Account for uncompressed rows in 'create_compressed_chunk'

`_timescaledb_internal.create_compressed_chunk` can be used to create a
compressed chunk with existing compressed data. It did not account for
the fact that the chunk can contain uncompressed data, in which case the
chunk status must be set to partial.

Fixes #5946

.unreleased/bugfix_5951

@@ -0,0 +1,4 @@
+Implements: #5951 _timescaledb_internal.create_compressed_chunk doesn't account for existing uncompressed rows
+
+Fixes: #5946
+

src/chunk.c

@@ -3674,22 +3674,28 @@ chunk_set_compressed_id_in_tuple(TupleInfo *ti, void *data)
 bool
 ts_chunk_set_compressed_chunk(Chunk *chunk, int32 compressed_chunk_id)
 {
+	bool success = false;
 	ScanKeyData scankey[1];
 	ScanKeyInit(&scankey[0],
 				Anum_chunk_idx_id,
 				BTEqualStrategyNumber,
 				F_INT4EQ,
 				Int32GetDatum(chunk->fd.id));
-	return chunk_scan_internal(CHUNK_ID_INDEX,
-							   scankey,
-							   1,
-							   chunk_check_ignorearg_dropped_filter,
-							   chunk_set_compressed_id_in_tuple,
-							   &compressed_chunk_id,
-							   0,
-							   ForwardScanDirection,
-							   RowExclusiveLock,
-							   CurrentMemoryContext) > 0;
+	success = chunk_scan_internal(CHUNK_ID_INDEX,
+								  scankey,
+								  1,
+								  chunk_check_ignorearg_dropped_filter,
+								  chunk_set_compressed_id_in_tuple,
+								  &compressed_chunk_id,
+								  0,
+								  ForwardScanDirection,
+								  RowExclusiveLock,
+								  CurrentMemoryContext) > 0;
+	if (success)
+	{
+		chunk->fd.status = ts_set_flags_32(chunk->fd.status, CHUNK_STATUS_COMPRESSED);
+	}
+	return success;
 }
 
 /*Assume permissions are already checked */

src/hypertable.c

@@ -1364,29 +1364,6 @@ hypertable_check_associated_schema_permissions(const char *schema_name, Oid user
 	return schema_oid;
 }
 
-static bool
-relation_has_tuples(Relation rel)
-{
-	TableScanDesc scandesc = table_beginscan(rel, GetActiveSnapshot(), 0, NULL);
-	TupleTableSlot *slot =
-		MakeSingleTupleTableSlot(RelationGetDescr(rel), table_slot_callbacks(rel));
-	bool hastuples = table_scan_getnextslot(scandesc, ForwardScanDirection, slot);
-
-	table_endscan(scandesc);
-	ExecDropSingleTupleTableSlot(slot);
-	return hastuples;
-}
-
-static bool
-table_has_tuples(Oid table_relid, LOCKMODE lockmode)
-{
-	Relation rel = table_open(table_relid, lockmode);
-	bool hastuples = relation_has_tuples(rel);
-
-	table_close(rel, lockmode);
-	return hastuples;
-}
-
 static bool
 table_is_logged(Oid table_relid)
 {
@@ -1641,7 +1618,7 @@ ts_hypertable_insert_blocker_trigger_add(PG_FUNCTION_ARGS)
 
 	ts_hypertable_permissions_check(relid, GetUserId());
 
-	if (table_has_tuples(relid, AccessShareLock))
+	if (ts_table_has_tuples(relid, AccessShareLock))
 		ereport(ERROR,
 				(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
 				 errmsg("hypertable \"%s\" has data in the root table", get_rel_name(relid)),
@@ -2208,7 +2185,7 @@ ts_hypertable_create_from_info(Oid table_relid, int32 hypertable_id, uint32 flag
 	/* Check that the table doesn't have any unsupported constraints */
 	hypertable_validate_constraints(table_relid, replication_factor);
 
-	table_has_data = relation_has_tuples(rel);
+	table_has_data = ts_relation_has_tuples(rel);
 
 	if ((flags & HYPERTABLE_CREATE_MIGRATE_DATA) == 0 && table_has_data)
 		ereport(ERROR,

src/utils.c

@@ -32,6 +32,7 @@
 #include <utils/fmgrprotos.h>
 #include <utils/lsyscache.h>
 #include <utils/relcache.h>
+#include <utils/snapmgr.h>
 #include <utils/syscache.h>
 
