def decode_metadata_response(cls, data):
"""
Decode bytes to a MetadataResponse
Params
======
data: bytes to decode
"""
((correlation_id, numBrokers), cur) = relative_unpack('>ii', data, 0)
# Broker info
brokers = {}
for i in range(numBrokers):
((nodeId, ), cur) = relative_unpack('>i', data, cur)
(host, cur) = read_short_string(data, cur)
((port,), cur) = relative_unpack('>i', data, cur)
brokers[nodeId] = BrokerMetadata(nodeId, host, port)
# Topic info
((num_topics,), cur) = relative_unpack('>i', data, cur)
topicMetadata = {}
for i in range(num_topics):
((topicError,), cur) = relative_unpack('>h', data, cur)
(topicName, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
partitionMetadata = {}
for j in range(num_partitions):
((partitionErrorCode, partition, leader, numReplicas), cur) = relative_unpack('>hiii', data, cur)
(replicas, cur) = relative_unpack('>%di' % numReplicas, data, cur)
((numIsr,), cur) = relative_unpack('>i', data, cur)
(isr, cur) = relative_unpack('>%di' % numIsr, data, cur)
partitionMetadata[partition] = PartitionMetadata(topicName, partition, leader, replicas, isr)
topicMetadata[topicName] = partitionMetadata
return (brokers, topicMetadata)
def _decode_message(cls, data, offset):
"""
Decode a single Message
The only caller of this method is decode_message_set_iter.
They are decoupled to support nested messages (compressed MessageSets).
The offset is actually read from decode_message_set_iter (it is part
of the MessageSet payload).
"""
if sys.version > '3':
((crc, magic, att), cur) = relative_unpack('>IBB', data, 0)
else:
((crc, magic, att), cur) = relative_unpack('>iBB', data, 0)
if crc != zlib.crc32(data[4:]):
raise ChecksumError("Message checksum failed")
(key, cur) = read_int_string(data, cur)
(value, cur) = read_int_string(data, cur)
codec = att & ATTRIBUTE_CODEC_MASK
if codec == CODEC_NONE:
yield (offset, Message(magic, att, key, value))
elif codec == CODEC_GZIP:
gz = gzip_decode(value)
for (offset, msg) in KafkaProtocol._decode_message_set_iter(gz):
yield (offset, msg)
elif codec == CODEC_SNAPPY:
snp = snappy_decode(value)
for (offset, msg) in KafkaProtocol._decode_message_set_iter(snp):
yield (offset, msg)
def decode_consumer_metadata_response(cls, data):
"""
Decode bytes to a ConsumerMetadataResponse
Arguments:
data: bytes to decode
"""
((correlation_id, error, nodeId), cur) = relative_unpack('>ihi', data, 0)
(host, cur) = read_short_string(data, cur)
((port,), cur) = relative_unpack('>i', data, cur)
return ConsumerMetadataResponse(error, nodeId, host, port)
def decode_metadata_response(cls, data):
"""
Decode bytes to a MetadataResponse
Params
======
data: bytes to decode
"""
((correlation_id, numbrokers), cur) = relative_unpack('>ii', data, 0)
# Broker info
brokers = {}
for i in range(numbrokers):
((nodeId, ), cur) = relative_unpack('>i', data, cur)
(host, cur) = read_short_string(data, cur)
((port,), cur) = relative_unpack('>i', data, cur)
brokers[nodeId] = BrokerMetadata(nodeId, host, port)
# Topic info
((num_topics,), cur) = relative_unpack('>i', data, cur)
topic_metadata = {}
for i in range(num_topics):
# NOTE: topic_error is discarded. Should probably be returned with
# the topic metadata.
((topic_error,), cur) = relative_unpack('>h', data, cur)
(topic_name, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
partition_metadata = {}
for j in range(num_partitions):
