistoryBackend uses a long-running Transaction that is committed
1178 // periodically, so this loop doesn't actually hit the disk too hard.
1179 for (std::set<uint32>::const_iterator it = ids.begin();
1180 it != ids.end(); ++it) {
1181 db_->RemoveDownload(*it);
1182 }
1183 ScheduleCommit();
1184 base::TimeTicks finished_removing = base::TimeTicks::Now();
1185 size_t downloads_count_after = db_->CountDownloads();
1186
1187 DCHECK_LE(downloads_count_after, downloads_count_before);
1188 if (downloads_count_after > downloads_count_before)
1189 return;
1190 size_t num_downloads_deleted = downloads_count_before - downloads_count_after;
1191 UMA_HISTOGRAM_COUNTS("Download.DatabaseRemoveDownloadsCount",
1192 num_downloads_deleted);
1193 base::TimeDelta micros = (1000 * (finished_removing - started_removing));
1194 UMA_HISTOGRAM_TIMES("Download.DatabaseRemoveDownloadsTime", micros);
1195 if (num_downloads_deleted > 0) {
1196 UMA_HISTOGRAM_TIMES("Download.DatabaseRemoveDownloadsTimePerRecord",
1197 (1000 * micros) / num_downloads_deleted);
1198 }
1199 DCHECK_GE(ids.size(), num_downloads_deleted);
1200 if (ids.size() < num_downloads_deleted)
1201 return;
1202 UMA_HISTOGRAM_COUNTS("Download.DatabaseRemoveDownloadsCountNotRemoved",
1203 ids.size() - num_downloads_deleted);
1204 }
1205
QueryHistory(const base::string16 & text_query,const QueryOptions & options,QueryResults * query_results)1206 void HistoryBackend::QueryHistory(const base::string16& text_query,
1207 const QueryOptions& options,
1208 QueryResults* query_results) {
1209 DCHECK(query_results);
1210 base::TimeTicks beginning_time = base::TimeTicks::Now();
1211 if (db_) {
1212 if (text_query.empty()) {
1213 // Basic history query for the main database.
1214 QueryHistoryBasic(options, query_results);
1215 } else {
1216 // Text history query.
1217 QueryHistoryText(text_query, options, query_results);
1218 }
1219 }
1220 UMA_HISTOGRAM_TIMES("History.QueryHistory",
1221 TimeTicks::Now() - beginning_time);
1222 }
1223
1224 // Basic time-based querying of history.
QueryHistoryBasic(const QueryOptions & options,QueryResults * result)1225 void HistoryBackend::QueryHistoryBasic(const QueryOptions& options,
1226 QueryResults* result) {
1227 // First get all visits.
1228 VisitVector visits;
1229 bool has_more_results = db_->GetVisibleVisitsInRange(options, &visits);
1230 DCHECK(static_cast<int>(visits.size()) <= options.EffectiveMaxCount());
1231
1232 // Now add them and the URL rows to the results.
1233 URLResult url_result;
1234 for (size_t i = 0; i < visits.size(); i++) {
1235 const VisitRow visit = visits[i];
1236
1237 // Add a result row for this visit, get the URL info from the DB.
1238 if (!db_->GetURLRow(visit.url_id, &url_result)) {
1239 VLOG(0) << "Failed to get id " << visit.url_id
1240 << " from history.urls.";
1241 continue; // DB out of sync and URL doesn't exist, try to recover.
1242 }
1243
1244 if (!url_result.url().is_valid()) {
1245 VLOG(0) << "Got invalid URL from history.urls with id "
1246 << visit.url_id << ": "
1247 << url_result.url().possibly_invalid_spec();
1248 continue; // Don't report invalid URLs in case of corruption.
1249 }
1250
1251 url_result.set_visit_time(visit.visit_time);
1252
1253 // Set whether the visit was blocked for a managed user by looking at the
1254 // transition type.
1255 url_result.set_blocked_visit(
1256 (visit.transition & ui::PAGE_TRANSITION_BLOCKED) != 0);
1257
1258 // We don't set any of the query-specific parts of the URLResult, since
1259 // snippets and stuff don't apply to basic querying.
1260 result->AppendURLBySwapping(&url_result);
1261 }
1262
1263 if (!has_more_results && options.begin_time <= first_recorded_time_)
1264 result->set_reached_beginning(true);
1265 }
1266
1267 // Text-based querying of history.
QueryHistoryText(const base::string16 & text_query,const QueryOptions & options,QueryResults * result)1268 void HistoryBackend::QueryHistoryText(const base::string16& text_query,
1269 const QueryOptions& options,
1270 QueryResults* result) {
1271 URLRows text_matches;
1272 db_->GetTextMatches(text_query, &text_matches);
1273
1274 std::vector<URLResult> matching_visits;
1275 VisitVector visits; // Declare outside loop to prevent re-construction.
1276 for (size_t i = 0; i < text_matches.size(); i++) {
1277 const URLRow& text_match = text_matches[i];
1278 // Get all visits for given URL match.
1279 db_->GetVisibleVisitsForURL(text_match.id(), options, &visits);
1280 for (size_t j = 0; j < visits.size(); j++) {
1281 URLResult url_result(text_match);
1282 url_result.set_visit_time(visits[j].visit_time);
1283 matching_visits.push_back(url_result);
1284 }
1285 }
1286
1287 std::sort(matching_visits.begin(), matching_visits.end(),
1288 URLResult::CompareVisitTime);
1289
1290 size_t max_results = options.max_count == 0 ?
1291 std::numeric_limits<size_t>::max() : static_cast<int>(options.max_count);
1292 for (std::vector<URLResult>::iterator it = matching_visits.begin();
1293 it != matching_visits.end() && result->size() < max_results; ++it) {
1294 result->AppendURLBySwapping(&(*it));
1295 }
1296
1297 if (matching_visits.size() == result->size() &&
1298 options.begin_time <= first_recorded_time_)
1299 result->set_reached_beginning(true);
1300 }
1301
QueryRedirectsFrom(const GURL & from_url,RedirectList * redirects)1302 void HistoryBackend::QueryRedirectsFrom(const GURL& from_url,
1303 RedirectList* redirects) {
1304 redirects->clear();
1305 if (!db_)
1306 return;
1307
1308 URLID from_url_id = db_->GetRowForURL(from_url, NULL);
1309 VisitID cur_visit = db_->GetMostRecentVisitForURL(from_url_id, NULL);
1310 if (!cur_visit)
1311 return; // No visits for URL.
1312
1313 GetRedirectsFromSpecificVisit(cur_visit, redirects);
1314 }
1315
QueryRedirectsTo(const GURL & to_url,RedirectList * redirects)1316 void HistoryBackend::QueryRedirectsTo(const GURL& to_url,
1317 RedirectList* redirects) {
1318 redirects->clear();
1319 if (!db_)
1320 return;
1321
1322 URLID to_url_id = db_->GetRowForURL(to_url, NULL);
1323 VisitID cur_visit = db_->GetMostRecentVisitForURL(to_url_id, NULL);
1324 if (!cur_visit)
1325 return; // No visits for URL.
1326
1327 GetRedirectsToSpecificVisit(cur_visit, redirects);
1328 }
1329
GetVisibleVisitCountToHost(const GURL & url,VisibleVisitCountToHostResult * result)1330 void HistoryBackend::GetVisibleVisitCountToHost(
1331 const GURL& url,
1332 VisibleVisitCountToHostResult* result) {
1333 result->count = 0;
1334 result->success = db_.get() &&
1335 db_->GetVisibleVisitCountToHost(
1336 url, &result->count, &result->first_visit);
1337 }
1338
QueryMostVisitedURLs(int result_count,int days_back,MostVisitedURLList * result)1339 void HistoryBackend::QueryMostVisitedURLs(int result_count,
1340 int days_back,
1341 MostVisitedURLList* result) {
1342 if (!db_)
1343 return;
1344
1345 ScopedVector<PageUsageData> data;
1346 db_->QuerySegmentUsage(
1347 base::Time::Now() - base::TimeDelta::FromDays(days_back),
1348 result_count,
1349 &data.get());
1350
1351 for (size_t i = 0; i < data.size(); ++i) {
1352 PageUsageData* current_data = data[i];
1353 RedirectList redirects;
1354 QueryRedirectsFrom(current_data->GetURL(), &redirects);
1355 MostVisitedURL url = MakeMostVisitedURL(*current_data, redirects);
1356 result->push_back(url);
1357 }
1358 }
1359
QueryFilteredURLs(int result_count,const history::VisitFilter & filter,bool extended_info,history::FilteredURLList * result)1360 void HistoryBackend::QueryFilteredURLs(int result_count,
1361 const history::VisitFilter& filter,
1362 bool extended_info,
1363 history::FilteredURLList* result) {
1364 DCHECK(result);
1365 base::Time request_start = base::Time::Now();
1366
1367 result->clear();
1368 if (!db_) {
1369 // No History Database - return an empty list.
1370 return;
1371 }
1372
1373 VisitVector visits;
1374 db_->GetDirectVisitsDuringTimes(filter, 0, &visits);
1375
1376 std::map<URLID, double> score_map;
1377 for (size_t i = 0; i < visits.size(); ++i) {
1378 score_map[visits[i].url_id] += filter.GetVisitScore(visits[i]);
1379 }
1380
1381 // TODO(georgey): experiment with visit_segment database granularity (it is
1382 // currently 24 hours) to use it directly instead of using visits database,
1383 // which is considerably slower.
