history_backend.cc revision 68043e1e95eeb07d5cae7aca370b26518b0867d6
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/autocomplete/history_url_provider.h" 27#include "chrome/browser/bookmarks/bookmark_service.h" 28#include "chrome/browser/chrome_notification_types.h" 29#include "chrome/browser/favicon/favicon_changed_details.h" 30#include "chrome/browser/history/download_row.h" 31#include "chrome/browser/history/history_db_task.h" 32#include "chrome/browser/history/history_notifications.h" 33#include "chrome/browser/history/history_publisher.h" 34#include "chrome/browser/history/in_memory_history_backend.h" 35#include "chrome/browser/history/page_collector.h" 36#include "chrome/browser/history/page_usage_data.h" 37#include "chrome/browser/history/select_favicon_frames.h" 38#include "chrome/browser/history/top_sites.h" 39#include "chrome/browser/history/typed_url_syncable_service.h" 40#include "chrome/browser/history/visit_filter.h" 41#include "chrome/common/chrome_constants.h" 42#include "chrome/common/importer/imported_favicon_usage.h" 43#include "chrome/common/url_constants.h" 44#include "grit/chromium_strings.h" 45#include "grit/generated_resources.h" 46#include "net/base/registry_controlled_domains/registry_controlled_domain.h" 47#include "sql/error_delegate_util.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 54using base::Time; 55using base::TimeDelta; 56using base::TimeTicks; 57 58/* The HistoryBackend consists of a number of 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 ArchivedDatabase (stores history older than 3 months) 67 URLDatabase (stores a list of URLs) 68 DownloadDatabase (stores a list of downloads) 69 VisitDatabase (stores a list of visits for the URLs) 70 71 (this does not store visit segments as they expire after 3 mos.) 72 73 ExpireHistoryBackend (manages moving things from HistoryDatabase to 74 the ArchivedDatabase and deleting) 75*/ 76 77namespace history { 78 79// How long we keep segment data for in days. Currently 3 months. 80// This value needs to be greater or equal to 81// MostVisitedModel::kMostVisitedScope but we don't want to introduce a direct 82// dependency between MostVisitedModel and the history backend. 83const int kSegmentDataRetention = 90; 84 85// How long we'll wait to do a commit, so that things are batched together. 86const int kCommitIntervalSeconds = 10; 87 88// The amount of time before we re-fetch the favicon. 89const int kFaviconRefetchDays = 7; 90 91// The maximum number of items we'll allow in the redirect list before 92// deleting some. 93const int kMaxRedirectCount = 32; 94 95// The number of days old a history entry can be before it is considered "old" 96// and is archived. 97const int kArchiveDaysThreshold = 90; 98 99#if defined(OS_ANDROID) 100// The maximum number of top sites to track when recording top page visit stats. 101const size_t kPageVisitStatsMaxTopSites = 50; 102#endif 103 104// Converts from PageUsageData to MostVisitedURL. |redirects| is a 105// list of redirects for this URL. Empty list means no redirects. 106MostVisitedURL MakeMostVisitedURL(const PageUsageData& page_data, 107 const RedirectList& redirects) { 108 MostVisitedURL mv; 109 mv.url = page_data.GetURL(); 110 mv.title = page_data.GetTitle(); 111 if (redirects.empty()) { 112 // Redirects must contain at least the target url. 113 mv.redirects.push_back(mv.url); 114 } else { 115 mv.redirects = redirects; 116 if (mv.redirects[mv.redirects.size() - 1] != mv.url) { 117 // The last url must be the target url. 118 mv.redirects.push_back(mv.url); 119 } 120 } 121 return mv; 122} 123 124// This task is run on a timer so that commits happen at regular intervals 125// so they are batched together. The important thing about this class is that 126// it supports canceling of the task so the reference to the backend will be 127// freed. The problem is that when history is shutting down, there is likely 128// to be one of these commits still pending and holding a reference. 129// 130// The backend can call Cancel to have this task release the reference. The 131// task will still run (if we ever get to processing the event before 132// shutdown), but it will not do anything. 133// 134// Note that this is a refcounted object and is not a task in itself. It should 135// be assigned to a RunnableMethod. 136// 137// TODO(brettw): bug 1165182: This should be replaced with a 138// base::WeakPtrFactory which will handle everything automatically (like we do 139// in ExpireHistoryBackend). 140class CommitLaterTask : public base::RefCounted<CommitLaterTask> { 141 public: 142 explicit CommitLaterTask(HistoryBackend* history_backend) 143 : history_backend_(history_backend) { 144 } 145 146 // The backend will call this function if it is being destroyed so that we 147 // release our reference. 148 void Cancel() { 149 history_backend_ = NULL; 150 } 151 152 void RunCommit() { 153 if (history_backend_.get()) 154 history_backend_->Commit(); 155 } 156 157 private: 158 friend class base::RefCounted<CommitLaterTask>; 159 160 ~CommitLaterTask() {} 161 162 scoped_refptr<HistoryBackend> history_backend_; 163}; 164 165// HistoryBackend -------------------------------------------------------------- 166 167HistoryBackend::HistoryBackend(const base::FilePath& history_dir, 168 int id, 169 Delegate* delegate, 170 BookmarkService* bookmark_service) 171 : delegate_(delegate), 172 id_(id), 173 history_dir_(history_dir), 174 scheduled_kill_db_(false), 175 expirer_(this, bookmark_service), 176 recent_redirects_(kMaxRedirectCount), 177 backend_destroy_message_loop_(NULL), 178 segment_queried_(false), 179 bookmark_service_(bookmark_service) { 180} 181 182HistoryBackend::~HistoryBackend() { 183 DCHECK(!scheduled_commit_.get()) << "Deleting without cleanup"; 184 ReleaseDBTasks(); 185 186#if defined(OS_ANDROID) 187 // Release AndroidProviderBackend before other objects. 188 android_provider_backend_.reset(); 189#endif 190 191 // First close the databases before optionally running the "destroy" task. 192 CloseAllDatabases(); 193 194 if (!backend_destroy_task_.is_null()) { 195 // Notify an interested party (typically a unit test) that we're done. 196 DCHECK(backend_destroy_message_loop_); 197 backend_destroy_message_loop_->PostTask(FROM_HERE, backend_destroy_task_); 198 } 199 200#if defined(OS_ANDROID) 201 sql::Connection::Delete(GetAndroidCacheFileName()); 202#endif 203} 204 205void HistoryBackend::Init(const std::string& languages, bool force_fail) { 206 if (!force_fail) 207 InitImpl(languages); 208 delegate_->DBLoaded(id_); 209 typed_url_syncable_service_.reset(new TypedUrlSyncableService(this)); 210 memory_pressure_listener_.reset(new base::MemoryPressureListener( 211 base::Bind(&HistoryBackend::OnMemoryPressure, base::Unretained(this)))); 212#if defined(OS_ANDROID) 213 PopulateMostVisitedURLMap(); 214#endif 215} 216 217void HistoryBackend::SetOnBackendDestroyTask(base::MessageLoop* message_loop, 218 const base::Closure& task) { 219 if (!backend_destroy_task_.is_null()) 220 DLOG(WARNING) << "Setting more than one destroy task, overriding"; 221 backend_destroy_message_loop_ = message_loop; 222 backend_destroy_task_ = task; 223} 224 225void HistoryBackend::Closing() { 226 // Any scheduled commit will have a reference to us, we must make it 227 // release that reference before we can be destroyed. 228 CancelScheduledCommit(); 229 230 // Release our reference to the delegate, this reference will be keeping the 231 // history service alive. 232 delegate_.reset(); 233} 234 235void HistoryBackend::NotifyRenderProcessHostDestruction(const void* host) { 236 tracker_.NotifyRenderProcessHostDestruction(host); 237} 238 239base::FilePath HistoryBackend::GetThumbnailFileName() const { 240 return history_dir_.Append(chrome::kThumbnailsFilename); 241} 242 243base::FilePath HistoryBackend::GetFaviconsFileName() const { 244 return history_dir_.Append(chrome::kFaviconsFilename); 245} 246 247base::FilePath HistoryBackend::GetArchivedFileName() const { 248 return history_dir_.Append(chrome::kArchivedHistoryFilename); 249} 250 251#if defined(OS_ANDROID) 252base::FilePath HistoryBackend::GetAndroidCacheFileName() const { 253 return history_dir_.Append(chrome::kAndroidCacheFilename); 254} 255#endif 256 257SegmentID HistoryBackend::GetLastSegmentID(VisitID from_visit) { 258 // Set is used to detect referrer loops. Should not happen, but can 259 // if the database is corrupt. 260 std::set<VisitID> visit_set; 261 VisitID visit_id = from_visit; 262 while (visit_id) { 263 VisitRow row; 264 if (!db_->GetRowForVisit(visit_id, &row)) 265 return 0; 266 if (row.segment_id) 267 return row.segment_id; // Found a visit in this change with a segment. 268 269 // Check the referrer of this visit, if any. 270 visit_id = row.referring_visit; 271 272 if (visit_set.find(visit_id) != visit_set.end()) { 273 NOTREACHED() << "Loop in referer chain, giving up"; 274 break; 275 } 276 visit_set.insert(visit_id); 277 } 278 return 0; 279} 280 281SegmentID HistoryBackend::UpdateSegments( 282 const GURL& url, 283 VisitID from_visit, 284 VisitID visit_id, 285 content::PageTransition transition_type, 286 const Time ts) { 287 if (!db_) 288 return 0; 289 290 // We only consider main frames. 291 if (!content::PageTransitionIsMainFrame(transition_type)) 292 return 0; 293 294 SegmentID segment_id = 0; 295 content::PageTransition t = 296 content::PageTransitionStripQualifier(transition_type); 297 298 // Are we at the beginning of a new segment? 299 // Note that navigating to an existing entry (with back/forward) reuses the 300 // same transition type. We are not adding it as a new segment in that case 301 // because if this was the target of a redirect, we might end up with 302 // 2 entries for the same final URL. Ex: User types google.net, gets 303 // redirected to google.com. A segment is created for google.net. On 304 // google.com users navigates through a link, then press back. That last 305 // navigation is for the entry google.com transition typed. We end up adding 306 // a segment for that one as well. So we end up with google.net and google.com 307 // in the segment table, showing as 2 entries in the NTP. 308 // Note also that we should still be updating the visit count for that segment 309 // which we are not doing now. It should be addressed when 310 // http://crbug.com/96860 is fixed. 311 if ((t == content::PAGE_TRANSITION_TYPED || 312 t == content::PAGE_TRANSITION_AUTO_BOOKMARK) && 313 (transition_type & content::PAGE_TRANSITION_FORWARD_BACK) == 0) { 314 // If so, create or get the segment. 315 std::string segment_name = db_->ComputeSegmentName(url); 316 URLID url_id = db_->GetRowForURL(url, NULL); 317 if (!url_id) 318 return 0; 319 320 if (!(segment_id = db_->GetSegmentNamed(segment_name))) { 321 if (!(segment_id = db_->CreateSegment(url_id, segment_name))) { 322 NOTREACHED(); 323 return 0; 324 } 325 } else { 326 // Note: if we update an existing segment, we update the url used to 327 // represent that segment in order to minimize stale most visited 328 // images. 329 db_->UpdateSegmentRepresentationURL(segment_id, url_id); 330 } 331 } else { 332 // Note: it is possible there is no segment ID set for this visit chain. 333 // This can happen if the initial navigation wasn't AUTO_BOOKMARK or 334 // TYPED. (For example GENERATED). In this case this visit doesn't count 335 // toward any segment. 336 if (!(segment_id = GetLastSegmentID(from_visit))) 337 return 0; 338 } 339 340 // Set the segment in the visit. 341 if (!db_->SetSegmentID(visit_id, segment_id)) { 342 NOTREACHED(); 343 return 0; 344 } 345 346 // Finally, increase the counter for that segment / day. 347 if (!db_->IncreaseSegmentVisitCount(segment_id, ts, 1)) { 348 NOTREACHED(); 349 return 0; 350 } 351 return segment_id; 352} 353 354void HistoryBackend::UpdateWithPageEndTime(const void* host, 355 int32 page_id, 356 const GURL& url, 357 Time end_ts) { 358 // Will be filled with the URL ID and the visit ID of the last addition. 359 VisitID visit_id = tracker_.GetLastVisit(host, page_id, url); 360 UpdateVisitDuration(visit_id, end_ts); 361} 362 363void HistoryBackend::UpdateVisitDuration(VisitID visit_id, const Time end_ts) { 364 if (!db_) 365 return; 366 367 // Get the starting visit_time for visit_id. 368 VisitRow visit_row; 369 if (db_->GetRowForVisit(visit_id, &visit_row)) { 370 // We should never have a negative duration time even when time is skewed. 371 visit_row.visit_duration = end_ts > visit_row.visit_time ? 372 end_ts - visit_row.visit_time : TimeDelta::FromMicroseconds(0); 373 db_->UpdateVisitRow(visit_row); 374 } 375} 376 377void HistoryBackend::AddPage(const HistoryAddPageArgs& request) { 378 if (!db_) 379 return; 380 381 // Will be filled with the URL ID and the visit ID of the last addition. 