Clean up handling of COLLATE clauses in index column definitions.

Ensure that COLLATE at the top level of an index expression is treated the
same as a grammatically separate COLLATE.  Fix bogus reverse-parsing logic
in pg_get_indexdef.

src/backend/commands/indexcmds.c

@@ -837,49 +837,62 @@ ComputeIndexAttrs(IndexInfo *indexInfo,
 			attcollation = attform->attcollation;
 			ReleaseSysCache(atttuple);
 		}
-		else if (attribute->expr && IsA(attribute->expr, Var) &&
-				 ((Var *) attribute->expr)->varattno != InvalidAttrNumber)
-		{
-			/* Tricky tricky, he wrote (column) ... treat as simple attr */
-			Var		   *var = (Var *) attribute->expr;
-
-			indexInfo->ii_KeyAttrNumbers[attn] = var->varattno;
-			atttype = get_atttype(relId, var->varattno);
-			attcollation = var->varcollid;
-		}
 		else
 		{
 			/* Index expression */
-			Assert(attribute->expr != NULL);
-			indexInfo->ii_KeyAttrNumbers[attn] = 0;		/* marks expression */
-			indexInfo->ii_Expressions = lappend(indexInfo->ii_Expressions,
-												attribute->expr);
-			atttype = exprType(attribute->expr);
-			attcollation = exprCollation(attribute->expr);
+			Node   *expr = attribute->expr;
 
-			/*
-			 * We don't currently support generation of an actual query plan
-			 * for an index expression, only simple scalar expressions; hence
-			 * these restrictions.
-			 */
-			if (contain_subplans(attribute->expr))
-				ereport(ERROR,
-						(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-						 errmsg("cannot use subquery in index expression")));
-			if (contain_agg_clause(attribute->expr))
-				ereport(ERROR,
-						(errcode(ERRCODE_GROUPING_ERROR),
-				errmsg("cannot use aggregate function in index expression")));
+			Assert(expr != NULL);
+			atttype = exprType(expr);
+			attcollation = exprCollation(expr);
 
 			/*
-			 * A expression using mutable functions is probably wrong, since
-			 * if you aren't going to get the same result for the same data
-			 * every time, it's not clear what the index entries mean at all.
+			 * Strip any top-level COLLATE clause.  This ensures that we treat
+			 * "x COLLATE y" and "(x COLLATE y)" alike.
 			 */
-			if (CheckMutability((Expr *) attribute->expr))
-				ereport(ERROR,
-						(errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
-						 errmsg("functions in index expression must be marked IMMUTABLE")));
+			while (IsA(expr, CollateExpr))
+				expr = (Node *) ((CollateExpr *) expr)->arg;
+
+			if (IsA(expr, Var) &&
+				((Var *) expr)->varattno != InvalidAttrNumber)
+			{
+				/*
+				 * User wrote "(column)" or "(column COLLATE something)".
+				 * Treat it like simple attribute anyway.
+				 */
+				indexInfo->ii_KeyAttrNumbers[attn] = ((Var *) expr)->varattno;
+			}
+			else
+			{
+				indexInfo->ii_KeyAttrNumbers[attn] = 0;	/* marks expression */
+				indexInfo->ii_Expressions = lappend(indexInfo->ii_Expressions,
+													expr);
+
+				/*
+				 * We don't currently support generation of an actual query
+				 * plan for an index expression, only simple scalar
+				 * expressions; hence these restrictions.
+				 */
+				if (contain_subplans(expr))
+					ereport(ERROR,
+							(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+							 errmsg("cannot use subquery in index expression")));
+				if (contain_agg_clause(expr))
+					ereport(ERROR,
+							(errcode(ERRCODE_GROUPING_ERROR),
+							 errmsg("cannot use aggregate function in index expression")));
+
+				/*
+				 * A expression using mutable functions is probably wrong,
+				 * since if you aren't going to get the same result for the
+				 * same data every time, it's not clear what the index entries
+				 * mean at all.
+				 */
+				if (CheckMutability((Expr *) expr))
+					ereport(ERROR,
+							(errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
+							 errmsg("functions in index expression must be marked IMMUTABLE")));
+			}
 		}
 
