table_id replace null to "na"

[aradhakrishnanGFDL]

Opening issue to reflect Ray's issues using the generated catalog in analysis.

It works sometimes, but segmentation faults at

d_ = catalog.search(**query_params).to_dataset_dict(progressbar=True)

I get that same warning. It comes from the fact that "table_id" can be either "fx" or blank, which pandas converts to NaN. So it's not sure if
table_id should be a numerical type or a string. Here is some output. The first time is succeeds, but the second time it fails:

(env) rlm:/local2/home/cleanup-fre-cli/tmp2> python3 catalog_test.py
/local2/home/cleanup-fre-cli/tmp2/env/lib/python3.10/site-packages/intake_esm/cat.py:249: DtypeWarning: Columns (6) have mixed types. Specify dtype option on import or set low_memory=False.
df = pd.read_csv(
(env) rlm:/local2/home/cleanup-fre-cli/tmp2> python3 catalog_test.py
/local2/home/cleanup-fre-cli/tmp2/env/lib/python3.10/site-packages/intake_esm/cat.py:249: DtypeWarning: Columns (6) have mixed types. Specify dtype option on import or set low_memory=False.
df = pd.read_csv(
Segmentation fault (core dumped)

[ceblanton]

Thanks Aparna! Interesting. Is the seg fault possibly unrelated to the table_id type issues?

Why would the seg fault not be repeatable?

[aradhakrishnanGFDL]

Not sure @ceblanton. In your catalog, you could change it to NA and see if Ray continues to see this issue or not to rule this out.