382 std::pair<URLID, VisitID> last_ids(0, tracker_.GetLastVisit( 383 request.id_scope, request.page_id, request.referrer)); 384 385 VisitID from_visit_id = last_ids.second; 386 387 // If a redirect chain is given, we expect the last item in that chain to be 388 // the final URL. 389 DCHECK(request.redirects.empty() || 390 request.redirects.back() == request.url); 391 392 // If the user is adding older history, we need to make sure our times 393 // are correct. 394 if (request.time < first_recorded_time_) 395 first_recorded_time_ = request.time; 396 397 content::PageTransition request_transition = request.transition; 398 content::PageTransition stripped_transition = 399 content::PageTransitionStripQualifier(request_transition); 400 bool is_keyword_generated = 401 (stripped_transition == content::PAGE_TRANSITION_KEYWORD_GENERATED); 402 403 // If the user is navigating to a not-previously-typed intranet hostname, 404 // change the transition to TYPED so that the omnibox will learn that this is 405 // a known host. 406 bool has_redirects = request.redirects.size() > 1; 407 if (content::PageTransitionIsMainFrame(request_transition) && 408 (stripped_transition != content::PAGE_TRANSITION_TYPED) && 409 !is_keyword_generated) { 410 const GURL& origin_url(has_redirects ? 411 request.redirects[0] : request.url); 412 if (origin_url.SchemeIs(content::kHttpScheme) || 413 origin_url.SchemeIs(content::kHttpsScheme) || 414 origin_url.SchemeIs(chrome::kFtpScheme)) { 415 std::string host(origin_url.host()); 416 size_t registry_length = 417 net::registry_controlled_domains::GetRegistryLength( 418 host, 419 net::registry_controlled_domains::EXCLUDE_UNKNOWN_REGISTRIES, 420 net::registry_controlled_domains::EXCLUDE_PRIVATE_REGISTRIES); 421 if (registry_length == 0 && !db_->IsTypedHost(host)) { 422 stripped_transition = content::PAGE_TRANSITION_TYPED; 423 request_transition = 424 content::PageTransitionFromInt( 425 stripped_transition | 426 content::PageTransitionGetQualifier(request_transition)); 427 } 428 } 429 } 430 431 if (!has_redirects) { 432 // The single entry is both a chain start and end. 433 content::PageTransition t = content::PageTransitionFromInt( 434 request_transition | 435 content::PAGE_TRANSITION_CHAIN_START | 436 content::PAGE_TRANSITION_CHAIN_END); 437 438 // No redirect case (one element means just the page itself). 439 last_ids = AddPageVisit(request.url, request.time, 440 last_ids.second, t, request.visit_source); 441 442 // Update the segment for this visit. KEYWORD_GENERATED visits should not 443 // result in changing most visited, so we don't update segments (most 444 // visited db). 445 if (!is_keyword_generated) { 446 UpdateSegments(request.url, from_visit_id, last_ids.second, t, 447 request.time); 448 449 // Update the referrer's duration. 450 UpdateVisitDuration(from_visit_id, request.time); 451 } 452 } else { 453 // Redirect case. Add the redirect chain. 454 455 content::PageTransition redirect_info = 456 content::PAGE_TRANSITION_CHAIN_START; 457 458 RedirectList redirects = request.redirects; 459 if (redirects[0].SchemeIs(chrome::kAboutScheme)) { 460 // When the redirect source + referrer is "about" we skip it. This 461 // happens when a page opens a new frame/window to about:blank and then 462 // script sets the URL to somewhere else (used to hide the referrer). It 463 // would be nice to keep all these redirects properly but we don't ever 464 // see the initial about:blank load, so we don't know where the 465 // subsequent client redirect came from. 466 // 467 // In this case, we just don't bother hooking up the source of the 468 // redirects, so we remove it. 469 redirects.erase(redirects.begin()); 470 } else if (request_transition & content::PAGE_TRANSITION_CLIENT_REDIRECT) { 471 redirect_info = content::PAGE_TRANSITION_CLIENT_REDIRECT; 472 // The first entry in the redirect chain initiated a client redirect. 473 // We don't add this to the database since the referrer is already 474 // there, so we skip over it but change the transition type of the first 475 // transition to client redirect. 476 // 477 // The referrer is invalid when restoring a session that features an 478 // https tab that redirects to a different host or to http. In this 479 // case we don't need to reconnect the new redirect with the existing 480 // chain. 481 if (request.referrer.is_valid()) { 482 DCHECK(request.referrer == redirects[0]); 483 redirects.erase(redirects.begin()); 484 485 // If the navigation entry for this visit has replaced that for the 486 // first visit, remove the CHAIN_END marker from the first visit. This 487 // can be called a lot, for example, the page cycler, and most of the 488 // time we won't have changed anything. 489 VisitRow visit_row; 490 if (request.did_replace_entry && 491 db_->GetRowForVisit(last_ids.second, &visit_row) && 492 visit_row.transition & content::PAGE_TRANSITION_CHAIN_END) { 493 visit_row.transition = content::PageTransitionFromInt( 494 visit_row.transition & ~content::PAGE_TRANSITION_CHAIN_END); 495 db_->UpdateVisitRow(visit_row); 496 } 497 } 498 } 499 500 for (size_t redirect_index = 0; redirect_index < redirects.size(); 501 redirect_index++) { 502 content::PageTransition t = 503 content::PageTransitionFromInt(stripped_transition | redirect_info); 504 505 // If this is the last transition, add a CHAIN_END marker 506 if (redirect_index == (redirects.size() - 1)) { 507 t = content::PageTransitionFromInt( 508 t | content::PAGE_TRANSITION_CHAIN_END); 509 } 510 511 // Record all redirect visits with the same timestamp. We don't display 512 // them anyway, and if we ever decide to, we can reconstruct their order 513 // from the redirect chain. 514 last_ids = AddPageVisit(redirects[redirect_index], 515 request.time, last_ids.second, 516 t, request.visit_source); 517 if (t & content::PAGE_TRANSITION_CHAIN_START) { 518 // Update the segment for this visit. 519 UpdateSegments(redirects[redirect_index], 520 from_visit_id, last_ids.second, t, request.time); 521 522 // Update the visit_details for this visit. 523 UpdateVisitDuration(from_visit_id, request.time); 524 } 525 526 // Subsequent transitions in the redirect list must all be server 527 // redirects. 528 redirect_info = content::PAGE_TRANSITION_SERVER_REDIRECT; 529 } 530 531 // Last, save this redirect chain for later so we can set titles & favicons 532 // on the redirected pages properly. 533 recent_redirects_.Put(request.url, redirects); 534 } 535 536 // TODO(brettw) bug 1140015: Add an "add page" notification so the history 537 // views can keep in sync. 538 539 // Add the last visit to the tracker so we can get outgoing transitions. 540 // TODO(evanm): Due to http://b/1194536 we lose the referrers of a subframe 541 // navigation anyway, so last_visit_id is always zero for them. But adding 542 // them here confuses main frame history, so we skip them for now. 543 if (stripped_transition != content::PAGE_TRANSITION_AUTO_SUBFRAME && 544 stripped_transition != content::PAGE_TRANSITION_MANUAL_SUBFRAME && 545 !is_keyword_generated) { 546 tracker_.AddVisit(request.id_scope, request.page_id, request.url, 547 last_ids.second); 548 } 549 550 if (page_collector_) 551 page_collector_->AddPageURL(request.url, request.time); 552 553 ScheduleCommit(); 554} 555 556void HistoryBackend::InitImpl(const std::string& languages) { 557 DCHECK(!db_) << "Initializing HistoryBackend twice"; 558 // In the rare case where the db fails to initialize a dialog may get shown 559 // the blocks the caller, yet allows other messages through. For this reason 560 // we only set db_ to the created database if creation is successful. That 561 // way other methods won't do anything as db_ is still NULL. 562 563 TimeTicks beginning_time = TimeTicks::Now(); 564 565 // Compute the file names. 566 base::FilePath history_name = history_dir_.Append(chrome::kHistoryFilename); 567 base::FilePath thumbnail_name = GetFaviconsFileName(); 568 base::FilePath archived_name = GetArchivedFileName(); 569 570 // Delete the old index database files which are no longer used. 571 DeleteFTSIndexDatabases(); 572 573 // History database. 574 db_.reset(new HistoryDatabase()); 575 576 // Unretained to avoid a ref loop with db_. 577 db_->set_error_callback( 578 base::Bind(&HistoryBackend::DatabaseErrorCallback, 579 base::Unretained(this))); 580 581 sql::InitStatus status = db_->Init(history_name); 582 switch (status) { 583 case sql::INIT_OK: 584 break; 585 case sql::INIT_FAILURE: { 586 // A NULL db_ will cause all calls on this object to notice this error 587 // and to not continue. If the error callback scheduled killing the 588 // database, the task it posted has not executed yet. Try killing the 589 // database now before we close it. 590 bool kill_db = scheduled_kill_db_; 591 if (kill_db) 592 KillHistoryDatabase(); 593 UMA_HISTOGRAM_BOOLEAN("History.AttemptedToFixProfileError", kill_db); 594 delegate_->NotifyProfileError(id_, status); 595 db_.reset(); 596 return; 597 } 598 default: 599 NOTREACHED(); 600 } 601 602 // Fill the in-memory database and send it back to the history service on the 603 // main thread. 604 InMemoryHistoryBackend* mem_backend = new InMemoryHistoryBackend; 605 if (mem_backend->Init(history_name, db_.get())) 606 delegate_->SetInMemoryBackend(id_, mem_backend); // Takes ownership of 607 // pointer. 608 else 609 delete mem_backend; // Error case, run without the in-memory DB. 610 db_->BeginExclusiveMode(); // Must be after the mem backend read the data. 611 612 // Create the history publisher which needs to be passed on to the thumbnail 613 // database for publishing history. 614 // TODO(shess): HistoryPublisher is being deprecated. I am still 615 // trying to track down who depends on it, meanwhile talk to me 616 // before removing interactions with it. http://crbug.com/294306 617 history_publisher_.reset(new HistoryPublisher()); 618 if (!history_publisher_->Init()) { 619 // The init may fail when there are no indexers wanting our history. 620 // Hence no need to log the failure. 621 history_publisher_.reset(); 622 } 623 624 // Collects page data for history_publisher_. 625 if (history_publisher_.get()) { 626 page_collector_.reset(new PageCollector()); 627 page_collector_->Init(history_publisher_.get()); 628 } 629 630 // Thumbnail database. 631 // TODO(shess): "thumbnail database" these days only stores 632 // favicons. Thumbnails are stored in "top sites". Consider 633 // renaming "thumbnail" references to "favicons" or something of the 634 // sort. 635 thumbnail_db_.reset(new ThumbnailDatabase()); 636 if (thumbnail_db_->Init(thumbnail_name, 637 db_.get()) != sql::INIT_OK) { 638 // Unlike the main database, we don't error out when the database is too 639 // new because this error is much less severe. Generally, this shouldn't 640 // happen since the thumbnail and main database versions should be in sync. 641 // We'll just continue without thumbnails & favicons in this case or any 642 // other error. 643 LOG(WARNING) << "Could not initialize the thumbnail database."; 644 thumbnail_db_.reset(); 645 } 646 647 // Archived database. 648 if (db_->needs_version_17_migration()) { 649 // See needs_version_17_migration() decl for more. In this case, we want 650 // to delete the archived database and need to do so before we try to 651 // open the file. We can ignore any error (maybe the file doesn't exist). 652 sql::Connection::Delete(archived_name); 653 } 654 archived_db_.reset(new ArchivedDatabase()); 655 if (!archived_db_->Init(archived_name)) { 656 LOG(WARNING) << "Could not initialize the archived database."; 657 archived_db_.reset(); 658 } 659 660 // Generate the history and thumbnail database metrics only after performing 661 // any migration work. 662 if (base::RandInt(1, 100) == 50) { 663 // Only do this computation sometimes since it can be expensive. 664 db_->ComputeDatabaseMetrics(history_name); 665 if (thumbnail_db_) 666 thumbnail_db_->ComputeDatabaseMetrics(); 667 } 668 669 // Tell the expiration module about all the nice databases we made. This must 670 // happen before db_->Init() is called since the callback ForceArchiveHistory 671 // may need to expire stuff. 672 // 673 // *sigh*, this can all be cleaned up when that migration code is removed. 674 // The main DB initialization should intuitively be first (not that it 675 // actually matters) and the expirer should be set last. 676 expirer_.SetDatabases(db_.get(), archived_db_.get(), thumbnail_db_.get()); 677 678 // Open the long-running transaction. 679 db_->BeginTransaction(); 680 if (thumbnail_db_) 681 thumbnail_db_->BeginTransaction(); 682 if (archived_db_) 683 archived_db_->BeginTransaction(); 684 685 // Get the first item in our database. 