def setUpClass(cls):
cls.bucket = "TEST_BUCKET"
conn = AwsConnectionManager(
AwsConnectionSettings(region="us-east-1",
secrets_manager=AwsSecretsManager(),
profile=None))
cls.s3 = S3Util(conn=conn, bucket=cls.bucket)
def test__list_source_files__should_list_the_correct_source_files_as_an_array(self):
test_bucket = "test"
test_source_key_prefix = "test_for_s3_to_cassandra"
aws_conn_settings = AwsConnectionSettings(
region="ap-southeast-2",
secrets_manager=AwsSecretsManager(),
profile=None)
s3 = S3Util(conn=AwsConnectionManager(aws_conn_settings), bucket=test_bucket)
s3.create_bucket()
for itr in range(3):
s3.upload_file(self.sample_file_location, f"{test_source_key_prefix}/{itr}.abc")
expected = [
f"{test_source_key_prefix}/0.abc",
f"{test_source_key_prefix}/1.abc",
f"{test_source_key_prefix}/2.abc", ]
util = S3ToCassandra(S3ToCassandraSettings(
source_bucket=test_bucket,
source_key_prefix=test_source_key_prefix,
source_connection_settings=aws_conn_settings,
destination_keyspace="test",
destination_table="test",
destination_table_primary_keys=[],
destination_table_options_statement="",
destination_batch_size=2,
destination_connection_settings=Mock(),
))
actual = util.list_source_files()
self.assertListEqual(actual, expected)
def test_the_etl_should_download_first_file_correctly(self):
test_bucket = "SOME_OTHER_BUCKET"
test_key = "some/key"
aws_conn = AwsConnectionSettings(region="ap-southeast-2",
secrets_manager=AwsSecretsManager(),
profile=None)
s3_util = S3Util(conn=AwsConnectionManager(aws_conn),
bucket=test_bucket)
df1 = pd.DataFrame(dict(A=range(10)),
index=pd.date_range('20130101',
periods=10,
freq='d'))
df2 = pd.DataFrame(dict(A=range(10, 20)),
index=pd.date_range('20130111',
periods=10,
freq='d'))
s3_util.create_bucket()
s3_util.upload_dataframe_as_parquet(df1, test_key, "df1")
s3_util.upload_dataframe_as_parquet(df2, test_key, "df2")
settings = S3ToDataFrameSettings(source_bucket=test_bucket,
source_key_prefix=test_key,
source_connection_settings=aws_conn)
etl = S3ToDataFrame(settings)
expected = pd.DataFrame(dict(A=range(20)),
index=pd.date_range('20130101',
periods=20,
freq='d'))
assert_frame_equal(expected, etl.get_all_files_as_data_frame())
def test_the_etl_should_list_files_correctly(self):
test_bucket = "TEST_BUCKET_ITR"
test_key = "some/key"
aws_conn = AwsConnectionSettings(region="ap-southeast-2",
secrets_manager=AwsSecretsManager(),
profile=None)
s3_util = S3Util(conn=AwsConnectionManager(aws_conn),
bucket=test_bucket)
df1 = pd.DataFrame(dict(A=range(10000)),
index=pd.date_range('20130101',
periods=10000,
freq='s'))
df2 = pd.DataFrame(dict(A=range(10000)),
index=pd.date_range('20140101',
periods=10000,
freq='s'))
df3 = pd.DataFrame(dict(A=range(10000)),
index=pd.date_range('20150101',
periods=10000,
freq='s'))
s3_util.create_bucket()
s3_util.upload_dataframe_as_parquet(df1, test_key, "df1")
s3_util.upload_dataframe_as_parquet(df2, test_key, "df2")
s3_util.upload_dataframe_as_parquet(df3, test_key, "df3")
settings = S3ToDataFrameSettings(source_bucket=test_bucket,
source_key_prefix=test_key,
source_connection_settings=aws_conn)
etl = S3ToDataFrame(settings)
assert_frame_equal(df1, etl.next())
assert_frame_equal(df2, etl.next())
assert_frame_equal(df3, etl.next())
def integration_test_should__load_sheet_to_athena__when_using_sheetUtil(
self):
with open('../resources/key-file.json', 'r') as f:
obj = json.load(f)
GoogleSheetToAthena(
GoogleSheetsToAthenaSettings(
source_workbook_url='https://docs.google.com/spreadsheets/d'
'/1W1vICFsHacKyrzCBLfsQM/edit?usp=sharing',
source_sheet='spec_example',
source_row_range=None,
source_field_names_row_number=5,
