Storing 1D DataFrame gives Pandas TypeError on retrieval about mdsynthesis HOT 4 CLOSED

This appears to be due to the defaults chosen for pandas object storage. The MDSynthesis.core.persistence.pdDataFile.add_data method first tries to store the object with data_columns=True, which will create an index that allows PyTables where queries to be performed on columns. However, this is a problem if column names start with numbers, and so the method falls back on not using data_columns=True if it fails.

This appears to work for all of the DataFrames in this test except for the 1D one. For all of those, an AttributeError is raised and caught as expected. But for that one, no AttributeError is raised.

I'll add a band-aid heuristic to catch this corner case, then I'll submit an issue to pandas to hopefully get it resolved there. Thanks for catching this!

from mdsynthesis.

I'm stealing these tests for the unittests, btw. :D

from mdsynthesis.

Hopefully there aren't any other weird corner cases that HDFStore doesn't catch. Otherwise, we'll just have to work around them.

from mdsynthesis.

Awesome, I just tried it and the issue is certainly resolved. I guess the lesson is to use sensible column names!

from mdsynthesis.