Use Python generators for test generation

[heinezen]

In the current implementation, the test generation logic puts all generated tests into a big list and passes this list to the engine for execution. This is fine for smaller test runs, but could result on a heavy memory footprint if runs contain a large number of tests or if individual TestCase objects is high. Generating the test
cases using generators could improve this situation, since tests would be generated on-the-fly.

It should still be possible to generate all tests at once, so that a run can be saved to file and executed later. The test engine should accept a list of tests or a list of generators.

Implementing Python generators would probably involve the following tasks:

• Redesign all existing TestCase.generate as generators (should be trivial)
• Store generators and pass them to the engine
• Test IDs need to be generated on-the-fly during test execution (a simple counter should be enough).
• Test results need to be stored after test execution, so that we can destroy the test object after execution to save memory.
• Tests should not be destroyed immediately, since the engine may want to execute it a second time. This can happen if the engine detects that the rate limit of a service has been reached and the last X tests have to be redone.

You can use the Python module tracemalloc to monitor memory usage.