Zhenxiao Luo created PARQUET-74:
-----------------------------------
Summary: Use thread local decoder cache in Binary
toStringUsingUTF8()
Key: PARQUET-74
URL: https://issues.apache.org/jira/browse/PARQUET-74
Project: Parquet
Issue Type: Improvement
Components: parquet-mr
Reporter: Zhenxiao Luo
In Binary toStringUsingUTF8(), new String(...) is slower because it
instantiates a new Decoder, while Charset#decode uses a thread-local decoder
cache, it is much faster.
ByteArraySliceBackedBinary is using Charset#decode, while,
ByteArrayBackedBinary and ByteBufferBackedBinary are still using new String()
--
This message was sent by Atlassian JIRA
(v6.2#6252)