source_field_types_row_number=4,
source_data_start_row_number=6,
source_connection_settings=GoogleApiConnectionSettings(
secrets_manager=GoogleApiSecretsManager(
source=DictKeyValueSource({"key_json": obj}),
key_json_var="key_json")),
manual_partition_key_value={
"column": "start_date",
"value": "2020-03-11"
},
target_database='dev',
target_table_name='test_sheets_example_v2',
target_s3_bucket='au-com-hipages-data-scratchpad',
target_s3_dir='sheets_example_v2',
target_connection_settings=AwsConnectionSettings(
region='us-east-1',
profile='default',
secrets_manager=None),
target_table_ddl_progress=True)).load_sheet_to_athena()
def test_should__copy_file_from_one_bucket_to_another__when_valid_locations_are_given(
self):
dest_bucket = "TEST_BUCKET_DEST"
conn = AwsConnectionManager(
AwsConnectionSettings(region="ap-southeast-2",
secrets_manager=AwsSecretsManager(),
profile=None))
s3_util_for_destination = S3Util(conn=conn, bucket=dest_bucket)
s3_util_for_source = self.s3
s3_util_for_source.create_bucket()
s3_util_for_destination.create_bucket()
tmp_file_path = "/tmp/testfile.txt"
dirname = os.path.dirname(tmp_file_path)
if not os.path.exists(dirname):
os.makedirs(dirname)
with open(tmp_file_path, "w+") as file:
file.write(str("Test file content"))
s3_util_for_source.upload_file(tmp_file_path, "test/testfile.txt")
s3_util_for_source.move_recursive_to_different_bucket(
source_key_prefix="test/",
destination_bucket_name=dest_bucket,
destination_key_prefix="{}/test_copy/".format(dest_bucket))
actual = s3_util_for_destination.read_lines_as_list("test_copy")[0]
expected = "Test file content"
self.assertEqual(actual, expected)
def create_batch_s3_uploader(batch_s3_uploader_config=None):
"""
Factory method to generate a kafka batch uploadxr
Args:
batch_s3_uploader_config (BatchS3UploaderConfig): Configuration
object for the s3 uploader
Returns (KafkaS3BatchExporter): Instantiated Exporter
"""
batch_s3_uploader_config = batch_s3_uploader_config \
if batch_s3_uploader_config is not None else BatchS3UploaderConfig()
aws_region = os.environ.get("AWS_DEFAULT_REGION")
conn = AwsConnectionManager(
AwsConnectionSettings(region=aws_region,
secrets_manager=AwsSecretsManager(),
profile=None))
s3_client = S3Util(conn, batch_s3_uploader_config.bucket)
return KafkaS3BatchExporter(batch_s3_uploader_config.root_path, s3_client,
batch_s3_uploader_config.ts_col_nm,
batch_s3_uploader_config.partition_key_nm)
def test_the_transformation_works(self):
aws_conn = AwsConnectionSettings(region="us-east-1",
secrets_manager=None,
profile="default")
execfile("../../secrets.py")
cassandra_conn_setting = Mock()
settings = AthenaToAdWordsOfflineConversionSettings(
source_database=os.getenv("dummy_athena_database"),
source_table=os.getenv("dummy_athena_table"),
source_connection_settings=aws_conn,
etl_identifier="xxxx",
destination_batch_size=100,
etl_state_manager_connection=cassandra_conn_setting,
etl_state_manager_keyspace="test",
transformation_column_mapping={
'abc': 'googleClickId',
'def1': 'conversionName',
'def2': 'conversionTime',
'def4': 'conversionValue'
},
destination_connection_settings=GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager()),
)
etl = AthenaToAdWordsOfflineConversion(settings)
df = DataFrame([{
"abc": "123",
"def1": "123",
"def2": "123",
"def3": "123",
"def4": "123",
"def5": "123",
}, {
"abc": "222",
"def1": "333",
"def2": "444",
"def3": "333",
"def4": "333",
"def5": "333",
}])
result = etl._data_frame_to_destination_dict(df)
expected = [{
'conversionName': '123',
'conversionTime': '123',
'conversionValue': '123',
'googleClickId': '123'
}, {
