def spark(self, line, cell="", local_ns=None):
"""Magic to execute spark remotely.
This magic allows you to create a Livy Scala or Python session against a Livy endpoint. Every session can
be used to execute either Spark code or SparkSQL code by executing against the SQL context in the session.
When the SQL context is used, the result will be a Pandas dataframe of a sample of the results.
If invoked with no subcommand, the cell will be executed against the specified session.
Subcommands
-----------
info
Display the available Livy sessions and other configurations for sessions.
add
Add a Livy session given a session name (-s), language (-l), and endpoint credentials.
The -k argument, if present, will skip adding this session if it already exists.
e.g. `%spark add -s test -l python -u https://sparkcluster.net/livy -t Kerberos -a u -p -k`
config
Override the livy session properties sent to Livy on session creation. All session creations will
contain these config settings from then on.
Expected value is a JSON key-value string to be sent as part of the Request Body for the POST /sessions
endpoint in Livy.
e.g. `%%spark config`
`{"driverMemory":"1000M", "executorCores":4}`
run
Run Spark code against a session.
e.g. `%%spark -s testsession` will execute the cell code against the testsession previously created
e.g. `%%spark -s testsession -c sql` will execute the SQL code against the testsession previously created
e.g. `%%spark -s testsession -c sql -o my_var` will execute the SQL code against the testsession
previously created and store the pandas dataframe created in the my_var variable in the
Python environment.
use
Run Spark code against a exist session
e.g. `%%spark use -u https://sparkcluster.net/livy -i 1 -l python` will execute the cell code against the existed session 1
e.g. `%%spark use -u https://sparkcluster.net/livy -i 1 -l python -c sql -l python` will execute the SQL code against the existed session 1
e.g. `%%spark use -u https://sparkcluster.net/livy -i 1 -l python -c sql -o my_var -l python` will execute the SQL code against the existed session 1
previously created and store the pandas dataframe created in the my_var variable in the
Python environment.
logs
Returns the logs for a given session.
e.g. `%spark logs -s testsession` will return the logs for the testsession previously created
delete
Delete a Livy session.
e.g. `%spark delete -s defaultlivy`
cleanup
Delete all Livy sessions created by the notebook. No arguments required.
e.g. `%spark cleanup`
"""
usage = "Please look at usage of %spark by executing `%spark?`."
user_input = line
args = parse_argstring_or_throw(self.spark, user_input)
subcommand = args.command[0].lower()
if args.auth is None:
args.auth = conf.get_auth_value(args.user, args.password)
else:
args.auth = args.auth
# info
if subcommand == "info":
if args.url is not None and args.id is not None:
endpoint = Endpoint(args.url, args.auth, args.user,
args.password)
info_sessions = self.spark_controller.get_all_sessions_endpoint_info(
endpoint)
self._print_endpoint_info(info_sessions, args.id)
else:
self._print_local_info()
# config
elif subcommand == "config":
conf.override(conf.session_configs.__name__, json.loads(cell))
# add
elif subcommand == "add":
if args.url is None:
self.ipython_display.send_error(
"Need to supply URL argument (e.g. -u https://example.com/livyendpoint)"
)
return
name = args.session
language = args.language
endpoint = Endpoint(args.url, args.auth, args.user, args.password)
skip = args.skip
properties = conf.get_session_properties(language)
self.spark_controller.add_session(name, endpoint, skip, properties)
# delete
elif subcommand == "delete":
if args.session is not None:
self.spark_controller.delete_session_by_name(args.session)
elif args.url is not None:
if args.id is None:
self.ipython_display.send_error(
"Must provide --id or -i option to delete session at endpoint from URL"
)
return
endpoint = Endpoint(args.url, args.auth, args.user,
