[YSQL] #1412: Fix included index columns

Summary:
This revision fixes an issue with the included index columns where they should be included as the
non-primary-key columns in the index.

Test Plan: yb_create_index

Reviewers: mihnea, neil, neha

Reviewed By: neha

Subscribers: srhickma, yql

Differential Revision: https://phabricator.dev.yugabyte.com/D6650

src/postgres/src/backend/commands/ybccmds.c

@@ -319,21 +319,33 @@ YBCCreateIndex(const char *indexName,
 		char			  *attname = NameStr(att->attname);
 		AttrNumber		  attnum   = att->attnum;			
 		const YBCPgTypeEntity *col_type = YBCDataTypeFromOidMod(attnum, att->atttypid);
+		bool			  is_key = (i < indexInfo->ii_NumIndexKeyAttrs);
 
-		if (!YBCDataTypeIsValidForKey(att->atttypid)) {
-			ereport(ERROR,
-							(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-							 errmsg("INDEX on column of type '%s' not yet supported",
-											YBPgTypeOidToStr(att->atttypid))));
+		if (is_key)
+		{
+			if (!YBCDataTypeIsValidForKey(att->atttypid))
+				ereport(ERROR,
+						(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+						 errmsg("INDEX on column of type '%s' not yet supported",
+								YBPgTypeOidToStr(att->atttypid))));
+
+			HandleYBStmtStatus(YBCPgCreateIndexAddColumn(handle,
+														 attname,
+														 attnum,
+														 col_type,
+														 is_hash,
+														 true /* is_range */), handle);
+			is_hash = false;
+		}
+		else
+		{
+			HandleYBStmtStatus(YBCPgCreateIndexAddColumn(handle,
+														 attname,
+														 attnum,
+														 col_type,
+														 false /* is_hash */,
+														 false /* is_range */), handle);
 		}
-		HandleYBStmtStatus(YBCPgCreateIndexAddColumn(handle,
-													 attname,
-													 attnum,
-													 col_type,
-													 is_hash,
-													 true /* is_range */), handle);
-
-		is_hash = false;
 	}
 
 	/* Create the index. */

src/postgres/src/backend/executor/ybcModifyTable.c

@@ -483,14 +483,14 @@ void YBCExecuteDeleteIndex(Relation index, Datum *values, bool *isnull, Datum yb
 	Oid            dboid    = YBCGetDatabaseOid(index);
 	Oid            relid    = RelationGetRelid(index);
 	TupleDesc      tupdesc  = RelationGetDescr(index);
-	int            natts    = RelationGetNumberOfAttributes(index);
+	int            nkeys    = IndexRelationGetNumberOfKeyAttributes(index);
 	YBCPgStatement ybc_stmt = NULL;
 
 	Assert(index->rd_rel->relkind == RELKIND_INDEX);
 
 	/* Create the DELETE request and add the values from the tuple. */
 	HandleYBStatus(YBCPgNewDelete(ybc_pg_session, dboid, relid, &ybc_stmt));
-	for (AttrNumber attnum = 1; attnum <= natts; attnum++)
+	for (AttrNumber attnum = 1; attnum <= nkeys; attnum++)
 	{
 		Oid   type_id = GetTypeId(attnum, tupdesc);
 		Datum value   = values[attnum - 1];
@@ -514,7 +514,7 @@ void YBCExecuteDeleteIndex(Relation index, Datum *values, bool *isnull, Datum yb
 
 	YBCPgExpr ybc_expr = YBCNewConstant(ybc_stmt, BYTEAOID, ybctid, false /* is_null */);
 	HandleYBStmtStatus(YBCPgDmlBindColumn(ybc_stmt, YBBaseTupleIdAttributeNumber, ybc_expr),
-			ybc_stmt);
+					   ybc_stmt);
 
 	/* Execute the delete and clean up. */
 	HandleYBStmtStatus(YBCExecWriteStmt(ybc_stmt, index), ybc_stmt);

src/postgres/src/test/regress/expected/yb_create_index.out

@@ -467,3 +467,89 @@ Indexes:
     "test_method_v_idx" lsm (v)
 