'conversionName': '333',
'conversionTime': '444',
'conversionValue': '333',
'googleClickId': '222'
}]
self.assertListEqual(result, expected)
def test__should__get_correct_estimations__with__etl_get_parallel_payloads(
self):
aws_setting = AwsConnectionSettings(
region="us-east-1",
secrets_manager=AwsSecretsManager(),
profile=None)
target_bucket = os.getenv('S3_TEST_BUCKET')
target_key_prefix = "something/test"
# Load secrets via env vars
execfile("../../secrets.py")
adwords_settings = GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager())
target_table = "test_adwords_to_athena_table_creation"
etl_settings = AdWordsToAthenaSettings(
source_query_fragment=ServiceQueryBuilder().Select('Id').OrderBy(
'Id'),
source_service="AdGroupAdService",
source_service_version="v201809",
source_connection_settings=adwords_settings,
target_bucket=target_bucket,
target_key_prefix=target_key_prefix,
target_connection_settings=aws_setting,
target_database="dev",
target_table=target_table,
target_table_ddl_progress=True,
is_partitioned_table=True,
partition_values=[("abc", "def"), ("pqr", 123)])
etl = AdWordsToAthena(etl_settings)
actual_payloads = etl.get_parallel_payloads(page_size=1000,
number_of_workers=3)
expected_payloads = [{
'number_of_pages': 393,
'page_size': 1000,
'start_index': 0,
'worker': 0
}, {
'number_of_pages': 393,
'page_size': 1000,
'start_index': 393000,
'worker': 1
}, {
'number_of_pages': 393,
'page_size': 1000,
'start_index': 786000,
'worker': 2
}]
self.assertListEqual(expected_payloads, actual_payloads)
etl.create_athena_table()
conn = AwsConnectionManager(aws_setting)
au = AthenaUtil("dev", conn)
actual = au.get_glue_table_metadata(target_table)
print(actual)
def test_full_integration_with_local_cassandra(self):
aws_conn = AwsConnectionSettings(region="us-east-1",
secrets_manager=None,
profile="default")
execfile("../../secrets.py")
compose = DockerCompose(filepath=os.path.dirname(base.__file__))
with compose:
host = compose.get_service_host("cassandra", 9042)
port = int(compose.get_service_port("cassandra", 9042))
cassandra_conn_setting = CassandraConnectionSettings(
cluster_ips=[host],
port=port,
load_balancing_policy=DCAwareRoundRobinPolicy(),
secrets_manager=CassandraSecretsManager(
source=DictKeyValueSource({
"CASSANDRA_USERNAME": "",
"CASSANDRA_PASSWORD": "",
})),
)
conn = verify_container_is_up(cassandra_conn_setting)
# conn.get_session('system').execute(""" DROP TABLE test.etl_sink_record_state""")
settings = AthenaToAdWordsOfflineConversionSettings(
source_database=os.getenv("dummy_athena_database"),
source_table=os.getenv("dummy_athena_table"),
source_connection_settings=aws_conn,
etl_identifier="test",
destination_batch_size=100,
etl_state_manager_connection=cassandra_conn_setting,
etl_state_manager_keyspace="test",
transformation_column_mapping={
'google_click_id': 'googleClickId',
'conversion_name': 'conversionName',
'conversion_time': 'conversionTime',
'conversion_value': 'conversionValue',
'conversion_currency_code': 'conversionCurrencyCode'
},
destination_connection_settings=GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager()),
)
etl = AthenaToAdWordsOfflineConversion(settings)
files_actual = etl.list_source_files()
#
# self.assertListEqual(files_actual, [])
# etl.upload_all()
act = etl.upload_all()
self.assertListEqual(act, [])
def test__should__create_s3_file_for_the_given_indices(self):
# Load secrets via env vars
execfile("../../secrets.py")
aws_setting = AwsConnectionSettings(
region="us-east-1",
secrets_manager=AwsSecretsManager(),
profile=None)
target_bucket = os.getenv('S3_TEST_BUCKET')
target_key_prefix = "something/test"
