Class SingleCellStreamingAddTest

java.lang.Object
ubic.gemma.core.util.test.BaseTest5
ubic.gemma.core.util.test.BaseDatabaseTest5
ubic.gemma.persistence.service.expression.experiment.SingleCellStreamingAddTest
@ContextConfiguration public class SingleCellStreamingAddTest extends BaseDatabaseTest5
Tests for the streaming overload of SingleCellExpressionExperimentService.addSingleCellDataVectors(ExpressionExperiment, QuantitationType, SingleCellDimension, Stream, String, boolean, boolean).

Mirrors the in-memory-H2 harness of SingleCellExpressionExperimentServiceTest, including its mock beans for the read/write service split. These exercise the single-pass path added to stop single-cell imports from materialising every vector before writing any of them.

The audit event is not asserted here. This context wires the impl directly with no AOP proxy, so the @Audited advice never fires; aspect coverage lives in AuditedAspectTest.

  • Constructor Details

    • SingleCellStreamingAddTest

      public SingleCellStreamingAddTest()

  • Method Details

    • setUp

      @BeforeEach public void setUp()

    • resetMocks

      @AfterEach public void resetMocks()

    • testStreamingAddPersistsAllVectors

      @Test public void testStreamingAddPersistsAllVectors()

    • testStreamingAddRecordsTheDimensionInTheLinkTable

      @Test public void testStreamingAddRecordsTheDimensionInTheLinkTable()
      The dimension must be findable through the SINGLE_CELL_DIMENSION_EXPERIMENT link table, as it is after the collection variant. HBCC_Cohort (93544), added through the streaming path, had no link row, so deleteExperiments never found its dimension and failed on BIO_ASSAYS_SC_FKC.

    • testStreamingAddMatchesCollectionVariant

      @Test public void testStreamingAddMatchesCollectionVariant()

    • testStreamingAddAppliesSparsityMetricsForPreferredQt

      @Test public void testStreamingAddAppliesSparsityMetricsForPreferredQt()

    • testUpdateSparsityMetricsCountsEverySample

      @Test public void testUpdateSparsityMetricsCountsEverySample()
      Recomputing the metrics from the stored vectors must give what the add computed.

      The streaming applyBioAssaySparsityMetrics behind updateSparsityMetrics incremented sampleIndex a second time inside its loop, so samples 1 and 3 were never counted and got 0 cells, 0 design elements and 0 cells by design element. Master fixed it in ee070b4f3c; it was never ported.

    • testStreamingAddDoesNotApplySparsityForNonPreferredQt

      @Test public void testStreamingAddDoesNotApplySparsityForNonPreferredQt()

    • testStreamingAddCreatesCellTypeFactorWhenPreferred

      @Test public void testStreamingAddCreatesCellTypeFactorWhenPreferred()

    • testStreamingAddRejectsEmptyStream

      @Test public void testStreamingAddRejectsEmptyStream()

    • testStreamingAddRejectsWrongQuantitationType

      @Test public void testStreamingAddRejectsWrongQuantitationType()

    • testStreamingAddRejectsWrongSingleCellDimension

      @Test public void testStreamingAddRejectsWrongSingleCellDimension()

    • testStreamingAddRejectsDuplicateQuantitationTypeName

      @Test public void testStreamingAddRejectsDuplicateQuantitationTypeName()

    • testStreamingAddConsumesSourceExactlyOnce

      @Test public void testStreamingAddConsumesSourceExactlyOnce()
      Single-pass consumption is the whole point of the streaming overload: if the implementation ever materialises the source internally (a toList() slipped into the path, say), the memory saving evaporates silently and only this test notices. The counting iterator records every pull.

    • testStreamingAddAcceptsPersistentDimensionFromPriorAdd

      @Test public void testStreamingAddAcceptsPersistentDimensionFromPriorAdd()
      Covers the scdJustCreated branch of the streaming validator: re-adding vectors against a dimension that is already wired to existing vectors must be allowed.