Question

I’m trying to unstack a dataframe perform operations on it (over time only) and then stack it back together like this:

import pandas as pd
import numpy as np
from itertools import *

time = pd.date_range(pd.datetime(2007,1,1),pd.datetime(2007,1,2))
slot = map(lambda n:"s-"+str(n),reversed(range(2)))
obj  = map(lambda n:"o-"+str(n),reversed(range(2)))
idx  = pd.MultiIndex.from_tuples(list(product(slot, obj, time)), names=['Ananas','Bananas','time']) #list(.) needed to get a length, should this really be needed?
data = np.random.rand(len(idx),4)

df = pd.DataFrame(data=data,index=idx, columns=['a','b','c','d']) #why is idx.size==0?

print df.to_string()
print "====="
unstacked = df.unstack(level=[0,1])
print unstacked.to_string()
print "====="
stacked = unstacked.stack(level=[2,1])
print stacked.to_string()

The problem is that multiindex is getting reversed after the operation, is there any easy way to make this work? Perhaps I’m misusing the stack from the start?

Answer 1

stack and unstack add level(s) to the end of the MultiIndex, this is not controllable. You can change the order of the levels in a MultiIndex with reorder_levels(): stacked.reorder_levels([2, 1, 0]) will give you the same MultiIndex levels order as in df