conn = AwsConnectionManager(aws_setting)
s3u = S3Util(conn=conn, bucket=target_bucket)
s3u.delete_recursive(target_key_prefix)
adwords_settings = GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager())
adword_to_s3_util = AdWordsToS3(settings=AdWordsToS3Settings(
source_query_fragment=ServiceQueryBuilder().Select(
# Attributes
'BaseAdGroupId',
'Id',
'CampaignId',
'CampaignName',
'Name',
'Status',
'ContentBidCriterionTypeGroup',
'BaseCampaignId',
'TrackingUrlTemplate',
'FinalUrlSuffix',
'UrlCustomParameters',
'AdGroupType').OrderBy('Id'),
source_service="AdGroupService",
source_service_version="v201809",
source_connection_settings=adwords_settings,
target_bucket=target_bucket,
target_key_prefix=target_key_prefix,
target_file_prefix=None,
target_connection_settings=aws_setting))
adword_to_s3_util.build_query(start_index=35000,
page_size=1000,
num_iterations=1)
adword_to_s3_util.transfer_all()
actual = s3u.get_keys(target_key_prefix)
expected = [
'tmp/test/hip_data_tools/adwords_to_s3/test/index_35000__35999.parquet'
]
self.assertListEqual(expected, actual)
def test__should__get_correct_estimations__with__etl_get_parallel_payloads(
self):
# Load secrets via env vars
execfile("../../secrets.py")
aws_setting = AwsConnectionSettings(
region="us-east-1",
secrets_manager=AwsSecretsManager(),
profile=None)
target_bucket = os.getenv('S3_TEST_BUCKET')
target_key_prefix = "something/test"
adwords_settings = GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager())
etl_settings = AdWordsToS3Settings(
source_query_fragment=ServiceQueryBuilder().Select('Id').OrderBy(
'Id'),
source_service="AdGroupAdService",
source_service_version="v201809",
source_connection_settings=adwords_settings,
target_bucket=target_bucket,
target_key_prefix=target_key_prefix,
target_file_prefix=None,
target_connection_settings=aws_setting)
etl = AdWordsToS3(etl_settings)
actual_payloads = etl.get_parallel_payloads(page_size=1000,
number_of_workers=3)
expected_payloads = [{
'worker': 0,
'start_index': 0,
'number_of_pages': 435,
'page_size': 1000
}, {
'worker': 1,
'start_index': 435000,
'number_of_pages': 435,
'page_size': 1000
}, {
'worker': 2,
'start_index': 870000,
'number_of_pages': 435,
'page_size': 1000
}]
self.assertListEqual(expected_payloads, actual_payloads)
def test__should__create_table__with__a_general_report(self):
aws_setting = AwsConnectionSettings(
region="us-east-1",
secrets_manager=AwsSecretsManager(),
profile=None)
target_bucket = os.getenv('S3_TEST_BUCKET')
target_key_prefix = "something/test"
# Load secrets via env vars
execfile("../../secrets.py")
adwords_settings = GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager())
target_table = "test_adwords_negative_report"
etl_settings = AdWordsReportsToAthenaSettings(
source_query=(ReportQueryBuilder().Select(
'AccountDescriptiveName', 'CampaignId', 'CampaignName',
'CampaignStatus', 'Id', 'KeywordMatchType', 'Criteria').From(
'CAMPAIGN_NEGATIVE_KEYWORDS_PERFORMANCE_REPORT').Build()),
source_include_zero_impressions=True,
source_connection_settings=adwords_settings,
target_bucket=target_bucket,
target_key_prefix=target_key_prefix,
target_connection_settings=aws_setting,
target_database="dev",
target_table=target_table,
target_table_ddl_progress=True,
is_partitioned_table=True,
partition_values=[("abc", "def"), ("pqr", 123)],
target_file_prefix="data",
transformation_field_type_mask=None)
etl = AdWordsReportsToAthena(etl_settings)
etl.transfer()
etl.create_athena_table()
etl.add_partitions()
au = AthenaUtil(database="dev",
conn=AwsConnectionManager(aws_setting),
output_bucket=os.environ["S3_TEST_BUCKET"])