686 db_->GetStartDate(&first_recorded_time_); 687 688 // Start expiring old stuff. 689 expirer_.StartArchivingOldStuff(TimeDelta::FromDays(kArchiveDaysThreshold)); 690 691#if defined(OS_ANDROID) 692 if (thumbnail_db_) { 693 android_provider_backend_.reset(new AndroidProviderBackend( 694 GetAndroidCacheFileName(), db_.get(), thumbnail_db_.get(), 695 bookmark_service_, delegate_.get())); 696 } 697#endif 698 699 HISTOGRAM_TIMES("History.InitTime", 700 TimeTicks::Now() - beginning_time); 701} 702 703void HistoryBackend::OnMemoryPressure( 704 base::MemoryPressureListener::MemoryPressureLevel memory_pressure_level) { 705 bool trim_aggressively = memory_pressure_level == 706 base::MemoryPressureListener::MEMORY_PRESSURE_CRITICAL; 707 if (db_) 708 db_->TrimMemory(trim_aggressively); 709 if (thumbnail_db_) 710 thumbnail_db_->TrimMemory(trim_aggressively); 711 if (archived_db_) 712 archived_db_->TrimMemory(trim_aggressively); 713} 714 715void HistoryBackend::CloseAllDatabases() { 716 if (db_) { 717 // Commit the long-running transaction. 718 db_->CommitTransaction(); 719 db_.reset(); 720 // Forget the first recorded time since the database is closed. 721 first_recorded_time_ = base::Time(); 722 } 723 if (thumbnail_db_) { 724 thumbnail_db_->CommitTransaction(); 725 thumbnail_db_.reset(); 726 } 727 if (archived_db_) { 728 archived_db_->CommitTransaction(); 729 archived_db_.reset(); 730 } 731} 732 733std::pair<URLID, VisitID> HistoryBackend::AddPageVisit( 734 const GURL& url, 735 Time time, 736 VisitID referring_visit, 737 content::PageTransition transition, 738 VisitSource visit_source) { 739 // Top-level frame navigations are visible, everything else is hidden 740 bool new_hidden = !content::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 content::PageTransition transition_type = 748 content::PageTransitionStripQualifier(transition); 749 if ((transition_type == content::PAGE_TRANSITION_TYPED && 750 !content::PageTransitionIsRedirect(transition)) || 751 transition_type == content::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 & content::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 (content::PageTransitionStripQualifier(transition) != 769 content::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 URLVisitedDetails* details = new URLVisitedDetails; 810 details->transition = transition; 811 details->row = url_info; 812 // TODO(meelapshah) Disabled due to potential PageCycler regression. 813 // Re-enable this. 814 // GetMostRecentRedirectsTo(url, &details->redirects); 815 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URL_VISITED, details); 816 } else { 817 VLOG(0) << "Failed to build visit insert statement: " 818 << "url_id = " << url_id; 819 } 820 821 return std::make_pair(url_id, visit_id); 822} 823 824void HistoryBackend::AddPagesWithDetails(const URLRows& urls, 825 VisitSource visit_source) { 826 if (!db_) 827 return; 828 829 scoped_ptr<URLsModifiedDetails> modified(new URLsModifiedDetails); 830 for (URLRows::const_iterator i = urls.begin(); i != urls.end(); ++i) { 831 DCHECK(!i->last_visit().is_null()); 832 833 // We will add to either the archived database or the main one depending on 834 // the date of the added visit. 835 URLDatabase* url_database; 836 VisitDatabase* visit_database; 837 if (IsExpiredVisitTime(i->last_visit())) { 838 if (!archived_db_) 839 return; // No archived database to save it to, just forget this. 840 url_database = archived_db_.get(); 841 visit_database = archived_db_.get(); 842 } else { 843 url_database = db_.get(); 844 visit_database = db_.get(); 845 } 846 847 URLRow existing_url; 848 URLID url_id = url_database->GetRowForURL(i->url(), &existing_url); 849 if (!url_id) { 850 // Add the page if it doesn't exist. 851 url_id = url_database->AddURL(*i); 852 if (!url_id) { 853 NOTREACHED() << "Could not add row to DB"; 854 return; 855 } 856 857 if (i->typed_count() > 0) { 858 modified->changed_urls.push_back(*i); 859 modified->changed_urls.back().set_id(url_id); // *i likely has |id_| 0. 860 } 861 } 862 863 // TODO(shess): I'm not sure this case needs to exist anymore. 864 if (page_collector_) { 865 page_collector_->AddPageData(i->url(), i->last_visit(), 866 i->title(), string16()); 867 } 868 869 // Sync code manages the visits itself. 870 if (visit_source != SOURCE_SYNCED) { 871 // Make up a visit to correspond to the last visit to the page. 872 VisitRow visit_info(url_id, i->last_visit(), 0, 873 content::PageTransitionFromInt( 874 content::PAGE_TRANSITION_LINK | 875 content::PAGE_TRANSITION_CHAIN_START | 876 content::PAGE_TRANSITION_CHAIN_END), 0); 877 if (!visit_database->AddVisit(&visit_info, visit_source)) { 878 NOTREACHED() << "Adding visit failed."; 879 return; 880 } 881 NotifyVisitObservers(visit_info); 882 883 if (visit_info.visit_time < first_recorded_time_) 884 first_recorded_time_ = visit_info.visit_time; 885 } 886 } 887 888 if (typed_url_syncable_service_.get()) 889 typed_url_syncable_service_->OnUrlsModified(&modified->changed_urls); 890 891 // Broadcast a notification for typed URLs that have been modified. This 892 // will be picked up by the in-memory URL database on the main thread. 893 // 894 // TODO(brettw) bug 1140015: Add an "add page" notification so the history 895 // views can keep in sync. 896 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URLS_MODIFIED, 897 modified.release()); 898 899 ScheduleCommit(); 900} 901 902bool HistoryBackend::IsExpiredVisitTime(const base::Time& time) { 903 return time < expirer_.GetCurrentArchiveTime(); 904} 905 906void HistoryBackend::SetPageTitle(const GURL& url, const string16& title) { 907 if (!db_) 908 return; 909 910 if (page_collector_) 911 page_collector_->AddPageTitle(url, title); 912 913 // Search for recent redirects which should get the same title. We make a 914 // dummy list containing the exact URL visited if there are no redirects so 915 // the processing below can be the same. 916 history::RedirectList dummy_list; 917 history::RedirectList* redirects; 918 RedirectCache::iterator iter = recent_redirects_.Get(url); 919 if (iter != recent_redirects_.end()) { 920 redirects = &iter->second; 921 922 // This redirect chain should have the destination URL as the last item. 923 DCHECK(!redirects->empty()); 924 DCHECK(redirects->back() == url); 925 } else { 926 // No redirect chain stored, make up one containing the URL we want so we 927 // can use the same logic below. 928 dummy_list.push_back(url); 929 redirects = &dummy_list; 930 } 931 932 scoped_ptr<URLsModifiedDetails> details(new URLsModifiedDetails); 933 for (size_t i = 0; i < redirects->size(); i++) { 934 URLRow row; 935 URLID row_id = db_->GetRowForURL(redirects->at(i), &row); 936 if (row_id && row.title() != title) { 937 row.set_title(title); 938 db_->UpdateURLRow(row_id, row); 939 details->changed_urls.push_back(row); 940 } 941 } 942 943 // Broadcast notifications for any URLs that have changed. This will 944 // update the in-memory database and the InMemoryURLIndex. 945 if (!details->changed_urls.empty()) { 946 if (typed_url_syncable_service_.get()) 947 typed_url_syncable_service_->OnUrlsModified(&details->changed_urls); 948 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URLS_MODIFIED, 949 details.release()); 950 ScheduleCommit(); 951 } 952} 953 954void HistoryBackend::AddPageNoVisitForBookmark(const GURL& url, 955 const string16& title) { 956 if (!db_) 957 return; 958 959 URLRow url_info(url); 960 URLID url_id = db_->GetRowForURL(url, &url_info); 961 if (url_id) { 962 // URL is already known, nothing to do. 963 return; 964 } 965 966 if (!title.empty()) { 967 url_info.set_title(title); 968 } else { 969 url_info.set_title(UTF8ToUTF16(url.spec())); 970 } 971 972 url_info.set_last_visit(Time::Now()); 973 // Mark the page hidden. If the user types it in, it'll unhide. 974 url_info.set_hidden(true); 975 976 db_->AddURL(url_info); 977} 978 979void HistoryBackend::IterateURLs( 980 const scoped_refptr<visitedlink::VisitedLinkDelegate::URLEnumerator>& 981 iterator) { 982 if (db_) { 983 HistoryDatabase::URLEnumerator e; 984 if (db_->InitURLEnumeratorForEverything(&e)) { 985 URLRow info; 986 while (e.GetNextURL(&info)) { 987 iterator->OnURL(info.url()); 988 } 989 iterator->OnComplete(true); // Success. 990 return; 991 } 992 } 993 iterator->OnComplete(false); // Failure. 994} 995 996bool HistoryBackend::GetAllTypedURLs(URLRows* urls) { 997 if (db_) 998 return db_->GetAllTypedUrls(urls); 999 return false; 1000} 1001 1002bool HistoryBackend::GetVisitsForURL(URLID id, VisitVector* visits) { 1003 if (db_) 1004 return db_->GetVisitsForURL(id, visits); 1005 return false; 1006} 1007 1008bool HistoryBackend::GetMostRecentVisitsForURL(URLID id, 1009 int max_visits, 1010 VisitVector* visits) { 1011 if (db_) 1012 return db_->GetMostRecentVisitsForURL(id, max_visits, visits); 1013 return false; 1014} 1015 1016bool HistoryBackend::UpdateURL(URLID id, const history::URLRow& url) { 1017 if (db_) 1018 return db_->UpdateURLRow(id, url); 1019 return false; 1020} 1021 1022bool HistoryBackend::AddVisits(const GURL& url, 1023 const std::vector<VisitInfo>& visits, 1024 VisitSource visit_source) { 1025 if (db_) { 1026 for (std::vector<VisitInfo>::const_iterator visit = visits.begin(); 1027 visit != visits.end(); ++visit) { 1028 if (!AddPageVisit( 1029 url, visit->first, 0, visit->second, visit_source).first) { 1030 return false; 1031 } 1032 } 1033 ScheduleCommit(); 1034 return true; 1035 } 1036 return false; 1037} 1038 1039bool HistoryBackend::RemoveVisits(const VisitVector& visits) { 1040 if (!db_) 1041 return false; 1042 1043 expirer_.ExpireVisits(visits); 1044 ScheduleCommit(); 1045 return true; 1046} 1047 1048bool HistoryBackend::GetVisitsSource(const VisitVector& visits, 1049 VisitSourceMap* sources) { 1050 if (!db_) 1051 return false; 1052 1053 db_->GetVisitsSource(visits, sources); 1054 return true; 1055} 1056 1057bool HistoryBackend::GetURL(const GURL& url, history::URLRow* url_row) { 1058 if (db_) 1059 return db_->GetRowForURL(url, url_row) != 0; 1060 return false; 1061} 1062 1063void HistoryBackend::QueryURL(scoped_refptr<QueryURLRequest> request, 1064 const GURL& url, 1065 bool want_visits) { 1066 if (request->canceled()) 1067 return; 1068 1069 bool success = false; 1070 URLRow* row = &request->value.a; 1071 VisitVector* visits = &request->value.b; 1072 if (db_) { 1073 if (db_->GetRowForURL(url, row)) { 1074 // Have a row. 1075 success = true; 1076 1077 // Optionally query the visits. 1078 if (want_visits) 1079 db_->GetVisitsForURL(row->id(), visits); 1080 } 1081 } 1082 request->ForwardResult(request->handle(), success, row, visits); 1083} 1084 1085TypedUrlSyncableService* HistoryBackend::GetTypedUrlSyncableService() const { 1086 return typed_url_syncable_service_.get(); 1087} 1088 1089// Segment usage --------------------------------------------------------------- 1090 1091void HistoryBackend::DeleteOldSegmentData() { 1092 if (db_) 1093 db_->DeleteSegmentData(Time::Now() - 1094 TimeDelta::FromDays(kSegmentDataRetention)); 1095} 1096 1097void HistoryBackend::QuerySegmentUsage( 1098 scoped_refptr<QuerySegmentUsageRequest> request, 1099 const Time from_time, 1100 int max_result_count) { 1101 if (request->canceled()) 1102 return; 1103 1104 if (db_) { 1105 db_->QuerySegmentUsage(from_time, max_result_count, &request->value.get()); 1106 1107 // If this is the first time we query segments, invoke 1108 // DeleteOldSegmentData asynchronously. We do this to cleanup old 1109 // entries. 1110 if (!segment_queried_) { 1111 segment_queried_ = true; 1112 base::MessageLoop::current()->PostTask( 1113 FROM_HERE, 1114 base::Bind(&HistoryBackend::DeleteOldSegmentData, this)); 1115 } 1116 } 1117 request->ForwardResult(request->handle(), &request->value.get()); 1118} 1119 1120void HistoryBackend::IncreaseSegmentDuration(const GURL& url, 1121 base::Time time, 1122 base::TimeDelta delta) { 1123 if (!db_) 1124 return; 1125 1126 const std::string segment_name(VisitSegmentDatabase::ComputeSegmentName(url)); 1127 SegmentID segment_id = db_->GetSegmentNamed(segment_name); 1128 if (!segment_id) { 1129 URLID url_id = db_->GetRowForURL(url, NULL); 1130 if (!url_id) 1131 return; 1132 segment_id = db_->CreateSegment(url_id, segment_name); 1133 if (!segment_id) 1134 return; 1135 } 1136 SegmentDurationID duration_id; 1137 base::TimeDelta total_delta; 1138 if (!db_->GetSegmentDuration(segment_id, time, &duration_id, 1139 &total_delta)) { 1140 db_->CreateSegmentDuration(segment_id, time, delta); 1141 return; 1142 } 1143 total_delta += delta; 1144 db_->SetSegmentDuration(duration_id, total_delta); 1145} 1146 1147void HistoryBackend::QuerySegmentDuration( 1148 scoped_refptr<QuerySegmentUsageRequest> request, 1149 const base::Time from_time, 1150 int max_result_count) { 1151 if (request->canceled()) 1152 return; 1153 1154 if (db_) { 1155 db_->QuerySegmentDuration(from_time, max_result_count, 1156 &request->value.get()); 1157 } 1158 request->ForwardResult(request->handle(), &request->value.get()); 1159} 1160 1161// Keyword visits -------------------------------------------------------------- 1162 1163void HistoryBackend::SetKeywordSearchTermsForURL(const GURL& url, 1164 TemplateURLID keyword_id, 1165 const string16& term) { 1166 if (!db_) 1167 return; 1168 1169 // Get the ID for this URL. 1170 URLRow url_row; 1171 if (!db_->GetRowForURL(url, &url_row)) { 1172 // There is a small possibility the url was deleted before the keyword 1173 // was added. Ignore the request. 