Test Beam Sum returns sensible result

tests/private_beam_test.py

@@ -4,6 +4,7 @@
 from apache_beam import pvalue
 from unittest.mock import patch
 import apache_beam.testing.util as beam_util
+from apache_beam.testing.test_pipeline import TestPipeline
 
 import pipeline_dp
 from pipeline_dp import private_beam
@@ -22,6 +23,11 @@ class PrivateBeamTest(unittest.TestCase):
     def privacy_id_extractor(x):
         return f"pid:{x}"
 
+    @staticmethod
+    def value_per_key_within_tolerance(expected, actual, tolerance):
+        return actual[0] == expected[0] and actual[1] >= expected[
+            1] - tolerance and actual[1] <= expected[1] + tolerance
+
     def test_make_private_transform_succeeds(self):
         runner = fn_api_runner.FnApiRunner()
         with beam.Pipeline(runner=runner) as pipeline:
@@ -154,6 +160,45 @@ def test_sum_calls_aggregate_with_params(self, mock_aggregate):
                 public_partitions=sum_params.public_partitions)
             self.assertEqual(args[1], params)
 
+    def test_sum_returns_sensible_result(self):
+        with TestPipeline() as pipeline:
+            # Arrange
+            col = [(u, "pk1", 100) for u in range(30)]
+            col += [(u + 30, "pk1", -100) for u in range(30)]
+            pcol = pipeline | 'Create produce' >> beam.Create(col)
+            # Use very high epsilon and delta to minimize noise and test
+            # flakiness.
+            budget_accountant = budget_accounting.NaiveBudgetAccountant(
+                total_epsilon=800, total_delta=0.999)
+            private_collection = (
+                pcol | 'Create private collection' >> private_beam.MakePrivate(
+                    budget_accountant=budget_accountant,
+                    privacy_id_extractor=lambda x: x[0]))
+
+            sum_params = aggregate_params.SumParams(
+                noise_kind=pipeline_dp.NoiseKind.GAUSSIAN,
+                max_partitions_contributed=2,
+                max_contributions_per_partition=3,
+                low=1,
+                high=2,
+                budget_weight=1,
+                partition_extractor=lambda x: x[1],
+                value_extractor=lambda x: x[2])
+
+            # Act
+            result = private_collection | private_beam.Sum(
+                sum_params=sum_params)
+            budget_accountant.compute_budgets()
+
+            # Assert
+            # This is a health check to validate that the result is sensible.
+            # Hence, we use a very large tolerance to reduce test flakiness.
+            beam_util.assert_that(
+                result,
+                beam_util.equal_to([("pk1", 90)],
+                                   equals_fn=lambda e, a: PrivateBeamTest.
+                                   value_per_key_within_tolerance(e, a, 10)))
+
     @patch('pipeline_dp.dp_engine.DPEngine.aggregate')
     def test_count_calls_aggregate_with_params(self, mock_aggregate):
         runner = fn_api_runner.FnApiRunner()