You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Because the hop window duplicates the rows and derives stream key with (input_pk, window start). It should not emit the rows with null time value which will generate the same window start value(null!) which can generate duplicate stream key.
The text was updated successfully, but these errors were encountered:
May need to go through all the table functions once? And udtf in the future
only for those relational operators which could duplicate multiple rows from one input row. I have thought about other operators(Expand, ProjectSet) and do not find the same issue.
Also, this bug reminds us a too-flexible UDTF is dangerous in our system. I think we can provide a UDTF that can only accept append-only stream(relation) and output append-only stream(relation) so that we can generate row_id for each row.
Because the hop window duplicates the rows and derives stream key with (input_pk, window start). It should not emit the rows with null time value which will generate the same window start value(null!) which can generate duplicate stream key.
The text was updated successfully, but these errors were encountered: