sqlite是支持write ahead logging(WAL)模式的,开启WAL模式可以提高写入数据库的速度,读和写之间不会阻塞,但是写与写之间依然是阻塞的,但是如果使用默认的TRUNCATE模式,当写入数据时会阻塞android中其他线程或者进程的读操作,并发降低。 相反,使用WAL可以提高并发。 由于使用WAL比ROLLBACK JOURNAL的模式减少了写的I/O,所以写入时速度较快,但是由于在读取数据时也需要读取WAL日志验证数据的正确性,所以读取数据相对要慢。 所以大家也要根据自己应用的场景去使用这种模式。
那么在android中如何开启WAL模式呢?
看SQLiteDatabase开启WAL的核心方法源码。
public boolean enableWriteAheadLogging() {
// make sure the database is not READONLY. WAL doesn't make sense for readonly-databases.
if (isReadOnly()) {
return false;
}
// acquire lock - no that no other thread is enabling WAL at the same time
lock();
try {
if (mConnectionPool != null) {
// already enabled
return true;
}
if (mPath.equalsIgnoreCase(MEMORY_DB_PATH)) {
Log.i(TAG, "can't enable WAL for memory databases.");
return false;
}
// make sure this database has NO attached databases because sqlite's write-ahead-logging
// doesn't work for databases with attached databases
if (mHasAttachedDbs) {
if (Log.isLoggable(TAG, Log.DEBUG)) {
Log.d(TAG,
"this database: " + mPath + " has attached databases. can't enable WAL.");
}
return false;
}
mConnectionPool = new DatabaseConnectionPool(this);
setJournalMode(mPath, "WAL");
return true;
} finally {
unlock();
}
}
在源码的注释中是这样写到:“This method enables parallel execution of queries from multiple threads on the same database.” 通过此方法可以支持多个线程并发查询一个数据库。 并在注释中给出了实例代码如下:
SQLiteDatabase db = SQLiteDatabase.openDatabase("db_filename",
cursorFactory,CREATE_IF_NECESSARY, myDatabaseErrorHandler);
db.enableWriteAheadLogging();
通过调用db.enableWriteAheadLogging即可开启WAL模式。 在enableWriteAheadLogging方法中关注的核心点:
mConnectionPool = new DatabaseConnectionPool(this);
setJournalMode(mPath, "WAL");
1.创建数据库连接池,由于要支持并发访问所以需要连接池的支持。
2.调用setJournalMode设置模式为WAL.
当开启了WAL模式之后,事务的开始需要注意,在源码的注释是这样写到。
Writers should use {@link #beginTransactionNonExclusive()} or
* {@link #beginTransactionWithListenerNonExclusive(SQLiteTransactionListener)}
调用者需要使用beginTransactionNonExclusive或者beginTransactionWithListenerNonExclusive来开始事务,也就是执行:BEGIN IMMEDIATE; 支持多并发。
注:关于EXCLUSIVE与IMMEDIATE模式请参考我的另一篇博客 http://blog.csdn.net/degwei/article/details/9672795
当开启了WAL模式更新数据时,会先将数据写入到*.db-wal文件中,而不是直接修改数据库文件,当执行checkpoint时或某个时间点才会将数据更新到数据库文件。当出现rollback也只是清除wal日志文件,而ROLLBACK JOURNAL模式,当数据有更新时,先将需要修改的数据备份到journal文件中,然后修改数据库文件,当发生rollback,从journal日志中取出数据,并修改数据库文件,然后清除journal日志。 从以上流程来看 WAL在数据更新上I/0量要小,所以写操作要快。
当开启了WAL模式磁盘中是这样的文件格式,当数据文件名为:test时 如下图:
图中红色部分为WAL的日志文件。
那么WAL日志中的数据何时更新到数据库文件中,刚才提到当手动执行checkpoint时或者由当前线程的某个时间点提交。
如何手动执行checkpoint,看SQLiteDatabase.endTransaction源码:
/**
* End a transaction. See beginTransaction for notes about how to use this and when transactions
* are committed and rolled back.
*/
public void endTransaction() {
verifyLockOwner();
try {
...
if (mTransactionIsSuccessful) {
execSQL(COMMIT_SQL);
// if write-ahead logging is used, we have to take care of checkpoint.
// TODO: should applications be given the flexibility of choosing when to
// trigger checkpoint?
// for now, do checkpoint after every COMMIT because that is the fastest
// way to guarantee that readers will see latest data.
// but this is the slowest way to run sqlite with in write-ahead logging mode.
if (this.mConnectionPool != null) {
execSQL("PRAGMA wal_checkpoint;");
if (SQLiteDebug.DEBUG_SQL_STATEMENTS) {
Log.i(TAG, "PRAGMA wal_Checkpoint done");
}
}
// log the transaction time to the Eventlog.
if (ENABLE_DB_SAMPLE) {
logTimeStat(getLastSqlStatement(), mTransStartTime, COMMIT_SQL);
}
} else {
...
}
} finally {
mTransactionListener = null;
unlockForced();
if (false) {
Log.v(TAG, "unlocked " + Thread.currentThread()
+ ", holdCount is " + mLock.getHoldCount());
}
}
}
在源码注释是这样写到:“if write-ahead logging is used, we have to take care of checkpoint.” 如果使用了WAL模式,那么就会执行checkpoint,当mConnectionPool != null时表示使用了WAL模式,也只有当WAL模式下才会有数据库连接池。 执行PRAGMA wal_checkpoint;即:将wal日志同步到数据库文件。 也就是当我们执行endTransaction时才会提交checkpoint。
在android中默认为TRUNCATE模式 , 请看如下源码:
public static SQLiteDatabase openDatabase(String path, CursorFactory factory, int flags,
DatabaseErrorHandler errorHandler) {
SQLiteDatabase sqliteDatabase = openDatabase(path, factory, flags, errorHandler,
(short) 0 /* the main connection handle */);
// set sqlite pagesize to mBlockSize
if (sBlockSize == 0) {
// TODO: "/data" should be a static final String constant somewhere. it is hardcoded
// in several places right now.
sBlockSize = new StatFs("/data").getBlockSize();
}
sqliteDatabase.setPageSize(sBlockSize);
sqliteDatabase.setJournalMode(path, "TRUNCATE");
// add this database to the list of databases opened in this process
synchronized(mActiveDatabases) {
mActiveDatabases.add(new WeakReference(sqliteDatabase));
}
return sqliteDatabase;
}
通过sqliteDatabase.setJournalMode(path, "TRUNCATE");设置为TRUNCATE模式。