// Copyright (c) 2011 The Chromium Authors. All rights reserved. // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. #ifndef CHROME_BROWSER_HISTORY_TEXT_DATABASE_MANAGER_H_ #define CHROME_BROWSER_HISTORY_TEXT_DATABASE_MANAGER_H_ #pragma once #include <set> #include <vector> #include "base/basictypes.h" #include "base/file_path.h" #include "base/gtest_prod_util.h" #include "base/string16.h" #include "base/task.h" #include "chrome/browser/history/history_types.h" #include "chrome/browser/history/text_database.h" #include "chrome/browser/history/query_parser.h" #include "chrome/browser/history/url_database.h" #include "content/common/mru_cache.h" namespace history { class HistoryPublisher; class VisitDatabase; // Manages a set of text databases representing different time periods. This // will page them in and out as necessary, and will manage queries for times // spanning multiple databases. // // It will also keep a list of partial changes, such as page adds and title and // body sets, all of which come in at different times for a given page. When // all data is received or enough time has elapsed since adding, the indexed // data will be comitted. // // This allows us to minimize inserts and modifications, which are slow for the // full text database, since each page's information is added exactly once. // // Note: be careful to delete the relevant entries from this uncommitted list // when clearing history or this information may get added to the database soon // after the clear. class TextDatabaseManager { public: // Tracks a set of changes (only deletes need to be supported now) to the // databases. This is opaque to the caller, but allows it to pass back a list // of all database that it has caused a change to. // // This is necessary for the feature where we optimize full text databases // which have changed as a result of the user deleting history via // OptimizeChangedDatabases. We want to do each affected database only once at // the end of the delete, but we don't want the caller to have to worry about // our internals. class ChangeSet { public: ChangeSet(); ~ChangeSet(); private: friend class TextDatabaseManager; typedef std::set<TextDatabase::DBIdent> DBSet; void Add(TextDatabase::DBIdent id) { changed_databases_.insert(id); } DBSet changed_databases_; }; // You must call Init() to complete initialization. // // |dir| is the directory that will hold the full text database files (there // will be many files named by their date ranges). // // The visit database is a pointer owned by the caller for the main database // (of recent visits). The visit database will be updated to refer to the // added text database entries. explicit TextDatabaseManager(const FilePath& dir, URLDatabase* url_database, VisitDatabase* visit_database); ~TextDatabaseManager(); // Must call before using other functions. If it returns false, no other // functions should be called. bool Init(const HistoryPublisher* history_publisher); // Returns the directory that holds the full text database files. const FilePath& GetDir() { return dir_; } // Allows scoping updates. This also allows things to go faster since every // page add doesn't need to be committed to disk (slow). Note that files will // still get created during a transaction. void BeginTransaction(); void CommitTransaction(); // Sets specific information for the given page to be added to the database. // In normal operation, URLs will be added as the user visits them, the titles // and bodies will come in some time after that. These changes will be // automatically coalesced and added to the database some time in the future // using AddPageData(). // // AddPageURL must be called for a given URL (+ its corresponding ID) before // either the title or body set. The visit ID specifies the visit that will // get updated to refer to the full text indexed information. The visit time // should be the time corresponding to that visit in the database. void AddPageURL(const GURL& url, URLID url_id, VisitID visit_id, base::Time visit_time); void AddPageTitle(const GURL& url, const string16& title); void AddPageContents(const GURL& url, const string16& body); // Adds the given data to the appropriate database file, returning true on // success. The visit database row identified by |visit_id| will be updated // to refer to the full text index entry. If the visit ID is 0, the visit // database will not be updated. bool AddPageData(const GURL& url, URLID url_id, VisitID visit_id, base::Time visit_time, const string16& title, const string16& body); // Deletes the instance of indexed data identified by the given time and URL. // Any changes will be tracked in the optional change set for use when calling // OptimizeChangedDatabases later. change_set can be NULL. void DeletePageData(base::Time time, const GURL& url, ChangeSet* change_set); // The text database manager keeps a list of changes that are made to the // file AddPageURL/Title/Body that may not be committed to the database yet. // This function removes entires from this list happening between the given // time range. It is called when the user clears their history for a time // range, and we don't want any of our data to "leak." If restrict_urls is // not empty, only changes on those URLs are deleted. // // Either or both times my be is_null to be unbounded in that direction. When // non-null, the range is [begin, end). void DeleteFromUncommitted(const std::set<GURL>& restrict_urls, base::Time begin, base::Time end); // Deletes all full text search data by removing the files from the disk. // This must be called OUTSIDE of a transaction since it actually deletes the // files rather than messing with the database. void DeleteAll(); // Calls optimize on all the databases identified in a given change set (see // the definition of ChangeSet above for more). Optimizing means that old data // will be removed rather than marked unused. void OptimizeChangedDatabases(const ChangeSet& change_set); // Executes the given query. See QueryOptions for more info on input. // // The results are filled into |results|, and the first time considered for // the output is in |first_time_searched| (see QueryResults for more). // // This function will return more than one match per URL if there is more than // one entry