 #include "compat/compat.h"
@@ -1348,3 +1349,26 @@ ts_map_attno(Oid src_rel, Oid dst_rel, AttrNumber attno)
 	pfree(attname);
 	return dst_attno;
 }
+
+bool
+ts_relation_has_tuples(Relation rel)
+{
+	TableScanDesc scandesc = table_beginscan(rel, GetActiveSnapshot(), 0, NULL);
+	TupleTableSlot *slot =
+		MakeSingleTupleTableSlot(RelationGetDescr(rel), table_slot_callbacks(rel));
+	bool hastuples = table_scan_getnextslot(scandesc, ForwardScanDirection, slot);
+
+	table_endscan(scandesc);
+	ExecDropSingleTupleTableSlot(slot);
+	return hastuples;
+}
+
+bool
+ts_table_has_tuples(Oid table_relid, LOCKMODE lockmode)
+{
+	Relation rel = table_open(table_relid, lockmode);
+	bool hastuples = ts_relation_has_tuples(rel);
+
+	table_close(rel, lockmode);
+	return hastuples;
+}

src/utils.h

@@ -215,6 +215,9 @@ extern TSDLLEXPORT void ts_copy_relation_acl(const Oid source_relid, const Oid t
 extern TSDLLEXPORT bool ts_data_node_is_available_by_server(const ForeignServer *server);
 extern TSDLLEXPORT bool ts_data_node_is_available(const char *node_name);
 
+extern TSDLLEXPORT bool ts_relation_has_tuples(Relation rel);
+extern TSDLLEXPORT bool ts_table_has_tuples(Oid table_relid, LOCKMODE lockmode);
+
 extern TSDLLEXPORT AttrNumber ts_map_attno(Oid src_rel, Oid dst_rel, AttrNumber attno);
 
 /*

tsl/src/compression/api.c

@@ -770,6 +770,7 @@ tsl_create_compressed_chunk(PG_FUNCTION_ARGS)
 	Chunk *compress_ht_chunk;
 	Cache *hcache;
 	CompressChunkCxt cxt;
+	bool chunk_was_compressed;
 
 	Assert(!PG_ARGISNULL(0));
 	Assert(!PG_ARGISNULL(1));
@@ -812,7 +813,15 @@ tsl_create_compressed_chunk(PG_FUNCTION_ARGS)
 										  numrows_pre_compression,
 										  numrows_post_compression);
 
+	chunk_was_compressed = ts_chunk_is_compressed(cxt.srcht_chunk);
 	ts_chunk_set_compressed_chunk(cxt.srcht_chunk, compress_ht_chunk->fd.id);
+	if (!chunk_was_compressed && ts_table_has_tuples(cxt.srcht_chunk->table_id, AccessShareLock))
+	{
+		/* The chunk was not compressed before it had the compressed chunk
+		 * attached to it, and it contains rows, so we set it to be partial.
+		 */
+		ts_chunk_set_partial(cxt.srcht_chunk);
+	}
 	ts_cache_release(hcache);
 
 	PG_RETURN_OID(chunk_relid);