 DROP TABLE test_method;
+-- Test include columns
+CREATE TABLE test_include (c1 int, c2 int, c3 int);
+INSERT INTO test_include VALUES (1, 1, 1), (1, 2, 2), (2, 2, 2), (3, 3, 3);
+-- Expect duplicate key error
+CREATE UNIQUE INDEX ON test_include (c1) include (c2);
+ERROR:  duplicate key value violates unique constraint "test_include_c1_c2_idx"
+DROP INDEX test_include_c1_c2_idx;
+DELETE FROM test_include WHERE c1 = 1 AND c2 = 2;
+CREATE UNIQUE INDEX ON test_include (c1) include (c2);
+EXPLAIN SELECT c1, c2 FROM test_include WHERE c1 = 1;
+                                            QUERY PLAN
+---------------------------------------------------------------------------------------------------
+ Index Only Scan using test_include_c1_c2_idx on test_include  (cost=0.00..4.01 rows=1000 width=8)
+   Index Cond: (c1 = 1)
+(2 rows)
+
+SELECT c1, c2 FROM test_include WHERE c1 = 1;
+ c1 | c2
+----+----
+  1 |  1
+(1 row)
+
+\d test_include
+            Table "public.test_include"
+ Column |  Type   | Collation | Nullable | Default
+--------+---------+-----------+----------+---------
+ c1     | integer |           |          |
+ c2     | integer |           |          |
+ c3     | integer |           |          |
+Indexes:
+    "test_include_c1_c2_idx" UNIQUE, lsm (c1) INCLUDE (c2)
+
+-- Verify the included column is updated in both the base table and the index. Use WHERE condition
+-- on c1 below to force the scan on the index vs. base table. Select the non-included column c3 in
+-- the other case to force the use of sequential scan on the base table.
+UPDATE test_include SET c2 = 22 WHERE c1 = 2;
+EXPLAIN SELECT c1, c2 FROM test_include WHERE c1 > 0 ORDER BY c2;
+                                               QUERY PLAN
+---------------------------------------------------------------------------------------------------------
+ Sort  (cost=53.84..56.34 rows=1000 width=8)
+   Sort Key: c2
+   ->  Index Only Scan using test_include_c1_c2_idx on test_include  (cost=0.00..4.01 rows=1000 width=8)
+         Index Cond: (c1 > 0)
+(4 rows)
+
+EXPLAIN SELECT * FROM test_include ORDER BY c2;
+                                 QUERY PLAN
+----------------------------------------------------------------------------
+ Sort  (cost=149.83..152.33 rows=1000 width=12)
+   Sort Key: c2
+   ->  Foreign Scan on test_include  (cost=0.00..100.00 rows=1000 width=12)
+(3 rows)
+
+SELECT c1, c2 FROM test_include WHERE c1 > 0 ORDER BY c2;
+ c1 | c2
+----+----
+  1 |  1
+  3 |  3
+  2 | 22
+(3 rows)
+
+SELECT * FROM test_include ORDER BY c2;
+ c1 | c2 | c3
+----+----+----
+  1 |  1 |  1
+  3 |  3 |  3
+  2 | 22 |  2
+(3 rows)
+
+UPDATE test_include SET c2 = NULL WHERE c1 = 1;
+UPDATE test_include SET c2 = 33 WHERE c2 = 3;
+DELETE FROM test_include WHERE c1 = 2;
+SELECT c1, c2 FROM test_include WHERE c1 > 0 ORDER BY c2;
+ c1 | c2
+----+----
+  3 | 33
+  1 |
+(2 rows)
+
+SELECT * FROM test_include ORDER BY c2;
+ c1 | c2 | c3
+----+----+----
+  3 | 33 |  3
+  1 |    |  1
+(2 rows)
+

src/postgres/src/test/regress/expected/yb_index_including.out

@@ -13,15 +13,15 @@ CREATE INDEX ON tbl_include_reg (c1, c2) INCLUDE (c1, c3);
 SELECT pg_get_indexdef(i.indexrelid)
 FROM pg_index i JOIN pg_class c ON i.indexrelid = c.oid
 WHERE i.indrelid = 'tbl_include_reg'::regclass ORDER BY c.relname;
-                                               pg_get_indexdef                                               
+                                               pg_get_indexdef
 -------------------------------------------------------------------------------------------------------------
  CREATE INDEX tbl_include_reg_c1_c2_c11_c3_idx ON public.tbl_include_reg USING lsm (c1, c2) INCLUDE (c1, c3)
  CREATE INDEX tbl_include_reg_idx ON public.tbl_include_reg USING lsm (c1, c2) INCLUDE (c3, c4)
 (2 rows)
 
