Merge unique ids for generated tasks into dev

Fixes #985.

CHANGELOG.rst

@@ -677,9 +677,13 @@ Bug Fixes
 * Fix aggregation of DSM-components
   `#1058 <https://github.com/openego/eGon-data/issues/1069>`_
 * Fix URL of TYNDP scenario dataset
+* Automatically generated tasks now get unique :code:`task_id`\s.
+  Fixes issue `#985`_ via PR `#986`_.
 
 .. _PR #692: https://github.com/openego/eGon-data/pull/692
 .. _#343: https://github.com/openego/eGon-data/issues/343
 .. _#556: https://github.com/openego/eGon-data/issues/556
 .. _#641: https://github.com/openego/eGon-data/issues/641
 .. _#669: https://github.com/openego/eGon-data/issues/669
+.. _#985: https://github.com/openego/eGon-data/issues/985
+.. _#986: https://github.com/openego/eGon-data/pull/986

setup.cfg

@@ -25,7 +25,7 @@ python_files =
     tests.py
 addopts =
     -ra
-    --strict
+    --strict-markers
     --ignore=docs/conf.py
     --ignore=setup.py
     --ignore=ci

src/egon/data/datasets/__init__.py

@@ -230,7 +230,7 @@ def __post_init__(self):
             # Explicitly create single final task, because we can't know
             # which of the multiple tasks finishes last.
             name = prefix(self)
-            name = name if name else f"{self.name}."
+            name = f"{name if name else f'{self.__module__}.'}{self.name}."
             update_version = PythonOperator(
                 task_id=f"{name}update-version",
                 # Do nothing, because updating will be added later.

tests/test_dataset_class.py

@@ -0,0 +1,45 @@
+from dataclasses import dataclass
+from typing import Union
+
+from airflow.models.dag import DAG
+
+from egon.data.datasets import Dataset, TaskGraph, Tasks
+
+
+def test_uniqueness_of_automatically_generated_final_dataset_task():
+    """Test that the generated final dataset task is named uniquely.
+
+    This is a regression test for issue #985. Having multiple `Dataset`s ending
+    in parallel tasks doesn't work if those `Dataset`s are in a module below
+    the `egon.data.datasets` package. In that case the code removing the module
+    name prefix from task ids and the code generating the final dataset task
+    which updates the dataset version once all parallel tasks have finished
+    interact in a way that generates non-distinct task ids so that tasks
+    generated later clobber the ones generated earlier. This leads to spurious
+    cycles and other inconsistencies and bugs in the graph.
+    """
+
+    noops = [(lambda: None) for _ in range(4)]
+    for i, noop in enumerate(noops):
+        noop.__name__ = f"noop-{i}"
+
+    @dataclass
+    class Dataset_1(Dataset):
+        name: str = "DS1"
+        version: str = "0.0.0"
+        tasks: Union[Tasks, TaskGraph] = ({noops[0], noops[1]},)
+
+    @dataclass
+    class Dataset_2(Dataset):
+        name: str = "DS2"
+        version: str = "0.0.0"
+        tasks: Union[Tasks, TaskGraph] = ({noops[2], noops[3]},)
+
+    Dataset_1.__module__ = "egon.data.datasets.test.datasets"
+    Dataset_2.__module__ = "egon.data.datasets.test.datasets"
+    with DAG(dag_id="Test-DAG", default_args={"start_date": "1111-11-11"}):
+        datasets = [Dataset_1(), Dataset_2()]
+    ids = [list(dataset.tasks)[-1] for dataset in datasets]
+    assert (
+        ids[0] != ids[1]
+    ), "Expected unique names for final tasks of distinct datasets."