]> granicus.if.org Git - postgresql/commit
Fix tuple routing in cases where tuple descriptors don't match.
authorRobert Haas <rhaas@postgresql.org>
Thu, 22 Dec 2016 22:31:52 +0000 (17:31 -0500)
committerRobert Haas <rhaas@postgresql.org>
Thu, 22 Dec 2016 22:36:37 +0000 (17:36 -0500)
commit2ac3ef7a01df859c62d0a02333b646d65eaec5ff
tree850532b8c6aac27df50cff9c45a5873b4c0da743
parent12bd7dd317e8f4346fb3507578aca790ede6ebea
Fix tuple routing in cases where tuple descriptors don't match.

The previous coding failed to work correctly when we have a
multi-level partitioned hierarchy where tables at successive levels
have different attribute numbers for the partition key attributes.  To
fix, have each PartitionDispatch object store a standalone
TupleTableSlot initialized with the TupleDesc of the corresponding
partitioned table, along with a TupleConversionMap to map tuples from
the its parent's rowtype to own rowtype.  After tuple routing chooses
a leaf partition, we must use the leaf partition's tuple descriptor,
not the root table's.  To that end, a dedicated TupleTableSlot for
tuple routing is now allocated in EState.

Amit Langote
src/backend/catalog/partition.c
src/backend/commands/copy.c
src/backend/executor/nodeModifyTable.c
src/include/catalog/partition.h
src/include/nodes/execnodes.h
src/test/regress/expected/insert.out
src/test/regress/sql/insert.sql