Dependency issues when running new catalog workflow

[akadlof]

Steps to reproduce

1. Clone repo, enter venv, and run pip install -e .
2. Update .\workflows\new.py to run on subset of provided sample data:

from stormhub.logger import initialize_logger
from stormhub.met.storm_catalog import new_catalog, new_collection

if __name__ == "__main__":
    initialize_logger()

    # Catalog Args
    root_dir = "catalogs/example-input-data"
    config_file = f"{root_dir}/config.json"
    catalog_id = "indian-creek"
    local_directory = f"{root_dir}"

    storm_catalog = new_catalog(
        catalog_id,
        config_file,
        local_directory=local_directory,
        catalog_description="Indian Creek Catalog",
    )

    # All Collection Args
    start_date = "2023-8-01"
    end_date = "2023-8-08"
    top_n_events = 440

    # Collection Args
    storm_duration_hours = 48
    min_precip_threshold = 2.5
    storm_collection = new_collection(
        storm_catalog,
        start_date,
        end_date,
        storm_duration_hours,
        min_precip_threshold,
        top_n_events,
        check_every_n_hours=6,
    )

4. Run python .\workflows\new.py

Expected behavior

The new catalog workflow runs without errors.

Current behavior

File "stormhub\stormhub\met\aorc\aorc.py", line 8, in <module> import s3fs ModuleNotFoundError: No module named 's3fs'

After pip installing s3fs (version 2024.12.0), I get:
File "stormhub\venv\Lib\site-packages\xarray\backends\plugins.py", line 202, in get_backend raise ValueError( ValueError: unrecognized engine 'zarr' must be one of your download engines: ['rasterio', 'store'].

After pip install of zarr, I get:
File "stormhub\venv\Lib\site-packages\zarr\storage\_common.py", line 316, in make_store_path raise TypeError(msg) TypeError: Unsupported type for store_like: 'FSMap'

Which is resolved by downgrading to version 2.18.4 (link to more info)

After downgrade of zarr, I get:
File "stormhub\venv\Lib\site-packages\xarray\namedarray\parallelcompat.py", line 110, in guess_chunkmanager raise ValueError( ValueError: unrecognized chunk manager dask - must be one of: []

Which is resolved by pip installing dask (version 2025.1.0)

At this point the workflow runs without errors.

Environment

• Windows 11
• Python 3.11