def test_portal_worker(self):
self._prepare_test()
map_task = dp_pb.MapTask()
map_task.output_base_dir = self._partition_output_dir
map_task.output_partition_num = self._output_partition_num
map_task.partition_id = 0
map_task.task_name = 'map_part_{}'.format(map_task.partition_id)
map_task.part_field = 'example_id'
map_task.data_portal_type = dp_pb.DataPortalType.Streaming
for partition_id in range(self._input_partition_num):
map_task.fpaths.append(self._get_input_fpath(partition_id))
# partitioner
task = dp_pb.NewTaskResponse()
task.map_task.CopyFrom(map_task)
self._portal_worker._run_map_task(task.map_task)
self._check_partitioner(task.map_task)
# merge
total_cnt = 0
for partition_id in range(self._output_partition_num):
reduce_task = dp_pb.ReduceTask()
reduce_task.map_base_dir = self._partition_output_dir
reduce_task.reduce_base_dir = self._merge_output_dir
reduce_task.partition_id = partition_id
reduce_task.task_name = 'reduce_part_{}'.format(partition_id)
self._portal_worker._run_reduce_task(reduce_task)
total_cnt += self._check_merge(reduce_task)
self.assertEqual(total_cnt,
self._partition_item_num * self._input_partition_num)
self._clean_up()
def _create_map_task(self, rank_id, partition_id):
assert self._processing_job is not None
job = self._processing_job
map_fpaths = []
for fpath in job.fpaths:
if hash(fpath) % self._output_partition_num == partition_id:
map_fpaths.append(fpath)
task_name = '{}-dp_portal_job_{:08}-part-{:04}-map'.format(
self._portal_manifest.name, job.job_id, partition_id
)
logging.info("Data portal worker-%d is allocated map task %s for "\
"partition %d of job %d. the map task has %d files"\
"-----------------\n", rank_id, task_name,
partition_id, job.job_id, len(map_fpaths))
for seq, fpath in enumerate(map_fpaths):
logging.info("%d. %s", seq, fpath)
logging.info("---------------------------------\n")
manifset = self._sync_portal_manifest()
return dp_pb.MapTask(task_name=task_name,
fpaths=map_fpaths,
output_base_dir=self._map_output_dir(job.job_id),
output_partition_num=self._output_partition_num,
partition_id=partition_id,
part_field=self._get_part_field(),
data_portal_type=manifset.data_portal_type)
def test_portal_worker(self):
self._prepare_test()
map_task = dp_pb.MapTask()
map_task.output_base_dir = self._partition_output_dir
map_task.output_partition_num = self._output_partition_num
map_task.partition_id = 0
for partition_id in range(self._input_partition_num):
map_task.fpaths.append(self._get_input_fpath(partition_id))
# partitioner
task = dp_pb.NewTaskResponse()
task.map_task.CopyFrom(map_task)
self._portal_worker._run_map_task(task.map_task)
self._check_partitioner(task.map_task)
# merge
total_cnt = 0
for partition_id in range(self._output_partition_num):
reduce_task = dp_pb.ReduceTask()
reduce_task.map_base_dir = self._partition_output_dir
reduce_task.reduce_base_dir = self._merge_output_dir
reduce_task.partition_id = partition_id
self._portal_worker._run_reduce_task(reduce_task)
total_cnt += self._check_merge(reduce_task)
self.assertEqual(total_cnt, self._partition_item_num * self._input_partition_num)
self._clean_up()
def _create_map_task(self, rank_id, partition_id):
assert self._processing_job is not None
job = self._processing_job
map_fpaths = []
for fpath in job.fpaths:
fname = path.basename(fpath)
if hash(fname) % self._output_partition_num == partition_id:
map_fpaths.append(fpath)
return dp_pb.MapTask(fpaths=map_fpaths,
output_base_dir=self._map_output_dir(job.job_id),
output_partition_num=self._output_partition_num,
partition_id=partition_id)
def _run_map_task(self):
map_task = dp_pb.MapTask()
map_task.output_base_dir = self._partition_output_dir
map_task.output_partition_num = self._output_partition_num
map_task.partition_id = 0
map_task.task_name = 'map_part_{}'.format(map_task.partition_id)
map_task.part_field = 'example_id'
map_task.data_portal_type = dp_pb.DataPortalType.Streaming
for partition_id in range(self._input_partition_num):
map_task.fpaths.append(self._get_input_fpath(partition_id))
# partitioner
task = dp_pb.NewTaskResponse()
task.map_task.CopyFrom(map_task)
self._portal_worker._run_map_task(task.map_task)
return task
