def testGetChunks(self):
self.assertEqual([[("a", "a")]], list(get_chunks([("a", "a")], 2**16)))
self.assertEqual([[("a", "a")]], list(get_chunks([("a", "a")], 1)))
a_100 = "a" * 100
self.assertEqual([[("a", a_100)],
[("a", a_100)],],
list(get_chunks([("a", a_100), ("a", a_100)], 150)))
self.assertEqual([[("a", a_100), ("a", a_100)],
[("a", a_100), ("a", a_100)]],
list(get_chunks([("a", a_100), ("a", a_100),
("a", a_100), ("a", a_100)], 300)))
def testGetChunks(self):
self.assertEqual([[("a", "a")]], list(get_chunks([("a", "a")], 2**16)))
self.assertEqual([[("a", "a")]], list(get_chunks([("a", "a")], 1)))
a_100 = "a" * 100
self.assertEqual([
[("a", a_100)],
[("a", a_100)],
], list(get_chunks([("a", a_100), ("a", a_100)], 150)))
self.assertEqual([[("a", a_100),
("a", a_100)], [("a", a_100), ("a", a_100)]],
list(
get_chunks([("a", a_100), ("a", a_100),
("a", a_100), ("a", a_100)], 300)))
def commit_to_queue_(self):
"""Adds all pending changes to the task queues for async commits
:return: Yields all shard names that need to be updated.
"""
pull = taskqueue.Queue('recordio-queue')
rpcs = []
key_values_not_added = RecordIORecords()
for shard_name, key_values in RecordIOShard.get_shards_for_key_values(
self.name, self.updates):
self.db_search += 1
if shard_name == None:
for entry in key_values:
key_values_not_added.insert(entry)
else:
for key_values_chunk in get_chunks(key_values,
MAX_TASKQUEUE_BATCH_SIZE):
payload = marshal.dumps(key_values_chunk, MARSHAL_VERSION)
rpc = pull.add_async(
taskqueue.Task(payload=payload,
method='PULL',
tag=shard_name))
rpcs.append((rpc, key_values_chunk, shard_name))
for rpc, key_values, shard_name in rpcs:
try:
rpc.get_result()
yield shard_name
except:
for entry in key_values:
key_values_not_added.insert(entry)
self.updates = key_values_not_added
if len(self.updates):
raise RecordIOWriterNotCompletedError(len(self.updates))
def commit_to_queue_(self):
"""Adds all pending changes to the task queues for async commits
:return: Yields all shard names that need to be updated.
"""
pull = taskqueue.Queue('recordio-queue')
rpcs = []
key_values_not_added = RecordIORecords()
for shard_name, key_values in RecordIOShard.get_shards_for_key_values(
self.name, self.updates):
self.db_search += 1
if shard_name == None:
for entry in key_values:
key_values_not_added.insert(entry)
else:
for key_values_chunk in get_chunks(key_values, MAX_TASKQUEUE_BATCH_SIZE):
payload = marshal.dumps(key_values_chunk, MARSHAL_VERSION)
rpc = pull.add_async(taskqueue.Task(payload=payload, method='PULL',
tag=shard_name))
rpcs.append((rpc, key_values_chunk, shard_name))
for rpc, key_values, shard_name in rpcs:
try:
rpc.get_result()
yield shard_name
except:
for entry in key_values:
key_values_not_added.insert(entry)
self.updates = key_values_not_added
if len(self.updates):
raise RecordIOWriterNotCompletedError(len(self.updates))
def get_zipped_chunks_(self, force_repackiging=False):
"""Returns all zipped chunks for this RecordIORecordsZipped.
:param force_repackiging: If the zip chunks should be unzipped and rezipped
to be more even.
:return: list of zipped chunks: (lo_entry, hi_entry, zipped_entries).
lo_entry and hi_entry ARE INCLUSIVE!
"""
if self.records_ or force_repackiging:
for records in recordio_chunks.get_chunks(self, ZIP_CHUNKS):
yield (records[0][:-1], records[-1][:-1],
zlib.compress(marshal.dumps(records, MARSHAL_VERSION),
COMPRESSION_LEVEL))
else:
for x in self.zipped_chunks_:
yield x
def get_zipped_chunks_(self, force_repackiging=False):
"""Returns all zipped chunks for this RecordIORecordsZipped.
:param force_repackiging: If the zip chunks should be unzipped and rezipped
to be more even.
:return: list of zipped chunks: (lo_entry, hi_entry, zipped_entries).
lo_entry and hi_entry ARE INCLUSIVE!
"""
if self.records_ or force_repackiging:
for records in recordio_chunks.get_chunks(self, ZIP_CHUNKS):
yield (records[0][:-1], records[-1][:-1],
zlib.compress(marshal.dumps(records, MARSHAL_VERSION),
COMPRESSION_LEVEL))
else:
for x in self.zipped_chunks_:
yield x
def commit_sync(self, retries=32, retry_timeout=1):
"""Applies all changes synchronously to the RecordIO.
:param retries: How many times a commit_sync should be retried in case of
datastore collisions.
