def _run(self):
red_out, out_url, fd_list = self.connect_output()
red_in = iter(ReduceReader(self))
params = self.params
if self.ext_reduce:
path = self.path('EXT_REDUCE')
external.prepare(self.reduce, self.ext_params, path)
self.reduce = FunctionType(external.ext_reduce.func_code,
globals=external.__dict__)
self.insert_globals([self.reduce])
Message("Starting reduce")
self.init(red_in, params)
self.reduce(red_in, red_out, params)
Message("Reduce done")
self.close_output(fd_list)
external.close_ext()
if self.save:
OutputURL(util.ddfs_save(self.blobs, self.jobname, self.master))
Message("Results pushed to DDFS")
else:
index, index_url = self.reduce_index
safe_update(index, {'%d %s' % (self.id, out_url): True})
OutputURL(index_url)
def _run(self):
entries = self.track_status(self, "%s entries reduced")
red_out, out_url, fd_list = self.connect_output()
params = self.params
if self.ext_reduce:
external.prepare(self.reduce, self.ext_params, self.path('ext.reduce'))
self.reduce = FunctionType(external.ext_reduce.func_code,
globals=external.__dict__)
self.insert_globals([self.reduce])
total_size = sum(size for fd, size, url in self.connected_inputs)
Message("Input is %s" % (util.format_size(total_size)))
self.init(entries, params)
self.reduce(entries, red_out, params)
self.close_output(fd_list)
external.close_ext()
if self.save:
OutputURL(util.ddfs_save(self.blobs, self.jobname, self.master))
Message("Results pushed to DDFS")
else:
index, index_url = self.reduce_index
safe_update(index, ['%d %s' % (self.id, out_url)])
OutputURL(index_url)
def op_map(job):
msg("Received a new map job!")
if len(Task.inputs) != 1:
err("Map can only handle one input. Got: %s" %
" ".join(Task.inputs))
global fun_reader, fun_writer, fun_partition
fun_reader = util.unpack(job['map_reader'], globals=globals())
fun_writer = util.unpack(job['map_writer'], globals=globals())
fun_partition = util.unpack(job['partition'], globals=globals())
global fun_init
if 'map_init' in job:
fun_init = util.unpack(job['map_init'], globals=globals())
global fun_map
if 'ext_map' in job:
if 'ext_params' in job:
map_params = job['ext_params']
else:
map_params = "0\n"
path = Task.path("EXT_MAP")
external.prepare(job['ext_map'], map_params, path)
fun_map = external.ext_map
else:
map_params = util.unpack(job['params'], globals=globals())
fun_map = util.unpack(job['map'], globals=globals())
global fun_combiner
if 'combiner' in job:
fun_combiner = util.unpack(job['combiner'], globals=globals())
init_common(job)
nr_part = max(1, Task.num_partitions)
if 'combiner' in job:
partitions = [MapOutput(i, map_params, fun_combiner)\
for i in range(nr_part)]
else:
partitions = [MapOutput(i, map_params) for i in range(nr_part)]
run_map(Task.inputs[0], partitions, map_params)
external.close_ext()
urls = {}
for i, p in enumerate(partitions):
p.close()
urls["%d %s" % (i, p.url())] = True
index, index_url = Task.map_index
safe_update(index, urls)
OutputURL(index_url)
def op_reduce(job):
msg("Received a new reduce job!")
do_sort = int(job['sort'])
mem_sort_limit = int(job['mem_sort_limit'])
global fun_init
if 'reduce_init' in job:
fun_init = util.unpack(job['reduce_init'], globals=globals())
global fun_reader, fun_writer
fun_reader = util.unpack(job['reduce_reader'], globals=globals())
fun_writer = util.unpack(job['reduce_writer'], globals=globals())
global fun_reduce
if 'ext_reduce' in job:
if "ext_params" in job:
red_params = job['ext_params']
else:
red_params = "0\n"
path = Task.path("EXT_MAP")
external.prepare(job['ext_reduce'], red_params, path)
fun_reduce = external.ext_reduce
else:
fun_reduce = util.unpack(job['reduce'], globals=globals())
red_params = util.unpack(job['params'], globals=globals())
init_common(job)
red_in = ReduceReader(Task.inputs, do_sort,
mem_sort_limit, red_params).iter()
red_out = ReduceOutput(red_params)
msg("Starting reduce")
fun_init(red_in, red_params)
fun_reduce(red_in, red_out, red_params)
msg("Reduce done")
red_out.close()
external.close_ext()
index, index_url = Task.reduce_index
safe_update(index, {"%d %s" % (Task.id, red_out.url()): True})
OutputURL(index_url)
def _run(self):
if len(self.inputs) != 1:
TaskFailed("Map can only handle one input. Got: %s" % ' '.join(self.inputs))
if self.ext_map:
external.prepare(self.map, self.ext_params, self.path('EXT_MAP'))
self.map = FunctionType(external.ext_map.func_code,
globals=external.__dict__)
self.insert_globals([self.map])
partitions = [MapOutput(self, i) for i in xrange(self.num_partitions)]
reader, sze, url = self.connect_input(self.inputs[0])
params = self.params
self.init(reader, params)
entries = (self.map(entry, params) for entry in reader)
for kvs in self.track_status(entries, "%s entries mapped"):
for k, v in kvs:
p = self.partition(k, self.num_partitions, params)
partitions[p].add(k, v)
external.close_ext()
urls = {}
for i, partition in enumerate(partitions):
partition.close()
urls['%d %s' % (i, partition.url)] = True
index, index_url = self.map_index
safe_update(index, urls)
if self.save and not self.reduce:
if self.ispartitioned:
TaskFailed("Storing partitioned outputs in DDFS is not yet supported")
else:
OutputURL(util.ddfs_save(self.blobs, self.jobname, self.master))
Message("Results pushed to DDFS")
else:
OutputURL(index_url)
def _run(self):
if len(self.inputs) != 1:
TaskFailed("Map takes 1 input, got: %s" % ' '.join(self.inputs))
if self.save and not self.reduce and self.ispartitioned:
TaskFailed("Storing partitioned outputs in DDFS is not yet supported")
if self.ext_map:
external.prepare(self.map, self.ext_params, self.path('ext.map'))
self.map = FunctionType(external.ext_map.func_code,
globals=external.__dict__)
self.insert_globals([self.map])
entries = self.track_status(self, "%s entries mapped")
params = self.params
outputs = [MapOutput(self, i)
for i in xrange(max(1, int(self.jobdict['partitions'])))]
self.init(entries, params)
for entry in entries:
for k, v in self.map(entry, params):
outputs[self.partition(k, len(outputs), params)].add(k, v)
external.close_ext()
index, index_url = self.map_index
for output in outputs:
output.close()
safe_update(index, ['%d %s' % (i, output.url)
for i, output in enumerate(outputs)])
if self.save and not self.reduce:
OutputURL(util.ddfs_save(self.blobs, self.jobname, self.master))
Message("Results pushed to DDFS")
else:
OutputURL(index_url)