args.password)
session_id = args.id
self.spark_controller.delete_session_by_id(
endpoint, session_id)
else:
self.ipython_display.send_error(
"Subcommand 'delete' requires a session name or a URL and session ID"
)
# cleanup
elif subcommand == "cleanup":
if args.url is not None:
endpoint = Endpoint(args.url, args.auth, args.user,
args.password)
self.spark_controller.cleanup_endpoint(endpoint)
else:
self.spark_controller.cleanup()
# logs
elif subcommand == "logs":
self.ipython_display.write(
self.spark_controller.get_logs(args.session))
# use
elif subcommand == "use":
if args.url is None:
self.ipython_display.send_error(
"Need to supply URL argument (e.g. -u https://example.com/livyendpoint)"
)
return
endpoint = Endpoint(args.url, args.auth, args.user, args.password)
properties = conf.get_session_properties(args.language)
session = self.spark_controller.tmp_session(
endpoint, properties, args.id)
coerce = get_coerce_value(args.coerce)
if args.context == CONTEXT_NAME_SPARK:
return self.execute_spark2(cell, args.output,
args.samplemethod, args.maxrows,
args.samplefraction, session,
coerce)
elif args.context == CONTEXT_NAME_SQL:
return self.execute_sqlquery2(cell, args.samplemethod,
args.maxrows,
args.samplefraction, session,
args.output, args.quiet, coerce)
else:
self.ipython_display.send_error(
"Context '{}' not found".format(args.context))
# run
elif len(subcommand) == 0:
coerce = get_coerce_value(args.coerce)
if args.context == CONTEXT_NAME_SPARK:
return self.execute_spark(cell, args.output, args.samplemethod,
args.maxrows, args.samplefraction,
args.session, coerce)
elif args.context == CONTEXT_NAME_SQL:
return self.execute_sqlquery(cell, args.samplemethod,
args.maxrows, args.samplefraction,
args.session, args.output,
args.quiet, coerce)
else:
self.ipython_display.send_error(
"Context '{}' not found".format(args.context))
# error
else:
self.ipython_display.send_error(
"Subcommand '{}' not found. {}".format(subcommand, usage))
def refresh_configuration(self):
credentials = getattr(conf, 'base64_kernel_' + self.language + '_credentials')()
(username, password, auth, url) = (credentials['username'], credentials['password'], credentials['auth'], credentials['url'])
self.endpoint = Endpoint(url, auth, username, password)
'googledataprocauthenticator.magics.dataprocmagics.DataprocMagics.self.db',
new_callable=PropertyMock,
return_value=mocked_db):
global magic, spark_controller, shell, ipython_display
magic = DataprocMagics(shell=None, widget=MagicMock())
magic.shell = shell = MagicMock()
magic.ipython_display = ipython_display = MagicMock()
magic.spark_controller = spark_controller = MagicMock()
def _teardown():
pass
stored_endpoints = ("http://url.com",
Endpoint("http://url.com", "default-credentials"))
get_session_id_to_name = {1234: 'my_session'}
sessions_mock = {'my_session': LivySession(http_client=MagicMock(), properties={"kind":SESSION_KIND_SPARK, \
"heartbeatTimeoutInSecond": 60}, ipython_display=ipython_display, session_id=1234)}
sessions_list_mock = [LivySession(http_client=MagicMock(), properties={"kind":SESSION_KIND_SPARK,\
"heartbeatTimeoutInSecond": 60}, ipython_display=ipython_display, session_id=1234)]
mocked_db = {
'autorestore/stored_endpoints': stored_endpoints,
'autorestore/get_session_id_to_name': get_session_id_to_name,
}
def make_credentials():
return credentials.Credentials(
token=None,
refresh_token='refresh',
def refresh_configuration(self):
self.endpoint = Endpoint("new_url", None)
def __init__(self, shell, data=None, spark_events=None):
super(TestKernelMagics, self).__init__(shell,
spark_events=spark_events)
self.language = constants.LANG_PYTHON
self.endpoint = Endpoint("url", None)
def test_equality():
assert_equals(Endpoint("http://url.com", AUTH_BASIC, "sdf", "w"),
Endpoint("http://url.com", AUTH_BASIC, "sdf", "w"))
assert_equals(Endpoint("http://url.com", NO_AUTH, "sdf", "w"),
Endpoint("http://url.com", NO_AUTH, "sdf", "w"))