# NOTE: partition_error_code is discarded. Should probably be
# returned with the partition metadata.
((partition_error_code, partition, leader, numReplicas), cur) = \
relative_unpack('>hiii', data, cur)
(replicas, cur) = relative_unpack(
'>%di' % numReplicas, data, cur)
((num_isr,), cur) = relative_unpack('>i', data, cur)
(isr, cur) = relative_unpack('>%di' % num_isr, data, cur)
partition_metadata[partition] = \
PartitionMetadata(
topic_name, partition, leader, replicas, isr)
topic_metadata[topic_name] = partition_metadata
return brokers, topic_metadata
def decode_metadata_response(cls, data):
"""
Decode bytes to a MetadataResponse
Params
======
data: bytes to decode
"""
((correlation_id, numbrokers), cur) = relative_unpack('>ii', data, 0)
# Broker info
brokers = []
for i in range(numbrokers):
((nodeId, ), cur) = relative_unpack('>i', data, cur)
(host, cur) = read_short_string(data, cur)
((port,), cur) = relative_unpack('>i', data, cur)
brokers.append(BrokerMetadata(nodeId, host, port))
# Topic info
((num_topics,), cur) = relative_unpack('>i', data, cur)
topic_metadata = []
for i in range(num_topics):
((topic_error,), cur) = relative_unpack('>h', data, cur)
(topic_name, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
partition_metadata = []
for j in range(num_partitions):
((partition_error_code, partition, leader, numReplicas), cur) = \
relative_unpack('>hiii', data, cur)
(replicas, cur) = relative_unpack(
'>%di' % numReplicas, data, cur)
((num_isr,), cur) = relative_unpack('>i', data, cur)
(isr, cur) = relative_unpack('>%di' % num_isr, data, cur)
partition_metadata.append(
PartitionMetadata(topic_name, partition, leader,
replicas, isr, partition_error_code)
)
topic_metadata.append(
TopicMetadata(topic_name, topic_error, partition_metadata)
)
return MetadataResponse(brokers, topic_metadata)
def _decode_message_set_iter(cls, data):
"""
Iteratively decode a MessageSet
Reads repeated elements of (offset, message), calling decode_message
to decode a single message. Since compressed messages contain futher
MessageSets, these two methods have been decoupled so that they may
recurse easily.
"""
cur = 0
read_message = False
while cur < len(data):
try:
((offset, ), cur) = relative_unpack('>q', data, cur)
(msg, cur) = read_int_string(data, cur)
for (offset, message) in KafkaProtocol._decode_message(msg, offset):
read_message = True
yield OffsetAndMessage(offset, message)
except BufferUnderflowError:
# NOTE: Not sure this is correct error handling:
# Is it possible to get a BUE if the message set is somewhere
# in the middle of the fetch response? If so, we probably have
# an issue that's not fetch size too small.
# Aren't we ignoring errors if we fail to unpack data by
# raising StopIteration()?
# If _decode_message() raises a ChecksumError, couldn't that
# also be due to the fetch size being too small?
if read_message is False:
# If we get a partial read of a message, but haven't
# yielded anything there's a problem
raise ConsumerFetchSizeTooSmall()
else:
raise StopIteration()
def _decode_message(cls, data, offset):
"""
Decode a single Message
The only caller of this method is decode_message_set_iter.
They are decoupled to support nested messages (compressed MessageSets).
The offset is actually read from decode_message_set_iter (it is part
of the MessageSet payload).
"""
((crc, magic, att), cur) = relative_unpack('>iBB', data, 0)
if crc != zlib.crc32(data[4:]):
raise ChecksumError("Message checksum failed")
(key, cur) = read_int_string(data, cur)
(value, cur) = read_int_string(data, cur)
codec = att & KafkaProtocol.ATTRIBUTE_CODEC_MASK
if codec == KafkaProtocol.CODEC_NONE:
yield (offset, Message(magic, att, key, value))
elif codec == KafkaProtocol.CODEC_GZIP:
gz = gzip_decode(value)
for (offset, msg) in KafkaProtocol._decode_message_set_iter(gz):
yield (offset, msg)
elif codec == KafkaProtocol.CODEC_SNAPPY:
snp = snappy_decode(value)
for (offset, msg) in KafkaProtocol._decode_message_set_iter(snp):
yield (offset, msg)
def _decode_message_set_iter(cls, data):
"""
Iteratively decode a MessageSet
Reads repeated elements of (offset, message), calling decode_message
to decode a single message. Since compressed messages contain futher
MessageSets, these two methods have been decoupled so that they may
recurse easily.
