Huang Kuan Hao created HDDS-16277:
-------------------------------------
Summary: Reuse CharsetEncoder/Decoder in StringCodec
Key: HDDS-16277
URL: https://issues.apache.org/jira/browse/HDDS-16277
Project: Apache Ozone
Issue Type: Sub-task
Reporter: Huang Kuan Hao
StringCodec is the key/value codec for many RocksDB tables (used across ~33
files). encode(...) allocates a fresh CharsetEncoder inside the returned
lambda, and decodeNoFallback/decodeWithFallback allocate a fresh CharsetDecoder
— on every encode/decode, i.e. per record on reads, writes and iteration:
final CoderResult result = newEncoder().encode(CharBuffer.wrap(string), buffer,
true);
...
return newDecoder().decode(buffer.asReadOnlyBuffer()).toString();
Fix: hold the coders in per-instance ThreadLocals (they are stateful and not
thread-safe), configured once, and reset() before each use:
private final ThreadLocal<CharsetEncoder> encoder =
ThreadLocal.withInitial(this::newEncoder);
private final ThreadLocal<CharsetDecoder> decoder =
ThreadLocal.withInitial(this::newDecoder);
...
final CoderResult result =
encoder.get().reset().encode(CharBuffer.wrap(string), buffer, true);
...
return decoder.get().reset().decode(buffer.asReadOnlyBuffer()).toString();
reset() clears only the coding state, not the configured
onMalformedInput/onUnmappableCharacter actions, so behavior is unchanged.
File: hadoop-hdds/common/.../hdds/utils/db/StringCodecBase.java
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
