Add validation for compatible schema evolution when partition fields or sort order fields reference schema fields

Author: gabeiglio

pyiceberg/partitioning.py

@@ -32,7 +32,8 @@
     model_validator,
 )
 
-from pyiceberg.schema import Schema
+from pyiceberg.exceptions import ValidationError
+from pyiceberg.schema import Schema, _index_parents
 from pyiceberg.transforms import (
     BucketTransform,
     DayTransform,
@@ -249,6 +250,36 @@ def partition_to_path(self, data: Record, schema: Schema) -> str:
         path = "/".join([field_str + "=" + value_str for field_str, value_str in zip(field_strs, value_strs, strict=True)])
         return path
 
+    def check_compatible(self, schema: Schema, allow_missing_fields: bool = False) -> None:
+        # if the underlying field is dropped, we cannot check they are compatible -- continue
+        schema_fields = schema._lazy_id_to_field
+        parents = _index_parents(schema)
+
+        def validate_parents_are_structs(field_id: int) -> None:
+            parent_id = parents.get(field_id)
+            while parent_id:
+                parent_type = schema.find_type(parent_id)
+                if not parent_type.is_struct:
+                    raise ValidationError("Invalid partition field parent: %s", parent_type)
+                parent_id = parents.get(parent_id)
+
+        for field in self.fields:
+            source_field = schema_fields.get(field.source_id)
+            if allow_missing_fields and source_field:
+                continue
+
+            if not isinstance(field.transform, VoidTransform):
+                if source_field:
+                    source_type = source_field.field_type
+                    if not source_type.is_primitive:
+                        raise ValidationError(f"Cannot partition by non-primitive source field: {source_type}")
+                    if not field.transform.can_transform(source_type):
+                        raise ValidationError(f"Invalid source type {source_type} for transform: {field.transform}")
+                    # The only valid parent types for a PartitionField are StructTypes. This must be checked recursively
+                    validate_parents_are_structs(field.source_id)
+                else:
+                    raise ValidationError(f"Cannot find source column for partition field: {field}")
+
 
 UNPARTITIONED_PARTITION_SPEC = PartitionSpec(spec_id=0)
 

pyiceberg/table/sorting.py

@@ -27,6 +27,7 @@
     model_validator,
 )
 
+from pyiceberg.exceptions import ValidationError
 from pyiceberg.schema import Schema
 from pyiceberg.transforms import IdentityTransform, Transform, parse_transform
 from pyiceberg.typedef import IcebergBaseModel
@@ -169,6 +170,17 @@ def __repr__(self) -> str:
         fields = f"{', '.join(repr(column) for column in self.fields)}, " if self.fields else ""
         return f"SortOrder({fields}order_id={self.order_id})"
 
+    def check_compatible(self, schema: Schema) -> None:
+        schema_ids = schema._lazy_id_to_field
+        for field in self.fields:
+            if source_field := schema_ids.get(field.source_id):
+                if not source_field.field_type.is_primitive:
+                    raise ValidationError(f"Cannot sort by non-primitive source field: {source_field}")
+                if not field.transform.can_transform(source_field.field_type):
+                    raise ValidationError(f"Invalid source type {source_field.field_type} for transform: {field.transform}")
+            else:
+                raise ValidationError(f"Cannot find source column for sort field: {field}")
+
 
 UNSORTED_SORT_ORDER_ID = 0
 UNSORTED_SORT_ORDER = SortOrder(order_id=UNSORTED_SORT_ORDER_ID)

pyiceberg/table/update/__init__.py

@@ -700,6 +700,12 @@ def update_table_metadata(
         if base_metadata.last_updated_ms == new_metadata.last_updated_ms:
             new_metadata = new_metadata.model_copy(update={"last_updated_ms": datetime_to_millis(datetime.now().astimezone())})
 
