def create(cls, schema_definition: SchemaDefinition) -> AbstractSchemaParser:
if schema_definition.is_avro():
return AvroSchemaParser()
if schema_definition.is_json():
return JsonSchemaParser()
assert False, "Parser not supported"
def test_equality_true(self):
# arrange
schema = SchemaDefinition.empty()
another_schema = SchemaDefinition.empty()
# act / assert
self.assertEqual(schema, another_schema)
def test_equality_inferred(self):
# arrange
schema = SchemaDefinition.empty()
another_schema = SchemaDefinition.empty()
another_schema.is_schema_inferred = True
# act / assert
self.assertNotEqual(schema, another_schema)
def test_equality_content(self):
# arrange
schema = SchemaDefinition.empty()
another_schema = SchemaDefinition.empty()
another_schema.schema_content = "another schema content"
# act / assert
self.assertNotEqual(schema, another_schema)
def convert_expectations(self, schema_obj: SchemaDefinition) -> SchemaDefinition:
"""
Method for converting the supported specifications to great expectation format
"""
if schema_obj.is_empty():
return schema_obj
content = schema_obj.get_content()
if self.get_property_keyword() not in content:
return schema_obj
self.convert_expectations_recursive(content)
return SchemaDefinition.create(json.dumps(content), schema_obj.is_schema_inferred)
def test_inspect_with_both_schema_formats(self):
# arrange
schema_json = DataLoader.load_schema_with_name("schema_diff_json.json")
schema_avro = DataLoader.load_schema_with_name("schema_diff_avro.json")
samples = DataLoader.load_samples()
# act
result_json = self.inspector.inspect(
samples, SchemaDefinition.create(schema_json, False))
result_avro = self.inspector.inspect(
samples, SchemaDefinition.create(schema_avro, False))
# assert
self.assertEqual(result_json, result_avro)
def test_inspect_with_multiple_expectations_asyncapi_style_json(self):
# arrange
schema = DataLoader.load_schema_with_name(
"schema_expectation_asyncapi_style_json.json")
samples = [
{
'random_integer': 1,
'random_string': 'id_1'
},
{
'random_integer': 2,
'random_string': 'foo'
}, # no match (string)
{
'random_integer': 3,
'random_string': 'id_3'
},
{
'random_integer': 4,
'random_string': 'id_4'
}, # no match (integer)
{
'random_integer': 5,
'random_string': 'foo'
}, # no match (integer, string)
]
# act
metrics = self.inspector.inspect(
samples, SchemaDefinition.create(schema, False))
# assert
self.assertAlmostEqual(6 / 10, metrics.attribute_integrity, 3)
def expand_schema(cls,
types: [],
required_types: [] = None,
expectations: dict = None) -> SchemaDefinition:
if required_types is None:
required_types = []
if expectations is None:
expectations = {}
properties = ""
for name, t in types:
type_description = f'"{t}"' if name in required_types else f'["{t}", "null"]'
properties += '{"name": "' + name + '","type": ' + type_description
if name in expectations.keys():
for key, value in expectations[name].items():
inner_value = f'"{value}"' if type(value) is str else value
properties += f', "{key}": {inner_value}'
properties += '},'
schema = r'''
{
"type": "record",
"name": "RandomMessage",
"namespace": "data.producer.random",
"fields": [
#PROPERTIES#
]
}
'''
schema_content = schema \
.replace("#PROPERTIES#", properties.rstrip(','))
return SchemaDefinition.create(schema_content, False)
def test_convert_with_unsupported_property(self):
# arrange
schema = '''
{
"fields": [
{
"name": "random_string",
"type": "string",
"foobar": "id_"
}
]
}
'''
expected_schema = '''
{
"fields": [
{
"name": "random_string",
"type": "string",
"foobar": "id_"
}
]
}
'''
# act
result = self.parser.convert_expectations(
SchemaDefinition.create(schema, False))
# assert
self.assertStingEqualAsDict(result.schema_content, expected_schema)
def test_convert_with_multiple_expectations(self):
# arrange
schema = '''
{
"fields": [
{
"name": "random_integer",
"type": "integer",
"minimum": 0,
"maximum": 3
},
{
"name": "random_string",
"type": "string",
"pattern": "id_"
}
]
}
'''
expected_schema = '''
{
"fields": [
{
"name": "random_integer",
"type": "integer",
"minimum": 0,
"maximum": 3,
"expectations": [
{
"kwargs": {
"min_value": 0,
"max_value": 3
},
"expectation_type": "expect_column_values_to_be_between"
}
]
},
{
"name": "random_string",
"type": "string",
"pattern": "id_",
"expectations": [
{
"kwargs": {
"regex": "id_"
},
"expectation_type": "expect_column_values_to_match_regex"
}
]
}
]
}
'''
# act
result = self.parser.convert_expectations(
SchemaDefinition.create(schema, False))