1384 ScopedVector<PageUsageData> data;
1385 data.reserve(score_map.size());
1386 for (std::map<URLID, double>::iterator it = score_map.begin();
1387 it != score_map.end(); ++it) {
1388 PageUsageData* pud = new PageUsageData(it->first);
1389 pud->SetScore(it->second);
1390 data.push_back(pud);
1391 }
1392
1393 // Limit to the top |result_count| results.
1394 std::sort(data.begin(), data.end(), PageUsageData::Predicate);
1395 if (result_count && implicit_cast<int>(data.size()) > result_count)
1396 data.resize(result_count);
1397
1398 for (size_t i = 0; i < data.size(); ++i) {
1399 URLRow info;
1400 if (db_->GetURLRow(data[i]->GetID(), &info)) {
1401 data[i]->SetURL(info.url());
1402 data[i]->SetTitle(info.title());
1403 }
1404 }
1405
1406 for (size_t i = 0; i < data.size(); ++i) {
1407 PageUsageData* current_data = data[i];
1408 FilteredURL url(*current_data);
1409
1410 if (extended_info) {
1411 VisitVector visits;
1412 db_->GetVisitsForURL(current_data->GetID(), &visits);
1413 if (visits.size() > 0) {
1414 url.extended_info.total_visits = visits.size();
1415 for (size_t i = 0; i < visits.size(); ++i) {
1416 url.extended_info.duration_opened +=
1417 visits[i].visit_duration.InSeconds();
1418 if (visits[i].visit_time > url.extended_info.last_visit_time) {
1419 url.extended_info.last_visit_time = visits[i].visit_time;
1420 }
1421 }
1422 // TODO(macourteau): implement the url.extended_info.visits stat.
1423 }
1424 }
1425 result->push_back(url);
1426 }
1427
1428 int delta_time = std::max(1, std::min(999,
1429 static_cast<int>((base::Time::Now() - request_start).InMilliseconds())));
1430 STATIC_HISTOGRAM_POINTER_BLOCK(
1431 "NewTabPage.SuggestedSitesLoadTime",
1432 Add(delta_time),
1433 base::LinearHistogram::FactoryGet("NewTabPage.SuggestedSitesLoadTime",
1434 1, 1000, 100, base::Histogram::kUmaTargetedHistogramFlag));
1435 }
1436
GetRedirectsFromSpecificVisit(VisitID cur_visit,history::RedirectList * redirects)1437 void HistoryBackend::GetRedirectsFromSpecificVisit(
1438 VisitID cur_visit, history::RedirectList* redirects) {
1439 // Follow any redirects from the given visit and add them to the list.
1440 // It *should* be impossible to get a circular chain here, but we check
1441 // just in case to avoid infinite loops.
1442 GURL cur_url;
1443 std::set<VisitID> visit_set;
1444 visit_set.insert(cur_visit);
1445 while (db_->GetRedirectFromVisit(cur_visit, &cur_visit, &cur_url)) {
1446 if (visit_set.find(cur_visit) != visit_set.end()) {
1447 NOTREACHED() << "Loop in visit chain, giving up";
1448 return;
1449 }
1450 visit_set.insert(cur_visit);
1451 redirects->push_back(cur_url);
1452 }
1453 }
1454
GetRedirectsToSpecificVisit(VisitID cur_visit,history::RedirectList * redirects)1455 void HistoryBackend::GetRedirectsToSpecificVisit(
1456 VisitID cur_visit,
1457 history::RedirectList* redirects) {
1458 // Follow redirects going to cur_visit. These are added to |redirects| in
1459 // the order they are found. If a redirect chain looks like A -> B -> C and
1460 // |cur_visit| = C, redirects will be {B, A} in that order.
1461 if (!db_)
1462 return;
1463
1464 GURL cur_url;
1465 std::set<VisitID> visit_set;
1466 visit_set.insert(cur_visit);
1467 while (db_->GetRedirectToVisit(cur_visit, &cur_visit, &cur_url)) {
1468 if (visit_set.find(cur_visit) != visit_set.end()) {
1469 NOTREACHED() << "Loop in visit chain, giving up";
1470 return;
1471 }
1472 visit_set.insert(cur_visit);
1473 redirects->push_back(cur_url);
1474 }
1475 }
1476
ScheduleAutocomplete(const base::Callback<void (history::HistoryBackend *,history::URLDatabase *)> & callback)1477 void HistoryBackend::ScheduleAutocomplete(const base::Callback<
1478 void(history::HistoryBackend*, history::URLDatabase*)>& callback) {
1479 callback.Run(this, db_.get());
1480 }
1481
DeleteFTSIndexDatabases()1482 void HistoryBackend::DeleteFTSIndexDatabases() {
1483 // Find files on disk matching the text databases file pattern so we can
1484 // quickly test for and delete them.
1485 base::FilePath::StringType filepattern =
1486 FILE_PATH_LITERAL("History Index *");
1487 base::FileEnumerator enumerator(
1488 history_dir_, false, base::FileEnumerator::FILES, filepattern);
1489 int num_databases_deleted = 0;
1490 base::FilePath current_file;
1491 while (!(current_file = enumerator.Next()).empty()) {
1492 if (sql::Connection::Delete(current_file))
1493 num_databases_deleted++;
1494 }
1495 UMA_HISTOGRAM_COUNTS("History.DeleteFTSIndexDatabases",
1496 num_databases_deleted);
1497 }
1498
GetFavicons(const std::vector<GURL> & icon_urls,int icon_types,const std::vector<int> & desired_sizes,std::vector<favicon_base::FaviconRawBitmapResult> * bitmap_results)1499 void HistoryBackend::GetFavicons(
1500 const std::vector<GURL>& icon_urls,
1501 int icon_types,
1502 const std::vector<int>& desired_sizes,
1503 std::vector<favicon_base::FaviconRawBitmapResult>* bitmap_results) {
1504 UpdateFaviconMappingsAndFetchImpl(NULL, icon_urls, icon_types, desired_sizes,
1505 bitmap_results);
1506 }
1507
GetLargestFaviconForURL(const GURL & page_url,const std::vector<int> & icon_types,int minimum_size_in_pixels,favicon_base::FaviconRawBitmapResult * favicon_bitmap_result)1508 void HistoryBackend::GetLargestFaviconForURL(
1509 const GURL& page_url,
1510 const std::vector<int>& icon_types,
1511 int minimum_size_in_pixels,
1512 favicon_base::FaviconRawBitmapResult* favicon_bitmap_result) {
1513 DCHECK(favicon_bitmap_result);
1514
1515 if (!db_ || !thumbnail_db_)
1516 return;
1517
1518 TimeTicks beginning_time = TimeTicks::Now();
1519
1520 std::vector<IconMapping> icon_mappings;
1521 if (!thumbnail_db_->GetIconMappingsForPageURL(page_url, &icon_mappings) ||
1522 icon_mappings.empty())
1523 return;
1524
1525 int required_icon_types = 0;
1526 for (std::vector<int>::const_iterator i = icon_types.begin();
1527 i != icon_types.end(); ++i) {
1528 required_icon_types |= *i;
1529 }
1530
1531 // Find the largest bitmap for each IconType placing in
1532 // |largest_favicon_bitmaps|.
1533 std::map<favicon_base::IconType, FaviconBitmap> largest_favicon_bitmaps;
1534 for (std::vector<IconMapping>::const_iterator i = icon_mappings.begin();
1535 i != icon_mappings.end(); ++i) {
1536 if (!(i->icon_type & required_icon_types))
1537 continue;
1538 std::vector<FaviconBitmapIDSize> bitmap_id_sizes;
1539 thumbnail_db_->GetFaviconBitmapIDSizes(i->icon_id, &bitmap_id_sizes);
1540 FaviconBitmap& largest = largest_favicon_bitmaps[i->icon_type];
1541 for (std::vector<FaviconBitmapIDSize>::const_iterator j =
1542 bitmap_id_sizes.begin(); j != bitmap_id_sizes.end(); ++j) {
1543 if (largest.bitmap_id == 0 ||
1544 (largest.pixel_size.width() < j->pixel_size.width() &&
1545 largest.pixel_size.height() < j->pixel_size.height())) {
1546 largest.icon_id = i->icon_id;
1547 largest.bitmap_id = j->bitmap_id;
1548 largest.pixel_size = j->pixel_size;
1549 }
1550 }
1551 }
1552 if (largest_favicon_bitmaps.empty())
1553 return;
1554
1555 // Find an icon which is larger than minimum_size_in_pixels in the order of
1556 // icon_types.
