Skip to content

Rerun selected dates

Declare a date-partitioned asset, inspect its selection contract, then plan and run only the reviewed slice.

daily.py
from kazeflow import DatePartitionDef, Flow, asset


@asset(partition_def=DatePartitionDef())
def publish_daily() -> None:
    print("published selected date")


flow = Flow(["publish_daily"])

kazeflow partitions daily.py
kazeflow plan daily.py --partition-key 2026-08-11
kazeflow run daily.py --partition-key 2026-08-11 --yes

DatePartitionDef strictly validates canonical YYYY-MM-DD input and normalizes it to a Python date before execution. An invalid date, a non-canonical date string, or omitted selection is rejected during preflight; publish_daily does not run.

For consecutive dates, use one explicit inclusive range rather than relying on an implicit current day or catalog:

kazeflow plan daily.py --partition-range 2026-08-11 2026-08-12
kazeflow run daily.py --partition-range 2026-08-11 2026-08-12 --yes

The equivalent Python selection remains bounded and explicit:

keys = DatePartitionDef().range("2026-08-11", "2026-08-12")
plan = flow.plan({"partition_keys": keys})

To plan no partition work on purpose, use --empty-partitions. This is distinct from omitting every selector, which is an error for a partitioned flow:

kazeflow plan daily.py --empty-partitions

See Select partitions deliberately for repeatable keys, custom falsey keys, domain compatibility, and portable-record sensitivity.