Create an explicit concept of collations that work for any encoding.

Use collencoding = -1 to represent such a collation in pg_collation.
We need this to make the "default" entry work sanely, and a later
patch will fix the C/POSIX entries to be represented this way instead
of duplicating them across all encodings.  All lookup operations now
search first for an entry that's database-encoding-specific, and then
for the same name with collencoding = -1.

Also some incidental code cleanup in collationcmds.c and pg_collation.c.

Author: tglsfdc

doc/src/sgml/catalogs.sgml

@@ -2132,7 +2132,8 @@
       <entry><structfield>collencoding</structfield></entry>
       <entry><type>int4</type></entry>
       <entry></entry>
-      <entry>Encoding to which the collation is applicable</entry>
+      <entry>Encoding in which the collation is applicable, or -1 if it
+       works for any encoding</entry>
      </row>
 
      <row>
@@ -2157,12 +2158,13 @@
    <structfield>collencoding</>, <structfield>collnamespace</>) not just
    (<structfield>collname</>, <structfield>collnamespace</>).
    <productname>PostgreSQL</productname> generally ignores all
-   collations not belonging to the current database's encoding; therefore
-   it is sufficient to use a qualified SQL name
+   collations that do not have <structfield>collencoding</> equal to
+   either the current database's encoding or -1, and creation of new
+   entries matching an entry with <structfield>collencoding</> = -1
+   is forbidden.  Therefore it is sufficient to use a qualified SQL name
    (<replaceable>schema</>.<replaceable>name</>) to identify a collation,
    even though this is not unique according to the catalog definition.
-   The current database's encoding is automatically used as an additional
-   lookup key.  The reason for defining the catalog this way is that
+   The reason for defining the catalog this way is that
    <application>initdb</> fills it in at cluster initialization time with
    entries for all locales available on the system, so it must be able to
    hold entries for all encodings that might ever be used in the cluster.

src/backend/catalog/information_schema.sql

@@ -448,7 +448,7 @@ CREATE VIEW collations AS
            CAST('NO PAD' AS character_data) AS pad_attribute
     FROM pg_collation c, pg_namespace nc
     WHERE c.collnamespace = nc.oid
-          AND collencoding = (SELECT encoding FROM pg_catalog.pg_database WHERE datname = pg_catalog.current_database());
+          AND collencoding IN (-1, (SELECT encoding FROM pg_database WHERE datname = current_database()));
 
 GRANT SELECT ON collations TO PUBLIC;
 
@@ -467,7 +467,7 @@ CREATE VIEW collation_character_set_applicability AS
            CAST(getdatabaseencoding() AS sql_identifier) AS character_set_name
     FROM pg_collation c, pg_namespace nc
     WHERE c.collnamespace = nc.oid
-          AND collencoding = (SELECT encoding FROM pg_catalog.pg_database WHERE datname = pg_catalog.current_database());
+          AND collencoding IN (-1, (SELECT encoding FROM pg_database WHERE datname = current_database()));
 
 GRANT SELECT ON collation_character_set_applicability TO PUBLIC;
 
@@ -2036,7 +2036,7 @@ CREATE VIEW usage_privileges AS
 
     WHERE u.oid = c.collowner
           AND c.collnamespace = n.oid
-          AND c.collencoding = (SELECT encoding FROM pg_catalog.pg_database WHERE datname = pg_catalog.current_database())
+          AND collencoding IN (-1, (SELECT encoding FROM pg_database WHERE datname = current_database()))
 
     UNION ALL
 

src/backend/catalog/namespace.c

@@ -1617,27 +1617,35 @@ OpfamilyIsVisible(Oid opfid)
  * CollationGetCollid
  *		Try to resolve an unqualified collation name.
  *		Returns OID if collation found in search path, else InvalidOid.
- *
- * This is essentially the same as RelnameGetRelid.
  */
 Oid
 CollationGetCollid(const char *collname)
 {
-	Oid			collid;
+	int32		dbencoding = GetDatabaseEncoding();
 	ListCell   *l;
 
 	recomputeNamespacePath();
 
 	foreach(l, activeSearchPath)
 	{
 		Oid			namespaceId = lfirst_oid(l);
+		Oid			collid;
 
 		if (namespaceId == myTempNamespace)
 			continue;			/* do not look in temp namespace */
 
