ad1happy2go commented on PR #19005:
URL: https://github.com/apache/hudi/pull/19005#issuecomment-4809054987

   Thanks @danny0405. My main point: **the table type shouldn't change the 
result of an incremental query** — COW vs MOR is a storage detail, not query 
semantics. Today they diverge: for the same data and the same incremental 
window, COW returns the correct fully-materialized rows while MOR returns wrong 
ones — only the changed columns for partial updates, and for 
`EVENT_TIME_ORDERING` a lower-ordering write can surface over the version that 
should win. That inconsistency is the bug this PR fixes.
   
   On the window concern: it isn't undefined — it's exactly the commit-time 
range of the incremental query (begin/end instants), the same window COW 
already uses. The PR doesn't merge "all history": it bounds the merge with an 
`InstantRange` to the window end and applies the commit-time predicate *after* 
the merge, so each record changed in the window comes back with its value as of 
the window end — identical to what COW returns for that same window. So I'd 
lean toward fixing MOR to match COW rather than steering users to COW, since 
incremental query is a supported read on MOR. Happy to walk through any 
specific scenario where the expected output seems ambiguous.
   


-- 
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]

Reply via email to