Given when people touch on large codebases, and run pytest they burn a significant period of time in the order of hours on test suites.
pytest-distributed
greatly simplify that process, in hence, we are creating an complementary plugin topytest-xdist
, with true parralelism at play
So let's begin!
-
True Parallelism: first thing that comes to mind here, is use of parallel algorithms.
- What if we can go beyond Parallel Task Queue implementation? Say, MapReduce - Divide test suites into micro-units (map) and aggregate results (reduce)
-
Dynamic Test Distribution: distribute tests across available CPU cores to optimize performance
Possible Strats for True Parallelism:
Test Distribution by Module?
Test Distribution by Dependency Graph - I would have to research on this...
Dynamic Work Stealing: a work-stealing algorithm where idle workers steal work from busy workers, ensuring optimal CPU core utilization.
- Customization Options: Give users access to customize settings based on their specific requirements but not too much power.
The project is considered success if:
- It could perform successful parallel test execution in the slightest way possible