+		/* Check for database-encoding-specific entry */
+		collid = GetSysCacheOid3(COLLNAMEENCNSP,
+								 PointerGetDatum(collname),
+								 Int32GetDatum(dbencoding),
+								 ObjectIdGetDatum(namespaceId));
+		if (OidIsValid(collid))
+			return collid;
+
+		/* Check for any-encoding entry */
 		collid = GetSysCacheOid3(COLLNAMEENCNSP,
 								 PointerGetDatum(collname),
-								 Int32GetDatum(GetDatabaseEncoding()),
+								 Int32GetDatum(-1),
 								 ObjectIdGetDatum(namespaceId));
 		if (OidIsValid(collid))
 			return collid;
@@ -2901,12 +2909,10 @@ get_collation_oid(List *name, bool missing_ok)
 {
 	char	   *schemaname;
 	char	   *collation_name;
+	int32		dbencoding = GetDatabaseEncoding();
 	Oid			namespaceId;
-	Oid			colloid = InvalidOid;
+	Oid			colloid;
 	ListCell   *l;
-	int			encoding;
-
-	encoding = GetDatabaseEncoding();
 
 	/* deconstruct the name list */
 	DeconstructQualifiedName(name, &schemaname, &collation_name);
@@ -2915,10 +2921,20 @@ get_collation_oid(List *name, bool missing_ok)
 	{
 		/* use exact schema given */
 		namespaceId = LookupExplicitNamespace(schemaname);
+
+		/* first try for encoding-specific entry, then any-encoding */
 		colloid = GetSysCacheOid3(COLLNAMEENCNSP,
 								  PointerGetDatum(collation_name),
-								  Int32GetDatum(encoding),
+								  Int32GetDatum(dbencoding),
 								  ObjectIdGetDatum(namespaceId));
+		if (OidIsValid(colloid))
+			return colloid;
+		colloid = GetSysCacheOid3(COLLNAMEENCNSP,
+								  PointerGetDatum(collation_name),
+								  Int32GetDatum(-1),
+								  ObjectIdGetDatum(namespaceId));
+		if (OidIsValid(colloid))
+			return colloid;
 	}
 	else
 	{
@@ -2934,20 +2950,26 @@ get_collation_oid(List *name, bool missing_ok)
 
 			colloid = GetSysCacheOid3(COLLNAMEENCNSP,
 									  PointerGetDatum(collation_name),
-									  Int32GetDatum(encoding),
+									  Int32GetDatum(dbencoding),
+									  ObjectIdGetDatum(namespaceId));
+			if (OidIsValid(colloid))
+				return colloid;
+			colloid = GetSysCacheOid3(COLLNAMEENCNSP,
+									  PointerGetDatum(collation_name),
+									  Int32GetDatum(-1),
 									  ObjectIdGetDatum(namespaceId));
 			if (OidIsValid(colloid))
 				return colloid;
 		}
 	}
 
 	/* Not found in path */
-	if (!OidIsValid(colloid) && !missing_ok)
+	if (!missing_ok)
 		ereport(ERROR,
 				(errcode(ERRCODE_UNDEFINED_OBJECT),
-				 errmsg("collation \"%s\" for current database encoding \"%s\" does not exist",
+				 errmsg("collation \"%s\" for encoding \"%s\" does not exist",
 						NameListToString(name), GetDatabaseEncodingName())));
-	return colloid;
+	return InvalidOid;
 }
 
 /*

src/backend/catalog/pg_collation.c

@@ -14,6 +14,7 @@
  */
 #include "postgres.h"
 
+#include "access/genam.h"
 #include "access/heapam.h"
 #include "access/sysattr.h"
 #include "catalog/dependency.h"
@@ -22,16 +23,13 @@
 #include "catalog/pg_collation.h"
 #include "catalog/pg_collation_fn.h"
 #include "catalog/pg_namespace.h"
-#include "catalog/pg_proc.h"
 #include "mb/pg_wchar.h"
-#include "miscadmin.h"
-#include "utils/acl.h"
 #include "utils/builtins.h"
 #include "utils/fmgroids.h"
-#include "utils/rel.h"
 #include "utils/syscache.h"
 #include "utils/tqual.h"
 
+
 /*
  * CollationCreate
  *
@@ -43,12 +41,11 @@ CollationCreate(const char *collname, Oid collnamespace,
 				int32 collencoding,
 				const char *collcollate, const char *collctype)
 {
-	int			i;
 	Relation	rel;
 	TupleDesc	tupDesc;
 	HeapTuple	tup;
-	bool		nulls[Natts_pg_collation];
 	Datum		values[Natts_pg_collation];
+	bool		nulls[Natts_pg_collation];
 	NameData	name_name, name_collate, name_ctype;
 	Oid			oid;
 	ObjectAddress myself,
@@ -60,7 +57,13 @@ CollationCreate(const char *collname, Oid collnamespace,
 	AssertArg(collcollate);
 	AssertArg(collctype);
 