:param retry_timeout: The amount of second to wait before the next retry.
"""
if not len(self.updates):
return
for attempt in range(retries + 1):
shard_does_not_exist = RecordIORecords()
for shard_name, key_values in RecordIOShard.get_shards_for_key_values(
self.name, self.updates):
self.db_search += 1
if shard_name == None and key_values:
logging.debug(
"RecordIO %s: No shard found for:\n%s -> %s" %
(self.name,
SPLIT_CHAR.join(RecordIOShard.entry_key(
key_values[0])), key_values[0][:-1]))
for entry in key_values:
shard_does_not_exist.insert(entry)
else:
lo_just_split = None
hi_just_split = None
for key_values_chunk in get_chunks(key_values,
MAX_WRITE_BATCH_SIZE):
if lo_just_split and hi_just_split and key_values_chunk:
if RecordIORecords.in_range(key_values_chunk[0],
lo=lo_just_split[0],
hi=lo_just_split[1]):
shard_name = RecordIOShard.key_name(
self.name,
lo=lo_just_split[0],
hi=lo_just_split[1])
elif RecordIORecords.in_range(key_values_chunk[0],
lo=hi_just_split[0],
hi=hi_just_split[1]):
shard_name = RecordIOShard.key_name(
self.name,
lo=hi_just_split[0],
hi=hi_just_split[1])
not_deleted = None
try:
not_deleted, lo_just_split, hi_just_split = self.commit_shard_(
shard_name, key_values_chunk)
except RecordIOShardDoesNotExistError:
logging.debug("Shard does not exist:\n" +
shard_name)
lo_just_split = None
hi_just_split = None
for entry in key_values_chunk:
shard_does_not_exist.insert(entry)
if not_deleted:
for to_delete_shard_name, to_delete_key_values in (
RecordIOShard.get_shards_for_key_values(
self.name, not_deleted)):
self.db_search += 1
try:
self.commit_shard_(to_delete_shard_name,
to_delete_key_values)
except RecordIOShardDoesNotExistError:
logging.debug("Shard does not exist:\n" +
shard_name)
for entry in to_delete_key_values:
shard_does_not_exist.insert(entry)
self.updates = shard_does_not_exist
if len(self.updates):
if attempt == retries:
raise RecordIOWriterNotCompletedError(len(self.updates))
else:
logging.debug("Commit attempt %d failed" % attempt)
time.sleep(retry_timeout)
else:
return
def commit_sync(self, retries=32, retry_timeout=1):
"""Applies all changes synchronously to the RecordIO.
:param retries: How many times a commit_sync should be retried in case of
datastore collisions.
:param retry_timeout: The amount of second to wait before the next retry.
"""
if not len(self.updates):
return
for attempt in range(retries + 1):
shard_does_not_exist = RecordIORecords()
for shard_name, key_values in RecordIOShard.get_shards_for_key_values(
self.name, self.updates):
self.db_search += 1
if shard_name == None and key_values:
logging.debug("RecordIO %s: No shard found for:\n%s -> %s" %
(self.name,
SPLIT_CHAR.join(RecordIOShard.entry_key(key_values[0])),
key_values[0][:-1]))
for entry in key_values:
shard_does_not_exist.insert(entry)
else:
lo_just_split = None
hi_just_split = None
for key_values_chunk in get_chunks(key_values, MAX_WRITE_BATCH_SIZE):
if lo_just_split and hi_just_split and key_values_chunk:
if RecordIORecords.in_range(key_values_chunk[0],
lo=lo_just_split[0],
hi=lo_just_split[1]):
shard_name = RecordIOShard.key_name(self.name,
lo=lo_just_split[0],
hi=lo_just_split[1])
elif RecordIORecords.in_range(key_values_chunk[0],
lo=hi_just_split[0],
hi=hi_just_split[1]):
shard_name = RecordIOShard.key_name(self.name,
lo=hi_just_split[0],
hi=hi_just_split[1])
not_deleted = None
try:
not_deleted, lo_just_split, hi_just_split = self.commit_shard_(
shard_name, key_values_chunk)
except RecordIOShardDoesNotExistError:
logging.debug("Shard does not exist:\n" + shard_name)
lo_just_split = None
hi_just_split = None
for entry in key_values_chunk:
shard_does_not_exist.insert(entry)
if not_deleted:
for to_delete_shard_name, to_delete_key_values in (
RecordIOShard.get_shards_for_key_values(
self.name, not_deleted)):
self.db_search += 1
try:
self.commit_shard_(to_delete_shard_name, to_delete_key_values)
except RecordIOShardDoesNotExistError:
logging.debug("Shard does not exist:\n" + shard_name)
for entry in to_delete_key_values:
shard_does_not_exist.insert(entry)
self.updates = shard_does_not_exist
if len(self.updates):
if attempt == retries:
raise RecordIOWriterNotCompletedError(len(self.updates))
else:
logging.debug("Commit attempt %d failed" % attempt)
time.sleep(retry_timeout)
else:
return