1174 return; 1175 } 1176 1177 db_->SetKeywordSearchTermsForURL(url_row.id(), keyword_id, term); 1178 1179 // details is deleted by BroadcastNotifications. 1180 KeywordSearchTermDetails* details = new KeywordSearchTermDetails; 1181 details->url = url; 1182 details->keyword_id = keyword_id; 1183 details->term = term; 1184 BroadcastNotifications( 1185 chrome::NOTIFICATION_HISTORY_KEYWORD_SEARCH_TERM_UPDATED, details); 1186 ScheduleCommit(); 1187} 1188 1189void HistoryBackend::DeleteAllSearchTermsForKeyword( 1190 TemplateURLID keyword_id) { 1191 if (!db_) 1192 return; 1193 1194 db_->DeleteAllSearchTermsForKeyword(keyword_id); 1195 // TODO(sky): bug 1168470. Need to move from archive dbs too. 1196 ScheduleCommit(); 1197} 1198 1199void HistoryBackend::GetMostRecentKeywordSearchTerms( 1200 scoped_refptr<GetMostRecentKeywordSearchTermsRequest> request, 1201 TemplateURLID keyword_id, 1202 const string16& prefix, 1203 int max_count) { 1204 if (request->canceled()) 1205 return; 1206 1207 if (db_) { 1208 db_->GetMostRecentKeywordSearchTerms(keyword_id, prefix, max_count, 1209 &(request->value)); 1210 } 1211 request->ForwardResult(request->handle(), &request->value); 1212} 1213 1214// Downloads ------------------------------------------------------------------- 1215 1216void HistoryBackend::GetNextDownloadId(uint32* next_id) { 1217 if (db_) 1218 db_->GetNextDownloadId(next_id); 1219} 1220 1221// Get all the download entries from the database. 1222void HistoryBackend::QueryDownloads(std::vector<DownloadRow>* rows) { 1223 if (db_) 1224 db_->QueryDownloads(rows); 1225} 1226 1227// Update a particular download entry. 1228void HistoryBackend::UpdateDownload(const history::DownloadRow& data) { 1229 if (!db_) 1230 return; 1231 db_->UpdateDownload(data); 1232 ScheduleCommit(); 1233} 1234 1235void HistoryBackend::CreateDownload(const history::DownloadRow& history_info, 1236 bool* success) { 1237 if (!db_) 1238 return; 1239 *success = db_->CreateDownload(history_info); 1240 ScheduleCommit(); 1241} 1242 1243void HistoryBackend::RemoveDownloads(const std::set<uint32>& ids) { 1244 if (!db_) 1245 return; 1246 size_t downloads_count_before = db_->CountDownloads(); 1247 base::TimeTicks started_removing = base::TimeTicks::Now(); 1248 // HistoryBackend uses a long-running Transaction that is committed 1249 // periodically, so this loop doesn't actually hit the disk too hard. 1250 for (std::set<uint32>::const_iterator it = ids.begin(); 1251 it != ids.end(); ++it) { 1252 db_->RemoveDownload(*it); 1253 } 1254 ScheduleCommit(); 1255 base::TimeTicks finished_removing = base::TimeTicks::Now(); 1256 size_t downloads_count_after = db_->CountDownloads(); 1257 1258 DCHECK_LE(downloads_count_after, downloads_count_before); 1259 if (downloads_count_after > downloads_count_before) 1260 return; 1261 size_t num_downloads_deleted = downloads_count_before - downloads_count_after; 1262 UMA_HISTOGRAM_COUNTS("Download.DatabaseRemoveDownloadsCount", 1263 num_downloads_deleted); 1264 base::TimeDelta micros = (1000 * (finished_removing - started_removing)); 1265 UMA_HISTOGRAM_TIMES("Download.DatabaseRemoveDownloadsTime", micros); 1266 if (num_downloads_deleted > 0) { 1267 UMA_HISTOGRAM_TIMES("Download.DatabaseRemoveDownloadsTimePerRecord", 1268 (1000 * micros) / num_downloads_deleted); 1269 } 1270 DCHECK_GE(ids.size(), num_downloads_deleted); 1271 if (ids.size() < num_downloads_deleted) 1272 return; 1273 UMA_HISTOGRAM_COUNTS("Download.DatabaseRemoveDownloadsCountNotRemoved", 1274 ids.size() - num_downloads_deleted); 1275} 1276 1277void HistoryBackend::QueryHistory(scoped_refptr<QueryHistoryRequest> request, 1278 const string16& text_query, 1279 const QueryOptions& options) { 1280 if (request->canceled()) 1281 return; 1282 1283 TimeTicks beginning_time = TimeTicks::Now(); 1284 1285 if (db_) { 1286 if (text_query.empty()) { 1287 // Basic history query for the main database. 1288 QueryHistoryBasic(db_.get(), db_.get(), options, &request->value); 1289 1290 // Now query the archived database. This is a bit tricky because we don't 1291 // want to query it if the queried time range isn't going to find anything 1292 // in it. 1293 // TODO(brettw) bug 1171036: do blimpie querying for the archived database 1294 // as well. 1295 // if (archived_db_.get() && 1296 // expirer_.GetCurrentArchiveTime() - TimeDelta::FromDays(7)) { 1297 } else { 1298 // Text history query. 1299 QueryHistoryText(db_.get(), db_.get(), text_query, options, 1300 &request->value); 1301 if (archived_db_.get() && 1302 expirer_.GetCurrentArchiveTime() >= options.begin_time) { 1303 QueryHistoryText(archived_db_.get(), archived_db_.get(), text_query, 1304 options, &request->value); 1305 } 1306 } 1307 } 1308 1309 request->ForwardResult(request->handle(), &request->value); 1310 1311 UMA_HISTOGRAM_TIMES("History.QueryHistory", 1312 TimeTicks::Now() - beginning_time); 1313} 1314 1315// Basic time-based querying of history. 1316void HistoryBackend::QueryHistoryBasic(URLDatabase* url_db, 1317 VisitDatabase* visit_db, 1318 const QueryOptions& options, 1319 QueryResults* result) { 1320 // First get all visits. 1321 VisitVector visits; 1322 bool has_more_results = visit_db->GetVisibleVisitsInRange(options, &visits); 1323 DCHECK(static_cast<int>(visits.size()) <= options.EffectiveMaxCount()); 1324 1325 // Now add them and the URL rows to the results. 1326 URLResult url_result; 1327 for (size_t i = 0; i < visits.size(); i++) { 1328 const VisitRow visit = visits[i]; 1329 1330 // Add a result row for this visit, get the URL info from the DB. 1331 if (!url_db->GetURLRow(visit.url_id, &url_result)) { 1332 VLOG(0) << "Failed to get id " << visit.url_id 1333 << " from history.urls."; 1334 continue; // DB out of sync and URL doesn't exist, try to recover. 1335 } 1336 1337 if (!url_result.url().is_valid()) { 1338 VLOG(0) << "Got invalid URL from history.urls with id " 1339 << visit.url_id << ": " 1340 << url_result.url().possibly_invalid_spec(); 1341 continue; // Don't report invalid URLs in case of corruption. 1342 } 1343 1344 // The archived database may be out of sync with respect to starring, 1345 // titles, last visit date, etc. Therefore, we query the main DB if the 1346 // current URL database is not the main one. 1347 if (url_db == db_.get()) { 1348 // Currently querying the archived DB, update with the main database to 1349 // catch any interesting stuff. This will update it if it exists in the 1350 // main DB, and do nothing otherwise. 1351 db_->GetRowForURL(url_result.url(), &url_result); 1352 } 1353 1354 url_result.set_visit_time(visit.visit_time); 1355 1356 // Set whether the visit was blocked for a managed user by looking at the 1357 // transition type. 1358 url_result.set_blocked_visit( 1359 (visit.transition & content::PAGE_TRANSITION_BLOCKED) != 0); 1360 1361 // We don't set any of the query-specific parts of the URLResult, since 1362 // snippets and stuff don't apply to basic querying. 1363 result->AppendURLBySwapping(&url_result); 1364 } 1365 1366 if (!has_more_results && options.begin_time <= first_recorded_time_) 1367 result->set_reached_beginning(true); 1368} 1369 1370// Text-based querying of history. 1371void HistoryBackend::QueryHistoryText(URLDatabase* url_db, 1372 VisitDatabase* visit_db, 1373 const string16& text_query, 1374 const QueryOptions& options, 1375 QueryResults* result) { 1376 URLRows text_matches; 1377 url_db->GetTextMatches(text_query, &text_matches); 1378 1379 std::vector<URLResult> matching_visits; 1380 VisitVector visits; // Declare outside loop to prevent re-construction. 1381 for (size_t i = 0; i < text_matches.size(); i++) { 1382 const URLRow& text_match = text_matches[i]; 1383 // Get all visits for given URL match. 1384 visit_db->GetVisitsForURLWithOptions(text_match.id(), options, &visits); 1385 for (size_t j = 0; j < visits.size(); j++) { 1386 URLResult url_result(text_match); 1387 url_result.set_visit_time(visits[j].visit_time); 1388 matching_visits.push_back(url_result); 1389 } 1390 } 1391 1392 std::sort(matching_visits.begin(), matching_visits.end(), 1393 URLResult::CompareVisitTime); 1394 1395 size_t max_results = options.max_count == 0 ? 1396 std::numeric_limits<size_t>::max() : static_cast<int>(options.max_count); 1397 for (std::vector<URLResult>::iterator it = matching_visits.begin(); 1398 it != matching_visits.end() && result->size() < max_results; ++it) { 1399 result->AppendURLBySwapping(&(*it)); 1400 } 1401 1402 if (matching_visits.size() == result->size() && 1403 options.begin_time <= first_recorded_time_) 1404 result->set_reached_beginning(true); 1405} 1406 1407// Frontend to GetMostRecentRedirectsFrom from the history thread. 1408void HistoryBackend::QueryRedirectsFrom( 1409 scoped_refptr<QueryRedirectsRequest> request, 1410 const GURL& url) { 1411 if (request->canceled()) 1412 return; 1413 bool success = GetMostRecentRedirectsFrom(url, &request->value); 1414 request->ForwardResult(request->handle(), url, success, &request->value); 1415} 1416 1417void HistoryBackend::QueryRedirectsTo( 1418 scoped_refptr<QueryRedirectsRequest> request, 1419 const GURL& url) { 1420 if (request->canceled()) 1421 return; 1422 bool success = GetMostRecentRedirectsTo(url, &request->value); 1423 request->ForwardResult(request->handle(), url, success, &request->value); 1424} 1425 1426void HistoryBackend::GetVisibleVisitCountToHost( 1427 scoped_refptr<GetVisibleVisitCountToHostRequest> request, 1428 const GURL& url) { 1429 if (request->canceled()) 1430 return; 1431 int count = 0; 1432 Time first_visit; 1433 const bool success = db_.get() && 1434 db_->GetVisibleVisitCountToHost(url, &count, &first_visit); 1435 request->ForwardResult(request->handle(), success, count, first_visit); 1436} 1437 1438void HistoryBackend::QueryTopURLsAndRedirects( 1439 scoped_refptr<QueryTopURLsAndRedirectsRequest> request, 1440 int result_count) { 1441 if (request->canceled()) 1442 return; 1443 1444 if (!db_) { 1445 request->ForwardResult(request->handle(), false, NULL, NULL); 1446 return; 1447 } 1448 1449 std::vector<GURL>* top_urls = &request->value.a; 1450 history::RedirectMap* redirects = &request->value.b; 1451 1452 ScopedVector<PageUsageData> data; 1453 db_->QuerySegmentUsage(base::Time::Now() - base::TimeDelta::FromDays(90), 1454 result_count, &data.get()); 1455 1456 for (size_t i = 0; i < data.size(); ++i) { 1457 top_urls->push_back(data[i]->GetURL()); 1458 RefCountedVector<GURL>* list = new RefCountedVector<GURL>; 1459 GetMostRecentRedirectsFrom(top_urls->back(), &list->data); 1460 (*redirects)[top_urls->back()] = list; 1461 } 1462 1463 request->ForwardResult(request->handle(), true, top_urls, redirects); 1464} 1465 1466// Will replace QueryTopURLsAndRedirectsRequest. 1467void HistoryBackend::QueryMostVisitedURLs( 1468 scoped_refptr<QueryMostVisitedURLsRequest> request, 1469 int result_count, 1470 int days_back) { 1471 if (request->canceled()) 1472 return; 1473 1474 if (!db_) { 1475 // No History Database - return an empty list. 1476 request->ForwardResult(request->handle(), MostVisitedURLList()); 1477 return; 1478 } 1479 1480 MostVisitedURLList* result = &request->value; 1481 QueryMostVisitedURLsImpl(result_count, days_back, result); 1482 request->ForwardResult(request->handle(), *result); 1483} 1484 1485void HistoryBackend::QueryFilteredURLs( 1486 scoped_refptr<QueryFilteredURLsRequest> request, 1487 int result_count, 1488 const history::VisitFilter& filter, 1489 bool extended_info) { 1490 if (request->canceled()) 1491 return; 1492 1493 base::Time request_start = base::Time::Now(); 1494 1495 if (!db_) { 1496 // No History Database - return an empty list. 1497 request->ForwardResult(request->handle(), FilteredURLList()); 1498 return; 1499 } 1500 1501 VisitVector visits; 1502 db_->GetDirectVisitsDuringTimes(filter, 0, &visits); 1503 1504 std::map<URLID, double> score_map; 1505 for (size_t i = 0; i < visits.size(); ++i) { 1506 score_map[visits[i].url_id] += filter.GetVisitScore(visits[i]); 1507 } 1508 1509 // TODO(georgey): experiment with visit_segment database granularity (it is 1510 // currently 24 hours) to use it directly instead of using visits database, 1511 // which is considerably slower. 