Fix load large data duplicates #148 #154
Conversation
There was a problem hiding this comment.
It looks good to me.
But, why don't you change the commit massage to note the reason before merging? For example,
title
Fix duplication handling for large data set(#148)
message
BTReaderInit() assumed that the case (itup->t_tid).ip_posid = 0 was anomalous when searching for the leftmost leaf page of a b-tree, but this is an old condition and has been removed.
With the commit (postgres/postgres@8224de4) to support the INCLUDE clause in indexes from PostgreSQL v11, new pivot tuples (e.g. b-tree root and intermediate nodes) were added, which could cause ip_posid to be 0.
Because they had forgotten to address this, errors sometimes occurred when running the process on B-trees that held some large data with pivot tuples.
|
Hey Guys, is this PR, a change that addressed this issue by any chance? #148 |
|
I reread the mikecaat's comment and I see #148 there, sorry! I missed that before. |
No description provided.