"""
cur = 0
read_message = False
while cur < len(data):
try:
((offset, ), cur) = relative_unpack('>q', data, cur)
(msg, cur) = read_int_string(data, cur)
for (offset, message) in KafkaProtocol._decode_message(msg, offset):
read_message = True
yield OffsetAndMessage(offset, message)
except BufferUnderflowError:
if read_message is False:
# If we get a partial read of a message, but haven't
# yielded anyhting there's a problem
raise ConsumerFetchSizeTooSmall()
else:
raise StopIteration()
def _decode_message_set_iter(cls, data):
"""
Iteratively decode a MessageSet
Reads repeated elements of (offset, message), calling decode_message
to decode a single message. Since compressed messages contain futher
MessageSets, these two methods have been decoupled so that they may
recurse easily.
"""
cur = 0
read_message = False
while cur < len(data):
try:
((offset, ), cur) = relative_unpack('>q', data, cur)
(msg, cur) = read_int_string(data, cur)
for (offset,
message) in KafkaProtocol._decode_message(msg, offset):
read_message = True
yield OffsetAndMessage(offset, message)
except BufferUnderflowError:
if read_message is False:
# If we get a partial read of a message, but haven't yielded anyhting
# there's a problem
raise ConsumerFetchSizeTooSmall()
else:
raise StopIteration()
def _decode_message_set_iter(cls, data):
"""
Iteratively decode a MessageSet
Reads repeated elements of (offset, message), calling decode_message
to decode a single message. Since compressed messages contain futher
MessageSets, these two methods have been decoupled so that they may
recurse easily.
"""
cur = 0
read_message = False
while cur < len(data):
try:
((offset, ), cur) = relative_unpack('>q', data, cur)
(msg, cur) = read_int_string(data, cur)
for (offset,
message) in KafkaProtocol._decode_message(msg, offset):
read_message = True
yield OffsetAndMessage(offset, message)
except BufferUnderflowError:
# NOTE: Not sure this is correct error handling:
# Is it possible to get a BUE if the message set is somewhere
# in the middle of the fetch response? If so, we probably have
# an issue that's not fetch size too small.
# Aren't we ignoring errors if we fail to unpack data by
# raising StopIteration()?
# If _decode_message() raises a ChecksumError, couldn't that
# also be due to the fetch size being too small?
if read_message is False:
# If we get a partial read of a message, but haven't
# yielded anything there's a problem
raise ConsumerFetchSizeTooSmall()
else:
raise StopIteration()
def _decode_message(cls, data, offset):
"""
Decode a single Message
The only caller of this method is decode_message_set_iter.
They are decoupled to support nested messages (compressed MessageSets).
The offset is actually read from decode_message_set_iter (it is part
of the MessageSet payload).