1557 FaviconBitmap largest_icon;
1558 for (std::vector<int>::const_iterator t = icon_types.begin();
1559 t != icon_types.end(); ++t) {
1560 for (std::map<favicon_base::IconType, FaviconBitmap>::const_iterator f =
1561 largest_favicon_bitmaps.begin();
1562 f != largest_favicon_bitmaps.end();
1563 ++f) {
1564 if (f->first & *t &&
1565 (largest_icon.bitmap_id == 0 ||
1566 (largest_icon.pixel_size.height() < f->second.pixel_size.height() &&
1567 largest_icon.pixel_size.width() < f->second.pixel_size.width()))) {
1568 largest_icon = f->second;
1569 }
1570 }
1571 if (largest_icon.pixel_size.width() > minimum_size_in_pixels &&
1572 largest_icon.pixel_size.height() > minimum_size_in_pixels)
1573 break;
1574 }
1575
1576 GURL icon_url;
1577 favicon_base::IconType icon_type;
1578 if (!thumbnail_db_->GetFaviconHeader(largest_icon.icon_id, &icon_url,
1579 &icon_type)) {
1580 return;
1581 }
1582
1583 base::Time last_updated;
1584 favicon_base::FaviconRawBitmapResult bitmap_result;
1585 bitmap_result.icon_url = icon_url;
1586 bitmap_result.icon_type = icon_type;
1587 if (!thumbnail_db_->GetFaviconBitmap(largest_icon.bitmap_id,
1588 &last_updated,
1589 &bitmap_result.bitmap_data,
1590 &bitmap_result.pixel_size)) {
1591 return;
1592 }
1593
1594 bitmap_result.expired = (Time::Now() - last_updated) >
1595 TimeDelta::FromDays(kFaviconRefetchDays);
1596 if (bitmap_result.is_valid())
1597 *favicon_bitmap_result = bitmap_result;
1598
1599 LOCAL_HISTOGRAM_TIMES("History.GetLargestFaviconForURL",
1600 TimeTicks::Now() - beginning_time);
1601 }
1602
GetFaviconsForURL(const GURL & page_url,int icon_types,const std::vector<int> & desired_sizes,std::vector<favicon_base::FaviconRawBitmapResult> * bitmap_results)1603 void HistoryBackend::GetFaviconsForURL(
1604 const GURL& page_url,
1605 int icon_types,
1606 const std::vector<int>& desired_sizes,
1607 std::vector<favicon_base::FaviconRawBitmapResult>* bitmap_results) {
1608 DCHECK(bitmap_results);
1609 GetFaviconsFromDB(page_url, icon_types, desired_sizes, bitmap_results);
1610 }
1611
GetFaviconForID(favicon_base::FaviconID favicon_id,int desired_size,std::vector<favicon_base::FaviconRawBitmapResult> * bitmap_results)1612 void HistoryBackend::GetFaviconForID(
1613 favicon_base::FaviconID favicon_id,
1614 int desired_size,
1615 std::vector<favicon_base::FaviconRawBitmapResult>* bitmap_results) {
1616 std::vector<favicon_base::FaviconID> favicon_ids;
1617 favicon_ids.push_back(favicon_id);
1618 std::vector<int> desired_sizes;
1619 desired_sizes.push_back(desired_size);
1620
1621 // Get results from DB.
1622 GetFaviconBitmapResultsForBestMatch(favicon_ids,
1623 desired_sizes,
1624 bitmap_results);
1625 }
1626
UpdateFaviconMappingsAndFetch(const GURL & page_url,const std::vector<GURL> & icon_urls,int icon_types,const std::vector<int> & desired_sizes,std::vector<favicon_base::FaviconRawBitmapResult> * bitmap_results)1627 void HistoryBackend::UpdateFaviconMappingsAndFetch(
1628 const GURL& page_url,
1629 const std::vector<GURL>& icon_urls,
1630 int icon_types,
1631 const std::vector<int>& desired_sizes,
1632 std::vector<favicon_base::FaviconRawBitmapResult>* bitmap_results) {
1633 UpdateFaviconMappingsAndFetchImpl(&page_url, icon_urls, icon_types,
1634 desired_sizes, bitmap_results);
1635 }
1636
MergeFavicon(const GURL & page_url,const GURL & icon_url,favicon_base::IconType icon_type,scoped_refptr<base::RefCountedMemory> bitmap_data,const gfx::Size & pixel_size)1637 void HistoryBackend::MergeFavicon(
1638 const GURL& page_url,
1639 const GURL& icon_url,
1640 favicon_base::IconType icon_type,
1641 scoped_refptr<base::RefCountedMemory> bitmap_data,
1642 const gfx::Size& pixel_size) {
1643 if (!thumbnail_db_ || !db_)
1644 return;
1645
1646 favicon_base::FaviconID favicon_id =
1647 thumbnail_db_->GetFaviconIDForFaviconURL(icon_url, icon_type, NULL);
1648
1649 if (!favicon_id) {
1650 // There is no favicon at |icon_url|, create it.
1651 favicon_id = thumbnail_db_->AddFavicon(icon_url, icon_type);
1652 }
1653
1654 std::vector<FaviconBitmapIDSize> bitmap_id_sizes;
1655 thumbnail_db_->GetFaviconBitmapIDSizes(favicon_id, &bitmap_id_sizes);
1656
1657 // If there is already a favicon bitmap of |pixel_size| at |icon_url|,
1658 // replace it.
1659 bool bitmap_identical = false;
1660 bool replaced_bitmap = false;
1661 for (size_t i = 0; i < bitmap_id_sizes.size(); ++i) {
1662 if (bitmap_id_sizes[i].pixel_size == pixel_size) {
1663 if (IsFaviconBitmapDataEqual(bitmap_id_sizes[i].bitmap_id, bitmap_data)) {
1664 thumbnail_db_->SetFaviconBitmapLastUpdateTime(
1665 bitmap_id_sizes[i].bitmap_id, base::Time::Now());
1666 bitmap_identical = true;
1667 } else {
1668 thumbnail_db_->SetFaviconBitmap(bitmap_id_sizes[i].bitmap_id,
1669 bitmap_data, base::Time::Now());
1670 replaced_bitmap = true;
1671 }
1672 break;
1673 }
1674 }
1675
1676 // Create a vector of the pixel sizes of the favicon bitmaps currently at
1677 // |icon_url|.
1678 std::vector<gfx::Size> favicon_sizes;
1679 for (size_t i = 0; i < bitmap_id_sizes.size(); ++i)
1680 favicon_sizes.push_back(bitmap_id_sizes[i].pixel_size);
1681
1682 if (!replaced_bitmap && !bitmap_identical) {
1683 // Set the preexisting favicon bitmaps as expired as the preexisting favicon
1684 // bitmaps are not consistent with the merged in data.
1685 thumbnail_db_->SetFaviconOutOfDate(favicon_id);
1686
1687 // Delete an arbitrary favicon bitmap to avoid going over the limit of
1688 // |kMaxFaviconBitmapsPerIconURL|.
1689 if (bitmap_id_sizes.size() >= kMaxFaviconBitmapsPerIconURL) {
1690 thumbnail_db_->DeleteFaviconBitmap(bitmap_id_sizes[0].bitmap_id);
1691 favicon_sizes.erase(favicon_sizes.begin());
1692 }
1693 thumbnail_db_->AddFaviconBitmap(favicon_id, bitmap_data, base::Time::Now(),
1694 pixel_size);
1695 favicon_sizes.push_back(pixel_size);
1696 }
1697
1698 // A site may have changed the favicons that it uses for |page_url|.
1699 // Example Scenario:
1700 // page_url = news.google.com
1701 // Initial State: www.google.com/favicon.ico 16x16, 32x32
1702 // MergeFavicon(news.google.com, news.google.com/news_specific.ico, ...,
1703 // ..., 16x16)
1704 //
1705 // Difficulties:
1706 // 1. Sync requires that a call to GetFaviconsForURL() returns the
1707 // |bitmap_data| passed into MergeFavicon().
1708 // - It is invalid for the 16x16 bitmap for www.google.com/favicon.ico to
1709 // stay mapped to news.google.com because it would be unclear which 16x16
1710 // bitmap should be returned via GetFaviconsForURL().
1711 //
1712 // 2. www.google.com/favicon.ico may be mapped to more than just
1713 // news.google.com (eg www.google.com).
1714 // - The 16x16 bitmap cannot be deleted from www.google.com/favicon.ico
1715 //
1716 // To resolve these problems, we copy all of the favicon bitmaps previously
1717 // mapped to news.google.com (|page_url|) and add them to the favicon at
1718 // news.google.com/news_specific.ico (|icon_url|). The favicon sizes for
1719 // |icon_url| are set to default to indicate that |icon_url| has incomplete
1720 // / incorrect data.
1721 // Difficulty 1: All but news.google.com/news_specific.ico are unmapped from
1722 // news.google.com
1723 // Difficulty 2: The favicon bitmaps for www.google.com/favicon.ico are not
1724 // modified.
1725
1726 std::vector<IconMapping> icon_mappings;
1727 thumbnail_db_->GetIconMappingsForPageURL(page_url, icon_type, &icon_mappings);
1728
1729 // Copy the favicon bitmaps mapped to |page_url| to the favicon at |icon_url|
1730 // till the limit of |kMaxFaviconBitmapsPerIconURL| is reached.
