Open shuai-xu opened 1 year ago
Can one of the admins verify this patch?
The main problem here is that, the 1st block comed into ColumnBuffer
could be non-nullable, but the later blocks come with nullable, and the real final schema shoul be nullable.
You can make an assumption that, if we meet a nullable column once, all results should be nullable. but we may have spill some blocks with non-nullable columns out into next stage before we meet the first block with nullable.
Changelog category (leave one):
Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):
This pr supports that when split blocks, they may have differnt nullable types as they may come from two different steams of union. It fix #436 .