class Data(Task):
def __init__(self, service: str, **kwargs):
self.chunk_size = 250000
self.table = None
self.watched_column = None
self.service_account = service_account.Credentials.from_service_account_file(
os.environ['MYSQL_BIG_QUERY_GOOGLE_AUTH'])
self.definitions = service_helpers.get_definitions(service)
self.sd_logger = StackDriverAdapter(self.service_account)
self.sd_logger.get_client()
self.sd_logger.create_logger(f"{self.definitions['service']}-etl")
self.pub_sub_client = PubSubAdapter(self.service_account)
self.pub_sub_client.get_subscriber()
self.pub_sub_client.set_subscription(
f"{self.definitions['service']}-etl-data")
self.big_query_client = BigQueryAdapter(self.service_account)
self.big_query_client.get_client()
self.big_query_client.set_data_set_ref(self.definitions['data_set'])
self.my_sql_client = MySqlAdapter(service)
super().__init__(**kwargs)
def check_message(self, message):
if 'table' in message:
self.table = message['table']
else:
error_message = 'Table was not included in the message'
self.sd_logger.error({'error': error_message}, {
'class': 'Data',
'method': 'check_message',
'table': self.table
})
raise RuntimeError(error_message)
if 'watched' in message:
self.watched_column = message['watched']
else:
error_message = 'Watched was not included in the message'
self.sd_logger.error({'error': error_message}, {
'class': 'Data',
'method': 'check_message',
'table': self.table
})
raise RuntimeError(error_message)
def get_schema_from_big_query(self) -> bool:
self.big_query_client.set_table_ref(self.table)
table_check = self.big_query_client.check_table()
if table_check:
return self.big_query_client.get_schema()
return table_check
def last_updated_data(self) -> Union[str, bool]:
query = f"SELECT MAX({self.watched_column}) as last_updated FROM {self.definitions['data_set']}.{self.table}"
result = self.big_query_client.query(query)
if result:
for value in result:
if value['last_updated']:
return value['last_updated']
else:
# return none to allow all records to be pulled at the start
return None
else:
self.sd_logger.critical(
self.big_query_client.errors, {
'class': 'Data',
'method': 'last_updated_data',
'table': self.table
})
return result
def get_number_of_records_to_import(self,
last_updated) -> Union[int, bool]:
result = self.my_sql_client.count_items_to_sync(
table=self.table,
watched_column=self.watched_column,
last_run=last_updated)
if self.my_sql_client.errors:
self.sd_logger.critical(
self.my_sql_client.errors, {
'class': 'Data',
'method': 'get_number_of_records_to_import',
'table': self.table
})
return result
def query_mysql_for_records(self, last_updated_date: str, limit: int,
offset: int) -> Union[list, bool]:
results = self.my_sql_client.get_records(
table=self.table,
watched_column=self.watched_column,
last_run=last_updated_date,
limit=limit,
offset=offset)
if not results:
self.sd_logger.critical(
self.my_sql_client.errors, {
'class': 'Data',
'method': 'get_number_of_records_to_import',
'table': self.table
})
return results
def load_mysql_data_into_data_frame(
self, data: list, schema: dict) -> Union[DataFrame, bool]:
if len(data[0]) is not len(schema.keys()):
self.sd_logger.critical(
{
'message': "Schema and data length mismatch",
'schema_length': len(schema.keys()),
'data_length': len(data[0])
}, {
'class': 'Data',
'method': 'load_mysql_data_into_data_frame',
'table': self.table
})
return False
df = pd.DataFrame.from_records(data, columns=schema.keys())
del data
return df
def transform_data_frame_to_match_big_query_schema(
self, data_frame: DataFrame,
schema: dict) -> Union[DataFrame, bool]:
try:
df = service_helpers.data_frame_to_schema(data_frame, schema)
except ValueError as e:
self.sd_logger.critical({'message': 'Error: {}'.format(e)}, {
'class': 'Data',
'method': 'transform_data_frame_to_match_big_query_schema',
'table': self.table
})
return False
return df
def append_data_frame_to_big_query(self, data_frame: DataFrame):
result = self.big_query_client.upload_data_frame(data_frame)
if result:
message = f"table:{self.table} | Records written: {data_frame.shape[0]}"