actual = au.run_query(query_string="""
select * from dev.test_adwords_negative_report limit 10
""",
return_result=True)
print(actual)
expected = 11
self.assertEqual(expected, len(actual["ResultSet"]["Rows"]))
def transfer_files():
"""
transfers files from one bucket to another
Returns: None
"""
# These Aws setting assume that you have your Aws Access keys in the Standard env vars
aws_setting = AwsConnectionSettings(region="ap-southeast-2",
secrets_manager=AwsSecretsManager(),
profile=None)
# If you want to use the Aws profiles stored by the swd cli tool, uncomment the following code:
# aws_setting = AwsConnectionSettings(
# region="us-east-1",
# secrets_manager=None,
# profile="default")
# Define the ETL instance
etl = S3ToS3FileCopy(
source=S3SourceSettings(
bucket="my_source_bucket",
key_prefix="source/prefix",
suffix=None,
connection_settings=aws_setting,
),
sink=S3SinkSettings(
bucket="my_target_bucket",
connection_settings=aws_setting,
),
transformers=[
AddTargetS3KeyTransformer(target_key_prefix="target/prefix")
],
)
# Check the files that will be transferred
files = etl.list_source_files()
print(files)
# If you want to transfer files in a loop
while etl.has_next():
etl.execute_next()
# Reset the source state
etl.reset_source()
# If you want to transfer all files sequentially
etl.execute_all()
def test__transfer_file__should_work(self):
aws_setting = AwsConnectionSettings(
region="us-east-1",
secrets_manager=AwsSecretsManager(),
profile=None)
source_bucket = "TEST_SOURCE_BUCKET"
target_bucket = "TEST_TARGET_BUCKET"
conn = AwsConnectionManager(aws_setting)
s3_util_for_destination = S3Util(conn=conn, bucket=target_bucket)
s3_util_for_source = S3Util(conn=conn, bucket=source_bucket)
s3_util_for_source.create_bucket()
s3_util_for_destination.create_bucket()
file = NamedTemporaryFile("w+", delete=False)
file.write(str("Test file content"))
s3_util_for_source.upload_file(file.name,
"source/prefix/test_file.txt")
file2 = NamedTemporaryFile("w+", delete=False)
file2.write(str("Test file content"))
s3_util_for_source.upload_file(file2.name,
"source/prefix/txt_file.parquet")
etl = S3ToS3FileCopy(
source=s3.S3SourceSettings(
bucket=source_bucket,
key_prefix="source/prefix",
suffix=".txt",
connection_settings=aws_setting,
),
sink=s3.S3SinkSettings(
bucket=target_bucket,
connection_settings=aws_setting,
),
transformers=[
AddTargetS3KeyTransformer(target_key_prefix="target/prefix")
],
)
etl.execute_next()
actual = s3_util_for_destination.get_keys("")
expected_destination_keys = ['target/prefix/test_file.txt']
self.assertListEqual(expected_destination_keys, actual)
def test_the_etl_should_work(self):
test_database = "xxx"
test_table = "xxx"
aws_conn = AwsConnectionSettings(region="ap-southeast-2",
secrets_manager=None,
profile="default")
settings = AthenaToDataFrameSettings(
source_database=test_database,
source_table=test_table,
source_connection_settings=aws_conn)
etl = AthenaToDataFrame(settings)
expected_s3_files = ['xxx']
self.assertEqual(expected_s3_files, etl.list_source_files())
self.assertEqual((47848, 28), etl.next().shape)
self.assertEqual((47848, 28), etl.get_all_files_as_data_frame())
def test__should__transfer_correct_amount_of_files__with__one_parallel_fragment(
self):
# Load secrets via env vars
execfile("../../secrets.py")
aws_setting = AwsConnectionSettings(
region="us-east-1",
secrets_manager=AwsSecretsManager(),
profile=None)
target_bucket = os.getenv('S3_TEST_BUCKET')
target_key_prefix = "tmp/test/hip_data_tools/adwords_to_s3/test"
conn = AwsConnectionManager(aws_setting)
s3u = S3Util(conn=conn, bucket=target_bucket)
s3u.delete_recursive(target_key_prefix)
