Mike Lin

It'd be quite safe to run this test only if there's a diff to the relevant source files under `apy/python/cellxgene_census/src/cellxgene_census/experimental/ml/huggingface`. Is there a straightforward way to effect that with some...

@cademirch Thanks for the report...I don't know exactly what the problem is, but I can describe what the "post-task chown" is meant to do, to give more context that may...

@a-frantz thanks for raising this, can I ask a clarification > only the root inputs of a WDL.Tree.Workflow get their parameter_meta parsed and stored in the resulting Python object. Can...

@a-frantz Thanks, makes sense. The `parameter_meta` dicts currently just reflect the corresponding sections of the workflow and task source code. They're syntax-level constructs not intended to automatically propagate from task...

@a-frantz in lieu of deciding what an API for this should look like, I put together this example with a ~10-line function that will resolve nested parameter_meta. Hope it helps!...

@vsmalladi Thanks a lot for compiling this, really helpful!

Hmm, this is supposed to work, so we'll need to look into why exactly it doesn't. * existing implementation: https://github.com/chanzuckerberg/miniwdl/blob/3ece03a6a5627eec21d82e4f21929ebd1db013d4/WDL/runtime/task_container.py#L350-L356 * existing test: https://github.com/chanzuckerberg/miniwdl/blob/3ece03a6a5627eec21d82e4f21929ebd1db013d4/tests/test_4taskrun.py#L837

@vsmalladi Is the problem here that the spec says the attribute is named `returnCodes` while the provided examples use `return_codes`? https://github.com/openwdl/wdl/blob/wdl-1.1/SPEC.md#returncodes

The spec certainly seems inconsistent with itself right now. As best I recall `returnCodes` was there in prose before the examples with `return_codes`. I'm sure the feature originated with Cromwell,...

@vsmalladi Does the spec say that the output struct *must not* explicate the null field?