Don't try to re-order the subcommands of CREATE SCHEMA

transformCreateSchemaStmtElements has always believed that it is
supposed to re-order the subcommands of CREATE SCHEMA into a safe
execution order.  However, it is nowhere near being capable of doing
that correctly.  Nor is there reason to think that it ever will be,
or that that is a well-defined requirement, or that there's any basis
in the SQL standard for it.  Moreover, the problem will get worse as
we add more subcommand types.  Let's just drop the whole idea and
execute the commands in the order given, which seems like a much less
astonishment-prone definition anyway.

Along the way, pass down a ParseState so that we can provide an
error cursor for the "wrong schema name" error, and fix
transformCreateSchemaStmtElements so that it doesn't scribble
on the parsetree passed to it.

Note: This will cause compatibility issue,
for example:
CREATE SCHEMA regress_schema_2
  CREATE VIEW abcd_view AS SELECT a FROM abcd
  CREATE TABLE abcd (a int);

With the patch, it will throw an error, whereas on HEAD it won’t.

Discussion: https://postgr.es/m/[email protected]

Author: tglsfdc

doc/src/sgml/ref/create_schema.sgml

@@ -193,12 +193,10 @@ CREATE VIEW hollywood.winners AS
   </para>
 
   <para>
-   The SQL standard specifies that the subcommands in <command>CREATE
-   SCHEMA</command> can appear in any order.  The present
-   <productname>PostgreSQL</productname> implementation does not
-   handle all cases of forward references in subcommands; it might
-   sometimes be necessary to reorder the subcommands in order to avoid
-   forward references.
+   <productname>PostgreSQL</productname> executes the subcommands
+   in <command>CREATE SCHEMA</command> in the order given.  Other
+   implementations may try to rearrange the subcommands into dependency
+   order, but that is hard if not impossible to do correctly.
   </para>
 
   <para>

src/backend/commands/extension.c

@@ -1773,14 +1773,17 @@ CreateExtensionInternal(char *extensionName,
 
 		if (!OidIsValid(schemaOid))
 		{
+			ParseState *pstate = make_parsestate(NULL);
 			CreateSchemaStmt *csstmt = makeNode(CreateSchemaStmt);
 
+			pstate->p_sourcetext = "(generated CREATE SCHEMA command)";
+
 			csstmt->schemaname = schemaName;
 			csstmt->authrole = NULL;	/* will be created by current user */
 			csstmt->schemaElts = NIL;
 			csstmt->if_not_exists = false;
-			CreateSchemaCommand(csstmt, "(generated CREATE SCHEMA command)",
-								-1, -1);
+
+			CreateSchemaCommand(pstate, csstmt, -1, -1);
 
 			/*
 			 * CreateSchemaCommand includes CommandCounterIncrement, so new

src/backend/commands/schemacmds.c

@@ -49,7 +49,7 @@ static void AlterSchemaOwner_internal(HeapTuple tup, Relation rel, Oid newOwnerI
  * a subquery.
  */
 Oid
-CreateSchemaCommand(CreateSchemaStmt *stmt, const char *queryString,
+CreateSchemaCommand(ParseState *pstate, CreateSchemaStmt *stmt,
 					int stmt_location, int stmt_len)
 {
 	const char *schemaName = stmt->schemaname;
@@ -189,12 +189,13 @@ CreateSchemaCommand(CreateSchemaStmt *stmt, const char *queryString,
 
 	/*
 	 * Examine the list of commands embedded in the CREATE SCHEMA command, and
-	 * reorganize them into a sequentially executable order with no forward
-	 * references.  Note that the result is still a list of raw parsetrees ---
-	 * we cannot, in general, run parse analysis on one statement until we
-	 * have actually executed the prior ones.
+	 * do preliminary transformations (mostly, verify that none are trying to
+	 * create objects outside the new schema).  Note that the result is still
+	 * a list of raw parsetrees --- we cannot, in general, run parse analysis
+	 * on one statement until we have actually executed the prior ones.
 	 */
-	parsetree_list = transformCreateSchemaStmtElements(stmt->schemaElts,
+	parsetree_list = transformCreateSchemaStmtElements(pstate,
+													   stmt->schemaElts,
 													   schemaName);
 
 	/*
@@ -219,7 +220,7 @@ CreateSchemaCommand(CreateSchemaStmt *stmt, const char *queryString,
 
 		/* do this step */
 		ProcessUtility(wrapper,
-					   queryString,
+					   pstate->p_sourcetext,
 					   false,
 					   PROCESS_UTILITY_SUBCOMMAND,
 					   NULL,

src/backend/parser/parse_utilcmd.c

@@ -95,18 +95,6 @@ typedef struct
 	bool		ofType;			/* true if statement contains OF typename */
 } CreateStmtContext;
 
