BUG: pandas.Interval-inherited class converted automatically to pandas.Interval

[PetitLepton]

Research

• I have searched the [pandas] tag on StackOverflow for similar questions.

• I have asked my usage related question on StackOverflow.

Link to question on StackOverflow

https://stackoverflow.com/questions/72044363/automatic-conversion-of-a-column-to-the-parent-class-in-pandas

Question about pandas

import pandas


class TimestampsInterval(pandas.Interval):
    def __init__(self, left: str, right: str, closed: str = "both") -> None:
        super().__init__(pandas.Timestamp(left), pandas.Timestamp(right), closed)

    @property
    def seconds(self) -> float:
        return self.length.seconds


df = pandas.DataFrame(
    data={
        "intervals": [
            TimestampsInterval("1970-01-01 00:00:00", "1970-01-01 00:00:01"),
        ],
    }
)

# I was expecting the DataFrame to keep the class intact but it converts the column 
# to the parent class pandas.Interval
try:
    df["intervals"].map(lambda x: x.seconds)
except AttributeError as e:
    print(e)  # 'pandas._libs.interval.Interval' object has no attribute 'seconds'
    print(df["intervals"].dtype)  # interval[datetime64[ns], both]

# If I "force" pandas not to convert it by embedding the interval into a list, then 
# the class is indeed kept
df = pandas.DataFrame(
    data={
        "intervals": [
            [TimestampsInterval("1970-01-01 00:00:00", "1970-01-01 00:00:01")],
        ],
    }
)

df["intervals"].map(lambda x: x[0].seconds)