1731 for (size_t i = 0; i < icon_mappings.size(); ++i) {
1732 if (favicon_sizes.size() >= kMaxFaviconBitmapsPerIconURL)
1733 break;
1734
1735 if (icon_mappings[i].icon_url == icon_url)
1736 continue;
1737
1738 std::vector<FaviconBitmap> bitmaps_to_copy;
1739 thumbnail_db_->GetFaviconBitmaps(icon_mappings[i].icon_id,
1740 &bitmaps_to_copy);
1741 for (size_t j = 0; j < bitmaps_to_copy.size(); ++j) {
1742 // Do not add a favicon bitmap at a pixel size for which there is already
1743 // a favicon bitmap mapped to |icon_url|. The one there is more correct
1744 // and having multiple equally sized favicon bitmaps for |page_url| is
1745 // ambiguous in terms of GetFaviconsForURL().
1746 std::vector<gfx::Size>::iterator it = std::find(favicon_sizes.begin(),
1747 favicon_sizes.end(), bitmaps_to_copy[j].pixel_size);
1748 if (it != favicon_sizes.end())
1749 continue;
1750
1751 // Add the favicon bitmap as expired as it is not consistent with the
1752 // merged in data.
1753 thumbnail_db_->AddFaviconBitmap(favicon_id,
1754 bitmaps_to_copy[j].bitmap_data, base::Time(),
1755 bitmaps_to_copy[j].pixel_size);
1756 favicon_sizes.push_back(bitmaps_to_copy[j].pixel_size);
1757
1758 if (favicon_sizes.size() >= kMaxFaviconBitmapsPerIconURL)
1759 break;
1760 }
1761 }
1762
1763 // Update the favicon mappings such that only |icon_url| is mapped to
1764 // |page_url|.
1765 bool mapping_changed = false;
1766 if (icon_mappings.size() != 1 || icon_mappings[0].icon_url != icon_url) {
1767 std::vector<favicon_base::FaviconID> favicon_ids;
1768 favicon_ids.push_back(favicon_id);
1769 SetFaviconMappingsForPageAndRedirects(page_url, icon_type, favicon_ids);
1770 mapping_changed = true;
1771 }
1772
1773 if (mapping_changed || !bitmap_identical)
1774 SendFaviconChangedNotificationForPageAndRedirects(page_url);
1775 ScheduleCommit();
1776 }
1777
SetFavicons(const GURL & page_url,favicon_base::IconType icon_type,const GURL & icon_url,const std::vector<SkBitmap> & bitmaps)1778 void HistoryBackend::SetFavicons(const GURL& page_url,
1779 favicon_base::IconType icon_type,
1780 const GURL& icon_url,
1781 const std::vector<SkBitmap>& bitmaps) {
1782 if (!thumbnail_db_ || !db_)
1783 return;
1784
1785 DCHECK_GE(kMaxFaviconBitmapsPerIconURL, bitmaps.size());
1786
1787 // Track whether the method modifies or creates any favicon bitmaps, favicons
1788 // or icon mappings.
1789 bool data_modified = false;
1790
1791 favicon_base::FaviconID icon_id =
1792 thumbnail_db_->GetFaviconIDForFaviconURL(icon_url, icon_type, NULL);
1793
1794 if (!icon_id) {
1795 icon_id = thumbnail_db_->AddFavicon(icon_url, icon_type);
1796 data_modified = true;
1797 }
1798
1799 data_modified |= SetFaviconBitmaps(icon_id, bitmaps);
1800
1801 std::vector<favicon_base::FaviconID> icon_ids(1u, icon_id);
1802 data_modified |=
1803 SetFaviconMappingsForPageAndRedirects(page_url, icon_type, icon_ids);
1804
1805 if (data_modified) {
1806 // Send notification to the UI as an icon mapping, favicon, or favicon
1807 // bitmap was changed by this function.
1808 SendFaviconChangedNotificationForPageAndRedirects(page_url);
1809 }
1810 ScheduleCommit();
1811 }
1812
SetFaviconsOutOfDateForPage(const GURL & page_url)1813 void HistoryBackend::SetFaviconsOutOfDateForPage(const GURL& page_url) {
1814 std::vector<IconMapping> icon_mappings;
1815
1816 if (!thumbnail_db_ ||
1817 !thumbnail_db_->GetIconMappingsForPageURL(page_url,
1818 &icon_mappings))
1819 return;
1820
1821 for (std::vector<IconMapping>::iterator m = icon_mappings.begin();
1822 m != icon_mappings.end(); ++m) {
1823 thumbnail_db_->SetFaviconOutOfDate(m->icon_id);
1824 }
1825 ScheduleCommit();
1826 }
1827
CloneFavicons(const GURL & old_page_url,const GURL & new_page_url)1828 void HistoryBackend::CloneFavicons(const GURL& old_page_url,
1829 const GURL& new_page_url) {
1830 if (!thumbnail_db_)
1831 return;
1832
1833 // Prevent cross-domain cloning.
1834 if (old_page_url.GetOrigin() != new_page_url.GetOrigin())
1835 return;
1836
1837 thumbnail_db_->CloneIconMappings(old_page_url, new_page_url);
1838 ScheduleCommit();
1839 }
1840
SetImportedFavicons(const std::vector<ImportedFaviconUsage> & favicon_usage)1841 void HistoryBackend::SetImportedFavicons(
1842 const std::vector<ImportedFaviconUsage>& favicon_usage) {
1843 if (!db_ || !thumbnail_db_)
1844 return;
1845
1846 Time now = Time::Now();
1847
1848 // Track all URLs that had their favicons set or updated.
1849 std::set<GURL> favicons_changed;
1850
1851 for (size_t i = 0; i < favicon_usage.size(); i++) {
1852 favicon_base::FaviconID favicon_id =
1853 thumbnail_db_->GetFaviconIDForFaviconURL(
1854 favicon_usage[i].favicon_url, favicon_base::FAVICON, NULL);
1855 if (!favicon_id) {
1856 // This favicon doesn't exist yet, so we create it using the given data.
1857 // TODO(pkotwicz): Pass in real pixel size.
1858 favicon_id = thumbnail_db_->AddFavicon(
1859 favicon_usage[i].favicon_url,
1860 favicon_base::FAVICON,
1861 new base::RefCountedBytes(favicon_usage[i].png_data),
1862 now,
1863 gfx::Size());
1864 }
1865
1866 // Save the mapping from all the URLs to the favicon.
1867 HistoryClient* history_client = GetHistoryClient();
1868 for (std::set<GURL>::const_iterator url = favicon_usage[i].urls.begin();
1869 url != favicon_usage[i].urls.end(); ++url) {
1870 URLRow url_row;
1871 if (!db_->GetRowForURL(*url, &url_row)) {
1872 // If the URL is present as a bookmark, add the url in history to
1873 // save the favicon mapping. This will match with what history db does
1874 // for regular bookmarked URLs with favicons - when history db is
1875 // cleaned, we keep an entry in the db with 0 visits as long as that
1876 // url is bookmarked.
1877 if (history_client && history_client->IsBookmarked(*url)) {
1878 URLRow url_info(*url);
1879 url_info.set_visit_count(0);
1880 url_info.set_typed_count(0);
1881 url_info.set_last_visit(base::Time());
1882 url_info.set_hidden(false);
1883 db_->AddURL(url_info);
1884 thumbnail_db_->AddIconMapping(*url, favicon_id);
1885 favicons_changed.insert(*url);
1886 }
1887 } else {
1888 if (!thumbnail_db_->GetIconMappingsForPageURL(
1889 *url, favicon_base::FAVICON, NULL)) {
1890 // URL is present in history, update the favicon *only* if it is not
1891 // set already.
1892 thumbnail_db_->AddIconMapping(*url, favicon_id);
1893 favicons_changed.insert(*url);
1894 }
1895 }
1896 }
1897 }
1898
1899 if (!favicons_changed.empty() && delegate_) {
1900 // Send the notification about the changed favicon URLs.
1901 delegate_->NotifyFaviconChanged(favicons_changed);
1902 }
1903 }
1904
UpdateFaviconMappingsAndFetchImpl(const GURL * page_url,const std::vector<GURL> & icon_urls,int icon_types,const std::vector<int> & desired_sizes,std::vector<favicon_base::FaviconRawBitmapResult> * bitmap_results)1905 void HistoryBackend::UpdateFaviconMappingsAndFetchImpl(
1906 const GURL* page_url,
1907 const std::vector<GURL>& icon_urls,
1908 int icon_types,
1909 const std::vector<int>& desired_sizes,
1910 std::vector<favicon_base::FaviconRawBitmapResult>* bitmap_results) {
1911 // If |page_url| is specified, |icon_types| must be either a single icon
1912 // type or icon types which are equivalent.
1913 DCHECK(!page_url || icon_types == favicon_base::FAVICON ||
1914 icon_types == favicon_base::TOUCH_ICON ||
1915 icon_types == favicon_base::TOUCH_PRECOMPOSED_ICON ||
1916 icon_types ==
1917 (favicon_base::TOUCH_ICON | favicon_base::TOUCH_PRECOMPOSED_ICON));
1918 bitmap_results->clear();
1919
1920 if (!thumbnail_db_) {
1921 return;
1922 }
1923
1924 std::vector<favicon_base::FaviconID> favicon_ids;
1925
1926 // The icon type for which the mappings will the updated and data will be
1927 // returned.