-/* State shared by transformCreateSchemaStmtElements and its subroutines */
-typedef struct
-{
-	const char *schemaname;		/* name of schema */
-	List	   *sequences;		/* CREATE SEQUENCE items */
-	List	   *tables;			/* CREATE TABLE items */
-	List	   *views;			/* CREATE VIEW items */
-	List	   *indexes;		/* CREATE INDEX items */
-	List	   *triggers;		/* CREATE TRIGGER items */
-	List	   *grants;			/* GRANT items */
-} CreateSchemaStmtContext;
-
 
 static void transformColumnDefinition(CreateStmtContext *cxt,
 									  ColumnDef *column);
@@ -133,7 +121,8 @@ static void transformCheckConstraints(CreateStmtContext *cxt,
 static void transformConstraintAttrs(CreateStmtContext *cxt,
 									 List *constraintList);
 static void transformColumnType(CreateStmtContext *cxt, ColumnDef *column);
-static void setSchemaName(const char *context_schema, char **stmt_schema_name);
+static void checkSchemaName(ParseState *pstate, const char *context_schema,
+							RangeVar *relation);
 static void transformPartitionCmd(CreateStmtContext *cxt, PartitionCmd *cmd);
 static List *transformPartitionRangeBounds(ParseState *pstate, List *blist,
 										   Relation parent);
@@ -4075,109 +4064,85 @@ transformColumnType(CreateStmtContext *cxt, ColumnDef *column)
  * transformCreateSchemaStmtElements -
  *	  analyzes the elements of a CREATE SCHEMA statement
  *
- * Split the schema element list from a CREATE SCHEMA statement into
- * individual commands and place them in the result list in an order
- * such that there are no forward references (e.g. GRANT to a table
- * created later in the list). Note that the logic we use for determining
- * forward references is presently quite incomplete.
+ * This is now somewhat vestigial: its only real responsibility is to complain
+ * if any of the elements are trying to create objects outside the new schema.
+ * We used to try to re-order the commands in a way that would work even if
+ * the user-written order would not, but that's too hard (perhaps impossible)
+ * to do correctly with not-yet-parse-analyzed commands.  Now we'll just
+ * execute the elements in the order given.
  *
  * "schemaName" is the name of the schema that will be used for the creation
- * of the objects listed, that may be compiled from the schema name defined
+ * of the objects listed.  It may be obtained from the schema name defined
  * in the statement or a role specification.
  *
- * SQL also allows constraints to make forward references, so thumb through
- * the table columns and move forward references to a posterior alter-table
- * command.
- *
  * The result is a list of parse nodes that still need to be analyzed ---
  * but we can't analyze the later commands until we've executed the earlier
  * ones, because of possible inter-object references.
- *
- * Note: this breaks the rules a little bit by modifying schema-name fields
- * within passed-in structs.  However, the transformation would be the same
- * if done over, so it should be all right to scribble on the input to this
- * extent.
  */
 List *
-transformCreateSchemaStmtElements(List *schemaElts, const char *schemaName)
+transformCreateSchemaStmtElements(ParseState *pstate, List *schemaElts,
+								  const char *schemaName)
 {
-	CreateSchemaStmtContext cxt;
-	List	   *result;
-	ListCell   *elements;
-
-	cxt.schemaname = schemaName;
-	cxt.sequences = NIL;
-	cxt.tables = NIL;
-	cxt.views = NIL;
-	cxt.indexes = NIL;
-	cxt.triggers = NIL;
-	cxt.grants = NIL;
+	List	   *elements = NIL;
+	ListCell   *lc;
 
 	/*
-	 * Run through each schema element in the schema element list. Separate
-	 * statements by type, and do preliminary analysis.
+	 * Run through each schema element in the schema element list.  Check
+	 * target schema names, and collect the list of actions to be done.
 	 */
-	foreach(elements, schemaElts)
+	foreach(lc, schemaElts)
 	{
-		Node	   *element = lfirst(elements);
+		Node	   *element = lfirst(lc);
 
 		switch (nodeTag(element))
 		{
 			case T_CreateSeqStmt:
 				{
 					CreateSeqStmt *elp = (CreateSeqStmt *) element;
 
-					setSchemaName(cxt.schemaname, &elp->sequence->schemaname);
-					cxt.sequences = lappend(cxt.sequences, element);
+					checkSchemaName(pstate, schemaName, elp->sequence);
+					elements = lappend(elements, element);
 				}
 				break;
 
 			case T_CreateStmt:
 				{
 					CreateStmt *elp = (CreateStmt *) element;
 
-					setSchemaName(cxt.schemaname, &elp->relation->schemaname);
-
-					/*
-					 * XXX todo: deal with constraints
-					 */
-					cxt.tables = lappend(cxt.tables, element);
+					checkSchemaName(pstate, schemaName, elp->relation);
+					elements = lappend(elements, element);
 				}
 				break;
 
 			case T_ViewStmt:
 				{
 					ViewStmt   *elp = (ViewStmt *) element;
 
-					setSchemaName(cxt.schemaname, &elp->view->schemaname);
-
-					/*
-					 * XXX todo: deal with references between views
-					 */
-					cxt.views = lappend(cxt.views, element);
+					checkSchemaName(pstate, schemaName, elp->view);
+					elements = lappend(elements, element);
 				}
 				break;
 
 			case T_IndexStmt:
 				{
 					IndexStmt  *elp = (IndexStmt *) element;
 
