Allow annotations to be attached to transforms via a context.

This API is offered both on the pipeline object itself, and also
as a (thread-local) top-level function  as one many not always
have an easy reference to the pipeline.

sdks/python/apache_beam/__init__.py

@@ -89,7 +89,7 @@
 from apache_beam import metrics
 from apache_beam import typehints
 from apache_beam import version
-from apache_beam.pipeline import Pipeline
+from apache_beam.pipeline import *
 from apache_beam.transforms import *
 from apache_beam.pvalue import PCollection
 from apache_beam.pvalue import Row

sdks/python/apache_beam/pipeline.py

@@ -54,6 +54,7 @@
 import re
 import shutil
 import tempfile
+import threading
 import unicodedata
 import uuid
 from collections import defaultdict
@@ -109,7 +110,7 @@
   from apache_beam.runners.runner import PipelineResult
   from apache_beam.transforms import environments
 
-__all__ = ['Pipeline', 'PTransformOverride']
+__all__ = ['Pipeline', 'transform_annotations']
 
 
 class Pipeline(HasDisplayData):
@@ -226,7 +227,9 @@ def __init__(
     self.runner = runner
     # Stack of transforms generated by nested apply() calls. The stack will
     # contain a root node as an enclosing (parent) node for top transforms.
-    self.transforms_stack = [AppliedPTransform(None, None, '', None)]
+    self.transforms_stack = [
+        AppliedPTransform(None, None, '', None, None, None)
+    ]
     # Set of transform labels (full labels) applied to the pipeline.
     # If a transform is applied and the full label is already in the set
     # then the transform will have to be cloned with a new label.
@@ -244,6 +247,7 @@ def __init__(
 
     self._display_data = display_data or {}
     self._error_handlers = []
+    self._annotations_stack = [{}]
 
   def display_data(self):
     # type: () -> Dict[str, Any]
@@ -268,6 +272,24 @@ def _current_transform(self):
     """Returns the transform currently on the top of the stack."""
     return self.transforms_stack[-1]
 
+  @contextlib.contextmanager
+  def transform_annotations(self, **annotations):
+    """A context manager for attaching annotations to a set of transforms.
+
+    All transforms applied while this context is active will have these
+    annotations attached. This includes  sub-transforms applied within
+    composite transforms.
+    """
+    self._annotations_stack.append({
+        **self._annotations_stack[-1], **encode_annotations(annotations)
+    })
+    yield
+    self._annotations_stack.pop()
+
+  def _current_annotations(self):
+    """Returns the set of annotations that should be used on apply."""
+    return {**_global_annotations_stack()[-1], **self._annotations_stack[-1]}
+
   def _root_transform(self):
     # type: () -> AppliedPTransform
 
@@ -316,7 +338,9 @@ def _replace_if_needed(self, original_transform_node):
               original_transform_node.parent,
               replacement_transform,
               original_transform_node.full_label,
-              original_transform_node.main_inputs)
+              original_transform_node.main_inputs,
+              None,
+              annotations=original_transform_node.annotations)
 
           # TODO(https://github.com/apache/beam/issues/21178): Merge rather
           # than override.
@@ -741,7 +765,12 @@ def apply(
             'returned %s from %s' % (transform, inputs, pvalueish))
 
     current = AppliedPTransform(
-        self._current_transform(), transform, full_label, inputs)
+        self._current_transform(),
+        transform,
+        full_label,
+        inputs,
+        None,
+        annotations=self._current_annotations())
     self._current_transform().add_part(current)
 
     try:
@@ -1014,7 +1043,7 @@ def from_runner_api(
       root_transform_id, = proto.root_transform_ids
       p.transforms_stack = [context.transforms.get_by_id(root_transform_id)]
     else:
-      p.transforms_stack = [AppliedPTransform(None, None, '', None)]
+      p.transforms_stack = [AppliedPTransform(None, None, '', None, None, None)]
     # TODO(robertwb): These are only needed to continue construction. Omit?
     p.applied_labels = {
         t.unique_name
@@ -1124,8 +1153,8 @@ def __init__(
       transform,  # type: Optional[ptransform.PTransform]
       full_label,  # type: str
       main_inputs,  # type: Optional[Mapping[str, Union[pvalue.PBegin, pvalue.PCollection]]]
-      environment_id=None,  # type: Optional[str]
-      annotations=None, # type: Optional[Dict[str, bytes]]
+      environment_id,  # type: Optional[str]
+      annotations, # type: Optional[Dict[str, bytes]]
   ):
     # type: (...) -> None
     self.parent = parent
@@ -1149,24 +1178,11 @@ def __init__(
         transform.get_resource_hints()) if transform else {
         }  # type: Dict[str, bytes]
 