1512 ScopedVector<PageUsageData> data; 1513 data.reserve(score_map.size()); 1514 for (std::map<URLID, double>::iterator it = score_map.begin(); 1515 it != score_map.end(); ++it) { 1516 PageUsageData* pud = new PageUsageData(it->first); 1517 pud->SetScore(it->second); 1518 data.push_back(pud); 1519 } 1520 1521 // Limit to the top |result_count| results. 1522 std::sort(data.begin(), data.end(), PageUsageData::Predicate); 1523 if (result_count && implicit_cast<int>(data.size()) > result_count) 1524 data.resize(result_count); 1525 1526 for (size_t i = 0; i < data.size(); ++i) { 1527 URLRow info; 1528 if (db_->GetURLRow(data[i]->GetID(), &info)) { 1529 data[i]->SetURL(info.url()); 1530 data[i]->SetTitle(info.title()); 1531 } 1532 } 1533 1534 FilteredURLList& result = request->value; 1535 for (size_t i = 0; i < data.size(); ++i) { 1536 PageUsageData* current_data = data[i]; 1537 FilteredURL url(*current_data); 1538 1539 if (extended_info) { 1540 VisitVector visits; 1541 db_->GetVisitsForURL(current_data->GetID(), &visits); 1542 if (visits.size() > 0) { 1543 url.extended_info.total_visits = visits.size(); 1544 for (size_t i = 0; i < visits.size(); ++i) { 1545 url.extended_info.duration_opened += 1546 visits[i].visit_duration.InSeconds(); 1547 if (visits[i].visit_time > url.extended_info.last_visit_time) { 1548 url.extended_info.last_visit_time = visits[i].visit_time; 1549 } 1550 } 1551 // TODO(macourteau): implement the url.extended_info.visits stat. 1552 } 1553 } 1554 result.push_back(url); 1555 } 1556 1557 int delta_time = std::max(1, std::min(999, 1558 static_cast<int>((base::Time::Now() - request_start).InMilliseconds()))); 1559 STATIC_HISTOGRAM_POINTER_BLOCK( 1560 "NewTabPage.SuggestedSitesLoadTime", 1561 Add(delta_time), 1562 base::LinearHistogram::FactoryGet("NewTabPage.SuggestedSitesLoadTime", 1563 1, 1000, 100, base::Histogram::kUmaTargetedHistogramFlag)); 1564 1565 request->ForwardResult(request->handle(), result); 1566} 1567 1568void HistoryBackend::QueryMostVisitedURLsImpl(int result_count, 1569 int days_back, 1570 MostVisitedURLList* result) { 1571 if (!db_) 1572 return; 1573 1574 ScopedVector<PageUsageData> data; 1575 db_->QuerySegmentUsage(base::Time::Now() - 1576 base::TimeDelta::FromDays(days_back), 1577 result_count, &data.get()); 1578 1579 for (size_t i = 0; i < data.size(); ++i) { 1580 PageUsageData* current_data = data[i]; 1581 RedirectList redirects; 1582 GetMostRecentRedirectsFrom(current_data->GetURL(), &redirects); 1583 MostVisitedURL url = MakeMostVisitedURL(*current_data, redirects); 1584 result->push_back(url); 1585 } 1586} 1587 1588void HistoryBackend::GetRedirectsFromSpecificVisit( 1589 VisitID cur_visit, history::RedirectList* redirects) { 1590 // Follow any redirects from the given visit and add them to the list. 1591 // It *should* be impossible to get a circular chain here, but we check 1592 // just in case to avoid infinite loops. 1593 GURL cur_url; 1594 std::set<VisitID> visit_set; 1595 visit_set.insert(cur_visit); 1596 while (db_->GetRedirectFromVisit(cur_visit, &cur_visit, &cur_url)) { 1597 if (visit_set.find(cur_visit) != visit_set.end()) { 1598 NOTREACHED() << "Loop in visit chain, giving up"; 1599 return; 1600 } 1601 visit_set.insert(cur_visit); 1602 redirects->push_back(cur_url); 1603 } 1604} 1605 1606void HistoryBackend::GetRedirectsToSpecificVisit( 1607 VisitID cur_visit, 1608 history::RedirectList* redirects) { 1609 // Follow redirects going to cur_visit. These are added to |redirects| in 1610 // the order they are found. If a redirect chain looks like A -> B -> C and 1611 // |cur_visit| = C, redirects will be {B, A} in that order. 1612 if (!db_) 1613 return; 1614 1615 GURL cur_url; 1616 std::set<VisitID> visit_set; 1617 visit_set.insert(cur_visit); 1618 while (db_->GetRedirectToVisit(cur_visit, &cur_visit, &cur_url)) { 1619 if (visit_set.find(cur_visit) != visit_set.end()) { 1620 NOTREACHED() << "Loop in visit chain, giving up"; 1621 return; 1622 } 1623 visit_set.insert(cur_visit); 1624 redirects->push_back(cur_url); 1625 } 1626} 1627 1628bool HistoryBackend::GetMostRecentRedirectsFrom( 1629 const GURL& from_url, 1630 history::RedirectList* redirects) { 1631 redirects->clear(); 1632 if (!db_) 1633 return false; 1634 1635 URLID from_url_id = db_->GetRowForURL(from_url, NULL); 1636 VisitID cur_visit = db_->GetMostRecentVisitForURL(from_url_id, NULL); 1637 if (!cur_visit) 1638 return false; // No visits for URL. 1639 1640 GetRedirectsFromSpecificVisit(cur_visit, redirects); 1641 return true; 1642} 1643 1644bool HistoryBackend::GetMostRecentRedirectsTo( 1645 const GURL& to_url, 1646 history::RedirectList* redirects) { 1647 redirects->clear(); 1648 if (!db_) 1649 return false; 1650 1651 URLID to_url_id = db_->GetRowForURL(to_url, NULL); 1652 VisitID cur_visit = db_->GetMostRecentVisitForURL(to_url_id, NULL); 1653 if (!cur_visit) 1654 return false; // No visits for URL. 1655 1656 GetRedirectsToSpecificVisit(cur_visit, redirects); 1657 return true; 1658} 1659 1660void HistoryBackend::ScheduleAutocomplete(HistoryURLProvider* provider, 1661 HistoryURLProviderParams* params) { 1662 // ExecuteWithDB should handle the NULL database case. 1663 provider->ExecuteWithDB(this, db_.get(), params); 1664} 1665 1666void HistoryBackend::DeleteFTSIndexDatabases() { 1667 // Find files on disk matching the text databases file pattern so we can 1668 // quickly test for and delete them. 1669 base::FilePath::StringType filepattern = 1670 FILE_PATH_LITERAL("History Index *"); 1671 base::FileEnumerator enumerator( 1672 history_dir_, false, base::FileEnumerator::FILES, filepattern); 1673 int num_databases_deleted = 0; 1674 base::FilePath current_file; 1675 while (!(current_file = enumerator.Next()).empty()) { 1676 if (sql::Connection::Delete(current_file)) 1677 num_databases_deleted++; 1678 } 1679 UMA_HISTOGRAM_COUNTS("History.DeleteFTSIndexDatabases", 1680 num_databases_deleted); 1681} 1682 1683void HistoryBackend::SetPageContents(const GURL& url, 1684 const string16& contents) { 1685 if (page_collector_) 1686 page_collector_->AddPageContents(url, contents); 1687} 1688 1689void HistoryBackend::GetFavicons( 1690 const std::vector<GURL>& icon_urls, 1691 int icon_types, 1692 int desired_size_in_dip, 1693 const std::vector<ui::ScaleFactor>& desired_scale_factors, 1694 std::vector<chrome::FaviconBitmapResult>* bitmap_results) { 1695 UpdateFaviconMappingsAndFetchImpl(NULL, icon_urls, icon_types, 1696 desired_size_in_dip, desired_scale_factors, 1697 bitmap_results); 1698} 1699 1700void HistoryBackend::GetFaviconsForURL( 1701 const GURL& page_url, 1702 int icon_types, 1703 int desired_size_in_dip, 1704 const std::vector<ui::ScaleFactor>& desired_scale_factors, 1705 std::vector<chrome::FaviconBitmapResult>* bitmap_results) { 1706 DCHECK(bitmap_results); 1707 GetFaviconsFromDB(page_url, icon_types, desired_size_in_dip, 1708 desired_scale_factors, bitmap_results); 1709} 1710 1711void HistoryBackend::GetFaviconForID( 1712 chrome::FaviconID favicon_id, 1713 int desired_size_in_dip, 1714 ui::ScaleFactor desired_scale_factor, 1715 std::vector<chrome::FaviconBitmapResult>* bitmap_results) { 1716 std::vector<chrome::FaviconID> favicon_ids; 1717 favicon_ids.push_back(favicon_id); 1718 std::vector<ui::ScaleFactor> desired_scale_factors; 1719 desired_scale_factors.push_back(desired_scale_factor); 1720 1721 // Get results from DB. 1722 GetFaviconBitmapResultsForBestMatch(favicon_ids, 1723 desired_size_in_dip, 1724 desired_scale_factors, 1725 bitmap_results); 1726} 1727 1728void HistoryBackend::UpdateFaviconMappingsAndFetch( 1729 const GURL& page_url, 1730 const std::vector<GURL>& icon_urls, 1731 int icon_types, 1732 int desired_size_in_dip, 1733 const std::vector<ui::ScaleFactor>& desired_scale_factors, 1734 std::vector<chrome::FaviconBitmapResult>* bitmap_results) { 1735 UpdateFaviconMappingsAndFetchImpl(&page_url, icon_urls, icon_types, 1736 desired_size_in_dip, desired_scale_factors, 1737 bitmap_results); 1738} 1739 1740void HistoryBackend::MergeFavicon( 1741 const GURL& page_url, 1742 const GURL& icon_url, 1743 chrome::IconType icon_type, 1744 scoped_refptr<base::RefCountedMemory> bitmap_data, 1745 const gfx::Size& pixel_size) { 1746 if (!thumbnail_db_ || !db_) 1747 return; 1748 1749 chrome::FaviconID favicon_id = 1750 thumbnail_db_->GetFaviconIDForFaviconURL(icon_url, icon_type, NULL); 1751 1752 if (!favicon_id) { 1753 // There is no favicon at |icon_url|, create it. 1754 favicon_id = thumbnail_db_->AddFavicon(icon_url, icon_type); 1755 } 1756 1757 std::vector<FaviconBitmapIDSize> bitmap_id_sizes; 1758 thumbnail_db_->GetFaviconBitmapIDSizes(favicon_id, &bitmap_id_sizes); 1759 1760 // If there is already a favicon bitmap of |pixel_size| at |icon_url|, 1761 // replace it. 1762 bool bitmap_identical = false; 1763 bool replaced_bitmap = false; 1764 for (size_t i = 0; i < bitmap_id_sizes.size(); ++i) { 1765 if (bitmap_id_sizes[i].pixel_size == pixel_size) { 1766 if (IsFaviconBitmapDataEqual(bitmap_id_sizes[i].bitmap_id, bitmap_data)) { 1767 thumbnail_db_->SetFaviconBitmapLastUpdateTime( 1768 bitmap_id_sizes[i].bitmap_id, base::Time::Now()); 1769 bitmap_identical = true; 1770 } else { 1771 thumbnail_db_->SetFaviconBitmap(bitmap_id_sizes[i].bitmap_id, 1772 bitmap_data, base::Time::Now()); 1773 replaced_bitmap = true; 1774 } 1775 break; 1776 } 1777 } 1778 1779 // Create a vector of the pixel sizes of the favicon bitmaps currently at 1780 // |icon_url|. 1781 std::vector<gfx::Size> favicon_sizes; 1782 for (size_t i = 0; i < bitmap_id_sizes.size(); ++i) 1783 favicon_sizes.push_back(bitmap_id_sizes[i].pixel_size); 1784 1785 if (!replaced_bitmap && !bitmap_identical) { 1786 // Set the preexisting favicon bitmaps as expired as the preexisting favicon 1787 // bitmaps are not consistent with the merged in data. 1788 thumbnail_db_->SetFaviconOutOfDate(favicon_id); 1789 1790 // Delete an arbitrary favicon bitmap to avoid going over the limit of 1791 // |kMaxFaviconBitmapsPerIconURL|. 1792 if (bitmap_id_sizes.size() >= kMaxFaviconBitmapsPerIconURL) { 1793 thumbnail_db_->DeleteFaviconBitmap(bitmap_id_sizes[0].bitmap_id); 1794 favicon_sizes.erase(favicon_sizes.begin()); 1795 } 1796 thumbnail_db_->AddFaviconBitmap(favicon_id, bitmap_data, base::Time::Now(), 1797 pixel_size); 1798 favicon_sizes.push_back(pixel_size); 1799 } 1800 1801 // A site may have changed the favicons that it uses for |page_url|. 1802 // Example Scenario: 1803 // page_url = news.google.com 1804 // Initial State: www.google.com/favicon.ico 16x16, 32x32 1805 // MergeFavicon(news.google.com, news.google.com/news_specific.ico, ..., 1806 // ..., 16x16) 1807 // 1808 // Difficulties: 1809 // 1. Sync requires that a call to GetFaviconsForURL() returns the 1810 // |bitmap_data| passed into MergeFavicon(). 1811 // - It is invalid for the 16x16 bitmap for www.google.com/favicon.ico to 1812 // stay mapped to news.google.com because it would be unclear which 16x16 1813 // bitmap should be returned via GetFaviconsForURL(). 1814 // 1815 // 2. www.google.com/favicon.ico may be mapped to more than just 1816 // news.google.com (eg www.google.com). 1817 // - The 16x16 bitmap cannot be deleted from www.google.com/favicon.ico 1818 // 1819 // To resolve these problems, we copy all of the favicon bitmaps previously 1820 // mapped to news.google.com (|page_url|) and add them to the favicon at 1821 // news.google.com/news_specific.ico (|icon_url|). The favicon sizes for 1822 // |icon_url| are set to default to indicate that |icon_url| has incomplete 1823 // / incorrect data. 1824 // Difficulty 1: All but news.google.com/news_specific.ico are unmapped from 1825 // news.google.com 1826 // Difficulty 2: The favicon bitmaps for www.google.com/favicon.ico are not 1827 // modified. 1828 1829 std::vector<IconMapping> icon_mappings; 1830 thumbnail_db_->GetIconMappingsForPageURL(page_url, icon_type, &icon_mappings); 1831 1832 // Copy the favicon bitmaps mapped to |page_url| to the favicon at |icon_url| 1833 // till the limit of |kMaxFaviconBitmapsPerIconURL| is reached. 1834 for (size_t i = 0; i < icon_mappings.size(); ++i) { 1835 if (favicon_sizes.size() >= kMaxFaviconBitmapsPerIconURL) 1836 break; 1837 1838 if (icon_mappings[i].icon_url == icon_url) 1839 continue; 1840 1841 std::vector<FaviconBitmap> bitmaps_to_copy; 1842 thumbnail_db_->GetFaviconBitmaps(icon_mappings[i].icon_id, 1843 &bitmaps_to_copy); 1844 for (size_t j = 0; j < bitmaps_to_copy.size(); ++j) { 1845 // Do not add a favicon bitmap at a pixel size for which there is already 1846 // a favicon bitmap mapped to |icon_url|. The one there is more correct 1847 // and having multiple equally sized favicon bitmaps for |page_url| is 1848 // ambiguous in terms of GetFaviconsForURL(). 