Index: src/func.c
==================================================================
--- src/func.c
+++ src/func.c
@@ -14,11 +14,11 @@
 **
 ** There is only one exported symbol in this file - the function
 ** sqliteRegisterBuildinFunctions() found at the bottom of the file.
 ** All other code has file scope.
 **
-** $Id: func.c,v 1.75 2004/06/21 06:50:28 danielk1977 Exp $
+** $Id: func.c,v 1.76 2004/06/24 00:20:05 danielk1977 Exp $
 */
 #include <ctype.h>
 #include <math.h>
 #include <stdlib.h>
 #include <assert.h>
@@ -969,12 +969,12 @@
 static void minmaxStep(sqlite3_context *context, int argc, sqlite3_value **argv){
   int max = 0;
   int cmp = 0;
   Mem *pArg  = (Mem *)argv[0];
   Mem *pBest = (Mem *)sqlite3_aggregate_context(context, sizeof(*pBest));
+  if( !pBest ) return;
 
-  if( !pBest || SQLITE_NULL==sqlite3_value_type(argv[0]) ) return;
   if( pBest->flags ){
     CollSeq *pColl = sqlite3GetFuncCollSeq(context);
     /* This step function is used for both the min() and max() aggregates,
     ** the only difference between the two being that the sense of the
     ** comparison is inverted. For the max() aggregate, the

Index: test/func.test
==================================================================
--- test/func.test
+++ test/func.test
@@ -9,11 +9,11 @@
 #
 #***********************************************************************
 # This file implements regression tests for SQLite library.  The
 # focus of this file is testing built-in functions.
 #
-# $Id: func.test,v 1.25 2004/06/23 01:05:27 danielk1977 Exp $
+# $Id: func.test,v 1.26 2004/06/24 00:20:05 danielk1977 Exp $
 
 set testdir [file dirname $argv0]
 source $testdir/tester.tcl
 
 # Create a table to work with.
@@ -249,11 +249,11 @@
 #
 do_test func-8.1 {
   execsql {
     SELECT sum(a), count(a), round(avg(a),2), min(a), max(a), count(*) FROM t2;
   }
-} {68236 3 22745.33 1 67890 5}
+} {68236 3 22745.33 {} 67890 5}
 do_test func-8.2 {
   execsql {
     SELECT max('z+'||a||'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOP') FROM t2;
   }
 } {z+67890abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOP}
@@ -260,11 +260,16 @@
 do_test func-8.3 {
   execsql {
     CREATE TEMP TABLE t3 AS SELECT a FROM t2 ORDER BY a DESC;
     SELECT min('z+'||a||'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOP') FROM t3;
   }
-} {z+1abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOP}
+} {{}}
+do_test func-8.4 {
+  execsql {
+    SELECT max('z+'||a||'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOP') FROM t3;
+  }
+} {z+67890abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOP}
 
 # How do you test the random() function in a meaningful, deterministic way?
 #
 do_test func-9.1 {
   execsql {

Index: test/minmax.test
==================================================================
--- test/minmax.test
+++ test/minmax.test
@@ -11,11 +11,11 @@
 # This file implements regression tests for SQLite library.  The
 # focus of this file is testing SELECT statements that contain
 # aggregate min() and max() functions and which are handled as
 # as a special case.
 #
-# $Id: minmax.test,v 1.9 2004/03/13 14:00:37 drh Exp $
+# $Id: minmax.test,v 1.10 2004/06/24 00:20:05 danielk1977 Exp $
 
 set testdir [file dirname $argv0]
 source $testdir/tester.tcl
 
 do_test minmax-1.0 {
@@ -125,11 +125,11 @@
 do_test minmax-4.1 {
   execsql {
     SELECT coalesce(min(x+0),-1), coalesce(max(x+0),-1) FROM
       (SELECT * FROM t1 UNION SELECT NULL as 'x', NULL as 'y')
   }
-} {1 20}
+} {-1 20}
 do_test minmax-4.2 {
   execsql {
     SELECT y, sum(x) FROM
       (SELECT null, y+1 FROM t1 UNION SELECT * FROM t1)
     GROUP BY y ORDER BY y;
@@ -273,7 +273,37 @@
     SELECT max(rowid) FROM (
       SELECT max(rowid) FROM t4 EXCEPT SELECT max(rowid) FROM t5
     )
   }
 } {{}}
+
+# If there is a NULL in an aggregate max(), ignore it.  If a NULL
+# occurs in an aggregate min(), then the result will be NULL because
+# NULL compares less than all other values.
+#
+do_test minmax-10.1 {
+  execsql {
+    CREATE TABLE t6(x);
+    INSERT INTO t6 VALUES(1);
+    INSERT INTO t6 VALUES(2);
+    INSERT INTO t6 VALUES(NULL);
+    SELECT coalesce(min(x),-1) FROM t6;
+  }
+} {-1}
+do_test minmax-10.2 {
+  execsql {
+    SELECT max(x) FROM t6;
+  }
+} {2}
+do_test minmax-10.3 {
+  execsql {
+    CREATE INDEX i6 ON t6(x);
+    SELECT coalesce(min(x),-1) FROM t6;
+  }
+} {-1}
+do_test minmax-10.4 {
+  execsql {
+    SELECT max(x) FROM t6;
+  }
+} {2}
 
 finish_test

Index: test/null.test
==================================================================
--- test/null.test
+++ test/null.test
@@ -98,11 +98,11 @@
 do_test null-3.1 {
   execsql {
     select count(*), count(b), count(c), sum(b), sum(c), 
            avg(b), avg(c), min(b), max(b) from t1;
   }
-} {7 4 6 2 3 0.5 0.5 0 1}
+} {7 4 6 2 3 0.5 0.5 {} 1}
 
 # Check to see how WHERE clauses handle NULL values.  A NULL value
 # is the same as UNKNOWN.  The WHERE clause should only select those
 # rows that are TRUE.  FALSE and UNKNOWN rows are rejected.
 #

Index: test/select1.test
==================================================================
--- test/select1.test
+++ test/select1.test
@@ -9,11 +9,11 @@
 #
 #***********************************************************************
 # This file implements regression tests for SQLite library.  The
 # focus of this file is testing the SELECT statement.
 #
-# $Id: select1.test,v 1.32 2004/05/28 11:37:29 danielk1977 Exp $
+# $Id: select1.test,v 1.33 2004/06/24 00:20:05 danielk1977 Exp $
 
 set testdir [file dirname $argv0]
 source $testdir/tester.tcl
 
 # Try to select on a non-existant table.
@@ -157,11 +157,11 @@
   set v [catch {execsql {SELECT MIN(f1,f2) FROM test1}} msg]
   lappend v [lsort $msg]
 } {0 {11 33}}
 do_test select1-2.8.1 {
   execsql {SELECT coalesce(min(a),'xyzzy') FROM t3}
-} {11}
+} {xyzzy}
 do_test select1-2.8.2 {
   execsql {SELECT min(coalesce(a,'xyzzy')) FROM t3}
 } {11}
 do_test select1-2.8.3 {
   execsql {SELECT min(b), min(b) FROM t4}