Why does .loc assignment with two sets of brackets result in NaN in a pandas.DataFrame?

Question

I have a DataFrame:

	name	age
0	Paul	25
1	John	27
2	Bill	23

I know that if I enter:

df[['name']] = df[['age']]

I'll get the following:

	name	age
0	25	25
1	27	27
2	23	23

But I expect the same result from the command:

df.loc[:, ['name']] = df.loc[:, ['age']]

But instead, I get this:

	name	age
0	NaN	25
1	NaN	27
2	NaN	23

For some reason, if I omit those square brackets [] around column names, I'll get exactly what I expected. That is the command:

df.loc[:, 'name'] = df.loc[:, 'age']

gives the right result:

	name	age
0	25	25
1	27	27
2	23	23

Why does two pairs of brackets with .loc result in NaN? Is it some sort of a bug or is it intended behaviour? I can't figure out the reason for such a behaviour.

Answer 1

From the Docs Pandas Data Alignment(emphasis mine):

pandas aligns all AXES when setting Series and DataFrame from .loc, and .iloc.

You can find this excerpt labelled with Warning under the Basics header. They have given an example to explain.

In [9]: df[['A', 'B']]
Out[9]: 
                   A         B
2000-01-01 -0.282863  0.469112
2000-01-02 -0.173215  1.212112
2000-01-03 -2.104569 -0.861849
2000-01-04 -0.706771  0.721555
2000-01-05  0.567020 -0.424972
2000-01-06  0.113648 -0.673690
2000-01-07  0.577046  0.404705
2000-01-08 -1.157892 -0.370647

In [10]: df.loc[:, ['B', 'A']] = df[['A', 'B']]

In [11]: df[['A', 'B']]
Out[11]: 
                   A         B
2000-01-01 -0.282863  0.469112
2000-01-02 -0.173215  1.212112
2000-01-03 -2.104569 -0.861849
2000-01-04 -0.706771  0.721555
2000-01-05  0.567020 -0.424972
2000-01-06  0.113648 -0.673690
2000-01-07  0.577046  0.404705
2000-01-08 -1.157892 -0.370647

From Docs(emphasis mine):

This will not modify df because the column alignment is before value assignment.

To Explicitly Avoid Automatic Alignment

Accessing the array can be useful when you need to do some operation without the index (to disable automatic alignment, for example).

The alignment comes into play when LHS and RHS are dataframes. To avoid alignment try using.

df.loc[:, ['B', 'A']] = df[['A', 'B']].to_numpy()

---

You have two cases at hand,

• .loc assignment with pd.DataFrame.
• .loc assignment with pd.Series in EDIT.

.loc assignment in pd.DataFrame

In pd.DataFrame has 2 axes index and columns. So, when you do

df.loc[:, ['name']] = df.loc[:, ['age']]

LHS has column A which doesn't align with RHS column B hence resulting in all NaN after assignment.

• Data Alignment and arthimetics

From Docs DataAlignment(emphasis mine)

Data alignment between DataFrame objects automatically align on both the columns and the index (row labels). Again, the resulting object will have the union of the column and row labels.

You can find this behaviour in most of the pandas' operations if not all. Example, addition, subtraction, multiplication etc. Non matching indices and columns are filled with NaN.

Example from Data Alignment and Arthimetics

df = pd.DataFrame(np.random.randn(10, 4), columns=["A", "B", "C", "D"])
df2 = pd.DataFrame(np.random.randn(7, 3), columns=["A", "B", "C"])

df + df2 

         A         B         C   D
0  0.045691 -0.014138  1.380871 NaN
1 -0.955398 -1.501007  0.037181 NaN
2 -0.662690  1.534833 -0.859691 NaN
3 -2.452949  1.237274 -0.133712 NaN
4  1.414490  1.951676 -2.320422 NaN
5 -0.494922 -1.649727 -1.084601 NaN
6 -1.047551 -0.748572 -0.805479 NaN
7       NaN       NaN       NaN NaN
8       NaN       NaN       NaN NaN
9       NaN       NaN       NaN NaN

To answer your comment

But why do column indexes need to match? I can see why one want row indexes to match, but why column indexes?