adwords_settings = GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager())
etl_settings = AdWordsToS3Settings(
source_query_fragment=ServiceQueryBuilder().Select('Id').OrderBy(
'Id'),
source_service="AdGroupAdService",
source_service_version="v201809",
source_connection_settings=adwords_settings,
target_bucket=target_bucket,
target_key_prefix=target_key_prefix,
target_file_prefix=None,
target_connection_settings=aws_setting)
etl = AdWordsToS3(etl_settings)
etl.build_query(start_index=0, page_size=5, num_iterations=2)
etl.transfer_all()
actual = s3u.get_keys(target_key_prefix)
print(actual)
expected = [
'tmp/test/hip_data_tools/adwords_to_s3/test/index_0__4.parquet',
'tmp/test/hip_data_tools/adwords_to_s3/test/index_5__9.parquet'
]
self.assertListEqual(expected, actual)
def test__transfer_all_files__should_work_without_transformers(self):
aws_setting = AwsConnectionSettings(
region="us-east-1",
secrets_manager=AwsSecretsManager(),
profile=None)
source_bucket = "TEST_SOURCE_BUCKET"
target_bucket = "TEST_TARGET_BUCKET"
conn = AwsConnectionManager(aws_setting)
s3_util_for_destination = S3Util(conn=conn, bucket=target_bucket)
s3_util_for_source = S3Util(conn=conn, bucket=source_bucket)
s3_util_for_source.create_bucket()
s3_util_for_destination.create_bucket()
for itr in range(10):
file = NamedTemporaryFile("w+", delete=False)
file.write(str("Test file content"))
s3_util_for_source.upload_file(
file.name, f"source/prefix/txt_file{itr}.parquet")
etl = S3ToS3FileCopy(
source=s3.S3SourceSettings(
bucket=source_bucket,
key_prefix="source/prefix",
suffix=None,
connection_settings=aws_setting,
),
sink=s3.S3SinkSettings(
bucket=target_bucket,
connection_settings=aws_setting,
),
)
etl.execute_all()
actual = s3_util_for_destination.get_keys("")
expected_destination_keys = [
f'source/prefix/txt_file{itr}.parquet' for itr in range(10)
]
self.assertListEqual(expected_destination_keys, actual)
def test_adwords_upload_with_duplicates_in_same_batch(self):
aws_conn = AwsConnectionSettings(region="us-east-1",
secrets_manager=None,
profile="default")
execfile("../../secrets.py")
compose = DockerCompose(filepath=os.path.dirname(base.__file__))
with compose:
host = compose.get_service_host("cassandra", 9042)
port = int(compose.get_service_port("cassandra", 9042))
cassandra_conn_setting = CassandraConnectionSettings(
cluster_ips=[host],
port=port,
load_balancing_policy=DCAwareRoundRobinPolicy(),
secrets_manager=CassandraSecretsManager(
source=DictKeyValueSource({
"CASSANDRA_USERNAME": "",
"CASSANDRA_PASSWORD": "",
})),
)
verify_container_is_up(cassandra_conn_setting)
settings = AthenaToAdWordsOfflineConversionSettings(
source_database=os.getenv("dummy_athena_database"),
source_table=os.getenv("dummy_athena_table"),
source_connection_settings=aws_conn,
etl_identifier="xxxx",
destination_batch_size=100,
etl_state_manager_connection=cassandra_conn_setting,
etl_state_manager_keyspace="test",
transformation_column_mapping={
'googleClickId': 'googleClickId',
'conversionName': 'conversionName',
'conversionTime': 'conversionTime',
'conversionValue': 'conversionValue',
'conversionCurrencyCode': 'conversionCurrencyCode'
},
destination_connection_settings=GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager()),
)
etl = AthenaToAdWordsOfflineConversion(settings)
test_df = DataFrame([
{
'googleClickId': 'xxx',
'conversionName': 'claim_attempts_testing',
'conversionTime': '20200309 074357 UTC',
'conversionValue': 17.0,
'conversionCurrencyCode': 'AUD',
},
{
'googleClickId':
"Cj0KCQiAqY3zBRDQARIsAJeCVxOIyZ8avQ0he3WIpHPwV6hRn"
"-8Y2gDrUBJcc95tDdLcE35TK1mhhmIaAgZGEALw_wcB",
