In Airflow, there is a built-in function, which we can use to find the Hive partition closest to the given date. However, it works only with partition identifiers in the YYYY-mm-dd format, so if you use a different partitioning method, this function will not help you.

To find the closest Hive partition, we should use the closest_ds_partition function:

1
2
3
4
5
6
7
8
9
from airflow.macros.hive import closest_ds_partition

closest_ds_partition(
    hive_table_name,
    the_date,
    before=True,
    schema='hive_schema',
    metastore_conn_id='metastore_connection_id'
)

Be careful with the before parameter. It has a weird behavior. As you may expect, True means a partition before the given date, False returns the partition after a given date, but when the before parameter is set to None it returns the closest partition, and it does not matter whether it is before or after the given date.

Please don’t follow this coding practice. Three value “boolean” logic is a terrible, terrible idea. It is way better to use an enum with descriptive names.