 \d tbl_include_reg_idx
   Index "public.tbl_include_reg_idx"
- Column |  Type   | Key? | Definition 
+ Column |  Type   | Key? | Definition
 --------+---------+------+------------
  c1     | integer | yes  | c1
  c2     | integer | yes  | c2
@@ -38,19 +38,19 @@ ALTER TABLE tbl_include_unique1 add UNIQUE (c1, c2) INCLUDE (c3, c4);
 SELECT pg_get_indexdef(i.indexrelid)
 FROM pg_index i JOIN pg_class c ON i.indexrelid = c.oid
 WHERE i.indrelid = 'tbl_include_unique1'::regclass ORDER BY c.relname;
-                                                      pg_get_indexdef                                                      
+                                                      pg_get_indexdef
 ---------------------------------------------------------------------------------------------------------------------------
  CREATE UNIQUE INDEX tbl_include_unique1_c1_c2_c3_c4_key ON public.tbl_include_unique1 USING lsm (c1, c2) INCLUDE (c3, c4)
  CREATE UNIQUE INDEX tbl_include_unique1_idx_unique ON public.tbl_include_unique1 USING lsm (c1, c2) INCLUDE (c3, c4)
 (2 rows)
 
--- TODO Enable after #1412
--- -- Unique index and unique constraint. Both must fail.
--- CREATE TABLE tbl_include_unique2 (c1 int, c2 int, c3 int, c4 int);
--- INSERT INTO tbl_include_unique2 SELECT 1, 2, 3*x, 4 FROM generate_series(1,10) AS x;
--- CREATE UNIQUE INDEX tbl_include_unique2_idx_unique ON tbl_include_unique2 using lsm (c1, c2) INCLUDE (c3, c4);
--- ALTER TABLE tbl_include_unique2 add UNIQUE (c1, c2) INCLUDE (c3, c4);
---
+-- Unique index and unique constraint. Both must fail.
+CREATE TABLE tbl_include_unique2 (c1 int, c2 int, c3 int, c4 int);
+INSERT INTO tbl_include_unique2 SELECT 1, 2, 3*x, 4 FROM generate_series(1,10) AS x;
+CREATE UNIQUE INDEX tbl_include_unique2_idx_unique ON tbl_include_unique2 using lsm (c1, c2) INCLUDE (c3, c4);
+ERROR:  duplicate key value violates unique constraint "tbl_include_unique2_idx_unique"
+ALTER TABLE tbl_include_unique2 add UNIQUE (c1, c2) INCLUDE (c3, c4);
+ERROR:  duplicate key value violates unique constraint "tbl_include_unique2_c1_c2_c3_c4_key"
 -- NOT SUPPORTED
 --
 -- -- PK constraint
@@ -80,32 +80,30 @@ ERROR:  This ALTER TABLE command is not yet supported.
 CREATE TABLE tbl (c1 int,c2 int, c3 int, c4 int,
 				CONSTRAINT covering UNIQUE(c1,c2) INCLUDE(c3,c4));
 SELECT indexrelid::regclass, indnatts, indnkeyatts, indisunique, indisprimary, indkey, indclass FROM pg_index WHERE indrelid = 'tbl'::regclass::oid;
- indexrelid | indnatts | indnkeyatts | indisunique | indisprimary | indkey  | indclass  
+ indexrelid | indnatts | indnkeyatts | indisunique | indisprimary | indkey  | indclass
 ------------+----------+-------------+-------------+--------------+---------+-----------
  covering   |        4 |           2 | t           | f            | 1 2 3 4 | 9942 9942
 (1 row)
 
 SELECT pg_get_constraintdef(oid), conname, conkey FROM pg_constraint WHERE conrelid = 'tbl'::regclass::oid;
-       pg_get_constraintdef       | conname  | conkey 
+       pg_get_constraintdef       | conname  | conkey
 ----------------------------------+----------+--------
  UNIQUE (c1, c2) INCLUDE (c3, c4) | covering | {1,2}
 (1 row)
 
