bugfixes: lazy validation, strategies (#550)

pandera/strategies.py

@@ -990,7 +990,7 @@ def _dataframe_strategy(draw):
         }
 
         row_strategy = None
-        if checks:
+        if row_strategy_checks:
             row_strategy = st.fixed_dictionaries(
                 {
                     col_name: make_row_strategy(col, row_strategy_checks)

tests/core/test_schemas.py

@@ -1248,6 +1248,81 @@ def test_lazy_dataframe_validation_nullable():
             )
 
 
+def test_lazy_dataframe_validation_with_checks():
+    """Test that all failure cases are reported for schemas with checks."""
+    schema = DataFrameSchema(
+        columns={
+            "analysis_path": Column(String),
+            "run_id": Column(String),
+            "sample_type": Column(String, Check.isin(["DNA", "RNA"])),
+            "sample_valid": Column(String, Check.isin(["Yes", "No"])),
+        },
+        strict=False,
+        coerce=True,
+    )
+
+    df = pd.DataFrame.from_dict(
+        {
+            "analysis_path": ["/", "/", "/", "/", "/"],
+            "run_id": ["1", "2", "3", "4", "5"],
+            "sample_type": ["DNA", "RNA", "DNA", "RNA", "RNA"],
+            "sample_valid": ["Yes", "YES", "YES", "NO", "NO"],
+        }
+    )
+
+    try:
+        schema(df, lazy=True)
+    except errors.SchemaErrors as err:
+        failure_case = err.failure_cases.failure_case.tolist()
+        assert failure_case == ["YES", "YES", "NO", "NO"]
+
+
+def test_lazy_dataframe_validation_nullable_with_checks():
+    """
+    Test that checks in non-nullable column failure cases are correctly
+    processed during lazy validation.
+    """
+    schema = DataFrameSchema(
+        {
+            "id": Column(
+                String,
+                checks=Check.str_matches(r"^ID[\d]{3}$"),
+                name="id",
+                required=True,
+                allow_duplicates=False,
+            )
+        }
+    )
+    df = pd.DataFrame({"id": ["ID001", None, "XXX"]})
+    try:
+        schema(df, lazy=True)
+    except errors.SchemaErrors as err:
+        expected_failure_cases = pd.DataFrame.from_dict(
+            {
+                0: {
+                    "schema_context": "Column",
+                    "column": "id",
+                    "check": "not_nullable",
+                    "check_number": None,
+                    "failure_case": None,
+                    "index": 1,
+                },
+                1: {
+                    "schema_context": "Column",
+                    "column": "id",
+                    "check": r"str_matches(re.compile('^ID[\\d]{3}$'))",
+                    "check_number": 0,
+                    "failure_case": "XXX",
+                    "index": 2,
+                },
+            },
+            orient="index",
+        ).astype({"check_number": object})
+        pd.testing.assert_frame_equal(
+            err.failure_cases, expected_failure_cases
+        )
+
+
 @pytest.mark.parametrize(
     "schema_cls, data",
     [

tests/strategies/test_strategies.py

@@ -794,35 +794,51 @@ def test_unsatisfiable_checks():
             schema.example(size=10)
 
 
-@pytest.fixture(scope="module")
-def schema_model():
-    """Schema model fixture."""
+class Schema(pa.SchemaModel):
+    """Schema model for strategy testing."""
 
-    class Schema(pa.SchemaModel):
-        """Schema model for strategy testing."""
+    col1: pa.typing.Series[int]
+    col2: pa.typing.Series[float]
+    col3: pa.typing.Series[str]
 
-        col1: pa.typing.Series[int]
-        col2: pa.typing.Series[float]
-        col3: pa.typing.Series[str]
 
-    return Schema
+@hypothesis.given(st.data())
+@hypothesis.settings(
+    suppress_health_check=[hypothesis.HealthCheck.too_slow],
+)
+def test_schema_model_strategy(data):
+    """Test that strategy can be created from a SchemaModel."""
+    strat = Schema.strategy(size=10)
+    sample_data = data.draw(strat)
+    Schema.validate(sample_data)
 
 
 @hypothesis.given(st.data())
 @hypothesis.settings(
     suppress_health_check=[hypothesis.HealthCheck.too_slow],
 )
-def test_schema_model_strategy(schema_model, data):
-    """Test that strategy can be created from a SchemaModel."""
-    strat = schema_model.strategy(size=10)
+def test_schema_model_strategy_df_check(data):
+    """Test that schema with custom checks produce valid data."""
+
+    class SchemaWithDFCheck(Schema):
+        """Schema with a custom dataframe-level check with no strategy."""
+
+        # pylint:disable=no-self-use
+        @pa.dataframe_check
+        @classmethod
+        def non_empty(cls, df: pd.DataFrame) -> bool:
+            """Checks that dataframe is not empty."""
+            return not df.empty
+
+    strat = SchemaWithDFCheck.strategy(size=10)
     sample_data = data.draw(strat)
-    schema_model.validate(sample_data)
+    Schema.validate(sample_data)
 
 
-def test_schema_model_example(schema_model):
+def test_schema_model_example():
     """Test that examples can be drawn from a SchemaModel."""
-    sample_data = schema_model.example(size=10)
-    schema_model.validate(sample_data)
+    sample_data = Schema.example(size=10)
+    Schema.validate(sample_data)
 
 
 def test_schema_component_with_no_pdtype():