-    if annotations is None and transform:
-
-      def annotation_to_bytes(key, a: Any) -> bytes:
-        if isinstance(a, bytes):
-          return a
-        elif isinstance(a, str):
-          return a.encode('ascii')
-        elif isinstance(a, message.Message):
-          return a.SerializeToString()
-        else:
-          raise TypeError(
-              'Unknown annotation type %r (type %s) for %s' % (a, type(a), key))
-
+    if transform:
       annotations = {
-          key: annotation_to_bytes(key, a)
-          for key,
-          a in transform.annotations().items()
+          **annotations or {}, **encode_annotations(transform.annotations())
       }
+
     self.annotations = annotations
 
   @property
@@ -1478,6 +1494,50 @@ def _merge_outer_resource_hints(self):
         part._merge_outer_resource_hints()
 
 
+def encode_annotations(annotations: Optional[Dict[str, Any]]):
+  """Encodes non-byte annotation values as bytes."""
+  if not annotations:
+    return {}
+
+  def annotation_to_bytes(key, a: Any) -> bytes:
+    if isinstance(a, bytes):
+      return a
+    elif isinstance(a, str):
+      return a.encode('ascii')
+    elif isinstance(a, message.Message):
+      return a.SerializeToString()
+    else:
+      raise TypeError(
+          'Unknown annotation type %r (type %s) for %s' % (a, type(a), key))
+
+  return {key: annotation_to_bytes(key, a) for (key, a) in annotations.items()}
+
+
+_global_annotations_stack_data = threading.local()
+
+
+def _global_annotations_stack():
+  try:
+    return _global_annotations_stack_data.stack
+  except AttributeError:
+    _global_annotations_stack_data.stack = [{}]
+    return _global_annotations_stack_data.stack
+
+
+@contextlib.contextmanager
+def transform_annotations(**annotations):
+  """A context manager for attaching annotations to a set of transforms.
+
+  All transforms applied while this context is active will have these
+  annotations attached. This includes  sub-transforms applied within
+  composite transforms.
+  """
+  cur_stack = _global_annotations_stack()
+  cur_stack.append({**cur_stack[-1], **encode_annotations(annotations)})
+  yield
+  cur_stack.pop()
+
+
 class PTransformOverride(metaclass=abc.ABCMeta):
   """For internal use only; no backwards-compatibility guarantees.
 

sdks/python/apache_beam/pipeline_test.py

@@ -1016,6 +1016,51 @@ def annotations(self):
             transform.annotations['proto'], some_proto.SerializeToString())
     self.assertEqual(seen, 2)
 
