Fix reporting of column typmods for multi-row VALUES constructs.

expandRTE() and get_rte_attribute_type() reported the exprType() and
exprTypmod() values of the expressions in the first row of the VALUES as
being the column type/typmod returned by the VALUES RTE.  That's fine for
the data type, since we coerce all expressions in a column to have the same
common type.  But we don't coerce them to have a common typmod, so it was
possible for rows after the first one to return values that violate the
claimed column typmod.  This leads to the incorrect result seen in bug
#14448 from Hassan Mahmood, as well as some other corner-case misbehaviors.

The desired behavior is the same as we use in other type-unification
cases: report the common typmod if there is one, but otherwise return -1
indicating no particular constraint.  It's cheap for transformValuesClause
to determine the common typmod while transforming a multi-row VALUES, but
it'd be less cheap for expandRTE() and get_rte_attribute_type() to
re-determine that info every time they're asked --- possibly a lot less
cheap, if the VALUES has many rows.  Therefore, the best fix is to record
the common typmods explicitly in a list in the VALUES RTE, as we were
already doing for column collations.  This looks quite a bit like what
we're doing for CTE RTEs, so we can save a little bit of space and code by
unifying the representation for those two RTE types.  They both now share
coltypes/coltypmods/colcollations fields.  (At some point it might seem
desirable to populate those fields for all RTE types; but right now it
looks like constructing them for other RTE types would add more code and
cycles than it would save.)

The RTE change requires a catversion bump, so this fix is only usable
in HEAD.  If we fix this at all in the back branches, the patch will
need to look quite different.

Report: https://postgr.es/m/20161205143037.4377.60754@wrigleys.postgresql.org
Discussion: https://postgr.es/m/27429.1480968538@sss.pgh.pa.us

src/backend/nodes/copyfuncs.c

@@ -2149,13 +2149,12 @@ _copyRangeTblEntry(const RangeTblEntry *from)
 	COPY_NODE_FIELD(functions);
 	COPY_SCALAR_FIELD(funcordinality);
 	COPY_NODE_FIELD(values_lists);
-	COPY_NODE_FIELD(values_collations);
 	COPY_STRING_FIELD(ctename);
 	COPY_SCALAR_FIELD(ctelevelsup);
 	COPY_SCALAR_FIELD(self_reference);
-	COPY_NODE_FIELD(ctecoltypes);
+	COPY_NODE_FIELD(coltypes);
-	COPY_NODE_FIELD(ctecoltypmods);
+	COPY_NODE_FIELD(coltypmods);
-	COPY_NODE_FIELD(ctecolcollations);
+	COPY_NODE_FIELD(colcollations);
 	COPY_NODE_FIELD(alias);
 	COPY_NODE_FIELD(eref);
 	COPY_SCALAR_FIELD(lateral);

src/backend/nodes/equalfuncs.c

@@ -2460,13 +2460,12 @@ _equalRangeTblEntry(const RangeTblEntry *a, const RangeTblEntry *b)
 	COMPARE_NODE_FIELD(functions);
 	COMPARE_SCALAR_FIELD(funcordinality);
 	COMPARE_NODE_FIELD(values_lists);
-	COMPARE_NODE_FIELD(values_collations);
 	COMPARE_STRING_FIELD(ctename);
 	COMPARE_SCALAR_FIELD(ctelevelsup);
 	COMPARE_SCALAR_FIELD(self_reference);
-	COMPARE_NODE_FIELD(ctecoltypes);
+	COMPARE_NODE_FIELD(coltypes);
-	COMPARE_NODE_FIELD(ctecoltypmods);
+	COMPARE_NODE_FIELD(coltypmods);
-	COMPARE_NODE_FIELD(ctecolcollations);
+	COMPARE_NODE_FIELD(colcollations);
 	COMPARE_NODE_FIELD(alias);
 	COMPARE_NODE_FIELD(eref);
 	COMPARE_SCALAR_FIELD(lateral);

