Why does the shape of the selection of my pandas dataframe is wrong

Question

I have a pandas DataFrame called df where df.shape is (53, 80) where indexes and columns are both int.

If I select the first row like this, I get :

df.loc[0].shape
(80,)

instead of :

(1,80)

But then df.loc[0:0].shape or df[0:1].shape both show the correct shape.

Answer 1

df.loc[0] returns a one-dimensional pd.Series object representing the data in a single row, extracted via indexing.

df.loc[0:0] returns a two-dimensional pd.DataFrame object representing the data in a dataframe with one row, extracted via slicing.

You can see this more clearly if you print the results of these operations:

import pandas as pd, numpy as np

df = pd.DataFrame(np.arange(9).reshape(3, 3))

res1 = df.loc[0]
res2 = df.loc[0:0]

print(type(res1), res1, sep='\n')

<class 'pandas.core.series.Series'>
0    0
1    1
2    2
Name: 0, dtype: int32

print(type(res2), res2, sep='\n')

<class 'pandas.core.frame.DataFrame'>
   0  1  2
0  0  1  2

The convention follows NumPy indexing / slicing. This is natural since Pandas is built on NumPy arrays.

arr = np.arange(9).reshape(3, 3)

print(arr[0].shape)    # (3,), i.e. 1-dimensional
print(arr[0:0].shape)  # (0, 3), i.e. 2-dimensional

Answer 2

When you call df.iloc[0], it is selecting first row and type is Series whereas, in other case df.iloc[0:0] it is slicing rows and is of type dataframe. And Series are according to pandas Series documentation :

One-dimensional ndarray with axis labels

whereas dataframe are Two-dimensional (pandas Dataframe documentation).

Try running following lines to see the difference:

print(type(df.iloc[0]))
# <class 'pandas.core.series.Series'>

print(type(df.iloc[0:0]))
# <class 'pandas.core.frame.DataFrame'>