+  def assertHasAnnotation(self, pipeline_proto, transform, key, value):
+    for transform_proto in pipeline_proto.components.transforms.values():
+      if transform_proto.unique_name == transform:
+        self.assertIn(key, transform_proto.annotations.keys())
+        self.assertEqual(transform_proto.annotations[key], value)
+        break
+    else:
+      self.fail(
+          "Unknown transform: %r not in %s" % (
+              transform,
+              sorted([
+                  t.unique_name
+                  for t in pipeline_proto.components.transforms.values()
+              ])))
+
+  def test_pipeline_context_annotations(self):
+    p = beam.Pipeline()
+    with p.transform_annotations(foo='first'):
+      pcoll = p | beam.Create([1, 2, 3]) | 'First' >> beam.Map(lambda x: x + 1)
+    with p.transform_annotations(foo='second'):
+      pcoll | 'Second' >> beam.Map(lambda x: x * 2)
+      with p.transform_annotations(foo='nested', another='more'):
+        pcoll | 'Nested' >> beam.Map(lambda x: x * 3)
+
+    proto = p.to_runner_api()
+    self.assertHasAnnotation(proto, 'First', 'foo', b'first')
+    self.assertHasAnnotation(proto, 'Second', 'foo', b'second')
+    self.assertHasAnnotation(proto, 'Nested', 'foo', b'nested')
+    self.assertHasAnnotation(proto, 'Nested', 'another', b'more')
+
+  def test_beam_context_annotations(self):
+    p = beam.Pipeline()
+    with beam.transform_annotations(foo='first'):
+      pcoll = p | beam.Create([1, 2, 3]) | 'First' >> beam.Map(lambda x: x + 1)
+    with beam.transform_annotations(foo='second'):
+      pcoll | 'Second' >> beam.Map(lambda x: x * 2)
+      with beam.transform_annotations(foo='nested', another='more'):
+        pcoll | 'Nested' >> beam.Map(lambda x: x * 3)
+
+    proto = p.to_runner_api()
+    self.assertHasAnnotation(proto, 'First', 'foo', b'first')
+    self.assertHasAnnotation(proto, 'Second', 'foo', b'second')
+    self.assertHasAnnotation(proto, 'Nested', 'foo', b'nested')
+    self.assertHasAnnotation(proto, 'Nested', 'another', b'more')
+
   def test_transform_ids(self):
     class MyPTransform(beam.PTransform):
       def expand(self, p):

sdks/python/apache_beam/runners/dataflow/dataflow_runner_test.py

@@ -272,7 +272,7 @@ def test_group_by_key_input_visitor_with_valid_inputs(self):
     pcoll3.element_type = typehints.KV[typehints.Any, typehints.Any]
     for pcoll in [pcoll1, pcoll2, pcoll3]:
       applied = AppliedPTransform(
-          None, beam.GroupByKey(), "label", {'pcoll': pcoll})
+          None, beam.GroupByKey(), "label", {'pcoll': pcoll}, None, None)
       applied.outputs[None] = PCollection(None)
       common.group_by_key_input_visitor().visit_transform(applied)
       self.assertEqual(
@@ -291,15 +291,19 @@ def test_group_by_key_input_visitor_with_invalid_inputs(self):
     for pcoll in [pcoll1, pcoll2]:
       with self.assertRaisesRegex(ValueError, err_msg):
         common.group_by_key_input_visitor().visit_transform(
-            AppliedPTransform(None, beam.GroupByKey(), "label", {'in': pcoll}))
+            AppliedPTransform(None, beam.GroupByKey(), "label", {'in': pcoll}),
+            None,
+            None)
 
   def test_group_by_key_input_visitor_for_non_gbk_transforms(self):
     p = TestPipeline()
     pcoll = PCollection(p)
     for transform in [beam.Flatten(), beam.Map(lambda x: x)]:
       pcoll.element_type = typehints.Any
       common.group_by_key_input_visitor().visit_transform(
-          AppliedPTransform(None, transform, "label", {'in': pcoll}))
+          AppliedPTransform(None, transform, "label", {'in': pcoll}),
+          None,
+          None)
       self.assertEqual(pcoll.element_type, typehints.Any)
 
   def test_flatten_input_with_visitor_with_single_input(self):
@@ -319,7 +323,8 @@ def _test_flatten_input_visitor(self, input_type, output_type, num_inputs):
     output_pcoll = PCollection(p)
     output_pcoll.element_type = output_type
 
-    flatten = AppliedPTransform(None, beam.Flatten(), "label", inputs)
+    flatten = AppliedPTransform(
+        None, beam.Flatten(), "label", inputs, None, None)
     flatten.add_output(output_pcoll, None)
     DataflowRunner.flatten_input_visitor().visit_transform(flatten)
     for _ in range(num_inputs):
@@ -357,7 +362,8 @@ def test_side_input_visitor(self):
         z: (x, y, z),
         beam.pvalue.AsSingleton(pc),
         beam.pvalue.AsMultiMap(pc))
-    applied_transform = AppliedPTransform(None, transform, "label", {'pc': pc})
+    applied_transform = AppliedPTransform(
+        None, transform, "label", {'pc': pc}, None, None)
     DataflowRunner.side_input_visitor().visit_transform(applied_transform)
     self.assertEqual(2, len(applied_transform.side_inputs))
     self.assertEqual(