'conversionName':
'claim_attempts_testing',
'conversionTime':
'20200309 074353 UTC',
'conversionValue':
17.0,
'conversionCurrencyCode':
'AUD',
},
{
'googleClickId':
"Cj0KCQiAqY3zBRDQARIsAJeCVxOIyZ8avQ0he3WIpHPwV6hRn"
"-8Y2gDrUBJcc95tDdLcE35TK1mhhmIaAgZGEALw_wcB",
'conversionName':
'claim_attempts_testing',
'conversionTime':
'20200309 074353 UTC', # Duplicate with same time
'conversionValue':
14.0,
'conversionCurrencyCode':
'AUD',
},
{
'googleClickId':
"Cj0KCQiAqY3zBRDQARIsAJeCVxOIyZ8avQ0he3WIpHPwV6hRn"
"-8Y2gDrUBJcc95tDdLcE35TK1mhhmIaAgZGEALw_wcB",
'conversionName':
'claim_attempts_testing',
'conversionTime':
'20200309 084353 UTC', # Duplicate with diff time
'conversionValue':
14.0,
'conversionCurrencyCode':
'AUD',
},
{
'googleClickId':
"EAIaIQobChMI6oiGy_vz5wIVkjUrCh3IcgAuEAAYASAAEgLRk_D_BwE",
'conversionName': "claim_attempts_testing",
'conversionTime': '20200309 023001 UTC',
'conversionValue': 17.0,
'conversionCurrencyCode': 'AUD',
},
])
actual = etl._process_data_frame(test_df)
expected = [ # The duplicate with same time has been Picked out as an issue
{
'error':
"Current State 'EtlStates.Processing' cannot transition to "
"'EtlStates.Processing'",
'data': {
'googleClickId':
'Cj0KCQiAqY3zBRDQARIsAJeCVxOIyZ8avQ0he3WIpHPwV6hRn'
'-8Y2gDrUBJcc95tDdLcE35TK1mhhmIaAgZGEALw_wcB',
'conversionName':
'claim_attempts_testing',
'conversionTime':
'20200309 074353 UTC',
'conversionValue':
14.0,
'conversionCurrencyCode':
'AUD'
}
},
]
self.assertListEqual(actual, expected)
def test_multiple_runs_of_same_data_and_verify_deduplication(self):
aws_conn = AwsConnectionSettings(region="us-east-1",
secrets_manager=None,
profile="default")
execfile("../../secrets.py")
compose = DockerCompose(filepath=os.path.dirname(base.__file__))
with compose:
host = compose.get_service_host("cassandra", 9042)
port = int(compose.get_service_port("cassandra", 9042))
cassandra_conn_setting = CassandraConnectionSettings(
cluster_ips=[host],
port=port,
load_balancing_policy=DCAwareRoundRobinPolicy(),
secrets_manager=CassandraSecretsManager(
source=DictKeyValueSource({
"CASSANDRA_USERNAME": "",
"CASSANDRA_PASSWORD": "",
})),
)
conn = verify_container_is_up(cassandra_conn_setting)
# conn.get_session('system').execute(""" DROP TABLE test.etl_sink_record_state""")
settings = AthenaToAdWordsOfflineConversionSettings(
source_database=os.getenv("dummy_athena_database"),
source_table=os.getenv("dummy_athena_table"),
source_connection_settings=aws_conn,
etl_identifier="test",
destination_batch_size=100,
etl_state_manager_connection=cassandra_conn_setting,
etl_state_manager_keyspace="test",
transformation_column_mapping={
'google_click_id': 'googleClickId',
'conversion_name': 'conversionName',
'conversion_time': 'conversionTime',
'conversion_value': 'conversionValue',
'conversion_currency_code': 'conversionCurrencyCode'
},
destination_connection_settings=GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id=os.getenv("adwords_client_customer_id"),
secrets_manager=GoogleAdWordsSecretsManager()),
)
etl = AthenaToAdWordsOfflineConversion(settings)
source_data = [
{
'google_click_id': 'theFirst',
'conversion_name': 'claim_attempts_testing',
'conversion_time': '20200309 074357 UTC',
'conversion_value': 17.0,
'conversion_currency_code': 'AUD',
},
{
'google_click_id': 'failedSecond',
'conversion_name': 'claim_attempts_testing',
'conversion_time': '20200309 074357 UTC',
'conversion_value': 17.0,
'conversion_currency_code': 'AUD',
},
]
test_df = DataFrame(source_data)
# Mock upload_conversions in AdWordsUtil so no actual data is transmitted
etl._upload_conversions = MagicMock(return_value=([
{
