syun64 commented on code in PR #219:
URL: https://github.com/apache/iceberg-python/pull/219#discussion_r1453579122
##########
pyiceberg/io/pyarrow.py:
##########
@@ -733,42 +854,178 @@ def _get_field_id(field: pa.Field) -> Optional[int]:
)
-class _ConvertToIceberg(PyArrowSchemaVisitor[Union[IcebergType, Schema]]):
- def _convert_fields(self, arrow_fields: Iterable[pa.Field], field_results:
List[Optional[IcebergType]]) -> List[NestedField]:
- fields = []
- for i, field in enumerate(arrow_fields):
- field_id = _get_field_id(field)
- field_doc = doc_str.decode() if (field.metadata and (doc_str :=
field.metadata.get(PYARROW_FIELD_DOC_KEY))) else None
- field_type = field_results[i]
- if field_type is not None and field_id is not None:
- fields.append(NestedField(field_id, field.name, field_type,
required=not field.nullable, doc=field_doc))
- return fields
-
- def schema(self, schema: pa.Schema, field_results:
List[Optional[IcebergType]]) -> Schema:
- return Schema(*self._convert_fields(schema, field_results))
-
- def struct(self, struct: pa.StructType, field_results:
List[Optional[IcebergType]]) -> IcebergType:
- return StructType(*self._convert_fields(struct, field_results))
-
- def list(self, list_type: pa.ListType, element_result:
Optional[IcebergType]) -> Optional[IcebergType]:
+class _HasIds(PyArrowSchemaVisitor[bool]):
+ def schema(self, schema: pa.Schema, struct_result: bool) -> bool:
+ return struct_result
+
+ def struct(self, struct: pa.StructType, field_results: List[bool]) -> bool:
+ return all(field_results)
+
+ def field(self, field: pa.Field, field_result: bool) -> bool:
+ return all([_get_field_id(field) is not None, field_result])
+
+ def list(self, list_type: pa.ListType, element_result: bool) -> bool:
element_field = list_type.value_field
element_id = _get_field_id(element_field)
- if element_result is not None and element_id is not None:
- return ListType(element_id, element_result, element_required=not
element_field.nullable)
- return None
+ return element_result and element_id is not None
- def map(
- self, map_type: pa.MapType, key_result: Optional[IcebergType],
value_result: Optional[IcebergType]
- ) -> Optional[IcebergType]:
+ def map(self, map_type: pa.MapType, key_result: bool, value_result: bool)
-> bool:
key_field = map_type.key_field
key_id = _get_field_id(key_field)
value_field = map_type.item_field
value_id = _get_field_id(value_field)
- if key_result is not None and value_result is not None and key_id is
not None and value_id is not None:
- return MapType(key_id, key_result, value_id, value_result,
value_required=not value_field.nullable)
- return None
+ return all([key_id is not None, value_id is not None, key_result,
value_result])
+
+ def primitive(self, primitive: pa.DataType) -> bool:
+ return True
+
+
+class _ConvertToIceberg(PyArrowSchemaVisitor[Union[IcebergType, Schema]]):
+ """Converts PyArrowSchema to Iceberg Schema. Applies the IDs from
name_mapping if provided."""
+
+ _field_names: List[str]
+ _name_mapping: Optional[NameMapping]
+
+ def __init__(self, name_mapping: Optional[NameMapping] = None) -> None:
+ self._field_names = []
+ self._name_mapping = name_mapping
+
+ def _current_path(self) -> str:
+ return ".".join(self._field_names)
+
+ def _field_id(self, field: pa.Field) -> int:
+ if self._name_mapping:
+ return self._name_mapping.find(self._current_path()).field_id
+ elif (field_id := _get_field_id(field)) is not None:
+ return field_id
+ else:
+ raise ValueError(f"Cannot convert {field} to Iceberg Field as
field_id is empty.")