self.sd_logger.info({'message': message}, {
'class': 'Data',
'method': 'append_data_frame_to_big_query',
'table': self.table
})
else:
self.sd_logger.critical(
self.big_query_client.errors, {
'class': 'Data',
'method': 'append_data_frame_to_big_query',
'table': self.table
})
return result
def write_df_to_storage(self, df: DataFrame) -> bool:
storage_client = StorageAdapter(self.service_account)
storage_client.get_client()
date_time_obj = datetime.datetime.utcnow()
location = f'error/csv/{self.table}/{date_time_obj.strftime("%m-%d-%Y_%H:%M:%S")}_UTC'
result = storage_client.write_string(
bucket=self.definitions['service'] + '-etl',
destination=location,
string=df.to_csv(),
encoding='text/csv')
if not result:
self.sd_logger.error(
storage_client.errors, {
'class': 'Data',
'method': 'write_df_to_storage',
'table': self.table
})
return False
self.sd_logger.info({'message': f"Failed CSV added to {location}"}, {
'class': 'Data',
'method': 'write_df_to_storage',
'table': self.table
})
return result
class Audit(Task):
def __init__(self, service: str, **kwargs):
self.table = None
self.watched_column = None
self.primary_id = None
self.sample_size = 1000
self.service_account = service_account.Credentials.from_service_account_file(
os.environ['MYSQL_BIG_QUERY_GOOGLE_AUTH'])
self.definitions = service_helpers.get_definitions(service)
self.sd_logger = StackDriverAdapter(self.service_account)
self.sd_logger.get_client()
self.sd_logger.create_logger(f"{self.definitions['service']}-etl")
self.pub_sub_client = PubSubAdapter(self.service_account)
self.pub_sub_client.get_subscriber()
self.pub_sub_client.set_subscription(
f"{self.definitions['service']}-etl")
self.big_query_client = BigQueryAdapter(self.service_account)
self.big_query_client.get_client()
self.big_query_client.set_data_set_ref(self.definitions['data_set'])
self.my_sql_client = MySqlAdapter()
super().__init__(**kwargs)
def extract_message_values(self, message: dict) -> bool:
if self.table is None:
if 'table' in message:
self.table = message['table']
else:
self.sd_logger.error(
{
'error': 'Key: table was not found in message',
'data': message
}, {
'class': 'Audit',
'method': 'extract_message_values'
})
return False
if 'watched' in message:
self.watched_column = message['watched']
else:
self.sd_logger.error(
{
'error': 'Key: watched was not found in message',
'data': message
}, {
'class': 'Audit',
'method': 'extract_message_values',
"table": self.table
})
return False
if 'primary_id' in message:
self.primary_id = message['primary_id']
else:
self.sd_logger.error(
{
'error': 'Key: primary_id was not found in message',
'data': message
}, {
'class': 'Audit',
'method': 'extract_message_values',
"table": self.table
})
return False
return True
def big_query_last_updated_data(self) -> Union[datetime.datetime, bool]:
query = f"SELECT MAX({self.watched_column}) as last_updated FROM {self.definitions['data_set']}.{self.table}"
result = self.big_query_client.query(query)
if result:
for value in result:
if value.last_updated:
return value.last_updated
else:
# use string of the start of unix time as the default time
return False
else:
self.sd_logger.critical(
self.big_query_client.errors, {
'class': 'Audit',
'method': 'big_query_last_updated_data',
'table': self.table
})
return result
def get_total_distinct_mysql(
self, last_updated: datetime.datetime) -> Union[int, bool]:
if last_updated:
records = self.my_sql_client.count_distinct(
table=self.table,
index=self.primary_id,
watched_column=self.watched_column,
last_updated=last_updated)
else:
records = self.my_sql_client.count_distinct(table=self.table,
index=self.primary_id)
if records is False:
self.sd_logger.error({'error': self.my_sql_client.errors}, {
'class': 'Audit',
'method': 'get_total_distinct_mysql',
"table": self.table
})
return records
def get_total_distinct_big_query(self) -> int:
query = [
"SELECT COUNT(*) as total_count", "FROM (",
f"select DISTINCT {self.primary_id}",
f"FROM {self.definitions['data_set']}.{self.table}", ")"
]
result = self.big_query_client.query(' '.join(query))
if not result:
self.sd_logger.error({'error': self.big_query_client.errors}, {
