diff options
author | Dana Powers <dana.powers@gmail.com> | 2017-10-05 14:19:52 -0700 |
---|---|---|
committer | GitHub <noreply@github.com> | 2017-10-05 14:19:52 -0700 |
commit | ffc7caef13a120f69788bcdd43ffa01468f575f9 (patch) | |
tree | 978b5a04e589c92124af9c5a0e32ccf24912e1c7 /kafka/consumer/fetcher.py | |
parent | cec1bdc9965b3d6729d4415e31b4dac04d603873 (diff) | |
download | kafka-python-ffc7caef13a120f69788bcdd43ffa01468f575f9.tar.gz |
Fix Fetcher.PartitionRecords to handle fetch_offset in the middle of compressed messageset (#1239)
Diffstat (limited to 'kafka/consumer/fetcher.py')
-rw-r--r-- | kafka/consumer/fetcher.py | 9 |
1 files changed, 7 insertions, 2 deletions
diff --git a/kafka/consumer/fetcher.py b/kafka/consumer/fetcher.py index b86c8ec..f552038 100644 --- a/kafka/consumer/fetcher.py +++ b/kafka/consumer/fetcher.py @@ -923,12 +923,17 @@ class Fetcher(six.Iterator): self._sensors.fetch_throttle_time_sensor.record(response.throttle_time_ms) self._sensors.fetch_latency.record((recv_time - send_time) * 1000) - class PartitionRecords(six.Iterator): + class PartitionRecords(object): def __init__(self, fetch_offset, tp, messages): self.fetch_offset = fetch_offset self.topic_partition = tp self.messages = messages - self.message_idx = 0 + # When fetching an offset that is in the middle of a + # compressed batch, we will get all messages in the batch. + # But we want to start 'take' at the fetch_offset + for i, msg in enumerate(messages): + if msg.offset == fetch_offset: + self.message_idx = i def discard(self): self.messages = None |