Add index scan to INSERT DML decompression

In order to verify constraints, we have to decompress
batches that could contain duplicates of the tuples
we are inserting. To find such batches, we use heap
scans which can be very expensive if the compressed
chunk contains a lot of tuples. Doing an index scan
makes much more sense in this scenario and will
give great performance benefits.

Additionally, we don't want to create the decompressor
until we determine we actually want to decompress a
batch so we try to lazily initialize it once a batch
is found.

.unreleased/pr_7048

@@ -0,0 +1 @@
+Implements: #7048 Add index scan to INSERT DML decompression

src/nodes/chunk_dispatch/chunk_insert_state.c

@@ -519,6 +519,7 @@ ts_chunk_insert_state_create(Oid chunk_relid, const ChunkDispatch *dispatch)
 	state->rel = rel;
 	state->result_relation_info = relinfo;
 	state->estate = dispatch->estate;
+	state->compressed_chunk_table_id = InvalidOid;
 	ts_set_compression_status(state, chunk);
 
 	if (relinfo->ri_RelationDesc->rd_rel->relhasindex && relinfo->ri_IndexRelationDescs == NULL)
@@ -634,7 +635,14 @@ ts_set_compression_status(ChunkInsertState *state, const Chunk *chunk)
 {
 	state->chunk_compressed = ts_chunk_is_compressed(chunk);
 	if (state->chunk_compressed)
+	{
 		state->chunk_partial = ts_chunk_is_partial(chunk);
+		if (!OidIsValid(state->compressed_chunk_table_id))
+		{
+			Chunk *comp = ts_chunk_get_by_id(chunk->fd.compressed_chunk_id, true);
+			state->compressed_chunk_table_id = comp->table_id;
+		}
+	}
 }
 
 extern void

src/nodes/chunk_dispatch/chunk_insert_state.h

@@ -55,6 +55,8 @@ typedef struct ChunkInsertState
 	/* for tracking compressed chunks */
 	bool chunk_compressed;
 	bool chunk_partial;
+
+	Oid compressed_chunk_table_id;
 } ChunkInsertState;
 
 typedef struct ChunkDispatch ChunkDispatch;