Temporal PyArrow types
At this point, we have reviewed many of the “first-class” data types built into pandas, while highlighting some rough edges and inconsistencies that plague them. Despite those issues, the types baked into pandas can take you a long way in your data journey.
But there are still cases where the pandas types are not suitable, with a common case being interoperability with databases. Most databases have distinct DATE and DATETIME types, so the fact that pandas only offers a DATETIME type can be disappointing to users fluent in SQL.
Fortunately, the Apache Arrow project defines a true DATE type. Starting in version 2.0, pandas users can start leveraging Arrow types exposed through the PyArrow library.
How to do it
To construct PyArrow types in pandas directly, you will always provide a dtype= argument of the pd.ArrowDtype(XXX) form, replacing XXX with the appropriate PyArrow type. The DATE type in PyArrow is called pa.date32():
ser...