'class': 'Audit',
'method': 'get_total_distinct_big_query',
"table": self.table
})
return False
for item in result:
return item.total_count
def audit_from_totals(self, total_mysql: int,
total_big_query: int) -> bool:
# check if the totals are the same
if total_mysql == total_big_query:
return True
# big_query total will be lower than the mysql most of the time
# allow it is be
passing_total = int(round((total_mysql * .99), 0))
if total_big_query >= passing_total:
return True
self.sd_logger.info(
{
'message': f"Audit totals did not match: {self.table}",
"data": {
"MySQL Total": total_mysql,
"Big Query Total": total_big_query,
"passing total": passing_total
}
}, {
'class': 'Audit',
'method': 'audit_from_totals',
"table": self.table
})
return False
def get_random_mysql(self, last_run: int = None) -> Union[list, bool]:
result = self.my_sql_client.get_random_records(
table=self.table,
limit=self.sample_size,
index=self.primary_id,
watched=self.watched_column,
last_updated=last_run)
if result is not False:
return result
self.sd_logger.error({'error': self.my_sql_client.errors}, {
'class': 'Audit',
'method': 'get_random_mysql',
"table": self.table
})
return False
def get_big_query_records_from_sample(self, sample: list) -> DataFrame:
primary_ids = self.primary_id.split(',')
or_statements = []
# get the same items from the mysql random query in big query
for item in sample:
pk_statements = []
for key in primary_ids:
if type(item[key]) is int:
pk_statements.append(f"`{key}` = {item[key]}")
else:
pk_statements.append(f"`{key}` = '{item[key]}'")
if type(item[self.watched_column]) is int:
watched_condition = f"AND `{self.watched_column}` = {item[self.watched_column]}"
else:
if item[self.watched_column] is None:
watched_condition = f"AND `{self.watched_column}` IS NULL"
else:
watched_condition = f"AND `{self.watched_column}` = '{item[self.watched_column]}'"
or_statements.append(
f"({' AND '.join(pk_statements)} {watched_condition})")
query = [
f"SELECT *", f"FROM {self.definitions['data_set']}.{self.table}",
f"WHERE {' OR '.join(or_statements)}", "ORDER BY"
]
order_by = []
for key in primary_ids:
order_by.append(f"{key} ASC")
query.append(', '.join(order_by))
result = self.big_query_client.query(' '.join(query))
if not result:
self.sd_logger.error({'error': self.big_query_client.errors}, {
'class': 'Audit',
'method': 'get_big_query_records_from_sample',
"table": self.table
})
return False
output = result.to_dataframe()
return output
def convert_mysql_to_data_frame(
self, mysql_data: list) -> Union[DataFrame, bool]:
self.big_query_client.set_table_ref(self.table)
bq_schema = self.big_query_client.get_schema()
schema = service_helpers.convert_schema_to_dict(bq_schema)
if len(mysql_data[0]) is not len(schema.keys()):
self.sd_logger.critical(
{
'message': "Schema and data length mismatch",
'schema_length': len(schema.keys()),
'data_length': len(mysql_data[0])
}, {
'class': 'Audit',
'method': 'convert_mysql_to_data_frame',
'table': self.table
})
return False
data_frame = pd.DataFrame.from_records(mysql_data,
columns=schema.keys())
try:
df = service_helpers.data_frame_to_schema(data_frame,
schema,
utc=True)
except ValueError as e:
self.sd_logger.critical({'message': 'Error: {}'.format(e)}, {
'class': 'Audit',
'method': 'transform_data_frame_to_match_big_query_schema',
'table': self.table
})
return False
return df
def audit_from_sample(self, mysql_data: DataFrame,
big_query_data: DataFrame) -> bool:
# test if data frames are equal - this will fail in most cases to do the poor data quality from sources
result = mysql_data.equals(big_query_data)
if result:
return True
# check if the number of returned rows are the same
# this check is useful because of the strictness of the queries that generate these DataFrames
if mysql_data.shape[0] == big_query_data.shape[0]:
return True
self.sd_logger.info(
{
'message': f"Audit Failure: {self.table}",
'DataFrame.equals': result,
'mysql.shape': mysql_data.shape[0],
'big_query.shape': big_query_data.shape[0]
}, {
'class': 'Audit',
'method': 'audit_from_sample',
'table': self.table
})
return False