def map_reduce_queryset(queryset, map_func, reduce_func, output_writer, *args,
**kwargs):
"""
Does a complete map-shuffle-reduce over the queryset
output_writer should be a mapreduce OutputWriter subclass
Returns the pipeline
"""
map_func = qualname(map_func)
reduce_func = qualname(reduce_func)
output_writer = qualname(output_writer)
options = extract_options(kwargs)
_shards = options.pop("_shards", None)
_job_name = options.pop("_job_name",
"Map reduce task over {}".format(queryset.model))
_queue_name = options.pop("_queue_name", 'default')
pipeline = MapreducePipeline(
_job_name,
map_func,
reduce_func,
qualname(DjangoInputReader),
output_writer,
mapper_params={
"input_reader": DjangoInputReader.params_from_queryset(queryset),
},
reducer_params={
'output_writer': options.pop("_output_writer_kwargs", {}) or {}
},
shards=_shards)
pipeline.start(queue_name=_queue_name)
return pipeline
def run_pipeline(self, pipeline, *args, **kwargs):
"""Runs the pipeline and returns outputs."""
require_slots_filled = kwargs.pop('_require_slots_filled', True)
task_retry = kwargs.pop('_task_retry', True)
pipeline.task_retry = task_retry
pipeline.start(*args, **kwargs)
while True:
task_list = self.get_tasks()
if not task_list:
break
print "DID TASKS"
for task in task_list:
self.run_task(task)
delete_tasks([task])
if require_slots_filled:
for slot_record in _SlotRecord.all():
self.assertEquals(_SlotRecord.FILLED, slot_record.status,
'_SlotRecord = %r' % slot_record.key())
for barrier_record in _BarrierRecord.all():
self.assertEquals(_BarrierRecord.FIRED, barrier_record.status,
'_BarrierRecord = %r' % barrier_record.key())
for pipeline_record in _PipelineRecord.all():
self.assertEquals(_PipelineRecord.DONE, pipeline_record.status,
'_PipelineRecord = %r' % pipeline_record.key())
return pipeline.__class__.from_id(pipeline.pipeline_id).outputs
def map_reduce_queryset(queryset, map_func, reduce_func, output_writer, *args, **kwargs):
"""
Does a complete map-shuffle-reduce over the queryset
output_writer should be a mapreduce OutputWriter subclass
Returns the pipeline
"""
map_func = qualname(map_func)
reduce_func = qualname(reduce_func)
output_writer = qualname(output_writer)
options = extract_options(kwargs)
_shards = options.pop("_shards", None)
_job_name = options.pop("_job_name", "Map reduce task over {}".format(queryset.model))
_queue_name = options.pop("_queue_name", 'default')
pipeline = MapreducePipeline(
_job_name,
map_func,
reduce_func,
qualname(DjangoInputReader),
output_writer,
mapper_params={
"input_reader": DjangoInputReader.params_from_queryset(queryset),
},
reducer_params={
'output_writer': options.pop("_output_writer_kwargs", {}) or {}
},
shards=_shards
)
pipeline.start(queue_name=_queue_name)
return pipeline
def run_pipeline(self, pipeline, *args, **kwargs):
"""Runs the pipeline and returns outputs."""
require_slots_filled = kwargs.pop('_require_slots_filled', True)
task_retry = kwargs.pop('_task_retry', True)
pipeline.task_retry = task_retry
pipeline.start(*args, **kwargs)
while True:
task_list = self.get_tasks()
if not task_list:
break
for task in task_list:
self.run_task(task)
delete_tasks([task])
if require_slots_filled:
for slot_record in _SlotRecord.all():
self.assertEquals(_SlotRecord.FILLED, slot_record.status,
'_SlotRecord = %r' % slot_record.key())
for barrier_record in _BarrierRecord.all():
self.assertEquals(_BarrierRecord.FIRED, barrier_record.status,
'_BarrierRecord = %r' % barrier_record.key())
for pipeline_record in _PipelineRecord.all():
self.assertEquals(
_PipelineRecord.DONE, pipeline_record.status,
'_PipelineRecord = %r' % pipeline_record.key())
return pipeline.__class__.from_id(pipeline.pipeline_id).outputs
def runner(type, input):
f = str(input.read())
separator = ',' if type == 'csv' else ' '
if type in ('csv', 'txt'):
for data in f.split('\n'):
data = data.split(separator)
if len(data) > 2:
pipeline.start(*data)
else:
for data in eval(f):
pipeline.start(data['email'], data['max_posts'], data['max_likes'])
print('>>> your config has been sent to the worker and will be done soon.')
def migrate_blob(blob_info, _mapper_params=None):
"""Starts a mapper pipeline to migrate single blob to cloud storage object.
Args:
blob_info: The blob to migrate.
_mapper_params: Allows injection of mapper parameters for testing.
Yields:
Various MapReduce counter operations.
