Re:A Roadmap for Rich Scientific Data Structures in Python
by wesm on Thursday July 21, 2011 @04:58PM
(#2791269)
To be honest I haven't really broached the hierarchical columns issue inside pandas. If someone can look and suggest an implementation strategy that doesn't interfere with the rest of the API I would be all for it.
If you have heterogeneous columns with possibly missing data basically pandas is the only game in town (did you see my diagram?? :) ). It's possible to get a numpy MaskedArray with structured dtype to function like you want but it's relatively tricky to do.
I'm not quite at the point where I'll be wrapping this into python. When I do I'll take a close look at pandas. Any other recommendations? I took a quick look at pytables, and pandas looks better for my app.