Let's take a look at the above example, if columns are not aligned how would you add two DataFrames? It makes sense to align them on columns and indices.

---

.loc assignment in pd.Series

pd.Series has only one axis i.e index. That is the reason why it worked when you did

df.loc[:, 'name'] = df.loc[:, 'age']

As pd.Series has only one axis, pandas tried to align index and it worked. Of course, if index doesn't align it results in NaN values.

From Docs Series Alignment(emphasis mine):

The result of an operation between unaligned Series will have the union of the indexes involved. If a label is not found in one Series or the other, the result will be marked as missing NaN.

Answer 2

That's because for the loc assignment all index axes are aligned, including the columns: Since age and name do not match, there is no data to assign, hence the NaNs.

You can make it work by renaming the columns:

df.loc[:, ["name"]] = df.loc[:, ["age"]].rename(columns={"age": "name"})

or by accessing the numpy array:

df.loc[:, ["name"]] = df.loc[:, ["age"]].values

Answer 3

When you use double brackets [[]] you are assigning a DataFrame. What you want is assign a (column) Series, and for that you use only one bracket [].

Here is some code:

import pandas as pd
df = pd.DataFrame({'name':['Paul','John','Bill'], 'age':[25,27,23]})
print('Inital Dataframe:\n',df)

df[['name']] = df[['age']]
print("\ndf[['name']] = df[['age']]\n",df)

print("df.loc[:, ['age']]:", type(df.loc[:, ['age']]))
print("df.loc[:, ['name']]:", type(df.loc[:, ['name']]))
df.loc[:, ['name']] = df.loc[:, ['age']]
print("\ndf.loc[:, ['name']] = df.loc[:, ['age']]\n",df)
    
print('=======================')
df = pd.DataFrame({'name':['Paul','John','Bill'], 'age':[25,27,23]})
print('Inital Dataframe:\n',df)

print("type(df.loc[:, 'age']):", type(df.loc[:, 'age']))
print("type(df.loc[:, 'name']):", type(df.loc[:, 'name']))
df.loc[:, 'name'] = df.loc[:, 'age']
print("\ndf.loc[:, 'name'] = df.loc[:, 'age']\n",df)

And the output:

Inital Dataframe:
    name  age
0  Paul   25
1  John   27
2  Bill   23

df[['name']] = df[['age']]
    name  age
0    25   25
1    27   27
2    23   23
df.loc[:, ['age']]: <class 'pandas.core.frame.DataFrame'>
df.loc[:, ['name']]: <class 'pandas.core.frame.DataFrame'>

df.loc[:, ['name']] = df.loc[:, ['age']]
    name   age
0   NaN  25.0
1   NaN  27.0
2   NaN  23.0
=======================
Inital Dataframe:
    name  age
0  Paul   25
1  John   27
2  Bill   23
type(df.loc[:, 'age']): <class 'pandas.core.series.Series'>
type(df.loc[:, 'name']): <class 'pandas.core.series.Series'>

df.loc[:, 'name'] = df.loc[:, 'age']
    name  age
0    25   25
1    27   27
2    23   23

However, here is another strange behaviour: Assigning the double brackets to difference variables, say df1 and df2, and then df1 = df2 works! Here is some more code:

df = pd.DataFrame({'name':['Paul','John','Bill'], 'age':[25,27,23]})
print('Inital Dataframe:\n',df)

df1 = df.loc[:, ['name']]
df2 = df.loc[:, ['age']]
print("\ndf1 = df.loc[:, ['name']]\n",df1)
print("\ndf2 = df.loc[:, ['age']]\n",df2)

df1=df2
print("\ndf1=df2\ndf1:\n",df1)

And the output:

Inital Dataframe:
    name  age
0  Paul   25
1  John   27
2  Bill   23

df1 = df.loc[:, ['name']]
    name
0  Paul
1  John
2  Bill

df2 = df.loc[:, ['age']]
    age
0   25
1   27
2   23

df1=df2
df1:
    age
0   25
1   27
2   23