Fixes for German sorting order.

Docs/manual.texi

@@ -748,7 +748,7 @@ is also available through the SQL interface as well.
 
 @item
 Full support for several different character sets, including
-ISO-8859-1 (Latin1), big5, ujis, and more.  For example, the
+ISO-8859-1 (Latin1), german, big5, ujis, and more.  For example, the
 Scandinavian characters `@ringaccent{a}', `@"a' and `@"o' are allowed
 in table and column names.
 
@@ -20442,6 +20442,35 @@ default-character-set=character-set-name
 
 but normally this is never needed.
 
+@menu
+* German character set::        
+@end menu
+
+@node German character set,  , Character sets, Character sets
+@subsubsection German character set
+
+To get German sorting order, you should start @code{mysqld} with
+@code{--default-character-set=latin_de}.  This will give you the following
+characteristics.
+
+When sorting and comparing string's the following mapping is done on the
+strings before doing the comparison:
+
+@example
+ä  ->  ae
+ö  ->  oe
+ü  ->  ue
+ß  ->  ss
+@end example
+
+All accented characters, except @code{'é'} and @code{É} are converted to
+their un-accented counterpart.  All letters are converted to uppercase.
+
+When comparing strings with @code{LIKE} the one -> two character mapping
+is not done. All letters are converted to uppercase. Accent are removed
+from all letters except: @code{Ü}, @code{ü}, @code{É}, @code{é}, @code{Ö},
+@code{ö}, @code{Ä} and @code{ä}.
+
 
 @node Languages, Adding character set, Character sets, Localization
 @subsection Non-English Error Messages
@@ -46753,6 +46782,8 @@ Our TODO section contains what we plan to have in 4.0. @xref{TODO MySQL 4.0}.
 
 @itemize @bullet
 @item
+New character set @code{latin_de} which provides correct German sorting.
+@item
 @code{TRUNCATE TABLE} and @code{DELETE FROM table_name} are now separate
 functions. One bonus is that @code{DELETE FROM table_name} now returns
 the number of deleted rows.

configure.in

@@ -1826,7 +1826,7 @@ CHARSETS_AVAILABLE="big5 cp1251 cp1257
                           latin1 latin1_de latin2 latin5 sjis swe7 tis620 ujis
                           usa7 win1250 win1251ukr"
 CHARSETS_DEPRECATED="win1251"
-DEFAULT_CHARSET=latin1_de
+DEFAULT_CHARSET=latin1
 AC_DIVERT_POP
 
 AC_ARG_WITH(charset,

myisam/mi_delete_all.c

@@ -15,7 +15,7 @@
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */
 
 /* Remove all rows from a MyISAM table */
-/* This only clears the status information;  The files are not truncated */
+/* This only clears the status information and truncates the data file */
 
 #include "myisamdef.h"
 
@@ -50,6 +50,8 @@ int mi_delete_all_rows(MI_INFO *info)
 
   myisam_log_command(MI_LOG_DELETE_ALL,info,(byte*) 0,0,0);
   VOID(_mi_writeinfo(info,WRITEINFO_UPDATE_KEYFILE));
+  if (my_chsize(info->dfile, 0, MYF(MY_WME)))
+    goto err;
   allow_break();			/* Allow SIGHUP & SIGINT */
   DBUG_RETURN(0);
 

myisam/mi_search.c

@@ -657,19 +657,19 @@ void _mi_dpointer(MI_INFO *info, uchar *buff, my_off_t pos)
 int _mi_compare_text(CHARSET_INFO *charset_info, uchar *a, uint a_length,
                      uchar *b, uint b_length, my_bool part_key)
 {
-  uint length= min(a_length,b_length);
-  uchar *end= a+ length;
   int flag;
 
