adamgleave / pytest-shard Goto Github PK

Hi,

I'd love to get the latest PR into our CI system. Any chance you could prepare a new release and upload that to Pypi?

Thanks a lot in advance!

It only says how to define shard numbers.

But how to tell it to submit jobs to the grid?
How to specify the submission command?

This command

$ pytest --shard-id=I --num-shards=N

only tells pytest what shard IDs are, but not how to submit individual jobs.

Thank you for the great project!

I wanted to ask couple of related questions.

1. I noticed that the hash(item.nodeid) doesn't give a good distribution on my workloads. I.e. if I have 8 tests, I often end up with one shard getting 6 and another shard getting 2.
2. Seems like item.nodeid has an absolute path to the test file, so it could be machine-specific (depending on the implementation). This makes it possible for a single test to run twice and vice-verse for a test not to be run at all.

So my question is -- can we use a simple round-robin algorithm? Something like

def filter_items_by_shard(
    items: Iterable[nodes.Node], shard_id: int, num_shards: int
) -> Sequence[nodes.Node]:
    """Computes `items` that should be tested in `shard_id` out of `num_shards` total shards."""
    return [item for idx, item in enumerate(items) if idx % num_shards == shard_id]

Is there any downside in it?

My understanding is that pytest has a deterministic order for the items, so it should be just fine.