def test_edit_user_readableYou can easily see that some pieces of this code (setups, actions) can be factored. But Ryan went a step further by reorganizing it into a matrix where the column headers represent the different setup cases, the row headers represent the action applied and the intersection defines the expected result for a given action and setup context.
some_setup_to_initialize_user_readable_context
some_action_here_edit
some_assertion_error_read
end
def test_edit_user_writable
some_setup_to_initialize_user_writable_context
some_action_here_edit
some_assertion_edit
end
def test_view_user_readable
some_setup_to_initialize_user_readable_context
some_action_here_view
some_other_assertion_view
end
def test_view_user_writable
some_setup_to_initialize_user_writable_context
some_action_here_view
some_other_assertion_view
end
The 4 test cases will look like this:
setups | :user_readable, | :user_writable | |
matrix | :edit, | :error_read, | :edit |
matrix | :view, | :view, | :view |
def test_#{action}_#{setup}So at the end the setup configuration will be factored at one place taking into parameters the different edge cases, the actions and assertions will also be put apart.
matrix_setup_configuration #{setup}.split(//) # global setup
matrix_setup_#{action} #{setup}, #{expected} # action setup + execution
matrix_test_#{expected}, #{setup} # expected verification
end
Ryan Davis gives a clear visual example of test cases before and after matrixization:
Before After Click the images to see the code in more detail.
The test matrix model is another demonstration of the DRY (Don't Repeat Yourself) process philosophy. It also easily makes it possible for non-developer to read and add tests by only modifying the matrix.