src/backend/nodes/outfuncs.c

@@ -2841,15 +2841,17 @@ _outRangeTblEntry(StringInfo str, const RangeTblEntry *node)
 			break;
 		case RTE_VALUES:
 			WRITE_NODE_FIELD(values_lists);
-			WRITE_NODE_FIELD(values_collations);
+			WRITE_NODE_FIELD(coltypes);
+			WRITE_NODE_FIELD(coltypmods);
+			WRITE_NODE_FIELD(colcollations);
 			break;
 		case RTE_CTE:
 			WRITE_STRING_FIELD(ctename);
 			WRITE_UINT_FIELD(ctelevelsup);
 			WRITE_BOOL_FIELD(self_reference);
-			WRITE_NODE_FIELD(ctecoltypes);
+			WRITE_NODE_FIELD(coltypes);
-			WRITE_NODE_FIELD(ctecoltypmods);
+			WRITE_NODE_FIELD(coltypmods);
-			WRITE_NODE_FIELD(ctecolcollations);
+			WRITE_NODE_FIELD(colcollations);
 			break;
 		default:
 			elog(ERROR, "unrecognized RTE kind: %d", (int) node->rtekind);

src/backend/nodes/readfuncs.c

@@ -1314,15 +1314,17 @@ _readRangeTblEntry(void)
 			break;
 		case RTE_VALUES:
 			READ_NODE_FIELD(values_lists);
-			READ_NODE_FIELD(values_collations);
+			READ_NODE_FIELD(coltypes);
+			READ_NODE_FIELD(coltypmods);
+			READ_NODE_FIELD(colcollations);
 			break;
 		case RTE_CTE:
 			READ_STRING_FIELD(ctename);
 			READ_UINT_FIELD(ctelevelsup);
 			READ_BOOL_FIELD(self_reference);
-			READ_NODE_FIELD(ctecoltypes);
+			READ_NODE_FIELD(coltypes);
-			READ_NODE_FIELD(ctecoltypmods);
+			READ_NODE_FIELD(coltypmods);
-			READ_NODE_FIELD(ctecolcollations);
+			READ_NODE_FIELD(colcollations);
 			break;
 		default:
 			elog(ERROR, "unrecognized RTE kind: %d",

src/backend/optimizer/plan/setrefs.c

@@ -396,10 +396,9 @@ add_rte_to_flat_rtable(PlannerGlobal *glob, RangeTblEntry *rte)
 	newrte->joinaliasvars = NIL;
 	newrte->functions = NIL;
 	newrte->values_lists = NIL;
-	newrte->values_collations = NIL;
+	newrte->coltypes = NIL;
-	newrte->ctecoltypes = NIL;
+	newrte->coltypmods = NIL;
-	newrte->ctecoltypmods = NIL;
+	newrte->colcollations = NIL;
-	newrte->ctecolcollations = NIL;
 	newrte->securityQuals = NIL;
 
 	glob->finalrtable = lappend(glob->finalrtable, newrte);

src/backend/parser/analyze.c

@@ -633,10 +633,11 @@ transformInsertStmt(ParseState *pstate, InsertStmt *stmt)
 		 * RTE.
 		 */
 		List	   *exprsLists = NIL;
-		List	   *collations = NIL;
+		List	   *coltypes = NIL;
+		List	   *coltypmods = NIL;
+		List	   *colcollations = NIL;
 		int			sublist_length = -1;
 		bool		lateral = false;
-		int			i;
 
 		Assert(selectStmt->intoClause == NULL);
 
@@ -703,11 +704,20 @@ transformInsertStmt(ParseState *pstate, InsertStmt *stmt)
 		}
 