-					setSchemaName(cxt.schemaname, &elp->relation->schemaname);
-					cxt.indexes = lappend(cxt.indexes, element);
+					checkSchemaName(pstate, schemaName, elp->relation);
+					elements = lappend(elements, element);
 				}
 				break;
 
 			case T_CreateTrigStmt:
 				{
 					CreateTrigStmt *elp = (CreateTrigStmt *) element;
 
-					setSchemaName(cxt.schemaname, &elp->relation->schemaname);
-					cxt.triggers = lappend(cxt.triggers, element);
+					checkSchemaName(pstate, schemaName, elp->relation);
+					elements = lappend(elements, element);
 				}
 				break;
 
 			case T_GrantStmt:
-				cxt.grants = lappend(cxt.grants, element);
+				elements = lappend(elements, element);
 				break;
 
 			default:
@@ -4186,32 +4151,40 @@ transformCreateSchemaStmtElements(List *schemaElts, const char *schemaName)
 		}
 	}
 
-	result = NIL;
-	result = list_concat(result, cxt.sequences);
-	result = list_concat(result, cxt.tables);
-	result = list_concat(result, cxt.views);
-	result = list_concat(result, cxt.indexes);
-	result = list_concat(result, cxt.triggers);
-	result = list_concat(result, cxt.grants);
-
-	return result;
+	return elements;
 }
 
 /*
- * setSchemaName
- *		Set or check schema name in an element of a CREATE SCHEMA command
+ * checkSchemaName
+ *		Check schema name in an element of a CREATE SCHEMA command
+ *
+ * It's okay if the command doesn't specify a target schema name, because
+ * CreateSchemaCommand will set up the default creation schema to be the
+ * new schema.  But if a target schema name is given, it had better match.
+ * We also have to check that the command doesn't say CREATE TEMP, since
+ * that would likewise put the object into the wrong schema.
  */
 static void
-setSchemaName(const char *context_schema, char **stmt_schema_name)
+checkSchemaName(ParseState *pstate, const char *context_schema,
+				RangeVar *relation)
 {
-	if (*stmt_schema_name == NULL)
-		*stmt_schema_name = unconstify(char *, context_schema);
-	else if (strcmp(context_schema, *stmt_schema_name) != 0)
+	if (relation->schemaname != NULL &&
+		strcmp(context_schema, relation->schemaname) != 0)
 		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_SCHEMA_DEFINITION),
-				 errmsg("CREATE specifies a schema (%s) "
+				errcode(ERRCODE_INVALID_SCHEMA_DEFINITION),
+				errmsg("CREATE specifies a schema (%s) "
 						"different from the one being created (%s)",
-						*stmt_schema_name, context_schema)));
+						relation->schemaname, context_schema),
+				parser_errposition(pstate, relation->location));
+
+	if (relation->relpersistence == RELPERSISTENCE_TEMP)
+	{
+		/* spell this error the same as in RangeVarAdjustRelationPersistence */
+		ereport(ERROR,
+				errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+				errmsg("cannot create temporary relation in non-temporary schema"),
+				parser_errposition(pstate, relation->location));
+	}
 }
 
 /*

src/backend/tcop/utility.c

@@ -1111,8 +1111,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				 * relation and attribute manipulation
 				 */
 			case T_CreateSchemaStmt:
-				CreateSchemaCommand((CreateSchemaStmt *) parsetree,
-									queryString,
+				CreateSchemaCommand(pstate, (CreateSchemaStmt *) parsetree,
 									pstmt->stmt_location,
 									pstmt->stmt_len);
 

src/include/commands/schemacmds.h

@@ -16,12 +16,9 @@
 #define SCHEMACMDS_H
 
 #include "catalog/objectaddress.h"
-#include "nodes/parsenodes.h"
-
-extern Oid	CreateSchemaCommand(CreateSchemaStmt *stmt,
-								const char *queryString,
-								int stmt_location, int stmt_len);
+#include "parser/parse_node.h"
 
+extern Oid	CreateSchemaCommand(ParseState *pstate, CreateSchemaStmt *stmt, int stmt_location, int stmt_len);
 extern ObjectAddress RenameSchema(const char *oldname, const char *newname);
 extern ObjectAddress AlterSchemaOwner(const char *name, Oid newOwnerId);
 extern void AlterSchemaOwner_oid(Oid schemaoid, Oid newOwnerId);

src/include/parser/parse_utilcmd.h

@@ -30,7 +30,8 @@ extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-extern List *transformCreateSchemaStmtElements(List *schemaElts,
+extern List *transformCreateSchemaStmtElements(ParseState *pstate,
+											   List *schemaElts,
 											   const char *schemaName);
 extern PartitionBoundSpec *transformPartitionBound(ParseState *pstate, Relation parent,
 												   PartitionBoundSpec *spec);