Add the ability for the core grammar to have more than one parse target.

This patch essentially allows gram.y to implement a family of related
syntax trees, rather than necessarily always parsing a list of SQL
statements.  raw_parser() gains a new argument, enum RawParseMode,
to say what to do.  As proof of concept, add a mode that just parses
a TypeName without any other decoration, and use that to greatly
simplify typeStringToTypeName().

In addition, invent a new SPI entry point SPI_prepare_extended() to
allow SPI users (particularly plpgsql) to get at this new functionality.
In hopes of making this the last variant of SPI_prepare(), set up its
additional arguments as a struct rather than direct arguments, and
promise that future additions to the struct can default to zero.
SPI_prepare_cursor() and SPI_prepare_params() can perhaps go away at
some point.

Discussion: https://postgr.es/m/4165684.1607707277@sss.pgh.pa.us

src/backend/parser/gram.y

@@ -384,7 +384,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 %type <node>	vacuum_relation
 %type <selectlimit> opt_select_limit select_limit limit_clause
 
-%type <list>	stmtblock stmtmulti
+%type <list>	parse_toplevel stmtmulti
 				OptTableElementList TableElementList OptInherit definition
 				OptTypedTableElementList TypedTableElementList
 				reloptions opt_reloptions
@@ -723,6 +723,15 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
  */
 %token		NOT_LA NULLS_LA WITH_LA
 
+/*
+ * The grammar likewise thinks these tokens are keywords, but they are never
+ * generated by the scanner.  Rather, they can be injected by parser.c as
+ * the initial token of the string (using the lookahead-token mechanism
+ * implemented there).  This provides a way to tell the grammar to parse
+ * something other than the usual list of SQL commands.
+ */
+%token		MODE_TYPE_NAME
+
 
 /* Precedence: lowest to highest */
 %nonassoc	SET				/* see relation_expr_opt_alias */
@@ -787,11 +796,20 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 
 /*
  *	The target production for the whole parse.
+ *
+ * Ordinarily we parse a list of statements, but if we see one of the
+ * special MODE_XXX symbols as first token, we parse something else.
+ * The options here correspond to enum RawParseMode, which see for details.
  */
-stmtblock:	stmtmulti
+parse_toplevel:
+			stmtmulti
 			{
 				pg_yyget_extra(yyscanner)->parsetree = $1;
 			}
+			| MODE_TYPE_NAME Typename
+			{
+				pg_yyget_extra(yyscanner)->parsetree = list_make1($2);
+			}
 		;
 
 /*

src/backend/parser/parse_coerce.c

@@ -1541,7 +1541,7 @@ select_common_typmod(ParseState *pstate, List *exprs, Oid common_type)
 
 	foreach(lc, exprs)
 	{
-		Node   *expr = (Node *) lfirst(lc);
+		Node	   *expr = (Node *) lfirst(lc);
 
 		/* Types must match */
 		if (exprType(expr) != common_type)
@@ -2380,7 +2380,8 @@ enforce_generic_type_consistency(const Oid *actual_arg_types,
 			if (!OidIsValid(elem_typeid))
 			{
 				/*
-				 * if we don't have an element type yet, use the one we just got
+				 * if we don't have an element type yet, use the one we just
+				 * got
 				 */
 				elem_typeid = range_typelem;
 			}

src/backend/parser/parse_type.c

@@ -719,13 +719,6 @@ pts_error_callback(void *arg)
 	const char *str = (const char *) arg;
 
 	errcontext("invalid type name \"%s\"", str);
-
-	/*
-	 * Currently we just suppress any syntax error position report, rather
-	 * than transforming to an "internal query" error.  It's unlikely that a
-	 * type name is complex enough to need positioning.
-	 */
-	errposition(0);
 }
 
 /*
@@ -737,11 +730,7 @@ pts_error_callback(void *arg)
 TypeName *
 typeStringToTypeName(const char *str)
 {
-	StringInfoData buf;
 	List	   *raw_parsetree_list;
-	SelectStmt *stmt;
-	ResTarget  *restarget;
-	TypeCast   *typecast;
 	TypeName   *typeName;
 	ErrorContextCallback ptserrcontext;
 
@@ -749,9 +738,6 @@ typeStringToTypeName(const char *str)
 	if (strspn(str, " \t\n\r\f") == strlen(str))
 		goto fail;
 
-	initStringInfo(&buf);
-	appendStringInfo(&buf, "SELECT NULL::%s", str);
-
 	/*
 	 * Setup error traceback support in case of ereport() during parse
 	 */
@@ -760,58 +746,18 @@ typeStringToTypeName(const char *str)
 	ptserrcontext.previous = error_context_stack;
 	error_context_stack = &ptserrcontext;
 
