[#3047] Added SPLIT INTO # TABLETS to CREATE INDEX DDL

Summary:
Extended the Postgres grammar to allow for SPLIT INTO # TABLETS in CREATE INDEX DDL
statments. Adding functionality for SPLIT AT is a WIP.

Test Plan:
Using yb-ctl performed CREATE INDEX ... SPLIT INTO # TABLETS. Confirmed using web UI.
Added to the TestPgRegressindex.java test:
```
./yb_build.sh release --java-test org.yb.pgsql.TestPgRegressIndex
```

Reviewers: neha, dmitry

Reviewed By: dmitry

Subscribers: yql

Differential Revision: https://phabricator.dev.yugabyte.com/D8583

src/postgres/src/backend/catalog/index.c

@@ -769,7 +769,8 @@ index_create(Relation heapRelation,
 			 bits16 constr_flags,
 			 bool allow_system_table_mods,
 			 bool is_internal,
-			 Oid *constraintId)
+			 Oid *constraintId,
+			 OptSplit *split_options)
 {
 	Oid			heapRelationId = RelationGetRelid(heapRelation);
 	Relation	pg_class;
@@ -969,7 +970,8 @@ index_create(Relation heapRelation,
 					   coloptions,
 					   reloptions,
 					   indexRelationId,
-					   heapRelation);
+					   heapRelation,
+					   split_options);
 	}
 
 	/*

src/postgres/src/backend/catalog/toasting.c

@@ -338,7 +338,7 @@ create_toast_table(Relation rel, Oid toastOid, Oid toastIndexOid,
 				 BTREE_AM_OID,
 				 rel->rd_rel->reltablespace,
 				 collationObjectId, classObjectId, coloptions, (Datum) 0,
-				 INDEX_CREATE_IS_PRIMARY, 0, true, true, NULL);
+				 INDEX_CREATE_IS_PRIMARY, 0, true, true, NULL, NULL);
 
 	heap_close(toast_rel, NoLock);
 

src/postgres/src/backend/commands/indexcmds.c

@@ -906,7 +906,7 @@ DefineIndex(Oid relationId,
 					 coloptions, reloptions,
 					 flags, constr_flags,
 					 allowSystemTableMods, !check_rights,
-					 &createdConstraintId);
+					 &createdConstraintId, stmt->split_options);
 
 	ObjectAddressSet(address, RelationRelationId, indexRelationId);
 

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

@@ -246,7 +246,7 @@ static void CreateTableHandleSplitOptions(YBCPgStatement handle,
 			}
 
 			/* Tell pggate about it */
-			YBCPgCreateTableSetNumTablets(handle, split_options->num_tablets);
+			HandleYBStatus(YBCPgCreateTableSetNumTablets(handle, split_options->num_tablets));
 			break;
 		case SPLIT_POINTS: ;
 			/* Number of columns used in the primary key */
@@ -368,7 +368,8 @@ static void CreateTableHandleSplitOptions(YBCPgStatement handle,
 					split_num++;
 				}
 
-				YBCPgCreateTableAddSplitRow(handle, split_columns, type_entities, (uint64_t *)datums);
+				HandleYBStatus(YBCPgCreateTableAddSplitRow(handle, split_columns,
+					type_entities, (uint64_t *)datums));
 			}
 
 			break;
@@ -602,14 +603,38 @@ YBCTruncateTable(Relation rel) {
 	list_free(indexlist);
 }
 