+    # Check correctness of partition spec, and sort order
+    new_metadata.spec().check_compatible(new_metadata.schema())
+
+    if sort_order := new_metadata.sort_order_by_id(new_metadata.default_sort_order_id):
+        sort_order.check_compatible(new_metadata.schema())
+
     if enforce_validation:
         return TableMetadataUtil.parse_obj(new_metadata.model_dump())
     else:

tests/conftest.py

@@ -70,6 +70,7 @@
 from pyiceberg.serializers import ToOutputFile
 from pyiceberg.table import FileScanTask, Table
 from pyiceberg.table.metadata import TableMetadataV1, TableMetadataV2, TableMetadataV3
+from pyiceberg.table.sorting import NullOrder, SortField, SortOrder
 from pyiceberg.transforms import DayTransform, IdentityTransform
 from pyiceberg.types import (
     BinaryType,
@@ -1894,6 +1895,11 @@ def test_partition_spec() -> Schema:
     )
 
 
+@pytest.fixture(scope="session")
+def test_sort_order() -> SortOrder:
+    return SortOrder(SortField(source_id=1, transform=IdentityTransform(), null_order=NullOrder.NULLS_FIRST))
+
+
 @pytest.fixture(scope="session")
 def generated_manifest_entry_file(
     avro_schema_manifest_entry: dict[str, Any], test_schema: Schema, test_partition_spec: PartitionSpec

tests/integration/test_catalog.py

@@ -33,6 +33,7 @@
     NoSuchNamespaceError,
     NoSuchTableError,
     TableAlreadyExistsError,
+    ValidationError,
 )
 from pyiceberg.io import WAREHOUSE
 from pyiceberg.partitioning import PartitionField, PartitionSpec
@@ -601,3 +602,56 @@ def test_register_table_existing(test_catalog: Catalog, table_schema_nested: Sch
     # Assert that registering the table again raises TableAlreadyExistsError
     with pytest.raises(TableAlreadyExistsError):
         test_catalog.register_table(identifier, metadata_location=table.metadata_location)
+
+
+@pytest.mark.integration
+@pytest.mark.parametrize("test_catalog", CATALOGS)
+def test_incompatible_partitioned_schema_evolution(
+    test_catalog: Catalog, test_schema: Schema, test_partition_spec: PartitionSpec, database_name: str, table_name: str
+) -> None:
+    if isinstance(test_catalog, HiveCatalog):
+        pytest.skip("HiveCatalog does not support schema evolution")
+
+    identifier = (database_name, table_name)
+    test_catalog.create_namespace(database_name)
+    table = test_catalog.create_table(identifier, test_schema, partition_spec=test_partition_spec)
+    assert test_catalog.table_exists(identifier)
+
+    with pytest.raises(ValidationError):
+        with table.update_schema() as update:
+            update.delete_column("VendorID")
+
+    # Assert column was not dropped
+    assert "VendorID" in table.schema().column_names
+
+    with table.transaction() as transaction:
+        with transaction.update_spec() as spec_update:
+            spec_update.remove_field("VendorID")
+
+        with transaction.update_schema() as schema_update:
+            schema_update.delete_column("VendorID")
+
+    assert table.spec() == PartitionSpec(PartitionField(2, 1001, DayTransform(), "tpep_pickup_day"), spec_id=1)
+    assert table.schema() == Schema(NestedField(2, "tpep_pickup_datetime", TimestampType(), False))
+
+
+@pytest.mark.integration
+@pytest.mark.parametrize("test_catalog", CATALOGS)
+def test_incompatible_sorted_schema_evolution(
+    test_catalog: Catalog, test_schema: Schema, test_sort_order: SortOrder, database_name: str, table_name: str
+) -> None:
+    if isinstance(test_catalog, HiveCatalog):
+        pytest.skip("HiveCatalog does not support schema evolution")
+
+    identifier = (database_name, table_name)
+    test_catalog.create_namespace(database_name)
+    table = test_catalog.create_table(identifier, test_schema, sort_order=test_sort_order)
+    assert test_catalog.table_exists(identifier)
+
+    with pytest.raises(ValidationError):
+        with table.update_schema() as update:
+            update.delete_column("VendorID")
+
+    assert table.schema() == Schema(
+        NestedField(1, "VendorID", IntegerType(), False), NestedField(2, "tpep_pickup_datetime", TimestampType(), False)
+    )

tests/table/test_partitioning.py

@@ -21,6 +21,7 @@
 
 import pytest
 
+from pyiceberg.exceptions import ValidationError
 from pyiceberg.partitioning import UNPARTITIONED_PARTITION_SPEC, PartitionField, PartitionSpec
 from pyiceberg.schema import Schema
 from pyiceberg.transforms import (
@@ -259,3 +260,36 @@ def test_deserialize_partition_field_v3() -> None:
 