 		/*

src/backend/utils/adt/ruleutils.c

@@ -794,7 +794,6 @@ pg_get_indexdef_worker(Oid indexrelid, int colno,
 	List	   *context;
 	Oid			indrelid;
 	int			keyno;
-	Oid			keycoltype;
 	Datum		indcollDatum;
 	Datum		indclassDatum;
 	Datum		indoptionDatum;
@@ -902,6 +901,8 @@ pg_get_indexdef_worker(Oid indexrelid, int colno,
 	{
 		AttrNumber	attnum = idxrec->indkey.values[keyno];
 		int16		opt = indoption->values[keyno];
+		Oid			keycoltype;
+		Oid			keycolcollation;
 
 		if (!colno)
 			appendStringInfoString(&buf, sep);
@@ -916,6 +917,7 @@ pg_get_indexdef_worker(Oid indexrelid, int colno,
 			if (!colno || colno == keyno + 1)
 				appendStringInfoString(&buf, quote_identifier(attname));
 			keycoltype = get_atttype(indrelid, attnum);
+			keycolcollation = get_attcollation(indrelid, attnum);
 		}
 		else
 		{
@@ -939,16 +941,18 @@ pg_get_indexdef_worker(Oid indexrelid, int colno,
 					appendStringInfo(&buf, "(%s)", str);
 			}
 			keycoltype = exprType(indexkey);
+			keycolcollation = exprCollation(indexkey);
 		}
 
 		if (!attrsOnly && (!colno || colno == keyno + 1))
 		{
-			Oid coll;
+			Oid		indcoll;
 
-			/* Add collation, if not default */
-			coll = indcollation->values[keyno];
-			if (coll && coll != DEFAULT_COLLATION_OID && coll != get_attcollation(indrelid, attnum))
-				appendStringInfo(&buf, " COLLATE %s", generate_collation_name((indcollation->values[keyno])));
+			/* Add collation, if not default for column */
+			indcoll = indcollation->values[keyno];
+			if (OidIsValid(indcoll) && indcoll != keycolcollation)
+				appendStringInfo(&buf, " COLLATE %s",
+								 generate_collation_name((indcoll)));
 
 			/* Add the operator class name, if not default */
 			get_opclass_name(indclass->values[keyno], keycoltype, &buf);
@@ -6646,7 +6650,8 @@ get_from_clause_coldeflist(List *names, List *types, List *typmods, List *collat
 						 quote_identifier(attname),
 						 format_type_with_typemod(atttypid, atttypmod));
 		if (attcollation && attcollation != DEFAULT_COLLATION_OID)
-			appendStringInfo(buf, " COLLATE %s", generate_collation_name(attcollation));
+			appendStringInfo(buf, " COLLATE %s",
+							 generate_collation_name(attcollation));
 		i++;
 	}
 