1849 std::vector<gfx::Size>::iterator it = std::find(favicon_sizes.begin(), 1850 favicon_sizes.end(), bitmaps_to_copy[j].pixel_size); 1851 if (it != favicon_sizes.end()) 1852 continue; 1853 1854 // Add the favicon bitmap as expired as it is not consistent with the 1855 // merged in data. 1856 thumbnail_db_->AddFaviconBitmap(favicon_id, 1857 bitmaps_to_copy[j].bitmap_data, base::Time(), 1858 bitmaps_to_copy[j].pixel_size); 1859 favicon_sizes.push_back(bitmaps_to_copy[j].pixel_size); 1860 1861 if (favicon_sizes.size() >= kMaxFaviconBitmapsPerIconURL) 1862 break; 1863 } 1864 } 1865 1866 // Update the favicon mappings such that only |icon_url| is mapped to 1867 // |page_url|. 1868 bool mapping_changed = false; 1869 if (icon_mappings.size() != 1 || icon_mappings[0].icon_url != icon_url) { 1870 std::vector<chrome::FaviconID> favicon_ids; 1871 favicon_ids.push_back(favicon_id); 1872 SetFaviconMappingsForPageAndRedirects(page_url, icon_type, favicon_ids); 1873 mapping_changed = true; 1874 } 1875 1876 if (mapping_changed || !bitmap_identical) 1877 SendFaviconChangedNotificationForPageAndRedirects(page_url); 1878 ScheduleCommit(); 1879} 1880 1881void HistoryBackend::SetFavicons( 1882 const GURL& page_url, 1883 chrome::IconType icon_type, 1884 const std::vector<chrome::FaviconBitmapData>& favicon_bitmap_data) { 1885 if (!thumbnail_db_ || !db_) 1886 return; 1887 1888 DCHECK(ValidateSetFaviconsParams(favicon_bitmap_data)); 1889 1890 // Build map of FaviconBitmapData for each icon url. 1891 typedef std::map<GURL, std::vector<chrome::FaviconBitmapData> > 1892 BitmapDataByIconURL; 1893 BitmapDataByIconURL grouped_by_icon_url; 1894 for (size_t i = 0; i < favicon_bitmap_data.size(); ++i) { 1895 const GURL& icon_url = favicon_bitmap_data[i].icon_url; 1896 grouped_by_icon_url[icon_url].push_back(favicon_bitmap_data[i]); 1897 } 1898 1899 // Track whether the method modifies or creates any favicon bitmaps, favicons 1900 // or icon mappings. 1901 bool data_modified = false; 1902 1903 std::vector<chrome::FaviconID> icon_ids; 1904 for (BitmapDataByIconURL::const_iterator it = grouped_by_icon_url.begin(); 1905 it != grouped_by_icon_url.end(); ++it) { 1906 const GURL& icon_url = it->first; 1907 chrome::FaviconID icon_id = 1908 thumbnail_db_->GetFaviconIDForFaviconURL(icon_url, icon_type, NULL); 1909 1910 if (!icon_id) { 1911 // TODO(pkotwicz): Remove the favicon sizes attribute from 1912 // ThumbnailDatabase::AddFavicon(). 1913 icon_id = thumbnail_db_->AddFavicon(icon_url, icon_type); 1914 data_modified = true; 1915 } 1916 icon_ids.push_back(icon_id); 1917 1918 if (!data_modified) 1919 SetFaviconBitmaps(icon_id, it->second, &data_modified); 1920 else 1921 SetFaviconBitmaps(icon_id, it->second, NULL); 1922 } 1923 1924 data_modified |= 1925 SetFaviconMappingsForPageAndRedirects(page_url, icon_type, icon_ids); 1926 1927 if (data_modified) { 1928 // Send notification to the UI as an icon mapping, favicon, or favicon 1929 // bitmap was changed by this function. 1930 SendFaviconChangedNotificationForPageAndRedirects(page_url); 1931 } 1932 ScheduleCommit(); 1933} 1934 1935void HistoryBackend::SetFaviconsOutOfDateForPage(const GURL& page_url) { 1936 std::vector<IconMapping> icon_mappings; 1937 1938 if (!thumbnail_db_ || 1939 !thumbnail_db_->GetIconMappingsForPageURL(page_url, 1940 &icon_mappings)) 1941 return; 1942 1943 for (std::vector<IconMapping>::iterator m = icon_mappings.begin(); 1944 m != icon_mappings.end(); ++m) { 1945 thumbnail_db_->SetFaviconOutOfDate(m->icon_id); 1946 } 1947 ScheduleCommit(); 1948} 1949 1950void HistoryBackend::CloneFavicons(const GURL& old_page_url, 1951 const GURL& new_page_url) { 1952 if (!thumbnail_db_) 1953 return; 1954 1955 // Prevent cross-domain cloning. 1956 if (old_page_url.GetOrigin() != new_page_url.GetOrigin()) 1957 return; 1958 1959 thumbnail_db_->CloneIconMappings(old_page_url, new_page_url); 1960 ScheduleCommit(); 1961} 1962 1963void HistoryBackend::SetImportedFavicons( 1964 const std::vector<ImportedFaviconUsage>& favicon_usage) { 1965 if (!db_ || !thumbnail_db_) 1966 return; 1967 1968 Time now = Time::Now(); 1969 1970 // Track all URLs that had their favicons set or updated. 1971 std::set<GURL> favicons_changed; 1972 1973 for (size_t i = 0; i < favicon_usage.size(); i++) { 1974 chrome::FaviconID favicon_id = thumbnail_db_->GetFaviconIDForFaviconURL( 1975 favicon_usage[i].favicon_url, chrome::FAVICON, NULL); 1976 if (!favicon_id) { 1977 // This favicon doesn't exist yet, so we create it using the given data. 1978 // TODO(pkotwicz): Pass in real pixel size. 1979 favicon_id = thumbnail_db_->AddFavicon( 1980 favicon_usage[i].favicon_url, 1981 chrome::FAVICON, 1982 new base::RefCountedBytes(favicon_usage[i].png_data), 1983 now, 1984 gfx::Size()); 1985 } 1986 1987 // Save the mapping from all the URLs to the favicon. 1988 BookmarkService* bookmark_service = GetBookmarkService(); 1989 for (std::set<GURL>::const_iterator url = favicon_usage[i].urls.begin(); 1990 url != favicon_usage[i].urls.end(); ++url) { 1991 URLRow url_row; 1992 if (!db_->GetRowForURL(*url, &url_row)) { 1993 // If the URL is present as a bookmark, add the url in history to 1994 // save the favicon mapping. This will match with what history db does 1995 // for regular bookmarked URLs with favicons - when history db is 1996 // cleaned, we keep an entry in the db with 0 visits as long as that 1997 // url is bookmarked. 1998 if (bookmark_service && bookmark_service_->IsBookmarked(*url)) { 1999 URLRow url_info(*url); 2000 url_info.set_visit_count(0); 2001 url_info.set_typed_count(0); 2002 url_info.set_last_visit(base::Time()); 2003 url_info.set_hidden(false); 2004 db_->AddURL(url_info); 2005 thumbnail_db_->AddIconMapping(*url, favicon_id); 2006 favicons_changed.insert(*url); 2007 } 2008 } else { 2009 if (!thumbnail_db_->GetIconMappingsForPageURL( 2010 *url, chrome::FAVICON, NULL)) { 2011 // URL is present in history, update the favicon *only* if it is not 2012 // set already. 2013 thumbnail_db_->AddIconMapping(*url, favicon_id); 2014 favicons_changed.insert(*url); 2015 } 2016 } 2017 } 2018 } 2019 2020 if (!favicons_changed.empty()) { 2021 // Send the notification about the changed favicon URLs. 2022 FaviconChangedDetails* changed_details = new FaviconChangedDetails; 2023 changed_details->urls.swap(favicons_changed); 2024 BroadcastNotifications(chrome::NOTIFICATION_FAVICON_CHANGED, 2025 changed_details); 2026 } 2027} 2028 2029void HistoryBackend::UpdateFaviconMappingsAndFetchImpl( 2030 const GURL* page_url, 2031 const std::vector<GURL>& icon_urls, 2032 int icon_types, 2033 int desired_size_in_dip, 2034 const std::vector<ui::ScaleFactor>& desired_scale_factors, 2035 std::vector<chrome::FaviconBitmapResult>* bitmap_results) { 2036 // If |page_url| is specified, |icon_types| must be either a single icon 2037 // type or icon types which are equivalent. 2038 DCHECK(!page_url || 2039 icon_types == chrome::FAVICON || 2040 icon_types == chrome::TOUCH_ICON || 2041 icon_types == chrome::TOUCH_PRECOMPOSED_ICON || 2042 icon_types == (chrome::TOUCH_ICON | chrome::TOUCH_PRECOMPOSED_ICON)); 2043 bitmap_results->clear(); 2044 2045 if (!thumbnail_db_) { 2046 return; 2047 } 2048 2049 std::vector<chrome::FaviconID> favicon_ids; 2050 2051 // The icon type for which the mappings will the updated and data will be 2052 // returned. 2053 chrome::IconType selected_icon_type = chrome::INVALID_ICON; 2054 2055 for (size_t i = 0; i < icon_urls.size(); ++i) { 2056 const GURL& icon_url = icon_urls[i]; 2057 chrome::IconType icon_type_out; 2058 const chrome::FaviconID favicon_id = 2059 thumbnail_db_->GetFaviconIDForFaviconURL( 2060 icon_url, icon_types, &icon_type_out); 2061 2062 if (favicon_id) { 2063 // Return and update icon mappings only for the largest icon type. As 2064 // |icon_urls| is not sorted in terms of icon type, clear |favicon_ids| 2065 // if an |icon_url| with a larger icon type is found. 2066 if (icon_type_out > selected_icon_type) { 2067 selected_icon_type = icon_type_out; 2068 favicon_ids.clear(); 2069 } 2070 if (icon_type_out == selected_icon_type) 2071 favicon_ids.push_back(favicon_id); 2072 } 2073 } 2074 2075 if (page_url && !favicon_ids.empty()) { 2076 bool mappings_updated = 2077 SetFaviconMappingsForPageAndRedirects(*page_url, selected_icon_type, 2078 favicon_ids); 2079 if (mappings_updated) { 2080 SendFaviconChangedNotificationForPageAndRedirects(*page_url); 2081 ScheduleCommit(); 2082 } 2083 } 2084 2085 GetFaviconBitmapResultsForBestMatch(favicon_ids, desired_size_in_dip, 2086 desired_scale_factors, bitmap_results); 2087} 2088 2089void HistoryBackend::SetFaviconBitmaps( 2090 chrome::FaviconID icon_id, 2091 const std::vector<chrome::FaviconBitmapData>& favicon_bitmap_data, 2092 bool* favicon_bitmaps_changed) { 2093 if (favicon_bitmaps_changed) 2094 *favicon_bitmaps_changed = false; 2095 2096 std::vector<FaviconBitmapIDSize> bitmap_id_sizes; 2097 thumbnail_db_->GetFaviconBitmapIDSizes(icon_id, &bitmap_id_sizes); 2098 2099 std::vector<chrome::FaviconBitmapData> to_add = favicon_bitmap_data; 2100 2101 for (size_t i = 0; i < bitmap_id_sizes.size(); ++i) { 2102 const gfx::Size& pixel_size = bitmap_id_sizes[i].pixel_size; 2103 std::vector<chrome::FaviconBitmapData>::iterator match_it = to_add.end(); 2104 for (std::vector<chrome::FaviconBitmapData>::iterator it = to_add.begin(); 2105 it != to_add.end(); ++it) { 2106 if (it->pixel_size == pixel_size) { 2107 match_it = it; 2108 break; 2109 } 2110 } 2111 2112 FaviconBitmapID bitmap_id = bitmap_id_sizes[i].bitmap_id; 2113 if (match_it == to_add.end()) { 2114 thumbnail_db_->DeleteFaviconBitmap(bitmap_id); 2115 2116 if (favicon_bitmaps_changed) 2117 *favicon_bitmaps_changed = true; 2118 } else { 2119 if (favicon_bitmaps_changed && 2120 !*favicon_bitmaps_changed && 2121 IsFaviconBitmapDataEqual(bitmap_id, match_it->bitmap_data)) { 2122 thumbnail_db_->SetFaviconBitmapLastUpdateTime( 2123 bitmap_id, base::Time::Now()); 2124 } else { 2125 thumbnail_db_->SetFaviconBitmap(bitmap_id, match_it->bitmap_data, 2126 base::Time::Now()); 2127 2128 if (favicon_bitmaps_changed) 2129 *favicon_bitmaps_changed = true; 2130 } 2131 to_add.erase(match_it); 2132 } 2133 } 2134 2135 for (size_t i = 0; i < to_add.size(); ++i) { 2136 thumbnail_db_->AddFaviconBitmap(icon_id, to_add[i].bitmap_data, 2137 base::Time::Now(), to_add[i].pixel_size); 2138 2139 if (favicon_bitmaps_changed) 2140 *favicon_bitmaps_changed = true; 2141 } 2142} 2143 2144bool HistoryBackend::ValidateSetFaviconsParams( 2145 const std::vector<chrome::FaviconBitmapData>& favicon_bitmap_data) const { 2146 typedef std::map<GURL, size_t> BitmapsPerIconURL; 2147 BitmapsPerIconURL num_bitmaps_per_icon_url; 2148 for (size_t i = 0; i < favicon_bitmap_data.size(); ++i) { 2149 if (!favicon_bitmap_data[i].bitmap_data.get()) 2150 return false; 2151 2152 const GURL& icon_url = favicon_bitmap_data[i].icon_url; 2153 if (!num_bitmaps_per_icon_url.count(icon_url)) 2154 num_bitmaps_per_icon_url[icon_url] = 1u; 2155 else 2156 ++num_bitmaps_per_icon_url[icon_url]; 2157 } 2158 2159 if (num_bitmaps_per_icon_url.size() > kMaxFaviconsPerPage) 2160 return false; 2161 2162 for (BitmapsPerIconURL::const_iterator it = num_bitmaps_per_icon_url.begin(); 2163 it != num_bitmaps_per_icon_url.end(); ++it) { 2164 if (it->second > kMaxFaviconBitmapsPerIconURL) 2165 return false; 2166 } 2167 return true; 2168} 2169 2170bool HistoryBackend::IsFaviconBitmapDataEqual( 2171 FaviconBitmapID bitmap_id, 2172 const scoped_refptr<base::RefCountedMemory>& new_bitmap_data) { 2173 if (!new_bitmap_data.get()) 2174 return false; 2175 2176 scoped_refptr<base::RefCountedMemory> original_bitmap_data; 2177 thumbnail_db_->GetFaviconBitmap(bitmap_id, 2178 NULL, 2179 &original_bitmap_data, 2180 NULL); 2181 return new_bitmap_data->Equals(original_bitmap_data); 2182} 2183 2184bool HistoryBackend::GetFaviconsFromDB( 2185 const GURL& page_url, 2186 int icon_types, 2187 int desired_size_in_dip, 2188 const std::vector<ui::ScaleFactor>& desired_scale_factors, 2189 std::vector<chrome::FaviconBitmapResult>* favicon_bitmap_results) { 2190 DCHECK(favicon_bitmap_results); 2191 favicon_bitmap_results->clear(); 2192 2193 if (!db_ || !thumbnail_db_) 2194 return false; 2195 2196 // Time the query. 2197 TimeTicks beginning_time = TimeTicks::Now(); 2198 2199 // Get FaviconIDs for |page_url| and one of |icon_types|. 