Android: multiple threads writing on SQLite database

Question

In the onCreate() method of my main activity I call the constructor of my dbManager, I call the open function that creates an instance of a SQLiteOpenHelper and than I call the getWritableDatabase() on it. Within the UIThread I add records to the database and save those records to an ArrayList. Two other threads check the ArrayList do stuff and than update the list and the database. Now I want to add a button in the UI to delete both database and list records using an AsyncTask. I've read that the SqliteOpenHelper object holds on one database connection. So if there is one helper instance there is only one db connection. This connection could be used from multiple threads and the SqliteDatabase object uses java locks to keep access serialize. So If I have multiple threads writing on the database, a thread will wait until the previous one has finished his operation?
Adding the new functionality (remove all) could create problems, because I have to avoid that one of the two threads may try to edit a record that no longer exists. How can I achieve my goal? Thanks.

Answer 1

The databese:

As long as you use only one helper, you are thread safe without a need to do anything.
In a multi threaded app, you'll need synchronized only on the creation of the helper. You can use transacion if you want to define a critical section (more than one atomic operation).
Wrap each call with getWritableDatabase() and close(false):

public void doSomething() {
    SQLiteDatabase tdb = getWritableDatabase();
    dbInstance.writeSomething(tdb, 1);
    close(false);
}

In this way I have multiple threads reading and writing to the database without any problem.

You app logic:

Use memory to keep track of object state, and use the database as storage only. So if one thread deletes rows from the database - you can immediately update your objects in memory, and handle your ui accordingly.
If your data is very large, than you can keep in memory only a SparseArray of dirty rows id.
It may be useful to synchronize some operations here.