-	raw_parsetree_list = raw_parser(buf.data);
+	raw_parsetree_list = raw_parser(str, RAW_PARSE_TYPE_NAME);
 
 	error_context_stack = ptserrcontext.previous;
 
-	/*
-	 * Make sure we got back exactly what we expected and no more; paranoia is
-	 * justified since the string might contain anything.
-	 */
-	if (list_length(raw_parsetree_list) != 1)
-		goto fail;
-	stmt = (SelectStmt *) linitial_node(RawStmt, raw_parsetree_list)->stmt;
-	if (stmt == NULL ||
-		!IsA(stmt, SelectStmt) ||
-		stmt->distinctClause != NIL ||
-		stmt->intoClause != NULL ||
-		stmt->fromClause != NIL ||
-		stmt->whereClause != NULL ||
-		stmt->groupClause != NIL ||
-		stmt->havingClause != NULL ||
-		stmt->windowClause != NIL ||
-		stmt->valuesLists != NIL ||
-		stmt->sortClause != NIL ||
-		stmt->limitOffset != NULL ||
-		stmt->limitCount != NULL ||
-		stmt->lockingClause != NIL ||
-		stmt->withClause != NULL ||
-		stmt->op != SETOP_NONE)
-		goto fail;
-	if (list_length(stmt->targetList) != 1)
-		goto fail;
-	restarget = (ResTarget *) linitial(stmt->targetList);
-	if (restarget == NULL ||
-		!IsA(restarget, ResTarget) ||
-		restarget->name != NULL ||
-		restarget->indirection != NIL)
-		goto fail;
-	typecast = (TypeCast *) restarget->val;
-	if (typecast == NULL ||
-		!IsA(typecast, TypeCast) ||
-		typecast->arg == NULL ||
-		!IsA(typecast->arg, A_Const))
-		goto fail;
+	/* We should get back exactly one TypeName node. */
+	Assert(list_length(raw_parsetree_list) == 1);
+	typeName = linitial_node(TypeName, raw_parsetree_list);
 
-	typeName = typecast->typeName;
-	if (typeName == NULL ||
-		!IsA(typeName, TypeName))
-		goto fail;
+	/* The grammar allows SETOF in TypeName, but we don't want that here. */
 	if (typeName->setof)
 		goto fail;
 
-	pfree(buf.data);
-
 	return typeName;
 
 fail:

src/backend/parser/parser.c

@@ -35,11 +35,11 @@ static char *str_udeescape(const char *str, char escape,
  * raw_parser
  *		Given a query in string form, do lexical and grammatical analysis.
  *
- * Returns a list of raw (un-analyzed) parse trees.  The immediate elements
- * of the list are always RawStmt nodes.
+ * Returns a list of raw (un-analyzed) parse trees.  The contents of the
+ * list have the form required by the specified RawParseMode.
  */
 List *
-raw_parser(const char *str)
+raw_parser(const char *str, RawParseMode mode)
 {
 	core_yyscan_t yyscanner;
 	base_yy_extra_type yyextra;
@@ -49,8 +49,22 @@ raw_parser(const char *str)
 	yyscanner = scanner_init(str, &yyextra.core_yy_extra,
 							 &ScanKeywords, ScanKeywordTokens);
 
-	/* base_yylex() only needs this much initialization */
-	yyextra.have_lookahead = false;
+	/* base_yylex() only needs us to initialize the lookahead token, if any */
+	if (mode == RAW_PARSE_DEFAULT)
+		yyextra.have_lookahead = false;
+	else
+	{
+		/* this array is indexed by RawParseMode enum */
+		static const int mode_token[] = {
+			0,					/* RAW_PARSE_DEFAULT */
+			MODE_TYPE_NAME		/* RAW_PARSE_TYPE_NAME */
+		};
+
+		yyextra.have_lookahead = true;
+		yyextra.lookahead_token = mode_token[mode];
+		yyextra.lookahead_yylloc = 0;
+		yyextra.lookahead_end = NULL;
+	}
 
 	/* initialize the bison parser */
 	parser_init(&yyextra);
@@ -104,7 +118,8 @@ base_yylex(YYSTYPE *lvalp, YYLTYPE *llocp, core_yyscan_t yyscanner)
 		cur_token = yyextra->lookahead_token;
 		lvalp->core_yystype = yyextra->lookahead_yylval;
 		*llocp = yyextra->lookahead_yylloc;
-		*(yyextra->lookahead_end) = yyextra->lookahead_hold_char;
+		if (yyextra->lookahead_end)
+			*(yyextra->lookahead_end) = yyextra->lookahead_hold_char;
 		yyextra->have_lookahead = false;
 	}
 	else