 		/*
-		 * Although we don't really need collation info, let's just make sure
+		 * Construct column type/typmod/collation lists for the VALUES RTE.
-		 * we provide a correctly-sized list in the VALUES RTE.
+		 * Every expression in each column has been coerced to the type/typmod
+		 * of the corresponding target column or subfield, so it's sufficient
+		 * to look at the exprType/exprTypmod of the first row.  We don't care
+		 * about the collation labeling, so just fill in InvalidOid for that.
 		 */
-		for (i = 0; i < sublist_length; i++)
+		foreach(lc, (List *) linitial(exprsLists))
-			collations = lappend_oid(collations, InvalidOid);
+		{
+			Node	   *val = (Node *) lfirst(lc);
+
+			coltypes = lappend_oid(coltypes, exprType(val));
+			coltypmods = lappend_int(coltypmods, exprTypmod(val));
+			colcollations = lappend_oid(colcollations, InvalidOid);
+		}
 
 		/*
 		 * Ordinarily there can't be any current-level Vars in the expression
@@ -722,7 +732,8 @@ transformInsertStmt(ParseState *pstate, InsertStmt *stmt)
 		/*
 		 * Generate the VALUES RTE
 		 */
-		rte = addRangeTableEntryForValues(pstate, exprsLists, collations,
+		rte = addRangeTableEntryForValues(pstate, exprsLists,
+										  coltypes, coltypmods, colcollations,
 										  NULL, lateral, true);
 		rtr = makeNode(RangeTblRef);
 		/* assume new rte is at end */
@@ -1274,7 +1285,9 @@ transformValuesClause(ParseState *pstate, SelectStmt *stmt)
 {
 	Query	   *qry = makeNode(Query);
 	List	   *exprsLists;
-	List	   *collations;
+	List	   *coltypes = NIL;
+	List	   *coltypmods = NIL;
+	List	   *colcollations = NIL;
 	List	  **colexprs = NULL;
 	int			sublist_length = -1;
 	bool		lateral = false;
@@ -1360,8 +1373,8 @@ transformValuesClause(ParseState *pstate, SelectStmt *stmt)
 
 	/*
 	 * Now resolve the common types of the columns, and coerce everything to
-	 * those types.  Then identify the common collation, if any, of each
+	 * those types.  Then identify the common typmod and common collation, if
-	 * column.
+	 * any, of each column.
 	 *
 	 * We must do collation processing now because (1) assign_query_collations
 	 * doesn't process rangetable entries, and (2) we need to label the VALUES
@@ -1372,11 +1385,12 @@ transformValuesClause(ParseState *pstate, SelectStmt *stmt)
 	 *
 	 * Note we modify the per-column expression lists in-place.
 	 */
-	collations = NIL;
 	for (i = 0; i < sublist_length; i++)
 	{
 		Oid			coltype;
+		int32		coltypmod = -1;
 		Oid			colcoll;
+		bool		first = true;
 
 		coltype = select_common_type(pstate, colexprs[i], "VALUES", NULL);
 
@@ -1386,11 +1400,24 @@ transformValuesClause(ParseState *pstate, SelectStmt *stmt)
 
 			col = coerce_to_common_type(pstate, col, coltype, "VALUES");
 			lfirst(lc) = (void *) col;
+			if (first)
+			{
+				coltypmod = exprTypmod(col);
+				first = false;
+			}
+			else
+			{
+				/* As soon as we see a non-matching typmod, fall back to -1 */
+				if (coltypmod >= 0 && coltypmod != exprTypmod(col))
+					coltypmod = -1;
+			}
 		}
 
 		colcoll = select_common_collation(pstate, colexprs[i], true);
 
-		collations = lappend_oid(collations, colcoll);
+		coltypes = lappend_oid(coltypes, coltype);
+		coltypmods = lappend_int(coltypmods, coltypmod);
+		colcollations = lappend_oid(colcollations, colcoll);
 	}
 
 	/*
@@ -1432,7 +1459,8 @@ transformValuesClause(ParseState *pstate, SelectStmt *stmt)
 	/*
 	 * Generate the VALUES RTE
 	 */
-	rte = addRangeTableEntryForValues(pstate, exprsLists, collations,
+	rte = addRangeTableEntryForValues(pstate, exprsLists,
+									  coltypes, coltypmods, colcollations,
 									  NULL, lateral, true);
 	addRTEtoQuery(pstate, rte, true, true, true);
 

src/backend/parser/parse_relation.c

@@ -1635,7 +1635,9 @@ addRangeTableEntryForFunction(ParseState *pstate,
 RangeTblEntry *
 addRangeTableEntryForValues(ParseState *pstate,
 							List *exprs,
-							List *collations,
+							List *coltypes,
+							List *coltypmods,
+							List *colcollations,
 							Alias *alias,
 							bool lateral,
 							bool inFromCl)
@@ -1652,7 +1654,9 @@ addRangeTableEntryForValues(ParseState *pstate,
 	rte->relid = InvalidOid;
 	rte->subquery = NULL;
 	rte->values_lists = exprs;
-	rte->values_collations = collations;
+	rte->coltypes = coltypes;
+	rte->coltypmods = coltypmods;
+	rte->colcollations = colcollations;
 	rte->alias = alias;
 
 	eref = alias ? copyObject(alias) : makeAlias(refname, NIL);
@@ -1822,9 +1826,9 @@ addRangeTableEntryForCTE(ParseState *pstate,
 					 parser_errposition(pstate, rv->location)));
 	}
 
-	rte->ctecoltypes = cte->ctecoltypes;
+	rte->coltypes = cte->ctecoltypes;
-	rte->ctecoltypmods = cte->ctecoltypmods;
+	rte->coltypmods = cte->ctecoltypmods;
-	rte->ctecolcollations = cte->ctecolcollations;
+	rte->colcollations = cte->ctecolcollations;
 
 	rte->alias = alias;
 	if (alias)
@@ -2153,46 +2157,6 @@ expandRTE(RangeTblEntry *rte, int rtindex, int sublevels_up,
 				}
 			}
 			break;
-		case RTE_VALUES:
-			{
-				/* Values RTE */
-				ListCell   *aliasp_item = list_head(rte->eref->colnames);
-				ListCell   *lcv;
-				ListCell   *lcc;
-
-				varattno = 0;
-				forboth(lcv, (List *) linitial(rte->values_lists),
-						lcc, rte->values_collations)
-				{
-					Node	   *col = (Node *) lfirst(lcv);
-					Oid			colcollation = lfirst_oid(lcc);
-
-					varattno++;
-					if (colnames)
-					{
-						/* Assume there is one alias per column */
-						char	   *label = strVal(lfirst(aliasp_item));
-
-						*colnames = lappend(*colnames,
-											makeString(pstrdup(label)));
-						aliasp_item = lnext(aliasp_item);
-					}
-
-					if (colvars)
-					{
-						Var		   *varnode;
-
-						varnode = makeVar(rtindex, varattno,
-										  exprType(col),
-										  exprTypmod(col),
-										  colcollation,
-										  sublevels_up);
-						varnode->location = location;
-						*colvars = lappend(*colvars, varnode);
-					}
-				}
-			}
-			break;
 		case RTE_JOIN:
 			{
 				/* Join RTE */
@@ -2262,17 +2226,19 @@ expandRTE(RangeTblEntry *rte, int rtindex, int sublevels_up,
 				}
 			}
 			break;
+		case RTE_VALUES:
 		case RTE_CTE:
 			{
+				/* Values or CTE RTE */
 				ListCell   *aliasp_item = list_head(rte->eref->colnames);
 				ListCell   *lct;
 				ListCell   *lcm;
 				ListCell   *lcc;
 
 				varattno = 0;
-				forthree(lct, rte->ctecoltypes,
+				forthree(lct, rte->coltypes,
-						 lcm, rte->ctecoltypmods,
+						 lcm, rte->coltypmods,
-						 lcc, rte->ctecolcollations)
+						 lcc, rte->colcollations)
 				{
 					Oid			coltype = lfirst_oid(lct);
 					int32		coltypmod = lfirst_int(lcm);
@@ -2285,7 +2251,8 @@ expandRTE(RangeTblEntry *rte, int rtindex, int sublevels_up,
 						/* Assume there is one alias per output column */
 						char	   *label = strVal(lfirst(aliasp_item));
 