src/test/regress/expected/collate.linux.utf8.out

@@ -747,19 +747,21 @@ SELECT a, (dup(b)).* FROM collate_test3 ORDER BY 2;
 CREATE INDEX collate_test1_idx1 ON collate_test1 (b);
 CREATE INDEX collate_test1_idx2 ON collate_test1 (b COLLATE "C");
 CREATE INDEX collate_test1_idx3 ON collate_test1 ((b COLLATE "C")); -- this is different grammatically
-CREATE INDEX collate_test1_idx4 ON collate_test1 (a COLLATE "C"); -- fail
+CREATE INDEX collate_test1_idx4 ON collate_test1 (((b||'foo') COLLATE "POSIX"));
+CREATE INDEX collate_test1_idx5 ON collate_test1 (a COLLATE "C"); -- fail
 ERROR:  collations are not supported by type integer
-CREATE INDEX collate_test1_idx5 ON collate_test1 ((a COLLATE "C")); -- fail
+CREATE INDEX collate_test1_idx6 ON collate_test1 ((a COLLATE "C")); -- fail
 ERROR:  collations are not supported by type integer
-LINE 1: ...ATE INDEX collate_test1_idx5 ON collate_test1 ((a COLLATE "C...
+LINE 1: ...ATE INDEX collate_test1_idx6 ON collate_test1 ((a COLLATE "C...
                                                              ^
-SELECT relname, pg_get_indexdef(oid) FROM pg_class WHERE relname LIKE 'collate_test%_idx%';
-      relname       |                                       pg_get_indexdef                                        
---------------------+----------------------------------------------------------------------------------------------
+SELECT relname, pg_get_indexdef(oid) FROM pg_class WHERE relname LIKE 'collate_test%_idx%' ORDER BY 1;
+      relname       |                                           pg_get_indexdef                                           
+--------------------+-----------------------------------------------------------------------------------------------------
  collate_test1_idx1 | CREATE INDEX collate_test1_idx1 ON collate_test1 USING btree (b)
  collate_test1_idx2 | CREATE INDEX collate_test1_idx2 ON collate_test1 USING btree (b COLLATE "C")
- collate_test1_idx3 | CREATE INDEX collate_test1_idx3 ON collate_test1 USING btree (((b COLLATE "C")) COLLATE "C")
-(3 rows)
+ collate_test1_idx3 | CREATE INDEX collate_test1_idx3 ON collate_test1 USING btree (b COLLATE "C")
+ collate_test1_idx4 | CREATE INDEX collate_test1_idx4 ON collate_test1 USING btree (((b || 'foo'::text)) COLLATE "POSIX")
+(4 rows)
 
 -- schema manipulation commands
 CREATE ROLE regress_test_role;

src/test/regress/expected/collate.out

@@ -524,21 +524,23 @@ SELECT a, (dup(b)).* FROM collate_test2 ORDER BY 2;
 
 -- indexes
 CREATE INDEX collate_test1_idx1 ON collate_test1 (b);
-CREATE INDEX collate_test1_idx2 ON collate_test1 (b COLLATE "C");
-CREATE INDEX collate_test1_idx3 ON collate_test1 ((b COLLATE "C")); -- this is different grammatically
-CREATE INDEX collate_test1_idx4 ON collate_test1 (a COLLATE "C"); -- fail
+CREATE INDEX collate_test1_idx2 ON collate_test1 (b COLLATE "POSIX");
+CREATE INDEX collate_test1_idx3 ON collate_test1 ((b COLLATE "POSIX")); -- this is different grammatically
+CREATE INDEX collate_test1_idx4 ON collate_test1 (((b||'foo') COLLATE "POSIX"));
+CREATE INDEX collate_test1_idx5 ON collate_test1 (a COLLATE "POSIX"); -- fail
 ERROR:  collations are not supported by type integer
-CREATE INDEX collate_test1_idx5 ON collate_test1 ((a COLLATE "C")); -- fail
+CREATE INDEX collate_test1_idx6 ON collate_test1 ((a COLLATE "POSIX")); -- fail
 ERROR:  collations are not supported by type integer
-LINE 1: ...ATE INDEX collate_test1_idx5 ON collate_test1 ((a COLLATE "C...
+LINE 1: ...ATE INDEX collate_test1_idx6 ON collate_test1 ((a COLLATE "P...
                                                              ^
-SELECT relname, pg_get_indexdef(oid) FROM pg_class WHERE relname LIKE 'collate_test%_idx%';
-      relname       |                                       pg_get_indexdef                                        
---------------------+----------------------------------------------------------------------------------------------
+SELECT relname, pg_get_indexdef(oid) FROM pg_class WHERE relname LIKE 'collate_test%_idx%' ORDER BY 1;
+      relname       |                                           pg_get_indexdef                                           
+--------------------+-----------------------------------------------------------------------------------------------------
  collate_test1_idx1 | CREATE INDEX collate_test1_idx1 ON collate_test1 USING btree (b)
- collate_test1_idx2 | CREATE INDEX collate_test1_idx2 ON collate_test1 USING btree (b)
- collate_test1_idx3 | CREATE INDEX collate_test1_idx3 ON collate_test1 USING btree (((b COLLATE "C")) COLLATE "C")
-(3 rows)
+ collate_test1_idx2 | CREATE INDEX collate_test1_idx2 ON collate_test1 USING btree (b COLLATE "POSIX")
+ collate_test1_idx3 | CREATE INDEX collate_test1_idx3 ON collate_test1 USING btree (b COLLATE "POSIX")
+ collate_test1_idx4 | CREATE INDEX collate_test1_idx4 ON collate_test1 USING btree (((b || 'foo'::text)) COLLATE "POSIX")
+(4 rows)
 
