Jim Crist-Harif

Results 482 comments of Jim Crist-Harif

Hi @JacksonMaxfield. With the DFE mapping refactor, mapping no longer is a layer on top of dask's `map` (Prefect walks its own `dag`, submitting components to the executor (in this...

Hi @xuevin, thanks for the reproducible example. I've marked this as an enhancement rather than a bug, as I don't think it breaks any of our existing guarantees. The way...

Yeah, that will gather all results locally before mapping, which will add a bunch of overhead if the results are large-ish (if they're not it's probably fine). The proper fix...

Hi @gryBox, as mentioned in https://github.com/PrefectHQ/prefect/issues/3291#issuecomment-690612823, this is currently expected behavior due to how mapped tasks are implemented. I think we should change this behavior though, so this is a...

I personally prefer the NumPy standard, but that's because I come from projects that use that formatting and am more used to it (I suspect many contributors in Prefect's space...

cc @znicholasbrown, who was helping me debug this earlier.

Looking through our test failure history, this no longer appears to be a flaky test. Closing.

I've started seeing this test failure again, this time with fastparquet alone (cc @martindurant). Example failure: https://github.com/dask/dask/runs/5872450945?check_suite_focus=true#step:6:21161. Traceback ```python _____________ test_create_metadata_file[fastparquet-fastparquet-a] _____________ [gw1] linux -- Python 3.9.12 /usr/share/miniconda3/envs/test-environment/bin/python tmpdir =...

:shrug: If you look at the traceback above, it looks like the indices differ (different missing values in each). Is the index here meaningful? Should `check_index=False` be passed here?