"""
params = _mapper_params or context.get().mapreduce_spec.mapper.params
bucket_name = params['bucket_name']
yield counters.Increment('BlobInfo_considered_for_migration')
blob_key_str = _get_blob_key_str(blob_info)
# dev_appserver's stubs store the GCS blobs in the same place as blobstore
# blobs. We'll skip these so our testing is cleaner.
if (appengine_config.IS_DEVSERVER
and blob_key_str.startswith('encoded_gs_file:')):
yield counters.Increment(
'BlobInfo_is_really_GCS_file_on_dev_appserver__skipping')
raise StopIteration()
# look up the blob_key in the migration table; if already migrated, skip it
already_mapped = models.BlobKeyMapping.build_key(blob_key_str).get()
if already_mapped:
yield counters.Increment('BlobInfo_previously_migrated')
raise StopIteration() # no work to do for this blob
# if the blob is "small", migrate it in-line
if blob_info.size <= config.config.DIRECT_MIGRATION_MAX_SIZE:
migrate_single_blob_inline(blob_info, bucket_name)
yield counters.Increment('BlobInfo_migrated_within_mapper')
# else start a full-scale pipeline to handle the blob migration
else:
pipeline = MigrateSingleBlobPipeline(blob_key_str, blob_info.filename,
blob_info.content_type,
bucket_name)
pipeline.start(queue_name=config.config.QUEUE_NAME)
yield counters.Increment('BlobInfo_migrated_via_secondary_pipeline')
yield counters.Increment('BlobInfo_migrated')
raise StopIteration()
def map_reduce_entities(kind_name, namespace, map_func, reduce_func,
output_writer, *args, **kwargs):
"""
Does a complete map-shuffle-reduce over the entities
output_writer should be a mapreduce OutputWriter subclass
_filters is an optional kwarg which will be passed directly to the input reader
Returns the pipeline
"""
map_func = qualname(map_func)
reduce_func = qualname(reduce_func)
output_writer = qualname(output_writer)
options = extract_options(kwargs, additional={"_filters"})
_shards = options.pop("_shards", None)
_job_name = options.pop("_job_name",
"Map reduce task over {}".format(kind_name))
_queue_name = options.pop("_queue_name", 'default')
pipeline = MapreducePipeline(
_job_name,
map_func,
reduce_func,
qualname(RawDatastoreInputReader),
output_writer,
mapper_params={
'input_reader': {
RawDatastoreInputReader.ENTITY_KIND_PARAM: kind_name,
RawDatastoreInputReader.NAMESPACE_PARAM: namespace,
RawDatastoreInputReader.FILTERS_PARAM:
options.pop("_filters", [])
},
},
reducer_params={
'output_writer': options.pop("_output_writer_kwargs", {}) or {}
},
shards=_shards)
pipeline.start(queue_name=_queue_name)
return pipeline
def map_reduce_entities(kind_name, namespace, map_func, reduce_func, output_writer, *args, **kwargs):
"""
Does a complete map-shuffle-reduce over the entities
output_writer should be a mapreduce OutputWriter subclass
_filters is an optional kwarg which will be passed directly to the input reader
Returns the pipeline
"""
map_func = qualname(map_func)
reduce_func = qualname(reduce_func)
output_writer = qualname(output_writer)
options = extract_options(kwargs, additional={"_filters"})
_shards = options.pop("_shards", None)
_job_name = options.pop("_job_name", "Map reduce task over {}".format(kind_name))
_queue_name = options.pop("_queue_name", 'default')
pipeline = MapreducePipeline(
_job_name,
map_func,
reduce_func,
qualname(RawDatastoreInputReader),
output_writer,
mapper_params={
'input_reader': {
RawDatastoreInputReader.ENTITY_KIND_PARAM: kind_name,
RawDatastoreInputReader.NAMESPACE_PARAM: namespace,
RawDatastoreInputReader.FILTERS_PARAM: options.pop("_filters", [])
},
},
reducer_params={
'output_writer': options.pop("_output_writer_kwargs", {}) or {}
},
shards=_shards
)
pipeline.start(queue_name=_queue_name)
return pipeline
def map_reduce_entities(kind_name, map_func, reduce_func, output_writer, *args,
**kwargs):
"""
Does a complete map-shuffle-reduce over the entities
output_writer should be a mapreduce OutputWriter subclass
Returns the pipeline
"""
map_func = qualname(map_func)
reduce_func = qualname(reduce_func)
output_writer = qualname(output_writer)
options = extract_options(kwargs)
_shards = options.pop("_shards", None)
_job_name = options.pop("_job_name",
"Map reduce task over {}".format(kind_name))
_queue_name = options.pop("_queue_name", 'default')
pipeline = MapreducePipeline(
_job_name,
map_func,
reduce_func,
qualname(RawDatastoreInputReader),
output_writer,
mapper_params={
'input_reader': {
RawDatastoreInputReader.ENTITY_KIND_PARAM: kind_name
},
},
reducer_params={
'output_writer': options.pop("_output_writer_kwargs", {}) or {}
},
shards=_shards)
pipeline.start(queue_name=_queue_name)
return pipeline
def run_job():
pipeline = TouchPipeline()
pipeline.start()
return 'Job started'
""" To start Pipeline, run this script from maya's script editor or a shelf button
Remember to alter the path below! """
import sys
path_to_pipeline = '/Users/liorbenhorin/Documents/Projects/2016/GitHub/pipeline2'
if not path_to_pipeline in sys.path:
sys.path.append(path_to_pipeline)
import pipeline
# reload(pipeline)
pipeline.start()