+static void CreateIndexHandleSplitOptions(YBCPgStatement handle,
+										OptSplit *split_options,
+										TupleDesc desc,
+										int16 * coloptions)
+{
+	/* Address both types of split options */
+	switch (split_options->split_type)
+	{
+		case NUM_TABLETS:
+			/* Make sure we have HASH columns */
+			if (!(coloptions[0] & INDOPTION_HASH)) {
+				ereport(ERROR, (errmsg("HASH columns must be present to "
+									   "split by number of tablets")));
+			}
+			/* Tell pggate about it */
+			HandleYBStatus(YBCPgCreateIndexSetNumTablets(handle, split_options->num_tablets));
+			break;
+		default:
+			ereport(ERROR, (errmsg("Illegal memory state for SPLIT options")));
+			break;
+	}
+}
+
 void
 YBCCreateIndex(const char *indexName,
 			   IndexInfo *indexInfo,
 			   TupleDesc indexTupleDesc,
 			   int16 *coloptions,
 			   Datum reloptions,
 			   Oid indexId,
-			   Relation rel)
+			   Relation rel,
+			   OptSplit *split_options)
 {
 	char *db_name	  = get_database_name(MyDatabaseId);
 	char *schema_name = get_namespace_name(RelationGetNamespace(rel));
@@ -680,6 +705,11 @@ YBCCreateIndex(const char *indexName,
 																						 is_nulls_first));
 	}
 
+	/* Handle SPLIT statement, if present */
+	if (split_options) {
+		CreateIndexHandleSplitOptions(handle, split_options, indexTupleDesc, coloptions);
+	}
+
 	/* Create the index. */
 	HandleYBStatus(YBCPgExecCreateIndex(handle));
 }

src/postgres/src/backend/nodes/copyfuncs.c

@@ -3481,6 +3481,7 @@ _copyIndexStmt(const IndexStmt *from)
 	COPY_SCALAR_FIELD(transformed);
 	COPY_SCALAR_FIELD(concurrent);
 	COPY_SCALAR_FIELD(if_not_exists);
+	COPY_NODE_FIELD(split_options);
 
 	return newnode;
 }

src/postgres/src/backend/parser/gram.y

@@ -7818,13 +7818,13 @@ defacl_privilege_target:
  *
  *		QUERY: CREATE INDEX
  *
- * Note: we cannot put TABLESPACE clause after WHERE clause unless we are
- * willing to make TABLESPACE a fully reserved word.
+ * Note: we cannot put TABLESPACE or SPLIT clause after WHERE clause unless we
+ * are willing to make TABLESPACE or SPLIT fully reserved words.
  *****************************************************************************/
 
 IndexStmt:	CREATE opt_unique INDEX opt_concurrently opt_index_name
 			ON relation_expr access_method_clause '(' yb_index_params ')'
-			opt_include opt_reloptions OptTableSpace where_clause
+			opt_include opt_reloptions OptTableSpace OptSplit where_clause
 				{
 					IndexStmt *n = makeNode(IndexStmt);
 					n->unique = $2;
@@ -7837,7 +7837,8 @@ IndexStmt:	CREATE opt_unique INDEX opt_concurrently opt_index_name
 					n->indexIncludingParams = $12;
 					n->options = $13;
 					n->tableSpace = $14;
-					n->whereClause = $15;
+					n->split_options = $15;
+					n->whereClause = $16;
 					n->excludeOpNames = NIL;
 					n->idxcomment = NULL;
 					n->indexOid = InvalidOid;
@@ -7852,7 +7853,7 @@ IndexStmt:	CREATE opt_unique INDEX opt_concurrently opt_index_name
 				}
 			| CREATE opt_unique INDEX opt_concurrently IF_P NOT EXISTS index_name
 			ON relation_expr access_method_clause '(' yb_index_params ')'
-			opt_include opt_reloptions OptTableSpace where_clause
+			opt_include opt_reloptions OptTableSpace OptSplit where_clause
 				{
 					IndexStmt *n = makeNode(IndexStmt);
 					n->unique = $2;
@@ -7865,7 +7866,8 @@ IndexStmt:	CREATE opt_unique INDEX opt_concurrently opt_index_name
 					n->indexIncludingParams = $15;
 					n->options = $16;
 					n->tableSpace = $17;
-					n->whereClause = $18;
+					n->split_options = $18;
+					n->whereClause = $19;
 					n->excludeOpNames = NIL;
 					n->idxcomment = NULL;
 					n->indexOid = InvalidOid;

src/postgres/src/include/catalog/index.h

@@ -70,7 +70,8 @@ extern Oid index_create(Relation heapRelation,
 			 bits16 constr_flags,
 			 bool allow_system_table_mods,
 			 bool is_internal,
-			 Oid *constraintId);
+			 Oid *constraintId,
+			 OptSplit *split_options);
 