# assert
self.assertStingEqualAsDict(result.schema_content, expected_schema)
def test_create_avro_parser(self):
# arrange
definition = SchemaDefinition.create(DataLoader.load_schema(), False)
# act
parser = SchemaParserFactory.create(definition)
# assert
self.assertIsInstance(parser, AvroSchemaParser)
def test_integrity_without_provided_schema(self) -> None:
# arrange
samples, _ = DataLoader.create_dummy_samples()
# act
empty_schema = SchemaDefinition.empty()
result = self.inspector.inspect(samples, empty_schema)
# assert
self.assertEqual(1.0, result.attribute_integrity)
self.assertEqual(.0, result.attribute_specification)
self.assertEqual(.5, result.attribute_quality_index)
def test_inspect_inferred(self) -> None:
# arrange
samples = DataLoader.load_samples()
# act
schema_definition = SchemaDefinition.create(DataLoader.load_schema())
result = self.inspector.inspect(samples, schema_definition)
# assert
self.assertEqual(1.0, result.attribute_integrity)
self.assertEqual(.0, result.attribute_specification)
self.assertEqual(.5, result.attribute_quality_index)
def test_inspect_with_inferred_schemas(self):
# arrange
schema = DataLoader.load_schema_with_name("schema_registry_json.json")
schema_definition = SchemaDefinition.create(schema, True)
samples = DataLoader.load_samples()
# act
result = self.inspector.inspect(samples, schema_definition)
# assert
self.assertEqual(1.0, result.attribute_integrity)
self.assertEqual(.0, result.attribute_specification)
self.assertEqual(.5, result.attribute_quality_index)
def test_convert_with_nested_expectations(self):
# arrange
schema = DataLoader.load_schema_with_name(
"schema_nested_expectation_json.json")
expected_schema = DataLoader.load_schema_with_name(
"schema_nested_expectation_result_json.json")
# act
result = self.parser.convert_expectations(
SchemaDefinition.create(schema, False))
# assert
self.assertStingEqualAsDict(result.schema_content, expected_schema)
def parse_schema(self, schema_definition: SchemaDefinition) -> SchemaParserResult:
"""
Template method
"""
# execute preprocessing step
schema_definition = self.convert_expectations(schema_definition)
schema_content = schema_definition.get_content()
# load types from schema
type_definitions, expectation_definitions = self.load_types_from_schema(schema_content)
# load required types from schema
required_types, type_expectations = self.load_required_types_from_schema(schema_content)
return SchemaParserResult(type_definitions, expectation_definitions, required_types, type_expectations)
def test_integrity_for_complex_type(self):
# arrange
schema = DataLoader.load_schema_with_name("schema_registry_avro.json")
samples = [
{
"timestamp": 1595601702,
"iss_position": {
"longitude": "-42.2948",
"latitude": "-40.3670"
},
"message": "success"
},
{
"timestamp": 1595601702,
"iss_position": {
"latitude": "-40.3670"
},
"message": "success"
},
{
"timestamp": "wrong",
"iss_position": {
"longitude": 666,
"latitude": "-40.0283"
},
"message": "success"
},
]
# act
result = self.inspector.inspect_attributes(
samples, SchemaDefinition.create(schema, False))
# assert - only message is not mandatory so 3 out of 12 (3*4) are missing or wrong
invalid_elements = 3
all_elements = 12
expected_integrity = (all_elements - invalid_elements) / all_elements
self.assertAlmostEqual(
expected_integrity, result.attribute_integrity, 3,
f"Integrity must be {expected_integrity * 100}%")
def test_various_types_do_not_throw_exceptions(self):
# arrange
schema = '''
{
"type": "record",
"name": "RandomData",
"namespace": "data.producer.random",
"fields": [
{
"name": "random_string",
"type": "string"
},
{
"name": "random_integer",
"type": "int"
},
{
"name": "random_float",
"type": "float"
},
{
"name": "random_boolean",
"type": "boolean"
}
]
}
'''
samples = [{
'random_string': 'wheyuugkwi',
'random_integer': 876,
'random_float': 0.2295482,
'random_boolean': False
}]
# act
metrics = self.inspector.inspect(
samples, SchemaDefinition.create(schema, False))
# assert
self.assertIsNotNone(metrics)
def test_specification_from_toeggelomat(self):
# arrange
samples = DataLoader.load_samples_from_file("samples_toeggelomat.json")
# act
schema = DataLoader.load_schema_with_name("schema_toeggelomat.json")
result = self.inspector.inspect(samples,
SchemaDefinition.create(schema, False))
# assert
self.assertEqual(53, len(result.attribute_details.keys()),
"There should be 53 keys in the dictionary")
for attribute_metric in result.attribute_details.keys():
self.assertEqual(
1.0, result.attribute_details[attribute_metric].