-						*colnames = lappend(*colnames, makeString(pstrdup(label)));
+						*colnames = lappend(*colnames,
+											makeString(pstrdup(label)));
 						aliasp_item = lnext(aliasp_item);
 					}
 
@@ -2296,6 +2263,8 @@ expandRTE(RangeTblEntry *rte, int rtindex, int sublevels_up,
 						varnode = makeVar(rtindex, varattno,
 										  coltype, coltypmod, colcoll,
 										  sublevels_up);
+						varnode->location = location;
+
 						*colvars = lappend(*colvars, varnode);
 					}
 				}
@@ -2654,22 +2623,6 @@ get_rte_attribute_type(RangeTblEntry *rte, AttrNumber attnum,
 								rte->eref->aliasname)));
 			}
 			break;
-		case RTE_VALUES:
-			{
-				/* Values RTE --- get type info from first sublist */
-				/* collation is stored separately, though */
-				List	   *collist = (List *) linitial(rte->values_lists);
-				Node	   *col;
-
-				if (attnum < 1 || attnum > list_length(collist))
-					elog(ERROR, "values list %s does not have attribute %d",
-						 rte->eref->aliasname, attnum);
-				col = (Node *) list_nth(collist, attnum - 1);
-				*vartype = exprType(col);
-				*vartypmod = exprTypmod(col);
-				*varcollid = list_nth_oid(rte->values_collations, attnum - 1);
-			}
-			break;
 		case RTE_JOIN:
 			{
 				/*
@@ -2685,13 +2638,14 @@ get_rte_attribute_type(RangeTblEntry *rte, AttrNumber attnum,
 				*varcollid = exprCollation(aliasvar);
 			}
 			break;
+		case RTE_VALUES:
 		case RTE_CTE:
 			{
-				/* CTE RTE --- get type info from lists in the RTE */
+				/* VALUES or CTE RTE --- get type info from lists in the RTE */
-				Assert(attnum > 0 && attnum <= list_length(rte->ctecoltypes));
+				Assert(attnum > 0 && attnum <= list_length(rte->coltypes));
-				*vartype = list_nth_oid(rte->ctecoltypes, attnum - 1);
+				*vartype = list_nth_oid(rte->coltypes, attnum - 1);
-				*vartypmod = list_nth_int(rte->ctecoltypmods, attnum - 1);
+				*vartypmod = list_nth_int(rte->coltypmods, attnum - 1);
-				*varcollid = list_nth_oid(rte->ctecolcollations, attnum - 1);
+				*varcollid = list_nth_oid(rte->colcollations, attnum - 1);
 			}
 			break;
 		default:

src/include/catalog/catversion.h

@@ -53,6 +53,6 @@
  */
 
 /*							yyyymmddN */
-#define CATALOG_VERSION_NO	201612071
+#define CATALOG_VERSION_NO	201612081
 
 #endif

src/include/nodes/parsenodes.h

@@ -927,7 +927,6 @@ typedef struct RangeTblEntry
 	 * Fields valid for a values RTE (else NIL):
 	 */
 	List	   *values_lists;	/* list of expression lists */
-	List	   *values_collations;		/* OID list of column collation OIDs */
 
 	/*
 	 * Fields valid for a CTE RTE (else NULL/zero):
@@ -935,9 +934,17 @@ typedef struct RangeTblEntry
 	char	   *ctename;		/* name of the WITH list item */
 	Index		ctelevelsup;	/* number of query levels up */
 	bool		self_reference; /* is this a recursive self-reference? */
-	List	   *ctecoltypes;	/* OID list of column type OIDs */
+
-	List	   *ctecoltypmods;	/* integer list of column typmods */
+	/*
-	List	   *ctecolcollations;		/* OID list of column collation OIDs */
+	 * Fields valid for values and CTE RTEs (else NIL):
+	 *
+	 * We need these for CTE RTEs so that the types of self-referential
+	 * columns are well-defined.  For VALUES RTEs, storing these explicitly
+	 * saves having to re-determine the info by scanning the values_lists.
+	 */
+	List	   *coltypes;		/* OID list of column type OIDs */
+	List	   *coltypmods;		/* integer list of column typmods */
+	List	   *colcollations;	/* OID list of column collation OIDs */
 
