I attach a version of my toast-slicing patch, against current CVS

(current as of a few hours ago.)

This patch:

1. Adds PG_GETARG_xxx_P_SLICE() macros and associated support routines.

2. Adds routines in src/backend/access/tuptoaster.c for fetching only
necessary chunks of a toasted value. (Modelled on latest changes to
assume chunks are returned in order).

3. Amends text_substr and bytea_substr to use new methods. It now
handles multibyte cases -and should still lead to a performance
improvement in the multibyte case where the substring is near the
beginning of the string.

4. Added new command: ALTER TABLE tabname ALTER COLUMN colname SET
STORAGE {PLAIN | EXTERNAL | EXTENDED | MAIN} to parser and documented in
alter-table.sgml. (NB I used ColId as the item type for the storage
mode string, rather than a new production - I hope this makes sense!).
All this does is sets attstorage for the specified column.

4. AlterTableAlterColumnStatistics is now AlterTableAlterColumnFlags and
handles both statistics and storage (it uses the subtype code to
distinguish). The previous version of my patch also re-arranged other
code in backend/commands/command.c but I have dropped that from this
patch.(I plan to return to it separately).

5. Documented new macros (and also the PG_GETARG_xxx_P_COPY macros) in
xfunc.sgml. ref/alter_table.sgml also contains documentation for ALTER
COLUMN SET STORAGE.

John Gray

1 files changed, 45 insertions, 42 deletions

diff --git a/src/backend/utils/adt/varlena.c b/src/backend/utils/adt/varlena.c
index 7e55bfa4ba1..b10ec7d6855 100644
--- a/src/backend/utils/adt/varlena.c
+++ b/src/backend/utils/adt/varlena.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/utils/adt/varlena.c,v 1.78 2001/11/19 19:15:07 momjian Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/utils/adt/varlena.c,v 1.79 2002/03/05 05:33:19 momjian Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -332,49 +332,71 @@ textcat(PG_FUNCTION_ARGS)
  * Changed behavior if starting position is less than one to conform to SQL92 behavior.
  * Formerly returned the entire string; now returns a portion.
  * - Thomas Lockhart 1998-12-10
+ * Now uses faster TOAST-slicing interface
+ * - John Gray 2002-02-22
  */
 Datum
 text_substr(PG_FUNCTION_ARGS)
 {
-	text	   *string = PG_GETARG_TEXT_P(0);
+	text	   *string;
 	int32		m = PG_GETARG_INT32(1);
 	int32		n = PG_GETARG_INT32(2);
-	text	   *ret;
-	int			len;
-
+	int32       sm;
+	int32       sn;
+	int         eml = 1;
 #ifdef MULTIBYTE
 	int			i;
+	int			len;
+	text	   *ret;
 	char	   *p;
-#endif
-
-	len = VARSIZE(string) - VARHDRSZ;
-#ifdef MULTIBYTE
-	len = pg_mbstrlen_with_len(VARDATA(string), len);
-#endif
-
-	/* starting position after the end of the string? */
-	if (m > len)
-	{
-		m = 1;
-		n = 0;
-	}
+#endif 
 
 	/*
 	 * starting position before the start of the string? then offset into
 	 * the string per SQL92 spec...
 	 */
-	else if (m < 1)
+	if (m < 1)
 	{
 		n += (m - 1);
 		m = 1;
 	}
+	/* Check for m > octet length is made in TOAST access routine */
 
 	/* m will now become a zero-based starting position */
+	sm = m - 1;
+	sn = n;
+
+#ifdef MULTIBYTE
+	eml = pg_database_encoding_max_length ();
+
+	if (eml > 1)
+	{
+		sm = 0;
+		sn = (m + n) * eml + 3; /* +3 to avoid mb characters overhanging slice end */
+	}
+#endif 
+
+	string = PG_GETARG_TEXT_P_SLICE (0, sm, sn);
+
+	if (eml == 1) 
+	{
+		PG_RETURN_TEXT_P (string);
+	}
+#ifndef MULTIBYTE
+	PG_RETURN_NULL();   /* notreached: suppress compiler warning */
+#endif
+#ifdef MULTIBYTE
+	len = pg_mbstrlen_with_len (VARDATA (string), sn - 3);
+
+	if (m > len)
+	{
+		m = 1;
+		n = 0;
+	}
 	m--;
 	if (((m + n) > len) || (n < 0))
 		n = (len - m);
 
-#ifdef MULTIBYTE
 	p = VARDATA(string);
 	for (i = 0; i < m; i++)
 		p += pg_mblen(p);
@@ -382,7 +404,6 @@ text_substr(PG_FUNCTION_ARGS)
 	for (i = 0; i < n; i++)
 		p += pg_mblen(p);
 	n = p - (VARDATA(string) + m);
-#endif
 
 	ret = (text *) palloc(VARHDRSZ + n);
 	VARATT_SIZEP(ret) = VARHDRSZ + n;
@@ -390,6 +411,7 @@ text_substr(PG_FUNCTION_ARGS)
 	memcpy(VARDATA(ret), VARDATA(string) + m, n);
 
 	PG_RETURN_TEXT_P(ret);
+#endif
 }
 
 /*
@@ -740,26 +762,14 @@ byteacat(PG_FUNCTION_ARGS)
 Datum
 bytea_substr(PG_FUNCTION_ARGS)
 {
-	bytea	   *string = PG_GETARG_BYTEA_P(0);
 	int32		m = PG_GETARG_INT32(1);
 	int32		n = PG_GETARG_INT32(2);
-	bytea	   *ret;
-	int			len;
-
-	len = VARSIZE(string) - VARHDRSZ;
-
-	/* starting position after the end of the string? */
-	if (m > len)
-	{
-		m = 1;
-		n = 0;
-	}
 
 	/*
 	 * starting position before the start of the string? then offset into
 	 * the string per SQL92 spec...
 	 */
-	else if (m < 1)
+	if (m < 1)
 	{
 		n += (m - 1);
 		m = 1;
@@ -767,15 +777,8 @@ bytea_substr(PG_FUNCTION_ARGS)
 
 	/* m will now become a zero-based starting position */
 	m--;
-	if (((m + n) > len) || (n < 0))
-		n = (len - m);
-
-	ret = (bytea *) palloc(VARHDRSZ + n);
-	VARATT_SIZEP(ret) = VARHDRSZ + n;
-
-	memcpy(VARDATA(ret), VARDATA(string) + m, n);
 
-	PG_RETURN_BYTEA_P(ret);
+	PG_RETURN_BYTEA_P(PG_GETARG_BYTEA_P_SLICE (0, m, n));
 }
 
 /*