def _set_workflow_config(self, yaml_file):
# Receives a yaml file path with Workflow configurations and sets appropriate values for properties in this class
log.info("Setting configurations from config file {0}".format(yaml_file))
try:
config = None
with open(yaml_file, "r") as ymlfile:
config = yaml.load(ymlfile)
self._source = config["source"]
self._destination = config["destination"]
self._io_reader = Factory.get_reader(self._source["type"], self._source)
self._io_writer = Factory.get_writer(
self._destination["type"], self._destination
)
# Set attributes for custom workflow properties
for key in config.keys():
if key not in self.DEFAULT_PARAMS:
setattr(self, key, config[key])
except Exception:
log.error(
"Error creating I/O reader and writer. Please check configurations in workflow config file at {0}".format(
yaml_file
)
)
raise
def __init__(self, name, source=None, destination=None):
# Initialize properties
self._source = None
self._destination = None
self._name = name
# Check to see if workflow yaml file exists. If so, set workflow configurations from file.
default_filepath = self._get_default_filepath(name)
backup_filepath = self._get_backup_filepath(name)
if os.path.exists(default_filepath):
log.info("Config file detected: {0}".format(default_filepath))
self._set_workflow_config(default_filepath)
elif os.path.exists(backup_filepath):
log.info("Config file detected: {0}".format(backup_filepath))
self._set_workflow_config(backup_filepath)
else:
log.info("No config file detected.")
# If source or destination are passed in as parameters, update source and dest configurations.
if source:
self._source = source
self._io_reader = Factory.get_reader(self._source["type"],
self._source)
if destination:
self._destination = destination
self._io_writer = Factory.get_writer(self._destination["type"],
self._destination)
def set_source(self, source):
"""
Set source.
:param source: dict of configuration parameters for data source (reader)
"""
self._source = source
self._io_reader = Factory.get_reader(self.source["type"], self.source)
def test_get_reader_orc(test_input_base_path, expected_df):
test_input_path = "%s/person.orc" % (test_input_base_path)
config = {
"type": "fs",
"input_path": test_input_path,
"usecols": ["firstname", "lastname", "gender"],
"input_format": "orc",
}
reader_from_factory = Factory.get_reader("fs", config)
fetched_df = reader_from_factory.fetch_data()
assert fetched_df.equals(expected_df)
def test_get_reader_json(tmpdir, expected_df):
fname = str(tmpdir.mkdir("tmp_test_fs_reader").join("person.json"))
cudf.io.json.to_json(expected_df, fname, orient="records")
config = {
"type": "fs",
"input_path": fname,
"orient": "records",
"input_format": "json",
}
reader_from_factory = Factory.get_reader("fs", config)
fetched_df = reader_from_factory.fetch_data()
assert fetched_df.equals(expected_df)
def test_get_reader_orc(tmpdir, expected_df):
fname = str(tmpdir.mkdir("tmp_test_fs_reader").join("person.orc"))
cudf.io.orc.to_orc(expected_df, fname)
config = {
"type": "fs",
"input_path": fname,
"usecols": ["firstname", "lastname", "gender"],
"input_format": "orc",
}
reader_from_factory = Factory.get_reader("fs", config)
fetched_df = reader_from_factory.fetch_data()
assert fetched_df.equals(expected_df)
def test_get_reader_text(test_input_base_path, expected_df):
test_input_path = "%s/person.csv" % (test_input_base_path)
config = {
"type": "fs",
"input_path": test_input_path,
"names": ["firstname", "lastname", "gender"],
"delimiter": ",",
"usecols": ["firstname", "lastname", "gender"],
"dtype": ["str", "str", "str"],
"header": 0,
"input_format": "text",
}
reader_from_factory = Factory.get_reader("fs", config)
fetched_df = reader_from_factory.fetch_data()
assert fetched_df.equals(expected_df)
def test_get_reader_csv(tmpdir, expected_df):
fname = tmpdir.mkdir("tmp_test_factory").join("person.csv")
expected_df.to_csv(fname, index=False)
config = {
"type": "fs",
"input_path": fname,
"names": ["firstname", "lastname", "gender"],
"delimiter": ",",
"usecols": ["firstname", "lastname", "gender"],
"dtype": ["str", "str", "str"],
"header": 0,
"input_format": "csv",
}
reader_from_factory = Factory.get_reader("fs", config)
fetched_df = reader_from_factory.fetch_data()
assert fetched_df.equals(expected_df)
def test_get_io_reader_fs(fs_reader_config):
reader = Factory.get_reader("fs", fs_reader_config)
expected_cls = FileSystemReader
assert isinstance(reader, expected_cls)
def test_get_io_reader_kafka(kafka_config):
reader = Factory.get_reader("kafka", kafka_config)
expected_cls = KafkaReader
assert isinstance(reader, expected_cls)
def set_source(self, source):
self._source = source
self._io_reader = Factory.get_reader(self.source["type"], self.source)