 #define	INDEX_CONSTR_CREATE_MARK_AS_PRIMARY	(1 << 0)
 #define	INDEX_CONSTR_CREATE_DEFERRABLE		(1 << 1)

src/postgres/src/include/commands/ybccmds.h

@@ -60,7 +60,8 @@ extern void YBCCreateIndex(const char *indexName,
 						   int16 *coloptions,
 						   Datum reloptions,
 						   Oid indexId,
-						   Relation rel);
+						   Relation rel,
+						   OptSplit *split_options);
 
 extern void YBCDropIndex(Oid relationId);
 

src/postgres/src/include/nodes/execnodes.h

@@ -142,6 +142,7 @@ typedef struct ExprState
  *		Am					Oid of index AM
  *		AmCache				private cache area for index AM
  *		Context				memory context holding this IndexInfo
+ *		SplitOptions		Options to split index into tablets. 
  *
  * ii_Concurrent, ii_BrokenHotChain, and ii_ParallelWorkers are used only
  * during index build; they're conventionally zeroed otherwise.

src/postgres/src/include/nodes/parsenodes.h

@@ -2784,6 +2784,8 @@ typedef struct IndexStmt
 	bool		transformed;	/* true when transformIndexStmt is finished */
 	bool		concurrent;		/* should this be a concurrent index build? */
 	bool		if_not_exists;	/* just do nothing if index already exists? */
+
+	OptSplit *split_options; /* SPLIT statement options */
 } IndexStmt;
 
 /* ----------------------

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

@@ -586,6 +586,36 @@ SELECT * FROM test_include ORDER BY c2;
   1 |    |  1
 (2 rows)
 
+-- Test SPLIT INTO
+CREATE TABLE test_split (
+  h1 int, h2 int, r1 int, r2 int, v1 int, v2 int,
+  PRIMARY KEY ((h1, h2) HASH, r1, r2));
+CREATE INDEX ON test_split (h2 HASH, r2, r1) SPLIT INTO 20 TABLETS;
+CREATE INDEX ON test_split ((r1,r2) HASH) SPLIT INTO 20 TABLETS;
+CREATE INDEX ON test_split (h2) SPLIT INTO 20 TABLETS;
+\d test_split
+             Table "public.test_split"
+ Column |  Type   | Collation | Nullable | Default 
+--------+---------+-----------+----------+---------
+ h1     | integer |           | not null | 
+ h2     | integer |           | not null | 
+ r1     | integer |           | not null | 
+ r2     | integer |           | not null | 
+ v1     | integer |           |          | 
+ v2     | integer |           |          | 
+Indexes:
+    "test_split_pkey" PRIMARY KEY, lsm ((h1, h2) HASH, r1, r2)
+    "test_split_h2_idx" lsm (h2 HASH)
+    "test_split_h2_r2_r1_idx" lsm (h2 HASH, r2, r1)
+    "test_split_r1_r2_idx" lsm ((r1, r2) HASH)
+
+-- These should fail
+CREATE INDEX ON test_split (r1 ASC) SPLIT INTO 20 TABLETS;
+ERROR:  HASH columns must be present to split by number of tablets
+CREATE INDEX ON test_split (h2 ASC, r1) SPLIT INTO 20 TABLETS;
+ERROR:  HASH columns must be present to split by number of tablets
+CREATE INDEX ON test_split (h1 HASH) SPLIT INTO 10000 TABLETS;
+ERROR:  Invalid argument: num_tablets exceeds system limit
 -- Test hash methods
 CREATE TABLE test_method (
   h1 int, h2 int, r1 int, r2 int, v1 int, v2 int,

src/postgres/src/test/regress/sql/yb_create_index.sql

@@ -217,6 +217,20 @@ DELETE FROM test_include WHERE c1 = 2;
 SELECT c1, c2 FROM test_include WHERE c1 > 0 ORDER BY c2;
 SELECT * FROM test_include ORDER BY c2;
 
