Closed lucalavezzo closed 9 months ago
Let's stick with storing it as metadata, we can re-evaluate if things break due to pandas instabilities. It would affect too many analyses to change it now, and it's less elegant; if it's not necessary I would prefer not doing it. The solution would be simple though, just store metadata as another dataframe in the same hdf5 file, which is easily supported by the current setup.
Seems like pandas doesn't like metadata and is unstable between versions. Let's just put it in the hdf5 file as a separate object?