"""
((crc, magic, att), cur) = relative_unpack('>iBB', data, 0)
if crc != zlib.crc32(data[4:]):
raise ChecksumError("Message checksum failed")
(key, cur) = read_int_string(data, cur)
(value, cur) = read_int_string(data, cur)
codec = att & ATTRIBUTE_CODEC_MASK
if codec == CODEC_NONE.mask:
yield (offset, Message(magic, att, key, value))
else:
decoders = filter(lambda c: c.mask == codec, ALL_CODECS)
if decoders:
message_set = decoders[0].decode(value)
for (offset, msg) in KafkaProtocol._decode_message_set_iter(message_set):
yield (offset, msg)
else:
raise UnsupportedCodecError('FILL ME IN')
def decode_metadata_response(cls, data):
"""
Decode bytes to a MetadataResponse
Arguments:
data: bytes to decode
"""
((correlation_id, numbrokers), cur) = relative_unpack('>ii', data, 0)
# Broker info
brokers = []
for i in range(numbrokers):
((nodeId, ), cur) = relative_unpack('>i', data, cur)
(host, cur) = read_short_string(data, cur)
((port,), cur) = relative_unpack('>i', data, cur)
brokers.append(BrokerMetadata(nodeId, host, port))
# Topic info
((num_topics,), cur) = relative_unpack('>i', data, cur)
topic_metadata = []
for i in range(num_topics):
((topic_error,), cur) = relative_unpack('>h', data, cur)
(topic_name, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
partition_metadata = []
for j in range(num_partitions):
((partition_error_code, partition, leader, numReplicas), cur) = \
relative_unpack('>hiii', data, cur)
(replicas, cur) = relative_unpack(
'>%di' % numReplicas, data, cur)
((num_isr,), cur) = relative_unpack('>i', data, cur)
(isr, cur) = relative_unpack('>%di' % num_isr, data, cur)
partition_metadata.append(
PartitionMetadata(topic_name, partition, leader,
replicas, isr, partition_error_code)
)
topic_metadata.append(
TopicMetadata(topic_name, topic_error, partition_metadata)
)
return MetadataResponse(brokers, topic_metadata)
def decode_offset_commit_response(cls, data):
"""
Decode bytes to an OffsetCommitResponse
Arguments:
data: bytes to decode
"""
((correlation_id,), cur) = relative_unpack('>i', data, 0)
((num_topics,), cur) = relative_unpack('>i', data, cur)
for i in xrange(num_topics):
(topic, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
for i in xrange(num_partitions):
((partition, error), cur) = relative_unpack('>ih', data, cur)
yield OffsetCommitResponse(topic, partition, error)
def decode_fetch_response(cls, data):
"""
Decode bytes to a FetchResponse
Params
======
data: bytes to decode
"""
((correlation_id, num_topics), cur) = relative_unpack('>ii', data, 0)
for i in range(num_topics):
(topic, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
for i in range(num_partitions):
((partition, error, highwater_mark_offset), cur) = relative_unpack('>ihq', data, cur)
(message_set, cur) = read_int_string(data, cur)
yield FetchResponse(topic, partition, error, highwater_mark_offset,
KafkaProtocol._decode_message_set_iter(message_set))
def parse_consumer_offset_message(self, message):
key = message.key
((key_schema,), cur) = relative_unpack(b'>h', key, 0)
if key_schema not in [0, 1]:
raise InvalidMessageException() # This is not an offset commit message
(group, cur) = read_short_string(key, cur)
(topic, cur) = read_short_string(key, cur)
((partition,), cur) = relative_unpack(b'>l', key, cur)
if message.value:
value = message.value
((value_schema,), cur) = relative_unpack(b'>h', value, 0)
if value_schema not in [0, 1]:
raise InvalidMessageException() # Unrecognized message value
((offset,), cur) = relative_unpack(b'>q', value, cur)
else:
offset = None # Offset was deleted
return group.decode(), topic.decode(), partition, offset
def parse_consumer_offset_message(self, message):
key = bytearray(message.key)
((key_schema,), cur) = relative_unpack('>h', key, 0)
if key_schema not in [0, 1]:
raise InvalidMessageException() # This is not an offset commit message
