In addition to the data itself, some node attributes can be stored as pickles in HDF files.
This seems to include the name of Pandas Series when None. According to ViTables, it is stored as a "Python" attribute which I fear means pickle.
My conclusion: HDF5 via Pandas is definitely a mess (see #1178 and others)
In addition to the data itself, some node attributes can be stored as pickles in HDF files.
This seems to include the name of Pandas Series when None. According to ViTables, it is stored as a "Python" attribute which I fear means pickle.
My conclusion: HDF5 via Pandas is definitely a mess (see #1178 and others)