class MRJob(object):
INPUT_PROTOCOL = protocol.JSONValueProtocol
INTERNAL_PROTOCOL = protocol.JSONProtocol
OUTPUT_PROTOCOL = protocol.JSONValueProtocol
def __init__(self, counters=None):
self._settings = self.settings()
self._steps = self.steps()
self._counters = MRCounter()
@classmethod
def run(cls):
mapreduce(cls)
@abstractmethod
def steps(self):
"""define steps necessary to run the job"""
@abstractmethod
def settings(self):
"""define settings"""
def increment_counter(self, group, counter, amount=1):
self._counters.incr(group, counter, amount)
def get_step(self, step_idx):
return self.steps()[step_idx]
def combine_counters(work_dir, n_map_shards, n_reduce_shards):
filenames = map(lambda (work_dir, shard):
os.path.join(work_dir, 'map-%d.counters' % shard),
zip([work_dir] * n_map_shards, range(n_map_shards)))
filenames += map(lambda (work_dir, shard):
os.path.join(work_dir, 'combine-%d.counters' % shard),
zip([work_dir] * n_map_shards, range(n_map_shards)))
filenames += map(lambda (work_dir, shard):
os.path.join(work_dir, 'reduce-%d.counters' % shard),
zip([work_dir] * n_reduce_shards, range(n_reduce_shards)))
return MRCounter.sum(
imap(MRCounter.deserialize,
read_files(filter(os.path.exists, filenames))))
def __init__(self, counters=None):
self._settings = self.settings()
self._steps = self.steps()
self._counters = MRCounter()
