How to assign a python object (such as a dictionary) to pandas column

Question

I would like to set a cell in a pandas dataframe equal to a dictionary for rows in which another column in that same row equals 1. I am using df.loc to filter the rows. Since my dictionary has two keys, it only works if the filtering done by df.loc also has two keys. If it doesn't have two keys, I get ValueError: Must have equal len keys and value when setting with an iterable.

I don't see why these two things are related.

import pandas as pd
df = pd.DataFrame(data=[[1,2], [0,3], [3,4]], columns=['Col1', 'Col2'])
#df = pd.DataFrame(data=[[1,2], [1,3], [3,4]], columns=['Col1', 'Col2'])

df.loc[df["Col1"]==1, "Col2"] = {'key1': 'A',
                                 'key2': 'B'}

print df

If I uncomment the third line of code, I would like to produce the below results.

   Col1                            Col2
0     1  {u'key2': u'B', u'key1': u'A'}
1     1  {u'key2': u'B', u'key1': u'A'}
2     3                               4

Before this gets marked as a duplicate, I have seen other questions regarding this pandas error, but none seem to solve this issue specifically.

Answer 1

IIUC, wrap the dictionary in a list, and pass it to loc:

df

   Col1  Col2
0     1     2
1     1     3
2     3     4

m = df['Col1'].eq(1)
df.loc[m, 'Col2'] = [{'a' : 1, 'b' : 2}] * m.sum()

df

   Col1              Col2
0     1  {'a': 1, 'b': 2}
1     1  {'a': 1, 'b': 2}
2     3                 4

This should apply to any result equally well. Just keep in mind that [] * n replicates the references, so you have the same dict object being assigned to multiple cells! Keep this in mind going forward.

There's an alternative if you want to avoid duplicating references - you can build a list with a list comprehension.

i = {'a' : 1, 'b' : 2}
df.loc[m, 'Col2'] = [i.copy() for _ in range(m.sum())]

If you have a nested dictionary, copy only performs a shallow copy, so use the copy module's deepcopy function instead:

from copy import deepcopy
df.loc[m, 'Col2'] = [deepcopy(i) for _ in range(m.sum())]