--- TODO Enable after #1412
--- -- ensure that constraint works
--- INSERT INTO tbl SELECT 1, 2, 3*x, 4 FROM generate_series(1,10) AS x;
---
+INSERT INTO tbl SELECT 1, 2, 3*x, 4 FROM generate_series(1,10) AS x;
+ERROR:  duplicate key value violates unique constraint "covering"
 DROP TABLE tbl;
 CREATE TABLE tbl (c1 int,c2 int, c3 int, c4 int,
 				CONSTRAINT covering PRIMARY KEY(c1,c2) INCLUDE(c3,c4));
 SELECT indexrelid::regclass, indnatts, indnkeyatts, indisunique, indisprimary, indkey, indclass FROM pg_index WHERE indrelid = 'tbl'::regclass::oid;
- indexrelid | indnatts | indnkeyatts | indisunique | indisprimary | indkey  | indclass  
+ indexrelid | indnatts | indnkeyatts | indisunique | indisprimary | indkey  | indclass
 ------------+----------+-------------+-------------+--------------+---------+-----------
  covering   |        4 |           2 | t           | t            | 1 2 3 4 | 9942 9942
 (1 row)
 
 SELECT pg_get_constraintdef(oid), conname, conkey FROM pg_constraint WHERE conrelid = 'tbl'::regclass::oid;
-         pg_get_constraintdef          | conname  | conkey 
+         pg_get_constraintdef          | conname  | conkey
 ---------------------------------------+----------+--------
  PRIMARY KEY (c1, c2) INCLUDE (c3, c4) | covering | {1,2}
 (1 row)
@@ -127,42 +125,41 @@ select * from tbl where (c1,c2,c3) < (2,5,1);
 (3 rows)
 
 select * from tbl where (c1,c2,c3) < (2,5,1);
- c1 | c2 | c3 | c4 
+ c1 | c2 | c3 | c4
 ----+----+----+----
-  1 |  2 |    | 
-  2 |  4 |    | 
+  1 |  2 |    |
+  2 |  4 |    |
 (2 rows)
 
 DROP TABLE tbl;
 CREATE TABLE tbl (c1 int,c2 int, c3 int, c4 int,
 				UNIQUE(c1,c2) INCLUDE(c3,c4));
 SELECT indexrelid::regclass, indnatts, indnkeyatts, indisunique, indisprimary, indkey, indclass FROM pg_index WHERE indrelid = 'tbl'::regclass::oid;
-     indexrelid      | indnatts | indnkeyatts | indisunique | indisprimary | indkey  | indclass  
+     indexrelid      | indnatts | indnkeyatts | indisunique | indisprimary | indkey  | indclass
 ---------------------+----------+-------------+-------------+--------------+---------+-----------
  tbl_c1_c2_c3_c4_key |        4 |           2 | t           | f            | 1 2 3 4 | 9942 9942
 (1 row)
 
 SELECT pg_get_constraintdef(oid), conname, conkey FROM pg_constraint WHERE conrelid = 'tbl'::regclass::oid;
-       pg_get_constraintdef       |       conname       | conkey 
+       pg_get_constraintdef       |       conname       | conkey
 ----------------------------------+---------------------+--------
  UNIQUE (c1, c2) INCLUDE (c3, c4) | tbl_c1_c2_c3_c4_key | {1,2}
 (1 row)
 
--- TODO Enable after #1412
--- -- ensure that constraint works
--- INSERT INTO tbl SELECT 1, 2, 3*x, 4 FROM generate_series(1,10) AS x;
---
+-- ensure that constraint works
+INSERT INTO tbl SELECT 1, 2, 3*x, 4 FROM generate_series(1,10) AS x;
+ERROR:  duplicate key value violates unique constraint "tbl_c1_c2_c3_c4_key"
 DROP TABLE tbl;
 CREATE TABLE tbl (c1 int,c2 int, c3 int, c4 int,
 				PRIMARY KEY(c1,c2) INCLUDE(c3,c4));
 SELECT indexrelid::regclass, indnatts, indnkeyatts, indisunique, indisprimary, indkey, indclass FROM pg_index WHERE indrelid = 'tbl'::regclass::oid;
- indexrelid | indnatts | indnkeyatts | indisunique | indisprimary | indkey  | indclass  
+ indexrelid | indnatts | indnkeyatts | indisunique | indisprimary | indkey  | indclass
 ------------+----------+-------------+-------------+--------------+---------+-----------
  tbl_pkey   |        4 |           2 | t           | t            | 1 2 3 4 | 9942 9942
 (1 row)
 