 --
 -- Clean up.  Many of these table names will be re-used if the user is

src/test/regress/sql/collate.linux.utf8.sql

@@ -231,11 +231,12 @@ SELECT a, (dup(b)).* FROM collate_test3 ORDER BY 2;
 CREATE INDEX collate_test1_idx1 ON collate_test1 (b);
 CREATE INDEX collate_test1_idx2 ON collate_test1 (b COLLATE "C");
 CREATE INDEX collate_test1_idx3 ON collate_test1 ((b COLLATE "C")); -- this is different grammatically
+CREATE INDEX collate_test1_idx4 ON collate_test1 (((b||'foo') COLLATE "POSIX"));
 
-CREATE INDEX collate_test1_idx4 ON collate_test1 (a COLLATE "C"); -- fail
-CREATE INDEX collate_test1_idx5 ON collate_test1 ((a COLLATE "C")); -- fail
+CREATE INDEX collate_test1_idx5 ON collate_test1 (a COLLATE "C"); -- fail
+CREATE INDEX collate_test1_idx6 ON collate_test1 ((a COLLATE "C")); -- fail
 
-SELECT relname, pg_get_indexdef(oid) FROM pg_class WHERE relname LIKE 'collate_test%_idx%';
+SELECT relname, pg_get_indexdef(oid) FROM pg_class WHERE relname LIKE 'collate_test%_idx%' ORDER BY 1;
 
 
 -- schema manipulation commands

src/test/regress/sql/collate.sql

@@ -178,13 +178,14 @@ SELECT a, (dup(b)).* FROM collate_test2 ORDER BY 2;
 -- indexes
 
 CREATE INDEX collate_test1_idx1 ON collate_test1 (b);
-CREATE INDEX collate_test1_idx2 ON collate_test1 (b COLLATE "C");
-CREATE INDEX collate_test1_idx3 ON collate_test1 ((b COLLATE "C")); -- this is different grammatically
+CREATE INDEX collate_test1_idx2 ON collate_test1 (b COLLATE "POSIX");
+CREATE INDEX collate_test1_idx3 ON collate_test1 ((b COLLATE "POSIX")); -- this is different grammatically
+CREATE INDEX collate_test1_idx4 ON collate_test1 (((b||'foo') COLLATE "POSIX"));
 
-CREATE INDEX collate_test1_idx4 ON collate_test1 (a COLLATE "C"); -- fail
-CREATE INDEX collate_test1_idx5 ON collate_test1 ((a COLLATE "C")); -- fail
+CREATE INDEX collate_test1_idx5 ON collate_test1 (a COLLATE "POSIX"); -- fail
+CREATE INDEX collate_test1_idx6 ON collate_test1 ((a COLLATE "POSIX")); -- fail
 
-SELECT relname, pg_get_indexdef(oid) FROM pg_class WHERE relname LIKE 'collate_test%_idx%';
+SELECT relname, pg_get_indexdef(oid) FROM pg_class WHERE relname LIKE 'collate_test%_idx%' ORDER BY 1;
 
 --
 -- Clean up.  Many of these table names will be re-used if the user is