 #ifdef USE_STRCOLL
   if (use_strcoll(charset_info))
   {
-    if ((flag = my_strnncoll(charset_info, a, a_length, b, b_length)))
-      return flag;
+    /* QQ: This needs to work with part keys at some point */
+    return my_strnncoll(charset_info, a, a_length, b, b_length);
   }
   else
 #endif
   {
+    uint length= min(a_length,b_length);
+    uchar *end= a+ length;
     uchar *sort_order=charset_info->sort_order;
     while (a < end)
       if ((flag= (int) sort_order[*a++] - (int) sort_order[*b++]))
@@ -768,8 +768,15 @@ int _mi_key_cmp(register MI_KEYSEG *keyseg, register uchar *a,
       }
       else
       {
-        uint length=(uint) (end-a);
-        if ((flag=_mi_compare_text(keyseg->charset,a,length,b,length,
+	uint length=(uint) (end-a), a_length=length, b_length=length;
+	if (!(nextflag & SEARCH_PREFIX))
+	{
+	  while (a_length && a[a_length-1] == ' ')
+	    a_length--;
+	  while (b_length && b[b_length-1] == ' ')
+	    b_length--;
+	}
+        if ((flag=_mi_compare_text(keyseg->charset,a,a_length,b,b_length,
                                    (my_bool) ((nextflag & SEARCH_PREFIX) &&
                                               next_key_length <= 0))))
           return ((keyseg->flag & HA_REVERSE_SORT) ? -flag : flag);

mysql-test/r/ctype_latin1_de.result

+a	b
+a	35
+ac	2
+ad	4
+	1
+ae	3
+	31
+aeae	33
+a	32
+aeb	6
+c	5
+e	28
+o	37
+oc	15
+od	18
+	14
+oe	17
+a	16
+oeb	20
+c	19
+o	30
+q	34
+s	21
+ss	22
+	23
+ssa	25
+a	27
+b	24
+ssc	26
+u	36
+uc	8
+ud	10
+ue	9
+	11
+ueb	12
+c	7
+uf	13
+u	29
+	38
+	39
+a	b
+a	35
+ac	2
+ad	4
+	1
+ae	3
+	31
+aeae	33
+a	32
+aeb	6
+c	5
+e	28
+o	37
+oc	15
+od	18
+	14
+oe	17
+a	16
+oeb	20
+c	19
+o	30
+q	34
+s	21
+ss	22
+	23
+ssa	25
+a	27
+b	24
+ssc	26
+u	36
+uc	8
+ud	10
+ue	9
+	11
+ueb	12
+c	7
+uf	13
+u	29
+	38
+	39
+a
+
+
+u
+uf
+c
+ueb
+
+ue
+ud
+uc
+u
+ssc
+b
+a
+ssa
+
+ss
+s
+q
+o
+c
+oeb
+a
+oe
+
+od
+oc
+o
+e
+c
+aeb
+a
+aeae
+
+ae
+
+ad
+ac
+a
+Table	Op	Msg_type	Msg_text
+test.t1	check	status	OK
+a	b
+a	16
+c	19
+o	30
+a	b
+	38
+	39
+a	b
+a	35
+ac	2
+ad	4
+ae	3
+aeae	33
+a	32
+aeb	6
+a	16
+ssa	25
+a	27
+a	b
+u	36
+uc	8
+ud	10
+ue	9
+ueb	12
+uf	13
+u	29
+a	b
+ss	22
+ssa	25
+ssc	26
+strcmp('','ae')	strcmp('ae','')	strcmp('aeq','q')	strcmp('q','aeq')
+0	0	0	0
+strcmp('ss','')	strcmp('','ss')	strcmp('s','sss')	strcmp('q','ssq')
+0	0	0	0
+strcmp('','af')	strcmp('a','')	strcmp('','aeq')	strcmp('','aeaeq')
+-1	-1	-1	-1
+strcmp('ss','a')	strcmp('','ssa')	strcmp('sa','sssb')	strcmp('s','')
+-1	-1	-1	-1
+strcmp('','o')	strcmp('','u')	strcmp('','oeb')
+-1	-1	-1
+strcmp('af','')	strcmp('','a')	strcmp('aeq','')	strcmp('aeaeq','')
+1	1	1	1
+strcmp('a','ss')	strcmp('ssa','')	strcmp('sssb','sa')	strcmp('','s')
+1	1	1	1
+strcmp('u','a')	strcmp('u','')
+1	1

