def get_context():
global global_tc
with lock:
if global_tc is None:
sparktkconf_dict = {
'spark.driver.memory': "3712m",
'spark.executor.instances': '5',
'spark.driver.cores': '1',
'spark.driver.extraJavaOptions': '-Xmx2688m',
'spark.driver.maxPermSize': '512m',
'spark.driver.maxResultSize': '2g',
'spark.dynamicAllocation.enabled': 'true',
'spark.dynamicAllocation.maxExecutors': '116',
'spark.dynamicAllocation.minExecutors': '1',
'spark.executor.cores': '1',
'spark.executor.extrajavaoptions': '-Xmx2688m',
'spark.executor.memory': '3200m',
'spark.shuffle.io.preferDirectBufs': 'false',
'spark.shuffle.service.enabled': 'true',
'spark.yarn.am.waitTime': '1000000',
'spark.yarn.driver.memoryOverhead': '384',
'spark.yarn.executor.memoryOverhead': '384',
'spark.eventLog.enabled': 'false',
'spark.sql.shuffle.partitions': '6'
}
if config.run_mode:
global_tc = stk.TkContext(master='yarn-client',
other_libs=[daaltk])
else:
global_tc = stk.TkContext(other_libs=[daaltk])
return global_tc
def get_context():
global global_tc
with lock:
if global_tc is None:
sparktkconf_dict = spark_context_config.get_spark_conf()
if config.run_mode:
global_tc = stk.TkContext(master='yarn-client',
extra_conf_dict=sparktkconf_dict,
py_files=udf_files)
else:
global_tc = stk.TkContext(py_files=udf_files)
return global_tc
def test_graph_example(self):
"""Documentation test for classifiers"""
# Get a context from the spark-tk library
tc = sparktk.TkContext()
# Graphs are composed of 2 sets, one of verticess, and one of edges
# that connect exactly two (possibly not distinct) verticees.
# The degree of a vertex is the number of edges attached to it
# Below we build a frame using a vertex list and an edge list.
vertex_frame = tc.frame.create(
[["vertex1"], ["vertex2"], ["vertex3"], ["vertex4"], ["vertex5"]],
[("id", str)])
edge_frame = tc.frame.create(
[["vertex2", "vertex3"], ["vertex2", "vertex1"],
["vertex2", "vertex4"], ["vertex2", "vertex5"]], [("src", str),
("dst", str)])
# The graph is a center vertex on vertex2, with 4 verticess each
# attached to the center vertex . This is known as a star graph, in
# this configuration it can be visualized as a plus sign
# To Create a graph first you define the vertices, and then the edges
graph = tc.graph.create(vertex_frame, edge_frame)
# get the degrees, which have known values
degrees = graph.degrees()
degree_list = degrees.take(5)
known_list = [[u'vertex4', 1], [u'vertex1', 1], [u'vertex5', 1],
[u'vertex2', 4], [u'vertex3', 1]]
self.assertItemsEqual(known_list, degree_list)
def get_context():
global global_tc
with lock:
if global_tc is None:
global_tc = stk.TkContext()
return global_tc
