catalog builder pkg usage updates #7
Conversation
| @@ -47,7 +47,8 @@ def create_data_catalog(path, output="data-catalog"): | |||
| yaml_path = Path(__file__).resolve().parent / "mdtf_timeslice_catalog.yaml" | |||
|
|
|||
| # Hack to stop click from exiting. | |||
| command = ["python3", "-m", "scripts.gen_intake_gfdl", str(path), output, | |||
| # TODO we have to move create catalog as its own pytest in the conda environment perhaps, and avoid this hardcoding | |||
| command = ["/usr/share/miniconda/envs/analysis-script-testing/bin/python", "-m", "catalogbuilder.scripts.gen_intake_gfdl", str(path), output, | |||
There was a problem hiding this comment.
When @ilaflott was digging into github CI pipelines last week, he found that one cannot run "conda activate" in the pipeline steps. I didn't understand why.. but you're clearly seeing the same rhing, as you have the conda activate commented out, and then use that python's path later.
There was a problem hiding this comment.
Yeah, it's kinda murky with the init file set up perhaps. We can however activate it, but have to use something like https://github.com/marketplace/actions/setup-miniconda. However, the new pipeline (see new PR) will use the new changes in the catalogbuilder and we will no longer need this particular line.
There was a problem hiding this comment.
Do we still want this PR? Sorry for leaving this hanging here. The hard-coded path to the python here would only be used in the CI environments in practice so we could live with it.
#6