Reduce allocations from Redis operations

[theigl]

RedisIndexedSessionRepository and RedisSessionExpirationPolicy create a lot of single-use proxies by relying on BoundHashOperations.

For instance in this case:

spring-session/spring-session-data-redis/src/main/java/org/springframework/session/data/redis/RedisIndexedSessionRepository.java

Lines 898 to 903 in f6f4c86

	private void saveDelta() {
	if (this.delta.isEmpty()) {
	return;
	}
	String sessionId = getId();
	getSessionBoundHashOperations(sessionId).putAll(this.delta);

spring-session/spring-session-data-redis/src/main/java/org/springframework/session/data/redis/RedisIndexedSessionRepository.java

Lines 733 to 735 in f6f4c86

	private BoundHashOperations<String, String, Object> getSessionBoundHashOperations(String sessionId) {
	String key = getSessionKey(sessionId);
	return this.sessionRedisOperations.boundHashOps(key);

This showed up in our production profiler because of hundreds of MB allocated in java.lang.reflect.Method:

If I understand the code correctly, this is done purely for convenience and could simply be replaced by the following code:

String key = getSessionKey(sessionId);
sessionRedisOperations.opsForHash().putAll(key, this.delta);

Or am I overlooking something?

[theigl]

@marcusdacoregio: Hi and sorry for pinging you directly.

I'd be willing to provide a PR for this if I get a preliminary OK on my suggestion. I think the performance optimization would be worth it.

[marcusdacoregio]

Hi @theigl, thanks for reaching out.

Unfortunately, I'm not in the project anymore so I won't be able to approve/merge your PR if submitted. I'll ping @rwinch here who might be able to guide you.

I'd say that the best thing to do is to create a copy implementation of RedisIndexedSessionRepository, apply your changes and measure it again, that will help the team analyze your suggestion better.