-	/* make sure there is no existing collation of same name */
+	/*
+	 * Make sure there is no existing collation of same name & encoding.
+	 *
+	 * This would be caught by the unique index anyway; we're just giving
+	 * a friendlier error message.  The unique index provides a backstop
+	 * against race conditions.
+	 */
 	if (SearchSysCacheExists3(COLLNAMEENCNSP,
 							  PointerGetDatum(collname),
 							  Int32GetDatum(collencoding),
@@ -70,18 +73,27 @@ CollationCreate(const char *collname, Oid collnamespace,
 				 errmsg("collation \"%s\" for encoding \"%s\" already exists",
 						collname, pg_encoding_to_char(collencoding))));
 
+	/*
+	 * Also forbid matching an any-encoding entry.  This test of course is
+	 * not backed up by the unique index, but it's not a problem since we
+	 * don't support adding any-encoding entries after initdb.
+	 */
+	if (SearchSysCacheExists3(COLLNAMEENCNSP,
+							  PointerGetDatum(collname),
+							  Int32GetDatum(-1),
+							  ObjectIdGetDatum(collnamespace)))
+		ereport(ERROR,
+				(errcode(ERRCODE_DUPLICATE_OBJECT),
+				 errmsg("collation \"%s\" already exists",
+						collname)));
+
 	/* open pg_collation */
 	rel = heap_open(CollationRelationId, RowExclusiveLock);
-	tupDesc = rel->rd_att;
-
-	/* initialize nulls and values */
-	for (i = 0; i < Natts_pg_collation; i++)
-	{
-		nulls[i] = false;
-		values[i] = (Datum) NULL;
-	}
+	tupDesc = RelationGetDescr(rel);
 
 	/* form a tuple */
+	memset(nulls, 0, sizeof(nulls));
+
 	namestrcpy(&name_name, collname);
 	values[Anum_pg_collation_collname - 1] = NameGetDatum(&name_name);
 	values[Anum_pg_collation_collnamespace - 1] = ObjectIdGetDatum(collnamespace);
@@ -101,8 +113,9 @@ CollationCreate(const char *collname, Oid collnamespace,
 	/* update the index if any */
 	CatalogUpdateIndexes(rel, tup);
 
+	/* set up dependencies for the new collation */
 	myself.classId = CollationRelationId;
-	myself.objectId = HeapTupleGetOid(tup);
+	myself.objectId = oid;
 	myself.objectSubId = 0;
 
 	/* create dependency on namespace */
@@ -120,7 +133,7 @@ CollationCreate(const char *collname, Oid collnamespace,
 
 	/* Post creation hook for new collation */
 	InvokeObjectAccessHook(OAT_POST_CREATE,
-						   CollationRelationId, HeapTupleGetOid(tup), 0);
+						   CollationRelationId, oid, 0);
 
 	heap_freetuple(tup);
 	heap_close(rel, RowExclusiveLock);
@@ -138,26 +151,28 @@ void
 RemoveCollationById(Oid collationOid)
 {
 	Relation	rel;
-	HeapTuple	tuple;
-	HeapScanDesc scan;
 	ScanKeyData scanKeyData;
+	SysScanDesc scandesc;
+	HeapTuple	tuple;
+
+	rel = heap_open(CollationRelationId, RowExclusiveLock);
 
 	ScanKeyInit(&scanKeyData,
 				ObjectIdAttributeNumber,
 				BTEqualStrategyNumber, F_OIDEQ,
 				ObjectIdGetDatum(collationOid));
 
-	/* open pg_collation */
-	rel = heap_open(CollationRelationId, RowExclusiveLock);
+	scandesc = systable_beginscan(rel, CollationOidIndexId, true,
+								  SnapshotNow, 1, &scanKeyData);
 
-	scan = heap_beginscan(rel, SnapshotNow,
-						  1, &scanKeyData);
+	tuple = systable_getnext(scandesc);
 
-	/* search for the target tuple */
-	if (HeapTupleIsValid(tuple = heap_getnext(scan, ForwardScanDirection)))
+	if (HeapTupleIsValid(tuple))
 		simple_heap_delete(rel, &tuple->t_self);
 	else
 		elog(ERROR, "could not find tuple for collation %u", collationOid);
-	heap_endscan(scan);
+
+	systable_endscan(scandesc);
+
 	heap_close(rel, RowExclusiveLock);
 }