ncar / mom6-tools Goto Github PK

Currently, ClimoGenerator returns the last time used in the calculation, e.g.:

It should return the averaged time instead.

mom6-tools relies on python scripts from CIME to get basic info from a simulation (casename, path to the output, etc). This makes mom6-tools vulnerable to changes in CIME; i.e., recent changes in cesm2_3_beta10 break mom6-tools. @klindsay28 has scripts that give the same info without relying on CIME. I propose we follow a similar approach in mom6-tools.

Right now ClimoGenerator only works if HIST_FILE_PREFIX can be found in diag_table. It would be useful if this tool also works with files that are not in diag_table (e.g., .cice.h.).

Currently stage_dset expects a list of fields to determine which files to read. There should be an alternative method that returns a dataset will all available fields from a given HIST_FILE_PREFIX.

Just putting this here so I remember to send a PR:

def preprocess_mom6_sections(ds):
    """
    Preprocess function for reading MOM6 field section output.
    
    Gets rid of dummy dimensions ("xh_sub01") and renames them to
    standard names ("xh")
    
    Intended use is the ``preprocess`` kwarg of xarray's ``open_mfdataset``.
    """

    # "standard" dimension names
    dims = ["xh", "xq", "yh", "yq"]

    for dim in dims:
        matches = [dsdim for dsdim in ds.dims if dim in dsdim]
        if not matches:
            continue
        if len(matches) > 1:
            [
                np.testing.assert_equal(
                    ds.coords[matches[0]].values, ds.coords[other].values
                )
                for other in matches[1:]
            ]

        ds = ds.rename({matches[0]: dim})
        for match in matches[1:]:
            ds = ds.drop_vars(match).rename({match: dim})

    return ds

I noticed that the documentation builds on RTD have been failing for the last few months due to the Command killed due to excessive memory consumption error. This is due to the fact that conda consumes a lot of memory, and Readthedocs sets a limit to how much memory can be use. One can request memory increase for a project by contacting the RTD (Readthedocs) via https://github.com/readthedocs/readthedocs.org/issues