def _build_scan_yml(self):
if not self.scan_yml_file and not self.scan_yml_dict and not self.scan_yml:
logger.error(f'No scan file specified')
return
elif self.scan_yml_file and not self.scan_yml_dict and not self.scan_yml:
scan_yml_file_str: Optional[str] = None
if isinstance(self.scan_yml_file, pathlib.PurePath):
scan_yml_file_str = str(self.scan_yml_file)
elif isinstance(self.scan_yml_file, str):
scan_yml_file_str = self.scan_yml_file
if not isinstance(scan_yml_file_str, str):
logger.error(
f'scan_builder.scan_yml_file must be str, but was {type(scan_yml_file_str)}: {scan_yml_file_str}'
)
elif self.file_system.is_readable_file(scan_yml_file_str):
scan_yml_str = self.file_system.file_read_as_str(
scan_yml_file_str)
if scan_yml_str:
self.scan_yml_dict = YamlHelper.parse_yaml(
scan_yml_str, scan_yml_file_str)
else:
logger.error(
f'Failed to parse scan yaml file: {scan_yml_file_str}')
if self.scan_yml_dict and not self.scan_yml:
from sodasql.scan.scan_yml_parser import ScanYmlParser
scan_yml_parser = ScanYmlParser(self.scan_yml_dict,
self.scan_yml_file)
scan_yml_parser.log()
self.parsers.append(scan_yml_parser)
self.scan_yml = scan_yml_parser.scan_yml
def _build_scan_yml(self):
file_system = FileSystemSingleton.INSTANCE
if not self.scan_yml_file and not self.scan_yml_dict and not self.scan_yml:
logging.error(f'No scan specified')
return
elif self.scan_yml_file and not self.scan_yml_dict and not self.scan_yml:
if not isinstance(self.scan_yml_file, str):
logging.error(
f'scan_builder.scan_yml_file must be str, but was {type(self.scan_yml_file)}: {self.scan_yml_file}'
)
elif file_system.is_readable_file(self.scan_yml_file):
scan_yml_str = self.file_system.file_read_as_str(
self.scan_yml_file)
if scan_yml_str:
self.scan_yml_dict = YamlHelper.parse_yaml(
scan_yml_str, self.scan_yml_file)
else:
logging.error(
f'Failed to file scan yaml file: {self.scan_yml_file}')
if self.scan_yml_dict and not self.scan_yml:
from sodasql.scan.scan_yml_parser import ScanYmlParser
scan_yml_parser = ScanYmlParser(self.scan_yml_dict,
self.scan_yml_file)
scan_yml_parser.log()
self.parsers.append(scan_yml_parser)
self.scan_yml = scan_yml_parser.scan_yml
def execute_metric(self,
warehouse: Warehouse,
metric: dict,
scan_dict: dict = None):
dialect = warehouse.dialect
if not scan_dict:
scan_dict = {}
if KEY_TABLE_NAME not in scan_dict:
scan_dict[KEY_TABLE_NAME] = self.default_test_table_name
scan_configuration_parser = ScanYmlParser(scan_dict, 'test-scan')
scan_configuration_parser.assert_no_warnings_or_errors()
scan = warehouse.create_scan(
scan_yml=scan_configuration_parser.scan_yml)
scan.close_warehouse = False
scan.execute()
fields: List[str] = []
group_by_column_names: List[str] = metric.get('groupBy')
if group_by_column_names:
for group_by_column in group_by_column_names:
fields.append(dialect.qualify_column_name(group_by_column))
column_name: str = metric.get('columnName')
qualified_column_name = dialect.qualify_column_name(column_name)
metric_type = metric['type']
if metric_type == Metric.ROW_COUNT:
fields.append('COUNT(*)')
if metric_type == Metric.MIN:
fields.append(f'MIN({qualified_column_name})')
elif metric_type == Metric.MAX:
fields.append(f'MAX({qualified_column_name})')
elif metric_type == Metric.SUM:
fields.append(f'SUM({qualified_column_name})')
sql = 'SELECT \n ' + ',\n '.join(fields) + ' \n' \
'FROM ' + scan.qualified_table_name
where_clauses = []
metric_filter = metric.get('filter')
if metric_filter:
where_clauses.append(dialect.sql_expression(metric_filter))
scan_column: ScanColumn = scan.scan_columns.get(column_name)
if scan_column and scan_column.non_missing_and_valid_condition:
where_clauses.append(scan_column.non_missing_and_valid_condition)
if where_clauses:
sql += '\nWHERE ' + '\n AND '.join(where_clauses)
if group_by_column_names:
