Testing SQLite database in Robolectric

Question

I'm trying to test a simple SQLite database using Robolectric in my Android application. I'm putting in some values, but when reading them back 0 rows are returned.

I'm using the SQLiteOpenHelper class to access the database.

// RequestCache extends SQLiteOpenHelper
RequestCache cache = new RequestCache(activity); 
SQLiteDatabase db = cache.getWritableDatabase();

// Write to DB
ContentValues values = new ContentValues();
values.put(REQUEST_TIMESTAMP, TEST_TIME); 
values.put(REQUEST_URL, TEST_URL);
db.insertOrThrow(TABLE_NAME, null, values);

// Read from DB and compare values      
Vector<Request> matchingRequests = new Vector<Request>();
db = cache.getReadableDatabase();
Cursor cursor = db.query(TABLE_NAME, SEARCH_URL_RETURN_COLUMNS, SEARCH_URL_WHERE, new String[] {url}, null, null, ORDER_BY, null);
int id = 0;

while(cursor.moveToNext()) {
    long timestamp = cursor.getLong(0);
    Request request = new Request(id++);
    request.setUrl(url);
    request.setCreationTimestamp(new Date(timestamp));
    matchingRequests.add(request);
}


// Assert that one row is returned
assertThat(matchingRequests.size(), equalTo(1));  // fails, size() returns 0

When debugging the code outside robolectric this works as expected. Am I doing anything wrong or is it not possible to test SQlite databases using Robolectric?

Answer 1

NOTE: This answer is outdated. If you are using Roboletric 2.X, please see https://stackoverflow.com/a/24578332/850787

The problem is that Robolectric's SQLiteDatabase is stored only in memory, so when you call getReadableDatabase or getWritableDatabase, the existing database will be overridden with new empty database.

I was running to the same problem and only solution I found was that I needed to fork the Robolectric project and added ShadowSQLiteOpenHelper to save database if same context is given two times. However the problem with my fork is that I had to 'disable' close()-function when contex is given because otherwise Connection.close() will destroy the database in memory. I have made pull request for it but it isn't merged to project yet.

But feel free to clone my version and it should fix your problem (If I understood it correctly :P ). It can be found on GitHub: https://github.com/waltsu/robolectric

Here is a example how to use the modification:

Context c = new Activity(); 
SQLiteOpenHelper helper = new SQLiteOpenHelper(c, "path", null, 1); 
SQLiteDatabase db = helper.getWritableDatabase(); 
// With the db write something to the database 
db.query(...); 
SQLiteOpenHelper helper2 = new SQLiteOpenHelper(c, "path", null, 1); 
SQLitedatabase db2 = helper2.getWritableDatabase(); 
// Now db and db2 is actually the same instance 
Cursor c = db2.query(...) ; // Fetch the data which was saved before 

Ofcourse you don't need to create new SQLiteOpenHelper, but that is just example that passing same context to two different SQLiteOpenHelper will give same database.

Answer 2

Robolectric 2.3 uses a real implementation of SQLite instead of a collection of shadows and fakes. Tests can now be written to verify real database behavior.

Answer 3

The code linked in the accepted answer did not work for me; it may be out of date. Or perhaps my setup is just different. I am using Robolectric 2.4 snapshot, which does not seem to include a ShadowSQLiteOpenHelper, unless I missed something. In any case, I figured out a solution. Here's what I did:

1. I created a class called ShadowSQLiteOpenHelper and copy-pasted the contents of the code linked above (https://github.com/waltsu/robolectric/blob/de2efdca39d26c5f18a3d278957b28a555119237/src/main/java/com/xtremelabs/robolectric/shadows/ShadowSQLiteOpenHelper.java), and fixed the imports.
2. Following the instructions on http://robolectric.org/custom-shadows/ for using custom shadows, I annotated my test class with @Config( shadows = { ShadowSQLiteOpenHelper.class } ). No need for a custom test runner.
3. In my tests, I instantiated my subclass of SQLiteOpenHelper, passing in a new Activity() as the Context (Robolectric happily takes care of that), and used it per usual.

Now, at this point, I noticed that an actual database file was getting created locally: after the first run of a test that used my SQLiteOpenHelper subclass, I kept getting an SQLiteException on subsequent tests because my table already existed; and I could see a file called "path" and a file called "path-journal" sitting in my local repository. This confused me because I was under the impression that the shadow class was using an in-memory database.

It turns out the offending line in the shadow class is:

database = SQLiteDatabase.openDatabase( "path", null, 0 );

in both getReadableDatabase() and getWriteableDatabase(). I knew that the real SQLiteOpenHelper could create an in-memory database, and after looking at the source to see how it's done, I replaced the above line with:

database = SQLiteDatabase.create( null );

after which, everything seems to work. I am able to insert rows and read them back.

Hopefully this helps someone else.

One weird thing that happened to me that isn't exactly related to this question, but might help someone further, is that I'm also using jmockit, and I had used it in one test in the same class; but for some reason, this caused my custom shadow class to not be used. I switched to Mockito for just that class, and it works fine.