2200 std::vector<IconMapping> icon_mappings; 2201 thumbnail_db_->GetIconMappingsForPageURL(page_url, icon_types, 2202 &icon_mappings); 2203 std::vector<chrome::FaviconID> favicon_ids; 2204 for (size_t i = 0; i < icon_mappings.size(); ++i) 2205 favicon_ids.push_back(icon_mappings[i].icon_id); 2206 2207 // Populate |favicon_bitmap_results| and |icon_url_sizes|. 2208 bool success = GetFaviconBitmapResultsForBestMatch(favicon_ids, 2209 desired_size_in_dip, desired_scale_factors, favicon_bitmap_results); 2210 UMA_HISTOGRAM_TIMES("History.GetFavIconFromDB", // historical name 2211 TimeTicks::Now() - beginning_time); 2212 return success && !favicon_bitmap_results->empty(); 2213} 2214 2215bool HistoryBackend::GetFaviconBitmapResultsForBestMatch( 2216 const std::vector<chrome::FaviconID>& candidate_favicon_ids, 2217 int desired_size_in_dip, 2218 const std::vector<ui::ScaleFactor>& desired_scale_factors, 2219 std::vector<chrome::FaviconBitmapResult>* favicon_bitmap_results) { 2220 favicon_bitmap_results->clear(); 2221 2222 if (candidate_favicon_ids.empty()) 2223 return true; 2224 2225 // Find the FaviconID and the FaviconBitmapIDs which best match 2226 // |desired_size_in_dip| and |desired_scale_factors|. 2227 // TODO(pkotwicz): Select bitmap results from multiple favicons once 2228 // content::FaviconStatus supports multiple icon URLs. 2229 chrome::FaviconID best_favicon_id = 0; 2230 std::vector<FaviconBitmapID> best_bitmap_ids; 2231 float highest_score = kSelectFaviconFramesInvalidScore; 2232 for (size_t i = 0; i < candidate_favicon_ids.size(); ++i) { 2233 std::vector<FaviconBitmapIDSize> bitmap_id_sizes; 2234 thumbnail_db_->GetFaviconBitmapIDSizes(candidate_favicon_ids[i], 2235 &bitmap_id_sizes); 2236 2237 // Build vector of gfx::Size from |bitmap_id_sizes|. 2238 std::vector<gfx::Size> sizes; 2239 for (size_t j = 0; j < bitmap_id_sizes.size(); ++j) 2240 sizes.push_back(bitmap_id_sizes[j].pixel_size); 2241 2242 std::vector<size_t> candidate_bitmap_indices; 2243 float score = 0; 2244 SelectFaviconFrameIndices(sizes, 2245 desired_scale_factors, 2246 desired_size_in_dip, 2247 &candidate_bitmap_indices, 2248 &score); 2249 if (score > highest_score) { 2250 highest_score = score; 2251 best_favicon_id = candidate_favicon_ids[i], 2252 best_bitmap_ids.clear(); 2253 for (size_t j = 0; j < candidate_bitmap_indices.size(); ++j) { 2254 size_t candidate_index = candidate_bitmap_indices[j]; 2255 best_bitmap_ids.push_back( 2256 bitmap_id_sizes[candidate_index].bitmap_id); 2257 } 2258 } 2259 } 2260 2261 // Construct FaviconBitmapResults from |best_favicon_id| and 2262 // |best_bitmap_ids|. 2263 GURL icon_url; 2264 chrome::IconType icon_type; 2265 if (!thumbnail_db_->GetFaviconHeader(best_favicon_id, &icon_url, 2266 &icon_type)) { 2267 return false; 2268 } 2269 2270 for (size_t i = 0; i < best_bitmap_ids.size(); ++i) { 2271 base::Time last_updated; 2272 chrome::FaviconBitmapResult bitmap_result; 2273 bitmap_result.icon_url = icon_url; 2274 bitmap_result.icon_type = icon_type; 2275 if (!thumbnail_db_->GetFaviconBitmap(best_bitmap_ids[i], 2276 &last_updated, 2277 &bitmap_result.bitmap_data, 2278 &bitmap_result.pixel_size)) { 2279 return false; 2280 } 2281 2282 bitmap_result.expired = (Time::Now() - last_updated) > 2283 TimeDelta::FromDays(kFaviconRefetchDays); 2284 if (bitmap_result.is_valid()) 2285 favicon_bitmap_results->push_back(bitmap_result); 2286 } 2287 return true; 2288} 2289 2290bool HistoryBackend::SetFaviconMappingsForPageAndRedirects( 2291 const GURL& page_url, 2292 chrome::IconType icon_type, 2293 const std::vector<chrome::FaviconID>& icon_ids) { 2294 if (!thumbnail_db_) 2295 return false; 2296 2297 // Find all the pages whose favicons we should set, we want to set it for 2298 // all the pages in the redirect chain if it redirected. 2299 history::RedirectList redirects; 2300 GetCachedRecentRedirects(page_url, &redirects); 2301 2302 bool mappings_changed = false; 2303 2304 // Save page <-> favicon associations. 2305 for (history::RedirectList::const_iterator i(redirects.begin()); 2306 i != redirects.end(); ++i) { 2307 mappings_changed |= SetFaviconMappingsForPage(*i, icon_type, icon_ids); 2308 } 2309 return mappings_changed; 2310} 2311 2312bool HistoryBackend::SetFaviconMappingsForPage( 2313 const GURL& page_url, 2314 chrome::IconType icon_type, 2315 const std::vector<chrome::FaviconID>& icon_ids) { 2316 DCHECK_LE(icon_ids.size(), kMaxFaviconsPerPage); 2317 bool mappings_changed = false; 2318 2319 // Two icon types are considered 'equivalent' if one of the icon types is 2320 // TOUCH_ICON and the other is TOUCH_PRECOMPOSED_ICON. 2321 // 2322 // Sets the icon mappings from |page_url| for |icon_type| to the favicons 2323 // with |icon_ids|. Mappings for |page_url| to favicons of type |icon_type| 2324 // whose FaviconID is not in |icon_ids| are removed. All icon mappings for 2325 // |page_url| to favicons of a type equivalent to |icon_type| are removed. 2326 // Remove any favicons which are orphaned as a result of the removal of the 2327 // icon mappings. 2328 2329 std::vector<chrome::FaviconID> unmapped_icon_ids = icon_ids; 2330 2331 std::vector<IconMapping> icon_mappings; 2332 thumbnail_db_->GetIconMappingsForPageURL(page_url, &icon_mappings); 2333 2334 for (std::vector<IconMapping>::iterator m = icon_mappings.begin(); 2335 m != icon_mappings.end(); ++m) { 2336 std::vector<chrome::FaviconID>::iterator icon_id_it = std::find( 2337 unmapped_icon_ids.begin(), unmapped_icon_ids.end(), m->icon_id); 2338 2339 // If the icon mapping already exists, avoid removing it and adding it back. 2340 if (icon_id_it != unmapped_icon_ids.end()) { 2341 unmapped_icon_ids.erase(icon_id_it); 2342 continue; 2343 } 2344 2345 if ((icon_type == chrome::TOUCH_ICON && 2346 m->icon_type == chrome::TOUCH_PRECOMPOSED_ICON) || 2347 (icon_type == chrome::TOUCH_PRECOMPOSED_ICON && 2348 m->icon_type == chrome::TOUCH_ICON) || (icon_type == m->icon_type)) { 2349 thumbnail_db_->DeleteIconMapping(m->mapping_id); 2350 2351 // Removing the icon mapping may have orphaned the associated favicon so 2352 // we must recheck it. This is not super fast, but this case will get 2353 // triggered rarely, since normally a page will always map to the same 2354 // favicon IDs. It will mostly happen for favicons we import. 2355 if (!thumbnail_db_->HasMappingFor(m->icon_id)) 2356 thumbnail_db_->DeleteFavicon(m->icon_id); 2357 mappings_changed = true; 2358 } 2359 } 2360 2361 for (size_t i = 0; i < unmapped_icon_ids.size(); ++i) { 2362 thumbnail_db_->AddIconMapping(page_url, unmapped_icon_ids[i]); 2363 mappings_changed = true; 2364 } 2365 return mappings_changed; 2366} 2367 2368void HistoryBackend::GetCachedRecentRedirects( 2369 const GURL& page_url, 2370 history::RedirectList* redirect_list) { 2371 RedirectCache::iterator iter = recent_redirects_.Get(page_url); 2372 if (iter != recent_redirects_.end()) { 2373 *redirect_list = iter->second; 2374 2375 // The redirect chain should have the destination URL as the last item. 2376 DCHECK(!redirect_list->empty()); 2377 DCHECK(redirect_list->back() == page_url); 2378 } else { 2379 // No known redirects, construct mock redirect chain containing |page_url|. 2380 redirect_list->push_back(page_url); 2381 } 2382} 2383 2384void HistoryBackend::SendFaviconChangedNotificationForPageAndRedirects( 2385 const GURL& page_url) { 2386 history::RedirectList redirect_list; 2387 GetCachedRecentRedirects(page_url, &redirect_list); 2388 2389 FaviconChangedDetails* changed_details = new FaviconChangedDetails; 2390 for (size_t i = 0; i < redirect_list.size(); ++i) 2391 changed_details->urls.insert(redirect_list[i]); 2392 2393 BroadcastNotifications(chrome::NOTIFICATION_FAVICON_CHANGED, 2394 changed_details); 2395} 2396 2397void HistoryBackend::Commit() { 2398 if (!db_) 2399 return; 2400 2401 // Note that a commit may not actually have been scheduled if a caller 2402 // explicitly calls this instead of using ScheduleCommit. Likewise, we 2403 // may reset the flag written by a pending commit. But this is OK! It 2404 // will merely cause extra commits (which is kind of the idea). We 2405 // could optimize more for this case (we may get two extra commits in 2406 // some cases) but it hasn't been important yet. 2407 CancelScheduledCommit(); 2408 2409 db_->CommitTransaction(); 2410 DCHECK(db_->transaction_nesting() == 0) << "Somebody left a transaction open"; 2411 db_->BeginTransaction(); 2412 2413 if (thumbnail_db_) { 2414 thumbnail_db_->CommitTransaction(); 2415 DCHECK(thumbnail_db_->transaction_nesting() == 0) << 2416 "Somebody left a transaction open"; 2417 thumbnail_db_->BeginTransaction(); 2418 } 2419 2420 if (archived_db_) { 2421 archived_db_->CommitTransaction(); 2422 archived_db_->BeginTransaction(); 2423 } 2424} 2425 2426void HistoryBackend::ScheduleCommit() { 2427 if (scheduled_commit_.get()) 2428 return; 2429 scheduled_commit_ = new CommitLaterTask(this); 2430 base::MessageLoop::current()->PostDelayedTask( 2431 FROM_HERE, 2432 base::Bind(&CommitLaterTask::RunCommit, scheduled_commit_.get()), 2433 base::TimeDelta::FromSeconds(kCommitIntervalSeconds)); 2434} 2435 2436void HistoryBackend::CancelScheduledCommit() { 2437 if (scheduled_commit_.get()) { 2438 scheduled_commit_->Cancel(); 2439 scheduled_commit_ = NULL; 2440 } 2441} 2442 2443void HistoryBackend::ProcessDBTaskImpl() { 2444 if (!db_) { 2445 // db went away, release all the refs. 2446 ReleaseDBTasks(); 2447 return; 2448 } 2449 2450 // Remove any canceled tasks. 2451 while (!db_task_requests_.empty() && db_task_requests_.front()->canceled()) { 2452 db_task_requests_.front()->Release(); 2453 db_task_requests_.pop_front(); 2454 } 2455 if (db_task_requests_.empty()) 2456 return; 2457 2458 // Run the first task. 2459 HistoryDBTaskRequest* request = db_task_requests_.front(); 2460 db_task_requests_.pop_front(); 2461 if (request->value->RunOnDBThread(this, db_.get())) { 2462 // The task is done. Notify the callback. 2463 request->ForwardResult(); 2464 // We AddRef'd the request before adding, need to release it now. 2465 request->Release(); 2466 } else { 2467 // Tasks wants to run some more. Schedule it at the end of current tasks. 2468 db_task_requests_.push_back(request); 2469 // And process it after an invoke later. 2470 base::MessageLoop::current()->PostTask( 2471 FROM_HERE, base::Bind(&HistoryBackend::ProcessDBTaskImpl, this)); 2472 } 2473} 2474 2475void HistoryBackend::ReleaseDBTasks() { 2476 for (std::list<HistoryDBTaskRequest*>::iterator i = 2477 db_task_requests_.begin(); i != db_task_requests_.end(); ++i) { 2478 (*i)->Release(); 2479 } 2480 db_task_requests_.clear(); 2481} 2482 2483//////////////////////////////////////////////////////////////////////////////// 2484// 2485// Generic operations 2486// 2487//////////////////////////////////////////////////////////////////////////////// 2488 2489void HistoryBackend::DeleteURLs(const std::vector<GURL>& urls) { 2490 expirer_.DeleteURLs(urls); 2491 2492 db_->GetStartDate(&first_recorded_time_); 2493 // Force a commit, if the user is deleting something for privacy reasons, we 2494 // want to get it on disk ASAP. 2495 Commit(); 2496} 2497 2498void HistoryBackend::DeleteURL(const GURL& url) { 2499 expirer_.DeleteURL(url); 2500 2501 db_->GetStartDate(&first_recorded_time_); 2502 // Force a commit, if the user is deleting something for privacy reasons, we 2503 // want to get it on disk ASAP. 2504 Commit(); 2505} 2506 2507void HistoryBackend::ExpireHistoryBetween( 2508 const std::set<GURL>& restrict_urls, 2509 Time begin_time, 2510 Time end_time) { 2511 if (!db_) 2512 return; 2513 2514 if (begin_time.is_null() && (end_time.is_null() || end_time.is_max()) && 2515 restrict_urls.empty()) { 2516 // Special case deleting all history so it can be faster and to reduce the 2517 // possibility of an information leak. 2518 DeleteAllHistory(); 2519 } else { 2520 // Clearing parts of history, have the expirer do the depend 2521 expirer_.ExpireHistoryBetween(restrict_urls, begin_time, end_time); 2522 2523 // Force a commit, if the user is deleting something for privacy reasons, 2524 // we want to get it on disk ASAP. 2525 Commit(); 2526 } 2527 2528 if (begin_time <= first_recorded_time_) 2529 db_->GetStartDate(&first_recorded_time_); 2530} 2531 2532void HistoryBackend::ExpireHistoryForTimes( 2533 const std::set<base::Time>& times, 2534 base::Time begin_time, base::Time end_time) { 2535 if (times.empty() || !db_) 2536 return; 2537 2538 DCHECK(*times.begin() >= begin_time) 2539 << "Min time is before begin time: " 2540 << times.begin()->ToJsTime() << " v.s. " << begin_time.ToJsTime(); 2541 DCHECK(*times.rbegin() < end_time) 2542 << "Max time is after end time: " 2543 << times.rbegin()->ToJsTime() << " v.s. " << end_time.ToJsTime(); 2544 2545 history::QueryOptions options; 2546 options.begin_time = begin_time; 2547 options.end_time = end_time; 2548 options.duplicate_policy = QueryOptions::KEEP_ALL_DUPLICATES; 2549 QueryResults results; 2550 QueryHistoryBasic(db_.get(), db_.get(), options, &results); 2551 2552 // 1st pass: find URLs that are visited at one of |times|. 