mysql-test/t/ctype_latin1_de-master.opt

+--default-character-set=latin1_de

mysql-test/t/ctype_latin1_de.test

+#
+# Test latin_de character set
+#
+drop table if exists t1;
+create table t1 (a char (20) not null, b int not null auto_increment, index (a,b),index(b));
+insert into t1 (a) values (''),('ac'),('ae'),('ad'),('c'),('aeb');
+insert into t1 (a) values ('c'),('uc'),('ue'),('ud'),(''),('ueb'),('uf');
+insert into t1 (a) values (''),('oc'),('a'),('oe'),('od'),('c'),('oeb');
+insert into t1 (a) values ('s'),('ss'),(''),('b'),('ssa'),('ssc'),('a');
+insert into t1 (a) values ('e'),('u'),('o'),(''),('a'),('aeae');
+insert into t1 (a) values ('q'),('a'),('u'),('o'),(''),('');
+select a,b from t1 order by a,b;
+select a,b from t1 order by upper(a),b;
+select a from t1 order by a desc;
+check table t1;
+select * from t1 where a like "%";
+select * from t1 where a like "%%";
+select * from t1 where a like "%%";
+select * from t1 where a like "%U%";
+select * from t1 where a like "%ss%";
+drop table t1;
+
+# The following should all be true
+select strcmp('','ae'),strcmp('ae',''),strcmp('aeq','q'),strcmp('q','aeq');
+select strcmp('ss',''),strcmp('','ss'),strcmp('s','sss'),strcmp('q','ssq');
+
+# The following should all return -1
+select strcmp('','af'),strcmp('a',''),strcmp('','aeq'),strcmp('','aeaeq');
+select strcmp('ss','a'),strcmp('','ssa'),strcmp('sa','sssb'),strcmp('s','');
+select strcmp('','o'),strcmp('','u'),strcmp('','oeb');
+
+# The following should all return 1
+select strcmp('af',''),strcmp('','a'),strcmp('aeq',''),strcmp('aeaeq','');
+select strcmp('a','ss'),strcmp('ssa',''),strcmp('sssb','sa'),strcmp('','s');
+select strcmp('u','a'),strcmp('u','');
+

sql/item_cmpfunc.cc

@@ -254,7 +254,7 @@ longlong Item_func_strcmp::val_int()
     null_value=1;
     return 0;
   }
-  int value=stringcmp(a,b);
+  int value= binary ? stringcmp(a,b) : sortcmp(a,b);
   null_value=0;
   return !value ? 0 : (value < 0 ? (longlong) -1 : (longlong) 1);
 }

strings/ctype-latin1_de.c

@@ -99,12 +99,10 @@ uchar to_upper_latin1_de[] = {
  * This is a simple latin1 mapping table, which maps all accented
  * characters to their non-accented equivalents.  Note: in this
  * table, 'ä' is mapped to 'A', 'ÿ' is mapped to 'Y', etc. - all
- * accented characters are treated the same way.
- *
- * SPECIAL NOTE: 'ß' (the sz ligature), which isn't really an
- * accented 's', is mapped to 'S', to simplify the sorting
- * functions.
+ * accented characters except the following are treated the same way.
+ * Ü, ü, É, é, Ö, ö, Ä, ä
  */
+
 uchar sort_order_latin1_de[] = {
     0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
    16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
@@ -118,10 +116,10 @@ uchar sort_order_latin1_de[] = {
   144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,
   160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,
   176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,
-   65, 65, 65, 65, 65, 65, 92, 67, 69, 69, 69, 69, 73, 73, 73, 73,
-   68, 78, 79, 79, 79, 79, 79,215,216, 85, 85, 85, 85, 89,222, 83,
-   65, 65, 65, 65, 65, 65, 92, 67, 69, 69, 69, 69, 73, 73, 73, 73,
-   68, 78, 79, 79, 79, 79, 79,247,216, 85, 85, 85, 85, 89,222, 89
+   65, 65, 65, 65,196, 65, 92, 67, 69,201, 69, 69, 73, 73, 73, 73,
+   68, 78, 79, 79, 79, 79,214,215,216, 85, 85, 85,220, 89,222,223,
+   65, 65, 65, 65,196, 65, 92, 67, 69,201, 69, 69, 73, 73, 73, 73,
+   68, 78, 79, 79, 79, 79,214,247,216, 85, 85, 85,220, 89,222, 89
 };
 