     field = PartitionField.model_validate_json(json_partition_spec)
     assert field == PartitionField(source_id=1, field_id=1000, transform=TruncateTransform(width=19), name="str_truncate")
+
+
+def test_incompatible_source_column_not_found() -> None:
+    schema = Schema(NestedField(1, "foo", IntegerType()), NestedField(2, "bar", IntegerType()))
+
+    spec = PartitionSpec(PartitionField(3, 1000, IdentityTransform(), "some_partition"))
+
+    with pytest.raises(ValidationError) as exc:
+        spec.check_compatible(schema)
+
+    assert "Cannot find source column for partition field: 1000: some_partition: identity(3)" in str(exc.value)
+
+
+def test_incompatible_non_primitive_type() -> None:
+    schema = Schema(NestedField(1, "foo", StructType()), NestedField(2, "bar", IntegerType()))
+
+    spec = PartitionSpec(PartitionField(1, 1000, IdentityTransform(), "some_partition"))
+
+    with pytest.raises(ValidationError) as exc:
+        spec.check_compatible(schema)
+
+    assert "Cannot partition by non-primitive source field: struct<>" in str(exc.value)
+
+
+def test_incompatible_transform_source_type() -> None:
+    schema = Schema(NestedField(1, "foo", IntegerType()), NestedField(2, "bar", IntegerType()))
+
+    spec = PartitionSpec(PartitionField(1, 1000, YearTransform(), "some_partition"))
+
+    with pytest.raises(ValidationError) as exc:
+        spec.check_compatible(schema)
+
+    assert "Invalid source type int for transform: year" in str(exc.value)

tests/table/test_sorting.py

@@ -20,6 +20,8 @@
 
 import pytest
 
+from pyiceberg.exceptions import ValidationError
+from pyiceberg.schema import Schema
 from pyiceberg.table.metadata import TableMetadataUtil
 from pyiceberg.table.sorting import (
     UNSORTED_SORT_ORDER,
@@ -28,7 +30,8 @@
     SortField,
     SortOrder,
 )
-from pyiceberg.transforms import BucketTransform, IdentityTransform, VoidTransform
+from pyiceberg.transforms import BucketTransform, IdentityTransform, VoidTransform, YearTransform
+from pyiceberg.types import IntegerType, NestedField, StructType
 
 
 @pytest.fixture
@@ -114,3 +117,36 @@ def test_serialize_sort_field_v3() -> None:
     expected = SortField(source_id=19, transform=IdentityTransform(), null_order=NullOrder.NULLS_FIRST)
     payload = '{"source-ids":[19],"transform":"identity","direction":"asc","null-order":"nulls-first"}'
     assert SortField.model_validate_json(payload) == expected
+
+
+def test_incompatible_source_column_not_found(sort_order: SortOrder) -> None:
+    schema = Schema(NestedField(1, "foo", IntegerType()), NestedField(2, "bar", IntegerType()))
+
+    with pytest.raises(ValidationError) as exc:
+        sort_order.check_compatible(schema)
+
+    assert "Cannot find source column for sort field: 19 ASC NULLS FIRST" in str(exc.value)
+
+
+def test_incompatible_non_primitive_type() -> None:
+    schema = Schema(NestedField(1, "foo", StructType()), NestedField(2, "bar", IntegerType()))
+
+    sort_order = SortOrder(SortField(source_id=1, transform=IdentityTransform(), null_order=NullOrder.NULLS_FIRST))
+
+    with pytest.raises(ValidationError) as exc:
+        sort_order.check_compatible(schema)
+
+    assert "Cannot sort by non-primitive source field: 1: foo: optional struct<>" in str(exc.value)
+
+
+def test_incompatible_transform_source_type() -> None:
+    schema = Schema(NestedField(1, "foo", IntegerType()), NestedField(2, "bar", IntegerType()))
+
+    sort_order = SortOrder(
+        SortField(source_id=1, transform=YearTransform(), null_order=NullOrder.NULLS_FIRST),
+    )
+
+    with pytest.raises(ValidationError) as exc:
+        sort_order.check_compatible(schema)
+
+    assert "Invalid source type int for transform: year" in str(exc.value)