attribute_specification,
f"Attribute specification must be 100% ({attribute_metric})")
self.assertEqual(
1.0,
result.attribute_details[attribute_metric].attribute_integrity,
f"Attribute integrity must be 100% ({attribute_metric})")
def test_convert(self):
# arrange
schema = '''
{
"properties": {
"random_string": {
"type": "string",
"minimum": 0,
"maximum": 10
}
}
}
'''
expected_schema = '''
{
"properties": {
"random_string": {
"type": "string",
"minimum": 0,
"maximum": 10,
"expectations": [
{
"kwargs": {
"min_value": 0,
"max_value": 10
},
"expectation_type": "expect_column_values_to_be_between"
}
]
}
}
}
'''
# act
result = self.parser.convert_expectations(
SchemaDefinition.create(schema, False))
# assert
self.assertStingEqualAsDict(result.schema_content, expected_schema)
def InspectQuality(self, request_iterator,
context: typing.Any) -> QualityMetrics:
all_samples = list()
schema = None
first_samples = None
is_schema_inferred = True
for batch in request_iterator:
samples = json.loads(batch.samples_json)
all_samples.append(samples)
schema = batch.schema_json
if not first_samples:
is_schema_inferred = batch.is_schema_inferred
first_samples = samples
message = profiler_pb2.InspectionDataStreamResponse()
inspector = QualityInspector()
schema_definition = SchemaDefinition.create(schema, is_schema_inferred)
try:
result = inspector.inspect(all_samples, schema_definition)
message.metric.attribute_quality_index = result.attribute_quality_index
message.metric.attribute_integrity = result.attribute_integrity
message.metric.attribute_specification = result.attribute_specification
for k, v in result.attribute_details.items():
message.metric.attribute_details.append(
domain_pb2.AttributeDetail(
name=k,
integrity=v.attribute_integrity,
specification=v.attribute_specification,
quality_index=v.attribute_quality_index))
except Exception as e:
logging.error(
f'Exception in inspection of quality (samples: {first_samples}, schema: {schema}, exception: {e})'
)
message.error.message = repr(e)
message.error.type = domain_pb2.InspectionError.Type.Value(
'UNKNOWN')
return message
def test_convert_with_min_expectation(self):
# arrange
schema = '''
{
"fields": [
{
"name": "random_integer",
"type": "integer",
"minimum": 0
}
]
}
'''
expected_schema = '''
{
"fields": [
{
"name": "random_integer",
"type": "integer",
"minimum": 0,
"expectations": [
{
"kwargs": {
"min_value": 0
},
"expectation_type": "expect_column_values_to_be_between"
}
]
}
]
}
'''
# act
result = self.parser.convert_expectations(
SchemaDefinition.create(schema, False))
# assert
self.assertStingEqualAsDict(result.schema_content, expected_schema)
def test_specification_with_partial_schema_and_inferred(self) -> None:
# arrange
samples = [{
"random_int": 1,
"random_string": "foo"
}, {
"random_int": 2,
"random_string": "bar"
}]
schema = '''
{
"$schema": "http://json-schema.org/schema#",
"type": "object",
"properties": {
"random_string": {
"type": "string"
},
"random_integer": {
"type": "integer"
}
},
"required": [
"random_integer",
"random_string"
]
}
'''
schema_definition = SchemaDefinition.create(schema, True)
# act
result = self.inspector.inspect(samples, schema_definition)
# assert
self.assertEqual(
.0, result.attribute_specification,
"Attribute specification is considered 0% when the schema is inferred"
)
def test_inspect_with_min_max_range_expectation(self):
# arrange
schema = DataLoader.load_schema_with_name("schema_with_min_max.json")
samples = [
{
'random_integer': 3
},
{
'random_integer': 11
},
{
'random_integer': 3
},
{
'random_integer': 8
},
{
'random_integer': 3
},
{
'random_integer': -5
},
{
'random_integer': 3
},
{
'random_integer': 10
},
]
# act
metrics = self.inspector.inspect(
samples, SchemaDefinition.create(schema, False))
# assert
self.assertEqual((6 / 8), metrics.attribute_integrity,
f"Attribute integrity must be {(6 / 8) * 100}%")
def test_integrity_on_attribute_level_with_expectations(self):
# arrange
schema = '''
{
"type": "record",
"name": "RandomData",
"namespace": "data.producer.random",
"fields": [
{
"name": "random_integer",
"type": "int",
"expectations": [
{
"kwargs": {
"min_value": 0,
"max_value": 10
},
"expectation_type": "expect_column_values_to_be_between"
}
]
},
{
"name": "random_string",
"type": "string",
"expectations": [
{
"kwargs": {
"regex": "id_"
},
"meta": {},
"expectation_type": "expect_column_values_to_match_regex"
}
]
}
]
}
'''
samples = [
{
'random_integer': 1,
'random_string': 'missing_id'
},
{
'random_integer': 11,
'random_string': 'id_1'
},
{
'random_integer': 3,
'random_string': 'missing_id'
},
]
# act
result = self.inspector.inspect(samples,
SchemaDefinition.create(schema, False))
# assert
attribute_details = result.attribute_details
self.assertAlmostEqual((3 / 6), result.attribute_integrity, 3,
"Attribute integrity is not correct")
self.assertTrue('random_integer' in attribute_details.keys(),
"Missing integrity for attribute random_integer")
self.assertTrue('random_string' in attribute_details.keys(),
"Missing integrity for attribute random_string")
self.assertAlmostEqual(
(2 / 3), attribute_details['random_integer'].attribute_integrity,
3, "Integrity of random_int is not correct")
self.assertAlmostEqual(
(1 / 3), attribute_details['random_string'].attribute_integrity, 3,
"Integrity of random_string is not correct")
