Strategy for testing

[charlesreid1]

Thinking forward to a strategy for running regression tests, we should devise a strategy for testing the workflow:

• Tests should be implemented as either short/light tests or long/heavy tests.

• Short/light tests are verification tests that use artificial data (e.g., all zeros or all ones); they ensure that the data-agnostic portions of the dahak workflow don't fail. (Note, this may be simply impossible for some tools or portions of workflows to run without "nearly" real data.)

• Meanwhile long/heavy tests would provide validation, running an entire workflow using a real (minimal) input dataset and reproducing a gold standard output. These require more compute power and would be performed more infrequently, but provide deeper validation that the way the pipeline handles data is free of defects.

[ctb]

+1