1928 favicon_base::IconType selected_icon_type = favicon_base::INVALID_ICON;
1929
1930 for (size_t i = 0; i < icon_urls.size(); ++i) {
1931 const GURL& icon_url = icon_urls[i];
1932 favicon_base::IconType icon_type_out;
1933 const favicon_base::FaviconID favicon_id =
1934 thumbnail_db_->GetFaviconIDForFaviconURL(
1935 icon_url, icon_types, &icon_type_out);
1936
1937 if (favicon_id) {
1938 // Return and update icon mappings only for the largest icon type. As
1939 // |icon_urls| is not sorted in terms of icon type, clear |favicon_ids|
1940 // if an |icon_url| with a larger icon type is found.
1941 if (icon_type_out > selected_icon_type) {
1942 selected_icon_type = icon_type_out;
1943 favicon_ids.clear();
1944 }
1945 if (icon_type_out == selected_icon_type)
1946 favicon_ids.push_back(favicon_id);
1947 }
1948 }
1949
1950 if (page_url && !favicon_ids.empty()) {
1951 bool mappings_updated =
1952 SetFaviconMappingsForPageAndRedirects(*page_url, selected_icon_type,
1953 favicon_ids);
1954 if (mappings_updated) {
1955 SendFaviconChangedNotificationForPageAndRedirects(*page_url);
1956 ScheduleCommit();
1957 }
1958 }
1959
1960 GetFaviconBitmapResultsForBestMatch(favicon_ids, desired_sizes,
1961 bitmap_results);
1962 }
1963
SetFaviconBitmaps(favicon_base::FaviconID icon_id,const std::vector<SkBitmap> & bitmaps)1964 bool HistoryBackend::SetFaviconBitmaps(favicon_base::FaviconID icon_id,
1965 const std::vector<SkBitmap>& bitmaps) {
1966 std::vector<FaviconBitmapIDSize> bitmap_id_sizes;
1967 thumbnail_db_->GetFaviconBitmapIDSizes(icon_id, &bitmap_id_sizes);
1968
1969 typedef std::pair<scoped_refptr<base::RefCountedBytes>, gfx::Size>
1970 PNGEncodedBitmap;
1971 std::vector<PNGEncodedBitmap> to_add;
1972 for (size_t i = 0; i < bitmaps.size(); ++i) {
1973 scoped_refptr<base::RefCountedBytes> bitmap_data(
1974 new base::RefCountedBytes);
1975 if (!gfx::PNGCodec::EncodeBGRASkBitmap(
1976 bitmaps[i], false, &bitmap_data->data())) {
1977 continue;
1978 }
1979 to_add.push_back(std::make_pair(
1980 bitmap_data, gfx::Size(bitmaps[i].width(), bitmaps[i].height())));
1981 }
1982
1983 bool favicon_bitmaps_changed = false;
1984 for (size_t i = 0; i < bitmap_id_sizes.size(); ++i) {
1985 const gfx::Size& pixel_size = bitmap_id_sizes[i].pixel_size;
1986 std::vector<PNGEncodedBitmap>::iterator match_it = to_add.end();
1987 for (std::vector<PNGEncodedBitmap>::iterator it = to_add.begin();
1988 it != to_add.end();
1989 ++it) {
1990 if (it->second == pixel_size) {
1991 match_it = it;
1992 break;
1993 }
1994 }
1995
1996 FaviconBitmapID bitmap_id = bitmap_id_sizes[i].bitmap_id;
1997 if (match_it == to_add.end()) {
1998 thumbnail_db_->DeleteFaviconBitmap(bitmap_id);
1999
2000 favicon_bitmaps_changed = true;
2001 } else {
2002 if (!favicon_bitmaps_changed &&
2003 IsFaviconBitmapDataEqual(bitmap_id, match_it->first)) {
2004 thumbnail_db_->SetFaviconBitmapLastUpdateTime(
2005 bitmap_id, base::Time::Now());
2006 } else {
2007 thumbnail_db_->SetFaviconBitmap(bitmap_id, match_it->first,
2008 base::Time::Now());
2009 favicon_bitmaps_changed = true;
2010 }
2011 to_add.erase(match_it);
2012 }
2013 }
2014
2015 for (size_t i = 0; i < to_add.size(); ++i) {
2016 thumbnail_db_->AddFaviconBitmap(icon_id, to_add[i].first,
2017 base::Time::Now(), to_add[i].second);
2018
2019 favicon_bitmaps_changed = true;
2020 }
2021 return favicon_bitmaps_changed;
2022 }
2023
IsFaviconBitmapDataEqual(FaviconBitmapID bitmap_id,const scoped_refptr<base::RefCountedMemory> & new_bitmap_data)2024 bool HistoryBackend::IsFaviconBitmapDataEqual(
2025 FaviconBitmapID bitmap_id,
2026 const scoped_refptr<base::RefCountedMemory>& new_bitmap_data) {
2027 if (!new_bitmap_data.get())
2028 return false;
2029
2030 scoped_refptr<base::RefCountedMemory> original_bitmap_data;
2031 thumbnail_db_->GetFaviconBitmap(bitmap_id,
2032 NULL,
2033 &original_bitmap_data,
2034 NULL);
2035 return new_bitmap_data->Equals(original_bitmap_data);
2036 }
2037
GetFaviconsFromDB(const GURL & page_url,int icon_types,const std::vector<int> & desired_sizes,std::vector<favicon_base::FaviconRawBitmapResult> * favicon_bitmap_results)2038 bool HistoryBackend::GetFaviconsFromDB(
2039 const GURL& page_url,
2040 int icon_types,
2041 const std::vector<int>& desired_sizes,
2042 std::vector<favicon_base::FaviconRawBitmapResult>* favicon_bitmap_results) {
2043 DCHECK(favicon_bitmap_results);
2044 favicon_bitmap_results->clear();
2045
2046 if (!db_ || !thumbnail_db_)
2047 return false;
2048
2049 // Time the query.
2050 TimeTicks beginning_time = TimeTicks::Now();
2051
2052 // Get FaviconIDs for |page_url| and one of |icon_types|.
2053 std::vector<IconMapping> icon_mappings;
2054 thumbnail_db_->GetIconMappingsForPageURL(page_url, icon_types,
2055 &icon_mappings);
2056 std::vector<favicon_base::FaviconID> favicon_ids;
2057 for (size_t i = 0; i < icon_mappings.size(); ++i)
2058 favicon_ids.push_back(icon_mappings[i].icon_id);
2059
2060 // Populate |favicon_bitmap_results| and |icon_url_sizes|.
2061 bool success = GetFaviconBitmapResultsForBestMatch(favicon_ids,
2062 desired_sizes, favicon_bitmap_results);
2063 UMA_HISTOGRAM_TIMES("History.GetFavIconFromDB", // historical name
2064 TimeTicks::Now() - beginning_time);
2065 return success && !favicon_bitmap_results->empty();
2066 }
2067
GetFaviconBitmapResultsForBestMatch(const std::vector<favicon_base::FaviconID> & candidate_favicon_ids,const std::vector<int> & desired_sizes,std::vector<favicon_base::FaviconRawBitmapResult> * favicon_bitmap_results)2068 bool HistoryBackend::GetFaviconBitmapResultsForBestMatch(
2069 const std::vector<favicon_base::FaviconID>& candidate_favicon_ids,
2070 const std::vector<int>& desired_sizes,
2071 std::vector<favicon_base::FaviconRawBitmapResult>* favicon_bitmap_results) {
2072 favicon_bitmap_results->clear();
2073
2074 if (candidate_favicon_ids.empty())
2075 return true;
2076
2077 // Find the FaviconID and the FaviconBitmapIDs which best match
2078 // |desired_size_in_dip| and |desired_scale_factors|.
2079 // TODO(pkotwicz): Select bitmap results from multiple favicons once
2080 // content::FaviconStatus supports multiple icon URLs.
2081 favicon_base::FaviconID best_favicon_id = 0;
2082 std::vector<FaviconBitmapID> best_bitmap_ids;
2083 float highest_score = kSelectFaviconFramesInvalidScore;
2084 for (size_t i = 0; i < candidate_favicon_ids.size(); ++i) {
2085 std::vector<FaviconBitmapIDSize> bitmap_id_sizes;
2086 thumbnail_db_->GetFaviconBitmapIDSizes(candidate_favicon_ids[i],
2087 &bitmap_id_sizes);
2088
2089 // Build vector of gfx::Size from |bitmap_id_sizes|.
2090 std::vector<gfx::Size> sizes;
2091 for (size_t j = 0; j < bitmap_id_sizes.size(); ++j)
2092 sizes.push_back(bitmap_id_sizes[j].pixel_size);
2093
2094 std::vector<size_t> candidate_bitmap_indices;
2095 float score = 0;
2096 SelectFaviconFrameIndices(sizes,
2097 desired_sizes,
2098 &candidate_bitmap_indices,
2099 &score);
2100 if (score > highest_score) {
2101 highest_score = score;
2102 best_favicon_id = candidate_favicon_ids[i],
2103 best_bitmap_ids.clear();
2104 for (size_t j = 0; j < candidate_bitmap_indices.size(); ++j) {
2105 size_t candidate_index = candidate_bitmap_indices[j];
2106 best_bitmap_ids.push_back(
2107 bitmap_id_sizes[candidate_index].bitmap_id);
2108 }
2109 }
2110 }
2111
2112 // Construct FaviconRawBitmapResults from |best_favicon_id| and
2113 // |best_bitmap_ids|.