+
+ def schema(self, schema: pa.Schema, struct_result: StructType) -> Schema:
+ return Schema(*struct_result.fields)
+
+ def struct(self, struct: pa.StructType, field_results: List[NestedField])
-> StructType:
+ return StructType(*field_results)
+
+ def field(self, field: pa.Field, field_result: IcebergType) -> NestedField:
+ field_id = self._field_id(field)
+ field_doc = doc_str.decode() if (field.metadata and (doc_str :=
field.metadata.get(PYARROW_FIELD_DOC_KEY))) else None
+ field_type = field_result
+ return NestedField(field_id, field.name, field_type, required=not
field.nullable, doc=field_doc)
+
+ def list(self, list_type: pa.ListType, element_result: IcebergType) ->
ListType:
+ element_field = list_type.value_field
+ self._field_names.append(LIST_ELEMENT_NAME)
+ element_id = self._field_id(element_field)
+ self._field_names.pop()
+ return ListType(element_id, element_result, element_required=not
element_field.nullable)
- def primitive(self, primitive: pa.DataType) -> IcebergType:
+ def map(self, map_type: pa.MapType, key_result: IcebergType, value_result:
IcebergType) -> MapType:
+ key_field = map_type.key_field
+ self._field_names.append(MAP_KEY_NAME)
+ key_id = self._field_id(key_field)
+ self._field_names.pop()
+ value_field = map_type.item_field
+ self._field_names.append(MAP_VALUE_NAME)
+ value_id = self._field_id(value_field)
+ self._field_names.pop()
+ return MapType(key_id, key_result, value_id, value_result,
value_required=not value_field.nullable)
+
+ def primitive(self, primitive: pa.DataType) -> PrimitiveType:
+ if pa.types.is_boolean(primitive):
+ return BooleanType()
+ elif pa.types.is_int32(primitive):
+ return IntegerType()
+ elif pa.types.is_int64(primitive):
+ return LongType()
+ elif pa.types.is_float32(primitive):
+ return FloatType()
+ elif pa.types.is_float64(primitive):
+ return DoubleType()
+ elif isinstance(primitive, pa.Decimal128Type):
+ primitive = cast(pa.Decimal128Type, primitive)
+ return DecimalType(primitive.precision, primitive.scale)
+ elif pa.types.is_string(primitive):
+ return StringType()
+ elif pa.types.is_date32(primitive):
+ return DateType()
+ elif isinstance(primitive, pa.Time64Type) and primitive.unit == "us":
+ return TimeType()
+ elif pa.types.is_timestamp(primitive):
+ primitive = cast(pa.TimestampType, primitive)
+ if primitive.unit == "us":
+ if primitive.tz == "UTC" or primitive.tz == "+00:00":
+ return TimestamptzType()
+ elif primitive.tz is None:
+ return TimestampType()
+ elif pa.types.is_binary(primitive):
+ return BinaryType()
+ elif pa.types.is_fixed_size_binary(primitive):
+ primitive = cast(pa.FixedSizeBinaryType, primitive)
+ return FixedType(primitive.byte_width)
+
+ raise TypeError(f"Unsupported type: {primitive}")
+
+ def before_field(self, field: pa.Field) -> None:
+ self._field_names.append(field.name)
+
+ def after_field(self, field: pa.Field) -> None:
+ self._field_names.pop()
+
+ def before_list_element(self, element: pa.Field) -> None:
+ self._field_names.append(LIST_ELEMENT_NAME)
+
+ def after_list_element(self, element: pa.Field) -> None:
+ self._field_names.pop()
+
+ def before_map_key(self, key: pa.Field) -> None:
+ self._field_names.append(MAP_KEY_NAME)
+
+ def after_map_key(self, element: pa.Field) -> None:
+ self._field_names.pop()
+
+ def before_map_value(self, value: pa.Field) -> None:
+ self._field_names.append(MAP_VALUE_NAME)
+
+ def after_map_value(self, element: pa.Field) -> None:
+ self._field_names.pop()
+
+
+class
_ConvertToIcebergWithFreshIds(PreOrderPyArrowSchemaVisitor[Union[IcebergType,
Schema]]):
Review Comment:
Makes sense @Fokko . Should we have a boolean attribute to drive this logic
in _ConvertToIceberg? I think as @rdblue pointed out in [this
comment](https://github.com/apache/iceberg-python/pull/219#issuecomment-1859273052),
falling back to assigning fresh IDs as a general fallback sounds like a
dangerous behavior that we would want to avoid. Having a boolean attribute will
make sure that we are only using this feature when creating a new table,
instead of avoid falling back when there are no IDs in existing Iceberg tables.
Should we have a boolean flag **_ignore_ids** which if True makes
_ConvertToIceberg ignore the existing ID assigning process, and instead assigns
all IDs as -1?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
