JingsongLi commented on PR #3264: URL: https://github.com/apache/paimon/pull/3264#issuecomment-2098132002
Hi @Aitozi , thanks for your contribution! Semantically speaking, supporting sequence in FIRST-ROW is very good, and it is this semantics. However, there are two issues: 1. Users will experience a significant performance gap, with significant differences in streaming read, streaming write, and batch reading, which can greatly affect tuning. Therefore, from this perspective, it is better to let them clearly use another approach. 2. There are many additional judgments in the code, and the previous assumptions about FISRT-ROW have been broken one by one, which will increase the complexity of the code. Here, I recommend another implementation method by introducing an option: `sequence.field.reverse`, flipping the comparison of sequences. The advantage of this is that it is relatively easy to modify. We can remind users of exception when using FIRST-ROW with sequence. If there are not many scenes, it is still recommended to use this low-cost approach, but if you think there are many scenes, we can tolerate these changes to FIRST-ROW. -- 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]