+-- Test SPLIT INTO
+CREATE TABLE test_split (
+  h1 int, h2 int, r1 int, r2 int, v1 int, v2 int,
+  PRIMARY KEY ((h1, h2) HASH, r1, r2));
+CREATE INDEX ON test_split (h2 HASH, r2, r1) SPLIT INTO 20 TABLETS;
+CREATE INDEX ON test_split ((r1,r2) HASH) SPLIT INTO 20 TABLETS;
+CREATE INDEX ON test_split (h2) SPLIT INTO 20 TABLETS;
+\d test_split
+
+-- These should fail
+CREATE INDEX ON test_split (r1 ASC) SPLIT INTO 20 TABLETS;
+CREATE INDEX ON test_split (h2 ASC, r1) SPLIT INTO 20 TABLETS;
+CREATE INDEX ON test_split (h1 HASH) SPLIT INTO 10000 TABLETS;
+
 -- Test hash methods
 CREATE TABLE test_method (
   h1 int, h2 int, r1 int, r2 int, v1 int, v2 int,

src/yb/yql/pggate/pggate.cc

@@ -658,6 +658,14 @@ Status PgApiImpl::CreateIndexAddColumn(PgStatement *handle, const char *attr_nam
       is_hash, is_range, is_desc, is_nulls_first);
 }
 
+Status PgApiImpl::CreateIndexSetNumTablets(PgStatement *handle, int32_t num_tablets) {
+  if (!PgStatement::IsValidStmt(handle, StmtOp::STMT_CREATE_INDEX)) {
+    // Invalid handle.
+    return STATUS(InvalidArgument, "Invalid statement handle");
+  }
+  return down_cast<PgCreateIndex*>(handle)->SetNumTablets(num_tablets);
+}
+
 Status PgApiImpl::ExecCreateIndex(PgStatement *handle) {
   if (!PgStatement::IsValidStmt(handle, StmtOp::STMT_CREATE_INDEX)) {
     // Invalid handle.

src/yb/yql/pggate/pggate.h

@@ -259,6 +259,8 @@ class PgApiImpl {
                                       const YBCPgTypeEntity *attr_type, bool is_hash,
                                       bool is_range, bool is_desc, bool is_nulls_first);
 
+  CHECKED_STATUS CreateIndexSetNumTablets(PgStatement *handle, int32_t num_tablets);
+
   CHECKED_STATUS ExecCreateIndex(PgStatement *handle);
 
   CHECKED_STATUS NewDropIndex(const PgObjectId& index_id,

src/yb/yql/pggate/ybc_pggate.cc

@@ -416,6 +416,10 @@ YBCStatus YBCPgCreateIndexAddColumn(YBCPgStatement handle, const char *attr_name
                                                  is_hash, is_range, is_desc, is_nulls_first));
 }
 
+YBCStatus YBCPgCreateIndexSetNumTablets(YBCPgStatement handle, int32_t num_tablets) {
+  return ToYBCStatus(pgapi->CreateIndexSetNumTablets(handle, num_tablets));
+}
+
 YBCStatus YBCPgExecCreateIndex(YBCPgStatement handle) {
   return ToYBCStatus(pgapi->ExecCreateIndex(handle));
 }

src/yb/yql/pggate/ybc_pggate.h

@@ -229,6 +229,8 @@ YBCStatus YBCPgCreateIndexAddColumn(YBCPgStatement handle, const char *attr_name
                                     const YBCPgTypeEntity *attr_type, bool is_hash, bool is_range,
                                     bool is_desc, bool is_nulls_first);
 
+YBCStatus YBCPgCreateIndexSetNumTablets(YBCPgStatement handle, int32_t num_tablets);
+
 YBCStatus YBCPgExecCreateIndex(YBCPgStatement handle);
 
 YBCStatus YBCPgNewDropIndex(YBCPgOid database_oid,