 #define L1_AE 196
@@ -132,6 +130,39 @@ uchar sort_order_latin1_de[] = {
 #define L1_ue 252
 #define L1_ss 223
 
+
+/*
+  Some notes about the following comparison rules:
+  By definition, my_strnncoll_latin_de must works exactly as if had called
+  my_strnxfrm_latin_de() on both strings and compared the result strings.
+
+  This means that:
+  Ä must also matches ÁE and Aè, because my_strxn_frm_latin_de() will convert
+  both to AE.
+
+  The other option would be to not do any accent removal in
+  sort_order_latin_de[] at all
+*/
+
+
+#define CHECK_S1_COMBO(ch1, ch2, str1, str1_end, res_if_str1_smaller, str2, fst, snd, accent)   \
+  /* Invariant: ch1 == fst == sort_order_latin1_de[accent] && ch1 != ch2 */ \
+  if (ch2 != accent)							\
+  {									\
+    ch1= fst;								\
+    goto normal;							\
+  }									\
+  if (str1 == str1_end)							\
+    return res_if_str1_smaller;						\
+  {									\
+     int diff = (int) sort_order_latin1_de[*str1] - snd;		\
+     if (diff)								\
+        return diff*(-(res_if_str1_smaller));				\
+      /* They are equal (e.g., "Ae" == 'ä') */				\
+     str1++;								\
+  }
+
+
 int my_strnncoll_latin1_de(const uchar * s1, int len1,
                            const uchar * s2, int len2)
 {
@@ -140,172 +171,71 @@ int my_strnncoll_latin1_de(const uchar * s1, int len1,
 