 SELECT pg_get_constraintdef(oid), conname, conkey FROM pg_constraint WHERE conrelid = 'tbl'::regclass::oid;
-         pg_get_constraintdef          | conname  | conkey 
+         pg_get_constraintdef          | conname  | conkey
 ---------------------------------------+----------+--------
  PRIMARY KEY (c1, c2) INCLUDE (c3, c4) | tbl_pkey | {1,2}
 (1 row)
@@ -193,14 +190,14 @@ DROP TABLE tbl;
 CREATE TABLE tbl (c1 int,c2 int, c3 int, c4 int);
 CREATE UNIQUE INDEX tbl_idx ON tbl using lsm(c1, c2, c3, c4);
 SELECT indexdef FROM pg_indexes WHERE tablename = 'tbl' ORDER BY indexname;
-                               indexdef                               
+                               indexdef
 ----------------------------------------------------------------------
  CREATE UNIQUE INDEX tbl_idx ON public.tbl USING lsm (c1, c2, c3, c4)
 (1 row)
 
 ALTER TABLE tbl DROP COLUMN c3;
 SELECT indexdef FROM pg_indexes WHERE tablename = 'tbl' ORDER BY indexname;
- indexdef 
+ indexdef
 ----------
 (0 rows)
 
@@ -213,14 +210,14 @@ DROP TABLE tbl;
 CREATE TABLE tbl (c1 int,c2 int, c3 int, c4 int);
 CREATE UNIQUE INDEX tbl_idx ON tbl using lsm(c1, c2) INCLUDE(c3,c4);
 SELECT indexdef FROM pg_indexes WHERE tablename = 'tbl' ORDER BY indexname;
-                                   indexdef                                    
+                                   indexdef
 -------------------------------------------------------------------------------
  CREATE UNIQUE INDEX tbl_idx ON public.tbl USING lsm (c1, c2) INCLUDE (c3, c4)
 (1 row)
 
 ALTER TABLE tbl DROP COLUMN c3;
 SELECT indexdef FROM pg_indexes WHERE tablename = 'tbl' ORDER BY indexname;
- indexdef 
+ indexdef
 ----------
 (0 rows)
 
@@ -232,20 +229,20 @@ DROP TABLE tbl;
  */
 CREATE TABLE tbl (c1 int,c2 int, c3 int, c4 int, UNIQUE(c1, c2) INCLUDE(c3,c4));
 SELECT indexdef FROM pg_indexes WHERE tablename = 'tbl' ORDER BY indexname;
-                                         indexdef                                          
+                                         indexdef
 -------------------------------------------------------------------------------------------
  CREATE UNIQUE INDEX tbl_c1_c2_c3_c4_key ON public.tbl USING lsm (c1, c2) INCLUDE (c3, c4)
 (1 row)
 
 ALTER TABLE tbl DROP COLUMN c3;
 SELECT indexdef FROM pg_indexes WHERE tablename = 'tbl' ORDER BY indexname;
- indexdef 
+ indexdef
 ----------
 (0 rows)
 
 ALTER TABLE tbl DROP COLUMN c1;
 SELECT indexdef FROM pg_indexes WHERE tablename = 'tbl' ORDER BY indexname;
- indexdef 
+ indexdef
 ----------
 (0 rows)
 
@@ -320,11 +317,11 @@ INSERT INTO tbl SELECT x, 2*x, 3*x, 4 FROM generate_series(1,10) AS x;
 CREATE UNIQUE INDEX tbl_idx_unique ON tbl using lsm(c1, c2) INCLUDE (c3,c4);
 UPDATE tbl SET c1 = 100 WHERE c1 = 2;
 UPDATE tbl SET c1 = 1 WHERE c1 = 3;
--- TODO Enable after #1412
--- -- should fail
--- UPDATE tbl SET c2 = 2 WHERE c1 = 1;
+UPDATE tbl SET c2 = 2 WHERE c1 = 1;
+ERROR:  duplicate key value violates unique constraint "tbl_idx_unique"
+-- Disabled due to issue https://github.com/YugaByte/yugabyte-db/issues/1448:
+-- UPDATE tbl SET c3 = 1;
 --
-UPDATE tbl SET c3 = 1;
 DELETE FROM tbl WHERE c1 = 5 OR c3 = 12;
 DROP TABLE tbl;
 -- NOT SUPPORTED