Details
-
Improvement
-
Status: Closed
-
Major
-
Resolution: Fixed
-
0.3.0
-
None
-
None
-
Reviewed
Description
LazySimpleSerDe currently serialize all data into a StringBuilder, and then convert it to String and then Text.
Even if the data is of type int/long/byte/short, we still do that unnecessary conversion.
We should directly serialize/append int/long/byte/short to a UTF-8 buffer.
This is a very simple change, but it is expected to save 2-3% of the time of a typical mapper (on a group-by query with some int/long columns), and this blocks HIVE-266.
Attachments
Attachments
Issue Links
- blocks
-
HIVE-266 Improve SerDe performance by using Text instead of String
- Closed