 	/*
 	 * Fields valid in all RTEs:

src/include/parser/parse_relation.h

@@ -85,7 +85,9 @@ extern RangeTblEntry *addRangeTableEntryForFunction(ParseState *pstate,
 							  bool inFromCl);
 extern RangeTblEntry *addRangeTableEntryForValues(ParseState *pstate,
 							List *exprs,
-							List *collations,
+							List *coltypes,
+							List *coltypmods,
+							List *colcollations,
 							Alias *alias,
 							bool lateral,
 							bool inFromCl);

src/test/regress/expected/create_view.out

@@ -288,6 +288,43 @@ SELECT relname, relkind, reloptions FROM pg_class
  mysecview4 | v       | {security_barrier=false}
 (4 rows)
 
+-- This test checks that proper typmods are assigned in a multi-row VALUES
+CREATE VIEW tt1 AS
+  SELECT * FROM (
+    VALUES
+       ('abc'::varchar(3), '0123456789', 42, 'abcd'::varchar(4)),
+       ('0123456789', 'abc'::varchar(3), 42.12, 'abc'::varchar(4))
+  ) vv(a,b,c,d);
+\d+ tt1
+                                View "testviewschm2.tt1"
+ Column |         Type         | Collation | Nullable | Default | Storage  | Description 
+--------+----------------------+-----------+----------+---------+----------+-------------
+ a      | character varying    |           |          |         | extended | 
+ b      | character varying    |           |          |         | extended | 
+ c      | numeric              |           |          |         | main     | 
+ d      | character varying(4) |           |          |         | extended | 
+View definition:
+ SELECT vv.a,
+    vv.b,
+    vv.c,
+    vv.d
+   FROM ( VALUES ('abc'::character varying(3),'0123456789'::character varying,42,'abcd'::character varying(4)), ('0123456789'::character varying,'abc'::character varying(3),42.12,'abc'::character varying(4))) vv(a, b, c, d);
+
+SELECT * FROM tt1;
+     a      |     b      |   c   |  d   
+------------+------------+-------+------
+ abc        | 0123456789 |    42 | abcd
+ 0123456789 | abc        | 42.12 | abc
+(2 rows)
+
+SELECT a::varchar(3) FROM tt1;
+  a  
+-----
+ abc
+ 012
+(2 rows)
+
+DROP VIEW tt1;
 -- Test view decompilation in the face of relation renaming conflicts
 CREATE TABLE tt1 (f1 int, f2 int, f3 text);
 CREATE TABLE tx1 (x1 int, x2 int, x3 text);

src/test/regress/sql/create_view.sql

@@ -224,6 +224,19 @@ SELECT relname, relkind, reloptions FROM pg_class
                      'mysecview3'::regclass, 'mysecview4'::regclass)
        ORDER BY relname;
 
+-- This test checks that proper typmods are assigned in a multi-row VALUES
+
+CREATE VIEW tt1 AS
+  SELECT * FROM (
+    VALUES
+       ('abc'::varchar(3), '0123456789', 42, 'abcd'::varchar(4)),
+       ('0123456789', 'abc'::varchar(3), 42.12, 'abc'::varchar(4))
+  ) vv(a,b,c,d);
+\d+ tt1
+SELECT * FROM tt1;
+SELECT a::varchar(3) FROM tt1;
+DROP VIEW tt1;
+
 -- Test view decompilation in the face of relation renaming conflicts
 
 CREATE TABLE tt1 (f1 int, f2 int, f3 text);