(group, cur) = read_short_string(key, cur)
(topic, cur) = read_short_string(key, cur)
((partition,), cur) = relative_unpack('>l', key, cur)
if message.value:
value = bytearray(message.value)
((value_schema,), cur) = relative_unpack('>h', value, 0)
if value_schema not in [0, 1]:
raise InvalidMessageException() # Unrecognized message value
((offset,), cur) = relative_unpack('>q', value, cur)
else:
offset = None # Offset was deleted
return str(group), str(topic), partition, offset
def decode_produce_response(cls, data):
"""
Decode bytes to a ProduceResponse
Params
======
data: bytes to decode
"""
((correlation_id, num_topics), cur) = relative_unpack('>ii', data, 0)
for i in range(num_topics):
((strlen,), cur) = relative_unpack('>h', data, cur)
topic = data[cur:cur+strlen]
cur += strlen
((num_partitions,), cur) = relative_unpack('>i', data, cur)
for i in range(num_partitions):
((partition, error, offset), cur) = relative_unpack('>ihq', data, cur)
yield ProduceResponse(topic, partition, error, offset)
def decode_offset_commit_response(cls, data):
"""
Decode bytes to an OffsetCommitResponse
Params
======
data: bytes to decode
"""
((correlation_id,), cur) = relative_unpack('>i', data, 0)
(client_id, cur) = read_short_string(data, cur)
((num_topics,), cur) = relative_unpack('>i', data, cur)
for i in xrange(num_topics):
(topic, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
for i in xrange(num_partitions):
((partition, error), cur) = relative_unpack('>ih', data, cur)
yield OffsetCommitResponse(topic, partition, error)
def decode_metadata_response(cls, data):
"""
Decode bytes to a MetadataResponse
Params
======
data: bytes to decode
"""
((correlation_id, numBrokers), cur) = relative_unpack('>ii', data, 0)
# Broker info
brokers = {}
for i in range(numBrokers):
((nodeId, ), cur) = relative_unpack('>i', data, cur)
(host, cur) = read_short_string(data, cur)
((port, ), cur) = relative_unpack('>i', data, cur)
brokers[nodeId] = BrokerMetadata(nodeId, host, port)
# Topic info
((num_topics, ), cur) = relative_unpack('>i', data, cur)
topicMetadata = {}
for i in range(num_topics):
((topicError, ), cur) = relative_unpack('>h', data, cur)
(topicName, cur) = read_short_string(data, cur)
((num_partitions, ), cur) = relative_unpack('>i', data, cur)
partitionMetadata = {}
for j in range(num_partitions):
((partitionErrorCode, partition, leader, numReplicas), cur) = \
relative_unpack('>hiii', data, cur)
(replicas, cur) = relative_unpack('>%di' % numReplicas, data,
cur)
((numIsr, ), cur) = relative_unpack('>i', data, cur)
(isr, cur) = relative_unpack('>%di' % numIsr, data, cur)
partitionMetadata[partition] = \
PartitionMetadata(topicName, partition, leader,
replicas, isr)
topicMetadata[topicName] = partitionMetadata
return (brokers, topicMetadata)
def decode_offset_response(cls, data):
"""
Decode bytes to an OffsetResponse
Params
======
data: bytes to decode
"""
((correlation_id, num_topics), cur) = relative_unpack('>ii', data, 0)
for i in range(num_topics):
(topic, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
for i in range(num_partitions):
((partition, error, num_offsets,), cur) = relative_unpack('>ihi', data, cur)
offsets = []
for j in range(num_offsets):
((offset,), cur) = relative_unpack('>q', data, cur)
offsets.append(offset)
yield OffsetResponse(topic, partition, error, tuple(offsets))
def decode_produce_response(cls, data):
"""
Decode bytes to a ProduceResponse
Params
======
data: bytes to decode
"""
((correlation_id, num_topics), cur) = relative_unpack('>ii', data, 0)
for i in range(num_topics):
((strlen, ), cur) = relative_unpack('>h', data, cur)
topic = data[cur:cur + strlen]
cur += strlen
((num_partitions, ), cur) = relative_unpack('>i', data, cur)
for i in range(num_partitions):
((partition, error, offset),