   while (s1 < e1 && s2 < e2)
   {
-    /* to_upper is used instead of sort_order, because we don't want
-     * 'Ä' to match "ÁE", only "AE".  This couples the to_upper and
-     * sort_order tables together, but that is acceptable. */
-    uchar c1 = to_upper_latin1_de[*s1];
-    uchar c2 = to_upper_latin1_de[*s2];
+    /*
+      Because sort_order_latin1_de doesn't convert 'Ä', Ü or ß we
+      can use it here.
+    */
+    uchar c1 = sort_order_latin1_de[*s1++];
+    uchar c2 = sort_order_latin1_de[*s2++];
     if (c1 != c2)
     {
-      switch (c1)
-      {
-
-#define CHECK_S1_COMBO(fst, snd, accent)                                  \
-  /* Invariant: c1 == fst == sort_order_latin1_de[accent] && c1 != c2 */  \
-  if (c2 == accent)                                                       \
-  {                                                                       \
-    if (s1 + 1 < e1)                                                      \
-    {                                                                     \
-      if (to_upper_latin1_de[*(s1 + 1)] == snd)                           \
-      {                                                                   \
-	/* They are equal (e.g., "Ae" == 'ä') */                          \
-	s1 += 2;                                                          \
-	s2 += 1;                                                          \
-      }                                                                   \
-      else                                                                \
-      {                                                                   \
-	int diff = sort_order_latin1_de[*(s1 + 1)] - snd;                 \
-	if (diff)                                                         \
-	  return diff;                                                    \
-	else                                                              \
-	  /* Comparison between, e.g., "AÉ" and 'Ä' */                    \
-	  return 1;                                                       \
-      }                                                                   \
-    }                                                                     \
-    else                                                                  \
-      return -1;                                                          \
-  }                                                                       \
-  else                                                                    \
-    /* The following should work even if c2 is [ÄÖÜß] */                  \
-    return fst - sort_order_latin1_de[c2]
-
+      switch (c1) {
       case 'A':
-	CHECK_S1_COMBO('A', 'E', L1_AE);
+	CHECK_S1_COMBO(c1, c2, s1, e1, -1, s2, 'A', 'E', L1_AE);
 	break;
       case 'O':
-	CHECK_S1_COMBO('O', 'E', L1_OE);
+	CHECK_S1_COMBO(c1, c2, s1, e1, -1, s2, 'O', 'E', L1_OE);
 	break;
       case 'U':
-	CHECK_S1_COMBO('U', 'E', L1_UE);
+	CHECK_S1_COMBO(c1, c2, s1, e1, -1, s2, 'U', 'E', L1_UE);
 	break;
       case 'S':
-	CHECK_S1_COMBO('S', 'S', L1_ss);
+	CHECK_S1_COMBO(c1, c2, s1, e1, -1, s2, 'S', 'S', L1_ss);
 	break;
-
-#define CHECK_S2_COMBO(fst, snd)                                          \
-  /* Invariant: sort_order_latin1_de[c1] == fst && c1 != c2 */            \
-  if (c2 == fst)                                                          \
-  {                                                                       \
-    if (s2 + 1 < e2)                                                      \
-    {                                                                     \
-      if (to_upper_latin1_de[*(s2 + 1)] == snd)                           \
-      {                                                                   \
-	/* They are equal (e.g., 'ä' == "Ae") */                          \
-	s1 += 1;                                                          \
-	s2 += 2;                                                          \
-      }                                                                   \
-      else                                                                \
-      {                                                                   \
-	int diff = sort_order_latin1_de[*(s1 + 1)] - snd;                 \
-	if (diff)                                                         \
-	  return diff;                                                    \
-	else                                                              \
-	  /* Comparison between, e.g., 'Ä' and "AÉ" */                    \
-	  return -1;                                                      \
-      }                                                                   \
-    }                                                                     \
-    else                                                                  \
-      return 1;                                                           \
-  }                                                                       \
-  else                                                                    \
-    /* The following should work even if c2 is [ÄÖÜß] */                  \
-    return fst - sort_order_latin1_de[c2]
-
       case L1_AE:
-	CHECK_S2_COMBO('A', 'E');
+	CHECK_S1_COMBO(c1, c2, s2, e2, 1, s1, 'A', 'E', 'A');
 	break;
       case L1_OE:
-	CHECK_S2_COMBO('O', 'E');
+	CHECK_S1_COMBO(c1, c2, s2, e2, 1, s1, 'O', 'E', 'O');
 	break;
       case L1_UE:
-	CHECK_S2_COMBO('U', 'E');
+	CHECK_S1_COMBO(c1, c2, s2, e2, 1, s1, 'U', 'E', 'U');
 	break;
       case L1_ss:
-	CHECK_S2_COMBO('S', 'S');
+	CHECK_S1_COMBO(c1, c2, s2, e2, 1, s1, 'S', 'S', 'S');
 	break;
       default:
+	/*
+	  Handle the case where 'c2' is a special character
+	  If this is true, we know that c1 can't match this character.
+	*/
+    normal:
 	switch (c2) {
 	case L1_AE:
+	  return  (int) c1 - (int) 'A';
 	case L1_OE:
+	  return  (int) c1 - (int) 'O';
 	case L1_UE:
+	  return  (int) c1 - (int) 'U';
 	case L1_ss:
-	  /* Make sure these do not match (e.g., "Ä" != "Á") */
-	  return sort_order_latin1_de[c1] - sort_order_latin1_de[c2];
-	  break;
+	  return  (int) c1 - (int) 'S';
 	default:
-	  if (sort_order_latin1_de[*s1] != sort_order_latin1_de[*s2])
-	    return sort_order_latin1_de[*s1] - sort_order_latin1_de[*s2];
-	  ++s1;
-	  ++s2;
-	  break;
+	{
+	  int diff= (int) c1 - (int) c2;
+	  if (diff)
+	    return diff;
 	}
 	break;
-
-#undef CHECK_S1_COMBO
-#undef CHECK_S2_COMBO
-
-      }
-    }
-    else
-    {
-      /* In order to consistently treat "ae" == 'ä', but to NOT allow
-       * "aé" == 'ä', we must look ahead here to ensure that the second
-       * letter in a combo really is the unaccented 'e' (or 's' for
-       * "ss") and is not an accented character with the same sort_order. */
-      ++s1;
-      ++s2;
-      if (s1 < e1 && s2 < e2)
-      {
-	switch (c1)
-	{
-	case 'A':
-	case 'O':
-	case 'U':
-	  if (sort_order_latin1_de[*s1] == 'E' &&
-	      to_upper_latin1_de[*s1] != 'E' &&
-	      to_upper_latin1_de[*s2] == 'E')
-	    /* Comparison between, e.g., "AÉ" and "AE" */
-	    return 1;
-	  if (sort_order_latin1_de[*s2] == 'E' &&
-	      to_upper_latin1_de[*s2] != 'E' &&
-	      to_upper_latin1_de[*s1] == 'E')
-	    /* Comparison between, e.g., "AE" and "AÉ" */
-	    return -1;
-	  break;
-	case 'S':
-	  if (sort_order_latin1_de[*s1] == 'S' &&
-	      to_upper_latin1_de[*s1] != 'S' &&
-	      to_upper_latin1_de[*s2] == 'S')
-	    /* Comparison between, e.g., "Sß" and "SS" */
-	    return 1;
-	  if (sort_order_latin1_de[*s2] == 'S' &&
-	      to_upper_latin1_de[*s2] != 'S' &&
-	      to_upper_latin1_de[*s1] == 'S')
-	    /* Comparison between, e.g., "SS" and "Sß" */
-	    return -1;
-	  break;
-	default:
-	  break;
 	}
       }
     }
   }
-
   /* A simple test of string lengths won't work -- we test to see
    * which string ran out first */
   return s1 < e1 ? 1 : s2 < e2 ? -1 : 0;
 }
 