2114 GURL icon_url;
2115 favicon_base::IconType icon_type;
2116 if (!thumbnail_db_->GetFaviconHeader(best_favicon_id, &icon_url,
2117 &icon_type)) {
2118 return false;
2119 }
2120
2121 for (size_t i = 0; i < best_bitmap_ids.size(); ++i) {
2122 base::Time last_updated;
2123 favicon_base::FaviconRawBitmapResult bitmap_result;
2124 bitmap_result.icon_url = icon_url;
2125 bitmap_result.icon_type = icon_type;
2126 if (!thumbnail_db_->GetFaviconBitmap(best_bitmap_ids[i],
2127 &last_updated,
2128 &bitmap_result.bitmap_data,
2129 &bitmap_result.pixel_size)) {
2130 return false;
2131 }
2132
2133 bitmap_result.expired = (Time::Now() - last_updated) >
2134 TimeDelta::FromDays(kFaviconRefetchDays);
2135 if (bitmap_result.is_valid())
2136 favicon_bitmap_results->push_back(bitmap_result);
2137 }
2138 return true;
2139 }
2140
SetFaviconMappingsForPageAndRedirects(const GURL & page_url,favicon_base::IconType icon_type,const std::vector<favicon_base::FaviconID> & icon_ids)2141 bool HistoryBackend::SetFaviconMappingsForPageAndRedirects(
2142 const GURL& page_url,
2143 favicon_base::IconType icon_type,
2144 const std::vector<favicon_base::FaviconID>& icon_ids) {
2145 if (!thumbnail_db_)
2146 return false;
2147
2148 // Find all the pages whose favicons we should set, we want to set it for
2149 // all the pages in the redirect chain if it redirected.
2150 history::RedirectList redirects;
2151 GetCachedRecentRedirects(page_url, &redirects);
2152
2153 bool mappings_changed = false;
2154
2155 // Save page <-> favicon associations.
2156 for (history::RedirectList::const_iterator i(redirects.begin());
2157 i != redirects.end(); ++i) {
2158 mappings_changed |= SetFaviconMappingsForPage(*i, icon_type, icon_ids);
2159 }
2160 return mappings_changed;
2161 }
2162
SetFaviconMappingsForPage(const GURL & page_url,favicon_base::IconType icon_type,const std::vector<favicon_base::FaviconID> & icon_ids)2163 bool HistoryBackend::SetFaviconMappingsForPage(
2164 const GURL& page_url,
2165 favicon_base::IconType icon_type,
2166 const std::vector<favicon_base::FaviconID>& icon_ids) {
2167 DCHECK_LE(icon_ids.size(), kMaxFaviconsPerPage);
2168 bool mappings_changed = false;
2169
2170 // Two icon types are considered 'equivalent' if one of the icon types is
2171 // TOUCH_ICON and the other is TOUCH_PRECOMPOSED_ICON.
2172 //
2173 // Sets the icon mappings from |page_url| for |icon_type| to the favicons
2174 // with |icon_ids|. Mappings for |page_url| to favicons of type |icon_type|
2175 // whose FaviconID is not in |icon_ids| are removed. All icon mappings for
2176 // |page_url| to favicons of a type equivalent to |icon_type| are removed.
2177 // Remove any favicons which are orphaned as a result of the removal of the
2178 // icon mappings.
2179
2180 std::vector<favicon_base::FaviconID> unmapped_icon_ids = icon_ids;
2181
2182 std::vector<IconMapping> icon_mappings;
2183 thumbnail_db_->GetIconMappingsForPageURL(page_url, &icon_mappings);
2184
2185 for (std::vector<IconMapping>::iterator m = icon_mappings.begin();
2186 m != icon_mappings.end(); ++m) {
2187 std::vector<favicon_base::FaviconID>::iterator icon_id_it = std::find(
2188 unmapped_icon_ids.begin(), unmapped_icon_ids.end(), m->icon_id);
2189
2190 // If the icon mapping already exists, avoid removing it and adding it back.
2191 if (icon_id_it != unmapped_icon_ids.end()) {
2192 unmapped_icon_ids.erase(icon_id_it);
2193 continue;
2194 }
2195
2196 if ((icon_type == favicon_base::TOUCH_ICON &&
2197 m->icon_type == favicon_base::TOUCH_PRECOMPOSED_ICON) ||
2198 (icon_type == favicon_base::TOUCH_PRECOMPOSED_ICON &&
2199 m->icon_type == favicon_base::TOUCH_ICON) ||
2200 (icon_type == m->icon_type)) {
2201 thumbnail_db_->DeleteIconMapping(m->mapping_id);
2202
2203 // Removing the icon mapping may have orphaned the associated favicon so
2204 // we must recheck it. This is not super fast, but this case will get
2205 // triggered rarely, since normally a page will always map to the same
2206 // favicon IDs. It will mostly happen for favicons we import.
2207 if (!thumbnail_db_->HasMappingFor(m->icon_id))
2208 thumbnail_db_->DeleteFavicon(m->icon_id);
2209 mappings_changed = true;
2210 }
2211 }
2212
2213 for (size_t i = 0; i < unmapped_icon_ids.size(); ++i) {
2214 thumbnail_db_->AddIconMapping(page_url, unmapped_icon_ids[i]);
2215 mappings_changed = true;
2216 }
2217 return mappings_changed;
2218 }
2219
GetCachedRecentRedirects(const GURL & page_url,history::RedirectList * redirect_list)2220 void HistoryBackend::GetCachedRecentRedirects(
2221 const GURL& page_url,
2222 history::RedirectList* redirect_list) {
2223 RedirectCache::iterator iter = recent_redirects_.Get(page_url);
2224 if (iter != recent_redirects_.end()) {
2225 *redirect_list = iter->second;
2226
2227 // The redirect chain should have the destination URL as the last item.
2228 DCHECK(!redirect_list->empty());
2229 DCHECK(redirect_list->back() == page_url);
2230 } else {
2231 // No known redirects, construct mock redirect chain containing |page_url|.
2232 redirect_list->push_back(page_url);
2233 }
2234 }
2235
SendFaviconChangedNotificationForPageAndRedirects(const GURL & page_url)2236 void HistoryBackend::SendFaviconChangedNotificationForPageAndRedirects(
2237 const GURL& page_url) {
2238 history::RedirectList redirect_list;
2239 GetCachedRecentRedirects(page_url, &redirect_list);
2240 if (!redirect_list.empty() && delegate_) {
2241 std::set<GURL> favicons_changed(redirect_list.begin(), redirect_list.end());
2242 delegate_->NotifyFaviconChanged(favicons_changed);
2243 }
2244 }
2245
Commit()2246 void HistoryBackend::Commit() {
2247 if (!db_)
2248 return;
2249
2250 // Note that a commit may not actually have been scheduled if a caller
2251 // explicitly calls this instead of using ScheduleCommit. Likewise, we
2252 // may reset the flag written by a pending commit. But this is OK! It
2253 // will merely cause extra commits (which is kind of the idea). We
2254 // could optimize more for this case (we may get two extra commits in
2255 // some cases) but it hasn't been important yet.
2256 CancelScheduledCommit();
2257
2258 db_->CommitTransaction();
2259 DCHECK(db_->transaction_nesting() == 0) << "Somebody left a transaction open";
2260 db_->BeginTransaction();
2261
2262 if (thumbnail_db_) {
2263 thumbnail_db_->CommitTransaction();
2264 DCHECK(thumbnail_db_->transaction_nesting() == 0) <<
2265 "Somebody left a transaction open";
2266 thumbnail_db_->BeginTransaction();
2267 }
2268 }
2269
ScheduleCommit()2270 void HistoryBackend::ScheduleCommit() {
2271 if (scheduled_commit_.get())
2272 return;
2273 scheduled_commit_ = new CommitLaterTask(this);
2274 base::MessageLoop::current()->PostDelayedTask(
2275 FROM_HERE,
2276 base::Bind(&CommitLaterTask::RunCommit, scheduled_commit_.get()),
2277 base::TimeDelta::FromSeconds(kCommitIntervalSeconds));
2278 }
2279
CancelScheduledCommit()2280 void HistoryBackend::CancelScheduledCommit() {
2281 if (scheduled_commit_.get()) {
2282 scheduled_commit_->Cancel();
2283 scheduled_commit_ = NULL;
2284 }
2285 }
2286
ProcessDBTaskImpl()2287 void HistoryBackend::ProcessDBTaskImpl() {
2288 if (!db_) {
2289 // db went away, release all the refs.
2290 STLDeleteContainerPointers(queued_history_db_tasks_.begin(),
2291 queued_history_db_tasks_.end());
2292 queued_history_db_tasks_.clear();
2293 return;
2294 }
2295
2296 // Remove any canceled tasks.
2297 while (!queued_history_db_tasks_.empty()) {
2298 QueuedHistoryDBTask* task = queued_history_db_tasks_.front();
2299 if (!task->is_canceled())
2300 break;
2301
2302 delete task;
2303 queued_history_db_tasks_.pop_front();
2304 }
2305 if (queued_history_db_tasks_.empty())
2306 return;
2307
2308 // Run the first task.