sql += '\nGROUP BY ' + ', '.join(group_by_column_names)
return warehouse.sql_fetchall(sql)
def scan(self,
scan_yml_dict: Optional[dict] = None,
variables: Optional[dict] = None) -> ScanResult:
if not scan_yml_dict:
scan_yml_dict = {}
if KEY_TABLE_NAME not in scan_yml_dict:
scan_yml_dict[KEY_TABLE_NAME] = self.default_test_table_name
logging.debug('Scan configuration \n' + json.dumps(scan_yml_dict, indent=2))
scan_configuration_parser = ScanYmlParser(scan_yml_dict, 'test-scan')
scan_configuration_parser.assert_no_warnings_or_errors()
scan = self.warehouse.create_scan(scan_yml=scan_configuration_parser.scan_yml,
variables=variables,
soda_server_client=self.mock_soda_server_client,
time=datetime.now().isoformat(timespec='seconds'))
scan.close_warehouse = False
return scan.execute()
def test_soda_server_client(self):
self.sql_recreate_table([f"name {self.dialect.data_type_integer}"],
["(1)", "(2)", "(3)", "(null)"])
scan_yml_dict = {
KEY_TABLE_NAME:
self.default_test_table_name,
KEY_METRIC_GROUPS: [
Metric.METRIC_GROUP_MISSING, Metric.METRIC_GROUP_VALIDITY,
Metric.METRIC_GROUP_DUPLICATES, Metric.METRIC_GROUP_STATISTICS,
Metric.METRIC_GROUP_LENGTH, Metric.METRIC_GROUP_PROFILING
],
'tests': [f'{Metric.ROW_COUNT} > 0'],
KEY_SQL_METRICS: [{
SQL_METRIC_KEY_SQL:
f'SELECT 0 AS zero FROM {self.default_test_table_name}',
SQL_METRIC_KEY_TESTS: ['zero == 0']
}],
KEY_COLUMNS: {
'name': {
COLUMN_KEY_TESTS: [
f'{Metric.MISSING_COUNT} < 1',
]
}
}
}
scan_configuration_parser = ScanYmlParser(scan_yml_dict, 'test-scan')
scan_configuration_parser.assert_no_warnings_or_errors()
soda_server_client = SodaServerClient(host='localhost',
port='5000',
protocol='http',
token='testtoken')
scan = self.warehouse.create_scan(
scan_yml=scan_configuration_parser.scan_yml,
soda_server_client=soda_server_client,
time=datetime.now().isoformat())
scan.close_warehouse = False
return scan.execute()
def test_invalid_column_metric(self):
parser = ScanYmlParser({
KEY_TABLE_NAME: 't',
KEY_METRICS: ['revenue']
}, 'Test scan')
log = parser.logs[0]
self.assertIn(WARNING, log.level)
self.assertIn('Invalid key', log.message)
self.assertIn('metrics', log.message)
self.assertIn('revenue', log.message)
def test_metrics_not_a_list(self):
parser = ScanYmlParser({
KEY_TABLE_NAME: 't',
KEY_METRICS: 'txt'
}, 'Test scan')
log = parser.logs[0]
self.assertIn(ERROR, log.level)
self.assertIn('Invalid metrics', log.message)
self.assertIn('list', log.message)
self.assertIn('str', log.message)
def test_invalid_valid_format(self):
parser = ScanYmlParser(
{
KEY_TABLE_NAME: 't',
KEY_COLUMNS: {
'col': {
'valid_format': 'buzz'
}
}
}, 'Test scan')
log = parser.logs[0]
self.assertIn(WARNING, log.level)
self.assertIn('Invalid', log.message)
self.assertIn('valid_format', log.message)
self.assertIn('buzz', log.message)
KEY_METRICS: [
'missing', 'validity', 'min', 'max', 'avg', 'sum', 'min_length',
'max_length', 'avg_length'
],
KEY_COLUMNS: {
'ID': {
KEY_METRICS: ['distinct', 'uniqueness'],
'tests': {
'nomissing': 'missing_percentage < 3.0',
'noinvalid': 'invalid_count == 0'
}
}
}
}
scan_configuration_parser = ScanYmlParser(scan_configuration_dict,
'demodata-scan')
scan_configuration_parser.assert_no_warnings_or_errors()
dialect = SqlTestCase.create_dialect('postgres')
warehouse_yml = WarehouseYml(dialect=dialect)
warehouse = Warehouse(warehouse_yml)
row = warehouse.sql_fetchone('SELECT MIN(date), MAX(date) FROM demodata')
min_date = row[0]
max_date = row[1]
scan_results = []
date = min_date
while date != max_date:
timeslice = datetime(year=date.year, month=date.month,
def test_table_name_required(self):
parser = ScanYmlParser({}, 'Test scan')
log = parser.logs[0]
self.assertIn(ERROR, log.level)
self.assertIn('table_name', log.message)
self.assertIn('does not exist', log.message)
