Following table highlights this on various aspects of PHPUnit.

PHPUnit belongs to xUnit family, same as that of the jUnit. PHPUnit not only provides various functions to assert values but also allows many complex operations possible through annotations. PHPUnit provides a lot of extensions, such as for database testing, performance testing. The Database testing extension provided by PHPUnit is very similar to that of what is provided by Dbunit. Both follow similar approaches of seeding the database initially, and then after code execution ensuring that expected data is found in the database. If one has some experience with mocking frameworks in Java such as EasyMock, then working with mock testing with PHPUnit would be a cakewalk. The comfort of migration extends to the level of acceptance testing, when one uses Selenium with PHPUnit.

PHPUnit is one of the many available testing frameworks in PHP world, and testing is one of the many areas where I see commonality between the two worlds(J2EE and PHP). The major reason of the commonality has come because of the inclusion of the best practices in the PHP world and also adapting helpful frameworks from the J2EE world. As I proceed further on my path ahead, I would be sharing my learning with you all.

Using same connection for entire test suite
If you are extending DatabaseTestCase then this class creates a connection to the database in setUp() and closes it in teardown() this can prove to be a big overhead and can cause considerable slowdown in running tests as the number of test cases increase. Maintaining a single connection for the entire test suite can prove to be beneficial in such cases.

If you are using JUnit 4 then you can use annotations for performing initial operations (@BeforeClass) and thus there is no need to extend the DatabaseTestCase and finally handling the closing of database connection.

@BeforeClass
public static void setUpBeforeClass() throws Exception {
handleSetUpOperation();
}

If several test cases are using a read only dataset as expected state of the database after running the test cases then loading this expected dataset file only once for all the test cases can help up in speeding up things.

Specifying a DTD for the dataset can act as a cure for some of the things
You can create a DTD by using the following code in DbUnit.

FlatDtdDataSet.write(connection.createDataSet(),
new FileOutputStream("test.dtd")); 

This DTD can be used now to
- Validate a dataset file against it.
- Providing null values to a column by omitting it for a row in dataset.

Keeping some reusable methods
A few reusable methods like asserting the database state after running tests with an expected state dataset file can be very useful. I am mentioning one of such methods here

    protected void assertTables(String filename, String[] tableNames)
    throws SQLException, Exception, DataSetException, IOException,
    DatabaseUnitException {
    IDataSet expectedDataSet = null;
    if (tableNames == null) {
    // Load expected data from an XML data set
    expectedDataSet = new FlatXmlDataSet(new File(filename));
    tableNames = expectedDataSet.getTableNames();
    } else {
    // Load subset of the complete dataset
    expectedDataSet = new FilteredDataSet(tableNames,
    new FlatXmlDataSet(new File(filename)));
    }
    // Fetch database data after executing your code
    IDataSet actualDataSet = getConnection().createDataSet(tableNames);
    ITable[] expectedTables = new CompositeTable[tableNames.length];
    for (int i = 0; i <tableNames.length; i++) {
    expectedTables[i] = new CompositeTable(expectedDataSet
    .getTableMetaData(tableNames[i]), actualDataSet
    .getTable(tableNames[i]));
    }
    Assertion.assertEquals(expectedDataSet, new CompositeDataSet(
    expectedTables));
    }

Above method assertTables can assert the current state of the database after running test cases with the state provided in the dataset file. This method is intelligent enough to handle both cases of asserting the database state with the expected dataset file by supplying table names (might be a subset of all the tables provided in the dataset file) or without the table names in which case all the tables provided in the dataset files are used. The method uses table meta-data provided in the dataset file to get state of the corresponding table from the database, thus to ignore column such as auto-generated column can be automatically avoided in comparison by not mentioning them in the dataset file.

Happy Red-Green-Refactor-ing.