tsl/test/expected/compression_create_compressed_table.out

@@ -0,0 +1,69 @@
+-- This file and its contents are licensed under the Timescale License.
+-- Please see the included NOTICE for copyright information and
+-- LICENSE-TIMESCALE for a copy of the license.
+\c :TEST_DBNAME :ROLE_SUPERUSER
+-- create compressed hypertable
+CREATE TABLE "public"."metrics" (
+    "time" timestamp with time zone NOT NULL,
+    "device_id" "text",
+    "val" double precision
+);
+SELECT create_hypertable('public.metrics', 'time');
+  create_hypertable   
+----------------------
+ (1,public,metrics,t)
+(1 row)
+
+ALTER TABLE metrics SET (timescaledb.compress, timescaledb.compress_orderby = 'time', timescaledb.compress_segmentby = 'device_id');
+-- insert uncompressed row into hypertable
+INSERT INTO metrics (time, device_id, val)
+VALUES('2023-05-01T00:00:00Z'::timestamptz, 1, 1.0);
+SELECT count(*) FROM _timescaledb_internal._hyper_1_1_chunk;
+ count 
+-------
+     1
+(1 row)
+
+-- compress these rows, we do this to get compressed row data for the test
+SELECT compress_chunk('_timescaledb_internal._hyper_1_1_chunk');
+             compress_chunk             
+----------------------------------------
+ _timescaledb_internal._hyper_1_1_chunk
+(1 row)
+
+-- create custom compressed chunk table
+CREATE TABLE "_timescaledb_internal"."custom_compressed_chunk"() INHERITS ("_timescaledb_internal"."_compressed_hypertable_2");
+-- copy compressed row from compressed table into custom compressed chunk table
+INSERT INTO "_timescaledb_internal"."custom_compressed_chunk" SELECT * FROM "_timescaledb_internal"."compress_hyper_2_2_chunk";
+-- decompress the rows, moving them back to uncompressed space
+SELECT decompress_chunk('"_timescaledb_internal"."_hyper_1_1_chunk"');
+            decompress_chunk            
+----------------------------------------
+ _timescaledb_internal._hyper_1_1_chunk
+(1 row)
+
+-- attach compressed chunk to parent chunk
+SELECT _timescaledb_internal.create_compressed_chunk(
+    '"_timescaledb_internal"."_hyper_1_1_chunk"'::TEXT::REGCLASS,
+    '"_timescaledb_internal"."custom_compressed_chunk"'::TEXT::REGCLASS,
+    8192,
+    8192,
+    16384,
+    8192,
+    8192,
+    16384,
+    1,
+    1
+);
+        create_compressed_chunk         
+----------------------------------------
+ _timescaledb_internal._hyper_1_1_chunk
+(1 row)
+
+-- select total rows in chunk (should be 2)
+SELECT count(*) FROM "_timescaledb_internal"."_hyper_1_1_chunk";
+ count 
+-------
+     2
+(1 row)
+

tsl/test/sql/CMakeLists.txt

@@ -13,6 +13,7 @@ set(TEST_FILES
     cagg_refresh.sql
     cagg_watermark.sql
     compressed_collation.sql
+    compression_create_compressed_table.sql
     compression_bgw.sql
     compression_conflicts.sql
     compression_insert.sql

tsl/test/sql/compression_create_compressed_table.sql

@@ -0,0 +1,50 @@
+-- This file and its contents are licensed under the Timescale License.
+-- Please see the included NOTICE for copyright information and
+-- LICENSE-TIMESCALE for a copy of the license.
+
+\c :TEST_DBNAME :ROLE_SUPERUSER
+
+-- create compressed hypertable
+CREATE TABLE "public"."metrics" (
+    "time" timestamp with time zone NOT NULL,
+    "device_id" "text",
+    "val" double precision
+);
+SELECT create_hypertable('public.metrics', 'time');
+ALTER TABLE metrics SET (timescaledb.compress, timescaledb.compress_orderby = 'time', timescaledb.compress_segmentby = 'device_id');
+
+-- insert uncompressed row into hypertable
+INSERT INTO metrics (time, device_id, val)
+VALUES('2023-05-01T00:00:00Z'::timestamptz, 1, 1.0);
+
+SELECT count(*) FROM _timescaledb_internal._hyper_1_1_chunk;
+
+-- compress these rows, we do this to get compressed row data for the test
+SELECT compress_chunk('_timescaledb_internal._hyper_1_1_chunk');
+
+-- create custom compressed chunk table
+CREATE TABLE "_timescaledb_internal"."custom_compressed_chunk"() INHERITS ("_timescaledb_internal"."_compressed_hypertable_2");
+
+-- copy compressed row from compressed table into custom compressed chunk table
+INSERT INTO "_timescaledb_internal"."custom_compressed_chunk" SELECT * FROM "_timescaledb_internal"."compress_hyper_2_2_chunk";
+
+-- decompress the rows, moving them back to uncompressed space
+SELECT decompress_chunk('"_timescaledb_internal"."_hyper_1_1_chunk"');
+
+-- attach compressed chunk to parent chunk
+SELECT _timescaledb_internal.create_compressed_chunk(
+    '"_timescaledb_internal"."_hyper_1_1_chunk"'::TEXT::REGCLASS,
+    '"_timescaledb_internal"."custom_compressed_chunk"'::TEXT::REGCLASS,
+    8192,
+    8192,
+    16384,
+    8192,
+    8192,
+    16384,
+    1,
+    1
+);
+
+-- select total rows in chunk (should be 2)
+SELECT count(*) FROM "_timescaledb_internal"."_hyper_1_1_chunk";
+