cur) = relative_unpack('>ihq', data, cur)
yield ProduceResponse(topic, partition, error, offset)
def decode_offset_fetch_response(cls, data):
"""
Decode bytes to an OffsetFetchResponse
Params
======
data: bytes to decode
"""
((correlation_id,), cur) = relative_unpack('>i', data, 0)
(client_id, cur) = read_short_string(data, cur)
((num_topics,), cur) = relative_unpack('>i', data, cur)
for i in range(num_topics):
(topic, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
for i in range(num_partitions):
((partition, offset), cur) = relative_unpack('>iq', data, cur)
(metadata, cur) = read_short_string(data, cur)
((error,), cur) = relative_unpack('>h', data, cur)
yield OffsetFetchResponse(topic, partition, offset, metadata, error)
def decode_offset_fetch_response(cls, data):
"""
Decode bytes to an OffsetFetchResponse
Arguments:
data: bytes to decode
"""
((correlation_id,), cur) = relative_unpack('>i', data, 0)
((num_topics,), cur) = relative_unpack('>i', data, cur)
for i in range(num_topics):
(topic, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
for i in range(num_partitions):
((partition, offset), cur) = relative_unpack('>iq', data, cur)
(metadata, cur) = read_short_string(data, cur)
((error,), cur) = relative_unpack('>h', data, cur)
yield OffsetFetchResponse(topic, partition, offset,
metadata, error)
def decode_fetch_response(cls, data):
"""
Decode bytes to a FetchResponse
Arguments:
data: bytes to decode
"""
((correlation_id, num_topics), cur) = relative_unpack('>ii', data, 0)
for _ in range(num_topics):
(topic, cur) = read_short_string(data, cur)
((num_partitions, ), cur) = relative_unpack('>i', data, cur)
for j in range(num_partitions):
((partition, error, highwater_mark_offset), cur) = \
relative_unpack('>ihq', data, cur)
(message_set, cur) = read_int_string(data, cur)
yield FetchResponse(
topic, partition, error, highwater_mark_offset,
KafkaProtocol._decode_message_set_iter(message_set))
def decode_offset_response(cls, data):
"""
Decode bytes to an OffsetResponse
Arguments:
data: bytes to decode
"""
((correlation_id, num_topics), cur) = relative_unpack('>ii', data, 0)
for i in range(num_topics):
(topic, cur) = read_short_string(data, cur)
((num_partitions,), cur) = relative_unpack('>i', data, cur)
for i in range(num_partitions):
((partition, error, num_offsets,), cur) = \
relative_unpack('>ihi', data, cur)
offsets = []
for j in range(num_offsets):
((offset,), cur) = relative_unpack('>q', data, cur)
offsets.append(offset)
yield OffsetResponse(topic, partition, error, tuple(offsets))
def _decode_message_set_iter(cls, data):
"""
Iteratively decode a MessageSet
Reads repeated elements of (offset, message), calling decode_message to decode a
single message. Since compressed messages contain futher MessageSets, these two methods
have been decoupled so that they may recurse easily.
"""
cur = 0
while cur < len(data):
try:
((offset, ), cur) = relative_unpack('>q', data, cur)
(msg, cur) = read_int_string(data, cur)
for (offset, message) in KafkaProtocol._decode_message(msg, offset):
yield OffsetAndMessage(offset, message)
except BufferUnderflowError: # If we get a partial read of a message, stop
raise StopIteration()
def _decode_message_set_iter(cls, data):
"""
Iteratively decode a MessageSet
Reads repeated elements of (offset, message), calling decode_message
to decode a single message. Since compressed messages contain futher
MessageSets, these two methods have been decoupled so that they may
recurse easily.
"""
cur = 0
while cur < len(data):
try:
((offset, ), cur) = relative_unpack('>q', data, cur)
(msg, cur) = read_int_string(data, cur)
for (offset, message) in KafkaProtocol._decode_message(msg,
offset):
yield OffsetAndMessage(offset, message)
except BufferUnderflowError:
# If we get a partial read of a message, stop
raise StopIteration()