'googleClickId': 'theFirst',
'conversionName': 'claim_attempts_testing',
'conversionTime': '20200309 074357 UTC',
'conversionValue': 17.0,
'conversionCurrencyCode': 'AUD',
},
], [
{
'fieldPath':
'operations[0].operand',
'fieldPathElements': [{
'field': 'operations',
'index': 0
}, {
'field': 'operand',
'index': None
}],
'trigger':
None,
'errorString':
'OfflineConversionError.UNPARSEABLE_GCLID',
'ApiError.Type':
'OfflineConversionError',
'reason':
'UNPARSEABLE_GCLID',
'data': {
'googleClickId': 'failedSecond',
'conversionName': 'claim_attempts_testing',
'conversionTime': '20200309 074357 UTC',
'conversionValue': 17.0,
'conversionCurrencyCode': 'AUD',
},
},
]))
# etl._process_data_frame(test_df)
first_actual = etl._process_data_frame(test_df)
self.assertListEqual(first_actual, [])
# Repeat process to cause Duplicates
actual = etl._process_data_frame(test_df)
# actual = etl.upload_next()
expected = [{
'data': {
'conversionCurrencyCode': 'AUD',
'conversionName': 'claim_attempts_testing',
'conversionTime': '20200309 074357 UTC',
'conversionValue': 17.0,
'googleClickId': 'theFirst'
},
'error': 'Current state is not Ready'
}, {
'data': {
'conversionCurrencyCode': 'AUD',
'conversionName': 'claim_attempts_testing',
'conversionTime': '20200309 074357 UTC',
'conversionValue': 17.0,
'googleClickId': 'failedSecond'
},
'error': 'Current state is not Ready'
}]
self.assertListEqual(actual, expected)
def test__should__create_table__with__geo_performance_report(self):
aws_setting = AwsConnectionSettings(
region="ap-southeast-2",
secrets_manager=AwsSecretsManager(
access_key_id_var="SOME_CUSTOM_AWS_ACCESS_KEY_ID",
secret_access_key_var="SOME_CUSTOM_AWS_SECRET_ACCESS_KEY",
use_session_token=True,
aws_session_token_var="SOME_CUSTOM_AWS_SESSION_TOKEN"),
profile=None)
target_bucket = "test-bucket"
target_key_prefix = "something/test"
# Load secrets via env vars
execfile("../../secrets.py")
adwords_settings = GoogleAdWordsConnectionSettings(
client_id=os.getenv("adwords_client_id"),
user_agent="Tester",
client_customer_id="1111111111",
secrets_manager=GoogleAdWordsSecretsManager())
target_table = "test_adwords_geo_performance_report"
etl_settings = AdWordsReportsToAthenaSettings(
source_query=(
ReportQueryBuilder().Select(
# Attributes
'AccountDescriptiveName',
'CampaignId',
'CityCriteriaId',
'CountryCriteriaId',
'CustomerDescriptiveName',
'ExternalCustomerId',
'IsTargetingLocation',
'MetroCriteriaId',
'MostSpecificCriteriaId',
'RegionCriteriaId',
# Segments
'Date',
# Metrics
'Impressions',
'Clicks',
'ConversionRate',
'Conversions',
'ConversionValue',
'Cost',
'CostPerConversion').From('GEO_PERFORMANCE_REPORT').During(
start_date="20200601", end_date="20200701").Build()),
source_include_zero_impressions=False,
source_connection_settings=adwords_settings,
target_bucket=target_bucket,
target_key_prefix=target_key_prefix,
target_connection_settings=aws_setting,
target_database="dev",
target_table=target_table,
target_table_ddl_progress=True,
is_partitioned_table=True,
partition_values=[("abc", "def"), ("pqr", 123)],
target_file_prefix="data",
transformation_field_type_mask={
"country__territory": np.int,
"region": np.int,
"most_specific_location": np.int
})
etl = AdWordsReportsToAthena(etl_settings)
etl.transfer()
etl.create_athena_table()
etl.add_partitions()
au = AthenaUtil(database="dev",
conn=AwsConnectionManager(aws_setting),
output_bucket=os.environ["S3_TEST_BUCKET"])
actual = au.run_query(query_string="""
select * from dev.test_adwords_geo_performance_report limit 10
""",
return_result=True)
print(actual)
expected = 11
self.assertEqual(expected, len(actual["ResultSet"]["Rows"]))
