Looping through MultiIndex in pandas

Question

I have a MultiIndexed DataFrame df1 and would like to iterate through it so that in each instance of the loop there is a DataFrame with a regular non-hierarchical index, which is a subset of df1 corresponding to the external index of the records. Ie if I have:

I want to receive

and then C1, C2, etc. I also don’t know what names will actually be (C1, etc., just placeholders here), so I just need to iterate over the number C _i that I have.

I stumbled with iterrows and various loops and did not get tangible results and don’t know how to proceed. I feel that a simple solution should exist, but could not find anything useful in the documentation, possibly due to my own lack of understanding.

Answer 1

Using the modified example from here

 In [30]: def mklbl(prefix,n): return ["%s%s" % (prefix,i) for i in range(n)] ....: In [31]: columns = MultiIndex.from_tuples([('a','foo'),('a','bar'), ('b','foo'),('b','bah')], names=['lvl0', 'lvl1']) In [33]: index = MultiIndex.from_product([mklbl('A',4),mklbl('B',2)]) In [34]: df = DataFrame(np.arange(len(index)*len(columns)).reshape((len(index),len(columns))), index=index, columns=columns).sortlevel().sortlevel(axis=1) In [35]: df Out[35]: lvl0 ab lvl1 bar foo bah foo A0 B0 1 0 3 2 B1 5 4 7 6 A1 B0 9 8 11 10 B1 13 12 15 14 A2 B0 17 16 19 18 B1 21 20 23 22 A3 B0 25 24 27 26 B1 29 28 31 30 In [36]: df.loc['A0'] Out[36]: lvl0 ab lvl1 bar foo bah foo B0 1 0 3 2 B1 5 4 7 6 In [37]: df.loc['A1'] Out[37]: lvl0 ab lvl1 bar foo bah foo B0 9 8 11 10 B1 13 12 15 14 

No loop required.

You can also select them to return the frame (with the original MI) for example df.loc[['A1']]

If you want to get the values ​​in the index:

 In [38]: df.index.get_level_values(0).unique() Out[38]: array(['A0', 'A1', 'A2', 'A3'], dtype=object) 

Answer 2

Are you trying to do something like this?

 for i in set(df.index): print df.loc[i].reset_index() 

• set(df.index) returns a set of unique tuples in your multi-index (hierarchical index).
• df.loc[i].reset_index() ... df.loc[i] , of course, returns a subset of the original source frame, and the .reset_index() converts the index into columns