Why the difference in checking for value of pd.DataFrame vs pd.Series if value in index?

I’m working with a pandas DataFrame and I noticed a difference in behavior when using the in operator.

Here’s an example to illustrate this:

import pandas as pd

df = pd.DataFrame({'a': [4, 5, 6], 'b': [7, 8, 9]})

print(1 in df)
print(type(df))

print(1 in df["a"])
print(type(df["a"]))

Output:

False
<class 'pandas.core.frame.DataFrame'>
True
<class 'pandas.core.series.Series'>

The most straightforward difference is that of course one object is a DataFrame, the other is a Series; nonetheless, I was not really expecting 1 to be found in the Index of the Series and this expression evaluating to True. Especially as it is False for the DataFrame.

Is there an explanation why I should have been expecting this?

>Solution :

In both cases the in operator calls __contains__ to test membership. Both pd.DataFrame and pd.Series are subclasses of NDFrame, which has this method defined as follows:

    def __contains__(self, key) -> bool:
        """True if the key is in the info axis"""
        return key in self._info_axis

So, under the hood the following happens:

print(df._info_axis)
# Index(['a', 'b'], dtype='object')

print(df.__contains__(1))
# False

# approx.: 1 in ['a', 'b']

print(df['a']._info_axis)
# RangeIndex(start=0, stop=3, step=1)

print(df['a'].__contains__(1))
# True

# approx.: 1 in [0, 1, 2]

I.e., the difference lies in the fact that a df uses df.columns as the ‘info axis’, while a series, such as df['a'], naturally must use series.index.