+
 int my_strnxfrm_latin1_de(uchar * dest, const uchar * src, int len, int srclen)
 {
   const uchar *dest_orig = dest;
@@ -313,22 +243,19 @@ int my_strnxfrm_latin1_de(uchar * dest, const uchar * src, int len, int srclen)
   const uchar *se = src + srclen;
   while (src < se && dest < de)
   {
-    switch (*src)
-    {
+    uchar chr=sort_order_latin1_de[*src];
+    switch (chr) {
     case L1_AE:
-    case L1_ae:
       *dest++ = 'A';
       if (dest < de)
 	*dest++ = 'E';
       break;
     case L1_OE:
-    case L1_oe:
       *dest++ = 'O';
       if (dest < de)
 	*dest++ = 'E';
       break;
     case L1_UE:
-    case L1_ue:
       *dest++ = 'U';
       if (dest < de)
 	*dest++ = 'E';
@@ -339,7 +266,7 @@ int my_strnxfrm_latin1_de(uchar * dest, const uchar * src, int len, int srclen)
 	*dest++ = 'S';
       break;
     default:
-      *dest++ = sort_order_latin1_de[*src];
+      *dest++= chr;
       break;
     }
     ++src;
@@ -347,6 +274,7 @@ int my_strnxfrm_latin1_de(uchar * dest, const uchar * src, int len, int srclen)
   return dest - dest_orig;
 }
 
+
 int my_strcoll_latin1_de(const uchar * s1, const uchar * s2)
 {
   /* XXX QQ: This should be fixed to not call strlen */