2309 scoped_ptr<QueuedHistoryDBTask> task(queued_history_db_tasks_.front());
2310 queued_history_db_tasks_.pop_front();
2311 if (task->Run(this, db_.get())) {
2312 // The task is done, notify the callback.
2313 task->DoneRun();
2314 } else {
2315 // The task wants to run some more. Schedule it at the end of the current
2316 // tasks, and process it after an invoke later.
2317 queued_history_db_tasks_.push_back(task.release());
2318 base::MessageLoop::current()->PostTask(
2319 FROM_HERE, base::Bind(&HistoryBackend::ProcessDBTaskImpl, this));
2320 }
2321 }
2322
2323 ////////////////////////////////////////////////////////////////////////////////
2324 //
2325 // Generic operations
2326 //
2327 ////////////////////////////////////////////////////////////////////////////////
2328
DeleteURLs(const std::vector<GURL> & urls)2329 void HistoryBackend::DeleteURLs(const std::vector<GURL>& urls) {
2330 expirer_.DeleteURLs(urls);
2331
2332 db_->GetStartDate(&first_recorded_time_);
2333 // Force a commit, if the user is deleting something for privacy reasons, we
2334 // want to get it on disk ASAP.
2335 Commit();
2336 }
2337
DeleteURL(const GURL & url)2338 void HistoryBackend::DeleteURL(const GURL& url) {
2339 expirer_.DeleteURL(url);
2340
2341 db_->GetStartDate(&first_recorded_time_);
2342 // Force a commit, if the user is deleting something for privacy reasons, we
2343 // want to get it on disk ASAP.
2344 Commit();
2345 }
2346
ExpireHistoryBetween(const std::set<GURL> & restrict_urls,Time begin_time,Time end_time)2347 void HistoryBackend::ExpireHistoryBetween(
2348 const std::set<GURL>& restrict_urls,
2349 Time begin_time,
2350 Time end_time) {
2351 if (!db_)
2352 return;
2353
2354 if (begin_time.is_null() && (end_time.is_null() || end_time.is_max()) &&
2355 restrict_urls.empty()) {
2356 // Special case deleting all history so it can be faster and to reduce the
2357 // possibility of an information leak.
2358 DeleteAllHistory();
2359 } else {
2360 // Clearing parts of history, have the expirer do the depend
2361 expirer_.ExpireHistoryBetween(restrict_urls, begin_time, end_time);
2362
2363 // Force a commit, if the user is deleting something for privacy reasons,
2364 // we want to get it on disk ASAP.
2365 Commit();
2366 }
2367
2368 if (begin_time <= first_recorded_time_)
2369 db_->GetStartDate(&first_recorded_time_);
2370 }
2371
ExpireHistoryForTimes(const std::set<base::Time> & times,base::Time begin_time,base::Time end_time)2372 void HistoryBackend::ExpireHistoryForTimes(
2373 const std::set<base::Time>& times,
2374 base::Time begin_time, base::Time end_time) {
2375 if (times.empty() || !db_)
2376 return;
2377
2378 DCHECK(*times.begin() >= begin_time)
2379 << "Min time is before begin time: "
2380 << times.begin()->ToJsTime() << " v.s. " << begin_time.ToJsTime();
2381 DCHECK(*times.rbegin() < end_time)
2382 << "Max time is after end time: "
2383 << times.rbegin()->ToJsTime() << " v.s. " << end_time.ToJsTime();
2384
2385 history::QueryOptions options;
2386 options.begin_time = begin_time;
2387 options.end_time = end_time;
2388 options.duplicate_policy = QueryOptions::KEEP_ALL_DUPLICATES;
2389 QueryResults results;
2390 QueryHistoryBasic(options, &results);
2391
2392 // 1st pass: find URLs that are visited at one of |times|.
2393 std::set<GURL> urls;
2394 for (size_t i = 0; i < results.size(); ++i) {
2395 if (times.count(results[i].visit_time()) > 0)
2396 urls.insert(results[i].url());
2397 }
2398 if (urls.empty())
2399 return;
2400
2401 // 2nd pass: collect all visit times of those URLs.
2402 std::vector<base::Time> times_to_expire;
2403 for (size_t i = 0; i < results.size(); ++i) {
2404 if (urls.count(results[i].url()))
2405 times_to_expire.push_back(results[i].visit_time());
2406 }
2407
2408 // Put the times in reverse chronological order and remove
2409 // duplicates (for expirer_.ExpireHistoryForTimes()).
2410 std::sort(times_to_expire.begin(), times_to_expire.end(),
2411 std::greater<base::Time>());
2412 times_to_expire.erase(
2413 std::unique(times_to_expire.begin(), times_to_expire.end()),
2414 times_to_expire.end());
2415
2416 // Expires by times and commit.
2417 DCHECK(!times_to_expire.empty());
2418 expirer_.ExpireHistoryForTimes(times_to_expire);
2419 Commit();
2420
2421 DCHECK(times_to_expire.back() >= first_recorded_time_);
2422 // Update |first_recorded_time_| if we expired it.
2423 if (times_to_expire.back() == first_recorded_time_)
2424 db_->GetStartDate(&first_recorded_time_);
2425 }
2426
ExpireHistory(const std::vector<history::ExpireHistoryArgs> & expire_list)2427 void HistoryBackend::ExpireHistory(
2428 const std::vector<history::ExpireHistoryArgs>& expire_list) {
2429 if (db_) {
2430 bool update_first_recorded_time = false;
2431
2432 for (std::vector<history::ExpireHistoryArgs>::const_iterator it =
2433 expire_list.begin(); it != expire_list.end(); ++it) {
2434 expirer_.ExpireHistoryBetween(it->urls, it->begin_time, it->end_time);
2435
2436 if (it->begin_time < first_recorded_time_)
2437 update_first_recorded_time = true;
2438 }
2439 Commit();
2440
2441 // Update |first_recorded_time_| if any deletion might have affected it.
2442 if (update_first_recorded_time)
2443 db_->GetStartDate(&first_recorded_time_);
2444 }
2445 }
2446
URLsNoLongerBookmarked(const std::set<GURL> & urls)2447 void HistoryBackend::URLsNoLongerBookmarked(const std::set<GURL>& urls) {
2448 if (!db_)
2449 return;
2450
2451 for (std::set<GURL>::const_iterator i = urls.begin(); i != urls.end(); ++i) {
2452 URLRow url_row;
2453 if (!db_->GetRowForURL(*i, &url_row))
2454 continue; // The URL isn't in the db; nothing to do.
2455
2456 VisitVector visits;
2457 db_->GetVisitsForURL(url_row.id(), &visits);
2458
2459 if (visits.empty())
2460 expirer_.DeleteURL(*i); // There are no more visits; nuke the URL.
2461 }
2462 }
2463
DatabaseErrorCallback(int error,sql::Statement * stmt)2464 void HistoryBackend::DatabaseErrorCallback(int error, sql::Statement* stmt) {
2465 if (!scheduled_kill_db_ && sql::IsErrorCatastrophic(error)) {
2466 scheduled_kill_db_ = true;
2467 // Don't just do the close/delete here, as we are being called by |db| and
2468 // that seems dangerous.
2469 // TODO(shess): Consider changing KillHistoryDatabase() to use
2470 // RazeAndClose(). Then it can be cleared immediately.
2471 base::MessageLoop::current()->PostTask(
2472 FROM_HERE,
2473 base::Bind(&HistoryBackend::KillHistoryDatabase, this));
2474 }
2475 }
2476
KillHistoryDatabase()2477 void HistoryBackend::KillHistoryDatabase() {
2478 scheduled_kill_db_ = false;
2479 if (!db_)
2480 return;
2481
2482 // Rollback transaction because Raze() cannot be called from within a
2483 // transaction.
2484 db_->RollbackTransaction();
2485 bool success = db_->Raze();
2486 UMA_HISTOGRAM_BOOLEAN("History.KillHistoryDatabaseResult", success);
2487
2488 #if defined(OS_ANDROID)
2489 // Release AndroidProviderBackend before other objects.
2490 android_provider_backend_.reset();
2491 #endif
2492
2493 // The expirer keeps tabs on the active databases. Tell it about the
2494 // databases which will be closed.
2495 expirer_.SetDatabases(NULL, NULL);
2496
2497 // Reopen a new transaction for |db_| for the sake of CloseAllDatabases().
2498 db_->BeginTransaction();
2499 CloseAllDatabases();
2500 }
2501
ProcessDBTask(scoped_ptr<HistoryDBTask> task,scoped_refptr<base::SingleThreadTaskRunner> origin_loop,const base::CancelableTaskTracker::IsCanceledCallback & is_canceled)2502 void HistoryBackend::ProcessDBTask(
2503 scoped_ptr<HistoryDBTask> task,
2504 scoped_refptr<base::SingleThreadTaskRunner> origin_loop,
2505 const base::CancelableTaskTracker::IsCanceledCallback& is_canceled) {
2506 bool scheduled = !queued_history_db_tasks_.empty();
2507 queued_history_db_tasks_.push_back(
2508 new QueuedHistoryDBTask(task.Pass(), origin_loop, is_canceled));
2509 if (!scheduled)
2510 ProcessDBTaskImpl();
2511 }
2512
BroadcastNotifications(int type,scoped_ptr<HistoryDetails> details)2513 void HistoryBackend::BroadcastNotifications(
2514 int type,
2515 scoped_ptr<HistoryDetails> details) {
2516 // |delegate_| may be NULL if |this| is in the process of closing (closed by
2517 // HistoryService -> HistoryBackend::Closing().
