gustavodemorais commented on PR #27508: URL: https://github.com/apache/flink/pull/27508#issuecomment-3953444724
Hey @ldadima, I synced with Timo yesterday and we came up with the optimization suggestion he mentioned. I think we have two ways of optimizing this further 1 - > Instead of always creating a new serializer, the simplest and most consistent approach would be to normalize all state keys to BinaryRowData using pre-initialized RowDataSerializer instances in MultiJoinStateViews.java, before accessing state. Then you could pass down to MultiJoinStateViews the state backend type and only do the extra work if necessary, for the heap state backend. 2 - Consider adjusting RowDataSerializer to incorporate the version you wanted to use, where we do not reuse the variable. Let's ask Timo what he thinks about that -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
