dstandish commented on code in PR #26208:
URL: https://github.com/apache/airflow/pull/26208#discussion_r965175687
##########
docs/apache-airflow/concepts/datasets.rst:
##########
@@ -15,33 +15,185 @@
specific language governing permissions and limitations
under the License.
-Datasets
-========
+Data-aware scheduling
+=====================
.. versionadded:: 2.4
-With datasets, instead of running a DAG on a schedule, a DAG can be configured
to run when a dataset has been updated.
+Quickstart
+----------
-To use this feature, define a dataset:
+In addition to scheduling DAGs based upon time, they can also be scheduled
based upon a task updating a dataset.
-.. exampleinclude:: /../../airflow/example_dags/example_datasets.py
- :language: python
- :start-after: [START dataset_def]
- :end-before: [END dataset_def]
+.. code-block:: python
-Then reference the dataset as a task outlet:
+ from airflow import Dataset
-.. exampleinclude:: /../../airflow/example_dags/example_datasets.py
- :language: python
- :dedent: 4
- :start-after: [START task_outlet]
- :end-before: [END task_outlet]
+ with DAG(...):
+ MyOperator(
+ # this task updates example.csv
+ outlets=[Dataset("s3://dataset-bucket/example.csv")],
+ ...,
+ )
-Finally, define a DAG and reference this dataset in the DAG's ``schedule``
argument:
-.. exampleinclude:: /../../airflow/example_dags/example_datasets.py
- :language: python
- :start-after: [START dag_dep]
- :end-before: [END dag_dep]
+ with DAG(
+ # this DAG should be run when example.csv is updated (by dag1)
+ schedule=[Dataset("s3://dataset-bucket/example.csv")],
+ ...,
+ ):
+ ...
-You can reference multiple datasets in the DAG's ``schedule`` argument. Once
there has been an update to all of the upstream datasets, the DAG will be
triggered. This means that the DAG will run as frequently as its
least-frequently-updated dataset.
+What is a "dataset"?
+--------------------
+
+An Airflow dataset is a stand-in for a logical grouping of data that flows
through multiple DAGs, possibly being changed or updated by each one. Datasets
are updated by upstream "producer" tasks, and dataset updates contribute to
scheduling downstream "consumer" DAGs.
+
+A dataset is a construct around a Uniform Resource Identifier (URI) that you
create in your code:
+
+.. code-block:: python
+
+ from airflow import Dataset
+
+ example_dataset = Dataset('s3://dataset-bucket/example.csv')
+
+Airflow treats the dataset URI as an opaque value intended to be
human-readable, and makes no assumptions about the content or location of the
data represented by the identifier. It is treated as a string, so any use of
regular expressions (eg ``input_\d+.csv``) or file glob patterns (eg
``input_2022*.csv``) as an attempt to create multiple datasets from one
declaration will not work.
Review Comment:
why: cus ... human readability is not necessarily a primary goal. if this
ultimately needs to line up with openlineage naming, then human readability may
not be a goal at all.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