2518 if (delegate_)
2519 delegate_->BroadcastNotifications(type, details.Pass());
2520 }
2521
NotifySyncURLsModified(URLRows * rows)2522 void HistoryBackend::NotifySyncURLsModified(URLRows* rows) {
2523 if (typed_url_syncable_service_.get())
2524 typed_url_syncable_service_->OnUrlsModified(rows);
2525 }
2526
NotifySyncURLsDeleted(bool all_history,bool expired,URLRows * rows)2527 void HistoryBackend::NotifySyncURLsDeleted(bool all_history,
2528 bool expired,
2529 URLRows* rows) {
2530 if (typed_url_syncable_service_.get())
2531 typed_url_syncable_service_->OnUrlsDeleted(all_history, expired, rows);
2532 }
2533
2534 // Deleting --------------------------------------------------------------------
2535
DeleteAllHistory()2536 void HistoryBackend::DeleteAllHistory() {
2537 // Our approach to deleting all history is:
2538 // 1. Copy the bookmarks and their dependencies to new tables with temporary
2539 // names.
2540 // 2. Delete the original tables. Since tables can not share pages, we know
2541 // that any data we don't want to keep is now in an unused page.
2542 // 3. Renaming the temporary tables to match the original.
2543 // 4. Vacuuming the database to delete the unused pages.
2544 //
2545 // Since we are likely to have very few bookmarks and their dependencies
2546 // compared to all history, this is also much faster than just deleting from
2547 // the original tables directly.
2548
2549 // Get the bookmarked URLs.
2550 std::vector<URLAndTitle> starred_urls;
2551 HistoryClient* history_client = GetHistoryClient();
2552 if (history_client)
2553 history_client->GetBookmarks(&starred_urls);
2554
2555 URLRows kept_urls;
2556 for (size_t i = 0; i < starred_urls.size(); i++) {
2557 URLRow row;
2558 if (!db_->GetRowForURL(starred_urls[i].url, &row))
2559 continue;
2560
2561 // Clear the last visit time so when we write these rows they are "clean."
2562 row.set_last_visit(Time());
2563 row.set_visit_count(0);
2564 row.set_typed_count(0);
2565 kept_urls.push_back(row);
2566 }
2567
2568 // Clear thumbnail and favicon history. The favicons for the given URLs will
2569 // be kept.
2570 if (!ClearAllThumbnailHistory(kept_urls)) {
2571 LOG(ERROR) << "Thumbnail history could not be cleared";
2572 // We continue in this error case. If the user wants to delete their
2573 // history, we should delete as much as we can.
2574 }
2575
2576 // ClearAllMainHistory will change the IDs of the URLs in kept_urls.
2577 // Therefore, we clear the list afterwards to make sure nobody uses this
2578 // invalid data.
2579 if (!ClearAllMainHistory(kept_urls))
2580 LOG(ERROR) << "Main history could not be cleared";
2581 kept_urls.clear();
2582
2583 db_->GetStartDate(&first_recorded_time_);
2584
2585 // Send out the notification that history is cleared. The in-memory database
2586 // will pick this up and clear itself.
2587 scoped_ptr<URLsDeletedDetails> details(new URLsDeletedDetails);
2588 details->all_history = true;
2589 NotifySyncURLsDeleted(true, false, NULL);
2590 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URLS_DELETED,
2591 details.PassAs<HistoryDetails>());
2592 }
2593
ClearAllThumbnailHistory(const URLRows & kept_urls)2594 bool HistoryBackend::ClearAllThumbnailHistory(const URLRows& kept_urls) {
2595 if (!thumbnail_db_) {
2596 // When we have no reference to the thumbnail database, maybe there was an
2597 // error opening it. In this case, we just try to blow it away to try to
2598 // fix the error if it exists. This may fail, in which case either the
2599 // file doesn't exist or there's no more we can do.
2600 sql::Connection::Delete(GetFaviconsFileName());
2601
2602 // Older version of the database.
2603 sql::Connection::Delete(GetThumbnailFileName());
2604 return true;
2605 }
2606
2607 // Urls to retain mappings for.
2608 std::vector<GURL> urls_to_keep;
2609 for (URLRows::const_iterator i = kept_urls.begin();
2610 i != kept_urls.end(); ++i) {
2611 urls_to_keep.push_back(i->url());
2612 }
2613
2614 // Isolate from any long-running transaction.
2615 thumbnail_db_->CommitTransaction();
2616 thumbnail_db_->BeginTransaction();
2617
2618 // TODO(shess): If this fails, perhaps the database should be razed
2619 // or deleted.
2620 if (!thumbnail_db_->RetainDataForPageUrls(urls_to_keep)) {
2621 thumbnail_db_->RollbackTransaction();
2622 thumbnail_db_->BeginTransaction();
2623 return false;
2624 }
2625
2626 #if defined(OS_ANDROID)
2627 // TODO (michaelbai): Add the unit test once AndroidProviderBackend is
2628 // avaliable in HistoryBackend.
2629 db_->ClearAndroidURLRows();
2630 #endif
2631
2632 // Vacuum to remove all the pages associated with the dropped tables. There
2633 // must be no transaction open on the table when we do this. We assume that
2634 // our long-running transaction is open, so we complete it and start it again.
2635 DCHECK(thumbnail_db_->transaction_nesting() == 1);
2636 thumbnail_db_->CommitTransaction();
2637 thumbnail_db_->Vacuum();
2638 thumbnail_db_->BeginTransaction();
2639 return true;
2640 }
2641
ClearAllMainHistory(const URLRows & kept_urls)2642 bool HistoryBackend::ClearAllMainHistory(const URLRows& kept_urls) {
2643 // Create the duplicate URL table. We will copy the kept URLs into this.
2644 if (!db_->CreateTemporaryURLTable())
2645 return false;
2646
2647 // Insert the URLs into the temporary table.
2648 for (URLRows::const_iterator i = kept_urls.begin(); i != kept_urls.end();
2649 ++i) {
2650 db_->AddTemporaryURL(*i);
2651 }
2652
2653 // Replace the original URL table with the temporary one.
2654 if (!db_->CommitTemporaryURLTable())
2655 return false;
2656
2657 // Delete the old tables and recreate them empty.
2658 db_->RecreateAllTablesButURL();
2659
2660 // Vacuum to reclaim the space from the dropped tables. This must be done
2661 // when there is no transaction open, and we assume that our long-running
2662 // transaction is currently open.
2663 db_->CommitTransaction();
2664 db_->Vacuum();
2665 db_->BeginTransaction();
2666 db_->GetStartDate(&first_recorded_time_);
2667
2668 return true;
2669 }
2670
GetHistoryClient()2671 HistoryClient* HistoryBackend::GetHistoryClient() {
2672 if (history_client_)
2673 history_client_->BlockUntilBookmarksLoaded();
2674 return history_client_;
2675 }
2676
NotifyVisitObservers(const VisitRow & visit)2677 void HistoryBackend::NotifyVisitObservers(const VisitRow& visit) {
2678 BriefVisitInfo info;
2679 info.url_id = visit.url_id;
2680 info.time = visit.visit_time;
2681 info.transition = visit.transition;
2682 // If we don't have a delegate yet during setup or shutdown, we will drop
2683 // these notifications.
2684 if (delegate_)
2685 delegate_->NotifyVisitDBObserversOnAddVisit(info);
2686 }
2687
2688 #if defined(OS_ANDROID)
PopulateMostVisitedURLMap()2689 void HistoryBackend::PopulateMostVisitedURLMap() {
2690 MostVisitedURLList most_visited_urls;
2691 QueryMostVisitedURLs(
2692 kPageVisitStatsMaxTopSites, kSegmentDataRetention, &most_visited_urls);
2693
2694 DCHECK_LE(most_visited_urls.size(), kPageVisitStatsMaxTopSites);
2695 for (size_t i = 0; i < most_visited_urls.size(); ++i) {
2696 most_visited_urls_map_[most_visited_urls[i].url] = i;
2697 for (size_t j = 0; j < most_visited_urls[i].redirects.size(); ++j)
2698 most_visited_urls_map_[most_visited_urls[i].redirects[j]] = i;
2699 }
2700 }
2701
RecordTopPageVisitStats(const GURL & url)2702 void HistoryBackend::RecordTopPageVisitStats(const GURL& url) {
2703 int rank = kPageVisitStatsMaxTopSites;
2704 std::map<GURL, int>::const_iterator it = most_visited_urls_map_.find(url);
2705 if (it != most_visited_urls_map_.end())
2706 rank = (*it).second;
2707 UMA_HISTOGRAM_ENUMERATION("History.TopSitesVisitsByRank",
2708 rank, kPageVisitStatsMaxTopSites + 1);
2709 }
2710 #endif
2711
2712 } // namespace history
2713