2553 std::set<GURL> urls; 2554 for (size_t i = 0; i < results.size(); ++i) { 2555 if (times.count(results[i].visit_time()) > 0) 2556 urls.insert(results[i].url()); 2557 } 2558 if (urls.empty()) 2559 return; 2560 2561 // 2nd pass: collect all visit times of those URLs. 2562 std::vector<base::Time> times_to_expire; 2563 for (size_t i = 0; i < results.size(); ++i) { 2564 if (urls.count(results[i].url())) 2565 times_to_expire.push_back(results[i].visit_time()); 2566 } 2567 2568 // Put the times in reverse chronological order and remove 2569 // duplicates (for expirer_.ExpireHistoryForTimes()). 2570 std::sort(times_to_expire.begin(), times_to_expire.end(), 2571 std::greater<base::Time>()); 2572 times_to_expire.erase( 2573 std::unique(times_to_expire.begin(), times_to_expire.end()), 2574 times_to_expire.end()); 2575 2576 // Expires by times and commit. 2577 DCHECK(!times_to_expire.empty()); 2578 expirer_.ExpireHistoryForTimes(times_to_expire); 2579 Commit(); 2580 2581 DCHECK(times_to_expire.back() >= first_recorded_time_); 2582 // Update |first_recorded_time_| if we expired it. 2583 if (times_to_expire.back() == first_recorded_time_) 2584 db_->GetStartDate(&first_recorded_time_); 2585} 2586 2587void HistoryBackend::ExpireHistory( 2588 const std::vector<history::ExpireHistoryArgs>& expire_list) { 2589 if (db_) { 2590 bool update_first_recorded_time = false; 2591 2592 for (std::vector<history::ExpireHistoryArgs>::const_iterator it = 2593 expire_list.begin(); it != expire_list.end(); ++it) { 2594 expirer_.ExpireHistoryBetween(it->urls, it->begin_time, it->end_time); 2595 2596 if (it->begin_time < first_recorded_time_) 2597 update_first_recorded_time = true; 2598 } 2599 Commit(); 2600 2601 // Update |first_recorded_time_| if any deletion might have affected it. 2602 if (update_first_recorded_time) 2603 db_->GetStartDate(&first_recorded_time_); 2604 } 2605} 2606 2607void HistoryBackend::URLsNoLongerBookmarked(const std::set<GURL>& urls) { 2608 if (!db_) 2609 return; 2610 2611 for (std::set<GURL>::const_iterator i = urls.begin(); i != urls.end(); ++i) { 2612 URLRow url_row; 2613 if (!db_->GetRowForURL(*i, &url_row)) 2614 continue; // The URL isn't in the db; nothing to do. 2615 2616 VisitVector visits; 2617 db_->GetVisitsForURL(url_row.id(), &visits); 2618 2619 if (visits.empty()) 2620 expirer_.DeleteURL(*i); // There are no more visits; nuke the URL. 2621 } 2622} 2623 2624void HistoryBackend::DatabaseErrorCallback(int error, sql::Statement* stmt) { 2625 if (!scheduled_kill_db_ && sql::IsErrorCatastrophic(error)) { 2626 scheduled_kill_db_ = true; 2627 // Don't just do the close/delete here, as we are being called by |db| and 2628 // that seems dangerous. 2629 // TODO(shess): Consider changing KillHistoryDatabase() to use 2630 // RazeAndClose(). Then it can be cleared immediately. 2631 base::MessageLoop::current()->PostTask( 2632 FROM_HERE, 2633 base::Bind(&HistoryBackend::KillHistoryDatabase, this)); 2634 } 2635} 2636 2637void HistoryBackend::KillHistoryDatabase() { 2638 scheduled_kill_db_ = false; 2639 if (!db_) 2640 return; 2641 2642 // Rollback transaction because Raze() cannot be called from within a 2643 // transaction. 2644 db_->RollbackTransaction(); 2645 bool success = db_->Raze(); 2646 UMA_HISTOGRAM_BOOLEAN("History.KillHistoryDatabaseResult", success); 2647 2648#if defined(OS_ANDROID) 2649 // Release AndroidProviderBackend before other objects. 2650 android_provider_backend_.reset(); 2651#endif 2652 2653 // The expirer keeps tabs on the active databases. Tell it about the 2654 // databases which will be closed. 2655 expirer_.SetDatabases(NULL, NULL, NULL); 2656 2657 // Reopen a new transaction for |db_| for the sake of CloseAllDatabases(). 2658 db_->BeginTransaction(); 2659 CloseAllDatabases(); 2660} 2661 2662void HistoryBackend::ProcessDBTask( 2663 scoped_refptr<HistoryDBTaskRequest> request) { 2664 DCHECK(request.get()); 2665 if (request->canceled()) 2666 return; 2667 2668 bool task_scheduled = !db_task_requests_.empty(); 2669 // Make sure we up the refcount of the request. ProcessDBTaskImpl will 2670 // release when done with the task. 2671 request->AddRef(); 2672 db_task_requests_.push_back(request.get()); 2673 if (!task_scheduled) { 2674 // No other tasks are scheduled. Process request now. 2675 ProcessDBTaskImpl(); 2676 } 2677} 2678 2679void HistoryBackend::BroadcastNotifications( 2680 int type, 2681 HistoryDetails* details_deleted) { 2682 // |delegate_| may be NULL if |this| is in the process of closing (closed by 2683 // HistoryService -> HistoryBackend::Closing(). 2684 if (delegate_) 2685 delegate_->BroadcastNotifications(type, details_deleted); 2686 else 2687 delete details_deleted; 2688} 2689 2690void HistoryBackend::NotifySyncURLsDeleted(bool all_history, 2691 bool archived, 2692 URLRows* rows) { 2693 if (typed_url_syncable_service_.get()) 2694 typed_url_syncable_service_->OnUrlsDeleted(all_history, archived, rows); 2695} 2696 2697// Deleting -------------------------------------------------------------------- 2698 2699void HistoryBackend::DeleteAllHistory() { 2700 // Our approach to deleting all history is: 2701 // 1. Copy the bookmarks and their dependencies to new tables with temporary 2702 // names. 2703 // 2. Delete the original tables. Since tables can not share pages, we know 2704 // that any data we don't want to keep is now in an unused page. 2705 // 3. Renaming the temporary tables to match the original. 2706 // 4. Vacuuming the database to delete the unused pages. 2707 // 2708 // Since we are likely to have very few bookmarks and their dependencies 2709 // compared to all history, this is also much faster than just deleting from 2710 // the original tables directly. 2711 2712 // Get the bookmarked URLs. 2713 std::vector<BookmarkService::URLAndTitle> starred_urls; 2714 BookmarkService* bookmark_service = GetBookmarkService(); 2715 if (bookmark_service) 2716 bookmark_service_->GetBookmarks(&starred_urls); 2717 2718 URLRows kept_urls; 2719 for (size_t i = 0; i < starred_urls.size(); i++) { 2720 URLRow row; 2721 if (!db_->GetRowForURL(starred_urls[i].url, &row)) 2722 continue; 2723 2724 // Clear the last visit time so when we write these rows they are "clean." 2725 row.set_last_visit(Time()); 2726 row.set_visit_count(0); 2727 row.set_typed_count(0); 2728 kept_urls.push_back(row); 2729 } 2730 2731 // Clear thumbnail and favicon history. The favicons for the given URLs will 2732 // be kept. 2733 if (!ClearAllThumbnailHistory(kept_urls)) { 2734 LOG(ERROR) << "Thumbnail history could not be cleared"; 2735 // We continue in this error case. If the user wants to delete their 2736 // history, we should delete as much as we can. 2737 } 2738 2739 // ClearAllMainHistory will change the IDs of the URLs in kept_urls. 2740 // Therefore, we clear the list afterwards to make sure nobody uses this 2741 // invalid data. 2742 if (!ClearAllMainHistory(kept_urls)) 2743 LOG(ERROR) << "Main history could not be cleared"; 2744 kept_urls.clear(); 2745 2746 // Delete archived history. 2747 if (archived_db_) { 2748 // Close the database and delete the file. 2749 archived_db_.reset(); 2750 base::FilePath archived_file_name = GetArchivedFileName(); 2751 sql::Connection::Delete(archived_file_name); 2752 2753 // Now re-initialize the database (which may fail). 2754 archived_db_.reset(new ArchivedDatabase()); 2755 if (!archived_db_->Init(archived_file_name)) { 2756 LOG(WARNING) << "Could not initialize the archived database."; 2757 archived_db_.reset(); 2758 } else { 2759 // Open our long-running transaction on this database. 2760 archived_db_->BeginTransaction(); 2761 } 2762 } 2763 2764 db_->GetStartDate(&first_recorded_time_); 2765 2766 // Send out the notification that history is cleared. The in-memory database 2767 // will pick this up and clear itself. 2768 URLsDeletedDetails* details = new URLsDeletedDetails; 2769 details->all_history = true; 2770 NotifySyncURLsDeleted(true, false, NULL); 2771 BroadcastNotifications(chrome::NOTIFICATION_HISTORY_URLS_DELETED, details); 2772} 2773 2774bool HistoryBackend::ClearAllThumbnailHistory(const URLRows& kept_urls) { 2775 if (!thumbnail_db_) { 2776 // When we have no reference to the thumbnail database, maybe there was an 2777 // error opening it. In this case, we just try to blow it away to try to 2778 // fix the error if it exists. This may fail, in which case either the 2779 // file doesn't exist or there's no more we can do. 2780 sql::Connection::Delete(GetFaviconsFileName()); 2781 2782 // Older version of the database. 2783 sql::Connection::Delete(GetThumbnailFileName()); 2784 return true; 2785 } 2786 2787 // Urls to retain mappings for. 2788 std::vector<GURL> urls_to_keep; 2789 for (URLRows::const_iterator i = kept_urls.begin(); 2790 i != kept_urls.end(); ++i) { 2791 urls_to_keep.push_back(i->url()); 2792 } 2793 2794 // Isolate from any long-running transaction. 2795 thumbnail_db_->CommitTransaction(); 2796 thumbnail_db_->BeginTransaction(); 2797 2798 // TODO(shess): If this fails, perhaps the database should be razed 2799 // or deleted. 2800 if (!thumbnail_db_->RetainDataForPageUrls(urls_to_keep)) { 2801 thumbnail_db_->RollbackTransaction(); 2802 thumbnail_db_->BeginTransaction(); 2803 return false; 2804 } 2805 2806#if defined(OS_ANDROID) 2807 // TODO (michaelbai): Add the unit test once AndroidProviderBackend is 2808 // avaliable in HistoryBackend. 2809 db_->ClearAndroidURLRows(); 2810#endif 2811 2812 // Vacuum to remove all the pages associated with the dropped tables. There 2813 // must be no transaction open on the table when we do this. We assume that 2814 // our long-running transaction is open, so we complete it and start it again. 2815 DCHECK(thumbnail_db_->transaction_nesting() == 1); 2816 thumbnail_db_->CommitTransaction(); 2817 thumbnail_db_->Vacuum(); 2818 thumbnail_db_->BeginTransaction(); 2819 return true; 2820} 2821 2822bool HistoryBackend::ClearAllMainHistory(const URLRows& kept_urls) { 2823 // Create the duplicate URL table. We will copy the kept URLs into this. 2824 if (!db_->CreateTemporaryURLTable()) 2825 return false; 2826 2827 // Insert the URLs into the temporary table. 2828 for (URLRows::const_iterator i = kept_urls.begin(); i != kept_urls.end(); 2829 ++i) { 2830 db_->AddTemporaryURL(*i); 2831 } 2832 2833 // Replace the original URL table with the temporary one. 2834 if (!db_->CommitTemporaryURLTable()) 2835 return false; 2836 2837 // Delete the old tables and recreate them empty. 2838 db_->RecreateAllTablesButURL(); 2839 2840 // Vacuum to reclaim the space from the dropped tables. This must be done 2841 // when there is no transaction open, and we assume that our long-running 2842 // transaction is currently open. 2843 db_->CommitTransaction(); 2844 db_->Vacuum(); 2845 db_->BeginTransaction(); 2846 db_->GetStartDate(&first_recorded_time_); 2847 2848 return true; 2849} 2850 2851BookmarkService* HistoryBackend::GetBookmarkService() { 2852 if (bookmark_service_) 2853 bookmark_service_->BlockTillLoaded(); 2854 return bookmark_service_; 2855} 2856 2857void HistoryBackend::NotifyVisitObservers(const VisitRow& visit) { 2858 BriefVisitInfo info; 2859 info.url_id = visit.url_id; 2860 info.time = visit.visit_time; 2861 info.transition = visit.transition; 2862 // If we don't have a delegate yet during setup or shutdown, we will drop 2863 // these notifications. 2864 if (delegate_) 2865 delegate_->NotifyVisitDBObserversOnAddVisit(info); 2866} 2867 2868#if defined(OS_ANDROID) 2869void HistoryBackend::PopulateMostVisitedURLMap() { 2870 MostVisitedURLList most_visited_urls; 2871 QueryMostVisitedURLsImpl(kPageVisitStatsMaxTopSites, kSegmentDataRetention, 2872 &most_visited_urls); 2873 2874 DCHECK_LE(most_visited_urls.size(), kPageVisitStatsMaxTopSites); 2875 for (size_t i = 0; i < most_visited_urls.size(); ++i) { 2876 most_visited_urls_map_[most_visited_urls[i].url] = i; 2877 for (size_t j = 0; j < most_visited_urls[i].redirects.size(); ++j) 2878 most_visited_urls_map_[most_visited_urls[i].redirects[j]] = i; 2879 } 2880} 2881 2882void HistoryBackend::RecordTopPageVisitStats(const GURL& url) { 2883 int rank = kPageVisitStatsMaxTopSites; 2884 std::map<GURL, int>::const_iterator it = most_visited_urls_map_.find(url); 2885 if (it != most_visited_urls_map_.end()) 2886 rank = (*it).second; 2887 UMA_HISTOGRAM_ENUMERATION("History.TopSitesVisitsByRank", 2888 rank, kPageVisitStatsMaxTopSites + 1); 2889} 2890#endif 2891 2892} // namespace history 2893