for that URL in the database. void GetTextMatches(const string16& query, const QueryOptions& options, std::vector<TextDatabase::Match>* results, base::Time* first_time_searched); private: // These tests call ExpireRecentChangesForTime to force expiration. FRIEND_TEST_ALL_PREFIXES(TextDatabaseManagerTest, InsertPartial); FRIEND_TEST_ALL_PREFIXES(TextDatabaseManagerTest, PartialComplete); FRIEND_TEST_ALL_PREFIXES(ExpireHistoryTest, DeleteURLAndFavicon); FRIEND_TEST_ALL_PREFIXES(ExpireHistoryTest, FlushRecentURLsUnstarred); FRIEND_TEST_ALL_PREFIXES(ExpireHistoryTest, FlushRecentURLsUnstarredRestricted); // Stores "recent stuff" that has happened with the page, since the page // visit, title, and body all come in at different times. class PageInfo { public: PageInfo(URLID url_id, VisitID visit_id, base::Time visit_time); ~PageInfo(); // Getters. URLID url_id() const { return url_id_; } VisitID visit_id() const { return visit_id_; } base::Time visit_time() const { return visit_time_; } const string16& title() const { return title_; } const string16& body() const { return body_; } // Setters, we can only update the title and body. void set_title(const string16& ttl); void set_body(const string16& bdy); // Returns true if both the title or body of the entry has been set. Since // both the title and body setters will "fix" empty strings to be a space, // these indicate if the setter was ever called. bool has_title() const { return !title_.empty(); } bool has_body() { return !body_.empty(); } // Returns true if this entry was added too long ago and we should give up // waiting for more data. The current time is passed in as an argument so we // can check many without re-querying the timer. bool Expired(base::TimeTicks now) const; private: URLID url_id_; VisitID visit_id_; // Time of the visit of the URL. This will be the value stored in the URL // and visit tables for the entry. base::Time visit_time_; // When this page entry was created. We have a cap on the maximum time that // an entry will be in the queue before being flushed to the database. base::TimeTicks added_time_; // Will be the string " " when they are set to distinguish set and unset. string16 title_; string16 body_; }; // Converts the given time to a database identifier or vice-versa. static TextDatabase::DBIdent TimeToID(base::Time time); static base::Time IDToTime(TextDatabase::DBIdent id); // Returns a text database for the given identifier or time. This file will // be created if it doesn't exist and |for_writing| is set. On error, // including the case where the file doesn't exist and |for_writing| // is false, it will return NULL. // // When |for_writing| is set, a transaction on the database will be opened // if there is a transaction open on this manager. // // The pointer will be tracked in the cache. The caller should not store it // or delete it since it will get automatically deleted as necessary. TextDatabase* GetDB(TextDatabase::DBIdent id, bool for_writing); TextDatabase* GetDBForTime(base::Time time, bool for_writing); // Populates the present_databases_ list based on which files are on disk. // When the list is already initialized, this will do nothing, so you can // call it whenever you want to ensure the present_databases_ set is filled. void InitDBList(); // Schedules a call to ExpireRecentChanges in the future. void ScheduleFlushOldChanges(); // Checks the recent_changes_ list and commits partial data that has been // around too long. void FlushOldChanges(); // Given "now," this will expire old things from the recent_changes_ list. // This is used as the backend for FlushOldChanges and is called directly // by the unit tests with fake times. void FlushOldChangesForTime(base::TimeTicks now); // Directory holding our index files. const FilePath dir_; // Non-owning pointers to the recent history databases for URLs and visits. URLDatabase* url_database_; VisitDatabase* visit_database_; // Lists recent additions that we have not yet filled out with the title and // body. Sorted by time, we will flush them when they are complete or have // been in the queue too long without modification. // // We kind of abuse the MRUCache because we never move things around in it // using Get. Instead, we keep them in the order they were inserted, since // this is the metric we use to measure age. The MRUCache gives us an ordered // list with fast lookup by URL. typedef MRUCache<GURL, PageInfo> RecentChangeList; RecentChangeList recent_changes_; // Nesting levels of transactions. Since sqlite only allows one open // transaction, we simulate nested transactions by mapping the outermost one // to a real transaction. Since this object never needs to do ROLLBACK, losing // the ability for all transactions to rollback is inconsequential. int transaction_nesting_; // The cache owns the TextDatabase pointers, they will be automagically // deleted when the cache entry is removed or expired. typedef OwningMRUCache<TextDatabase::DBIdent, TextDatabase*> DBCache; DBCache db_cache_; // Tells us about the existence of database files on disk. All existing // databases will be in here, and non-existant ones will not, so we don't // have to check the disk every time. // // This set is populated LAZILY by InitDBList(), you should call that function // before accessing the list. // // Note that iterators will work on the keys in-order. Normally, reverse // iterators will be used to iterate the keys in reverse-order. typedef std::set<TextDatabase::DBIdent> DBIdentSet; DBIdentSet present_databases_; bool present_databases_loaded_; // Set by InitDBList when populated. // Lists all databases with open transactions. These will have to be closed // when the transaction is committed. DBIdentSet open_transactions_; QueryParser query_parser_; // Generates tasks for our periodic checking of expired "recent changes". ScopedRunnableMethodFactory<TextDatabaseManager> factory_; // This object is created and managed by the history backend. We maintain an // opaque pointer to the object for our use. // This can be NULL if there are no indexers registered to receive indexing // data from us. const HistoryPublisher* history_publisher_; DISALLOW_COPY_AND_ASSIGN(TextDatabaseManager); }; } // namespace history #endif // CHROME_BROWSER_HISTORY_TEXT_DATABASE_MANAGER_H_