def create_config_table(self, big_query_client: BigQueryAdapter):
big_query_client.set_data_set_ref('mysql_sync')
big_query_client.set_table_ref('data_sources')
# return true if table already exists
if big_query_client.check_table():
return True
schema = [
bigquery.SchemaField('service',
'STRING',
description="Service Name"),
bigquery.SchemaField('data_set',
'STRING',
description="Big Query Data Set"),
bigquery.SchemaField('host',
'STRING',
description="MySQL host connection"),
bigquery.SchemaField('user',
'STRING',
description="MySQL connection user"),
bigquery.SchemaField('password',
'STRING',
description="MySQL connection password"),
bigquery.SchemaField('database',
'STRING',
description="MySQL Database"),
bigquery.SchemaField('last_run',
'TIMESTAMP',
description="Last produce date"),
]
result = big_query_client.create_table(schema)
if result:
self.sd_logger.info({'message': f"Created Config table"}, {
'class': 'PrefectInstallBigQuery',
'method': 'create_config_table'
})
else:
self.sd_logger.warning(big_query_client.errors, {
'class': 'PrefectInstallBigQuery',
'method': 'create_config_table'
})
return result
def create_tracking_table(self, big_query_client: BigQueryAdapter):
big_query_client.set_data_set_ref(self.definitions['data_set'])
big_query_client.set_table_ref('sync_tracking_table')
# return true if table already exists
if big_query_client.check_table():
return True
schema = [
bigquery.SchemaField(
'table',
'STRING',
description="Tracked Table Name"
),
bigquery.SchemaField(
'watched',
'STRING',
description="Column to watch to minimize the number of records loaded per sync"
),
bigquery.SchemaField(
'primary_id',
'STRING',
description="Primary Id Column(s)"
),
bigquery.SchemaField(
'synchronize',
'BOOLEAN',
description="Flag to Synchronize the table"
)
]
result = big_query_client.create_table(schema)
if result:
self.sd_logger.info(
{'message': f"Created tracking table"},
{'class': 'PrefectAddServiceBigQuery', 'method': 'create_tracking_table'}
)
else:
self.sd_logger.warning(
big_query_client.errors,
{'class': 'PrefectAddServiceBigQuery', 'method': 'create_tracking_table'}
)
return result
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 Schema(Task):
def __init__(self, service: str, **kwargs):
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-schema")
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'])
super().__init__(**kwargs)
def get_mysql_schema(self, table: str) -> Union[list, bool]:
mysql_client = MySqlAdapter(self.definitions['service'])
columns = mysql_client.mysql_table_definition(table)
if not columns:
self.sd_logger.error(mysql_client.errors, {
'class': 'Schema',
'method': 'get_mysql_schema',
'table': table
})
return False
return columns
@staticmethod
def organized_mysql_schema(schema: list) -> list:
organized_column_data = []
for column in schema:
organized_column_data.append(
service_helpers.label_mysql_table_definitions(column))
return organized_column_data
@staticmethod
def convert_mysql_to_big_query_schema(schema: list) -> list:
return service_helpers.generate_bq_schema_from_mysql(schema)
def store_mysql_schema(self, schema: list, table: str) -> bool:
encoded_schema = json.dumps(schema)
storage_client = StorageAdapter(self.service_account)
storage_client.get_client()
date_time_obj = datetime.utcnow()
result = storage_client.write_string(
bucket=self.definitions['service'] + '-etl',
destination=
f'schema/{table}/{date_time_obj.strftime("%m-%d-%Y_%H:%M:%S")}_UTC',
string=encoded_schema)
if not result:
self.sd_logger.error(storage_client.errors, {
'class': 'Schema',
'method': 'store_mysql_schema',
'table': table
})
return False
return result
def check_table_exists(self, table):
self.big_query_client.set_table_ref(table)
return self.big_query_client.check_table()
def get_current_schema(self, table: str):
self.big_query_client.set_table_ref(table)
return self.big_query_client.get_schema()
@staticmethod
def compare_schema(new_schema, current_schema) -> bool:
# first check is total number of items
if len(new_schema) is not len(current_schema):
return False
# compare column names and types
schema_matches = True
for x in range(len(new_schema)):
exists_in_current = False
if current_schema[x].name == new_schema[x].name:
if current_schema[x].field_type == new_schema[x].field_type:
exists_in_current = True
else:
exists_in_current = True
if not exists_in_current:
schema_matches = False
return schema_matches
def create_table(self, table: str, schema: list) -> bool:
self.big_query_client.set_table_ref(table)
result = self.big_query_client.create_table(schema=schema,
overwrite=True)
if result:
self.sd_logger.info({'message': f"Table {table} Created"}, {
'class': 'Schema',
'method': 'create_table',
'table': table
})
else:
self.sd_logger.error(self.big_query_client.errors, {
'class': 'Schema',
'method': 'create_table',
'table': table
})
return result
@staticmethod
def acknowledge_message(message):
message.ack()
def copy_bq_table(self, table: str):
self.big_query_client.set_table_ref(table)
copy_table_ref = self.big_query_client.table_ref
date_time_obj = datetime.utcnow()
destination_str = f'{table}_{date_time_obj.strftime("%m_%d_%Y")}'
self.big_query_client.set_table_ref(destination_str)
destination_table_ref = self.big_query_client.table_ref
result = self.big_query_client.copy_table(
copy_table=copy_table_ref, destination_table=destination_table_ref)
if result:
self.sd_logger.warning(
{'message': f"Table {table} copied to {destination_str}"}, {
'class': 'Schema',
'method': 'copy_bq_table',
'table': table
})
else:
self.sd_logger.error(self.big_query_client.errors, {
'class': 'Schema',
'method': 'copy_bq_table',
'table': table
})
return destination_str
def backup_table_to_storage(self, table):
self.big_query_client.set_table_ref(table)
copy_table_ref = self.big_query_client.table_ref
date_time_obj = datetime.utcnow()
destination = f'gs://{self.definitions["service"]}-etl/data/{table}/{date_time_obj.strftime("%m-%d-%Y_%H:%M:%S")}_UTC_*.avro'
result = self.big_query_client.export_table_to_storage(
table=copy_table_ref, destination=destination)
if result:
self.sd_logger.info(
{'message': f"Table {table} exported to {destination}"}, {
'class': 'Schema',
'method': 'backup_table_to_storage'
})
else:
self.sd_logger.error(self.big_query_client.errors, {
'class': 'Schema',
'method': 'backup_table_to_storage'
})
return result
def delete_table(self, table):
self.big_query_client.set_table_ref(table)
result = self.big_query_client.delete_table()
if result:
self.sd_logger.info({'message': f"Table {table} deleted"}, {
'class': 'Schema',
'method': 'delete_table'
})
else:
self.sd_logger.error(self.big_query_client.errors, {
'class': 'Schema',
'method': 'delete_table'
})
return result