Closed FyzHsn closed 2 years ago
It's possible that during L218 of electron.py, particularly in def iter() when consructing the electron we copy the metadata self.metadata.copy() which would also include the callbefore & callafter deps with return value keys defined, which would in turn try to inject the files kwarg. We may want to disable this for internally generated electrons or figure out a way to make the injected keywords optional.
PR #965 might be relevant here.
Environment
What is happening?
When transferring files and returning the results as a list / tuple / dict, we get an error saying:
The UI shows the following:
How can we reproduce the issue?
Add a train.csv and test.csv in the login node in beehive under your account. Accordingly replace the filepaths in the example workflow shown below.
What should happen?
The iterables should be unpacked without any errors. Also, note that for a non-iterable return, there is no issue.
Any suggestions?
It's possible that during L218 of electron.py, particularly in
def __iter__()
when consructing the electron we copy the metadataself.metadata.copy()
which would also include the callbefore & callafter deps with return value keys defined, which would in turn try to inject thefiles
kwarg. We may want to disable this for internally generated electrons or figure out a way to make the injected keywords optional.