MDEV-371 Unique Index for long columns

post-merge fixes

mysql-test/main/long_unique.result

@@ -1387,13 +1387,13 @@ create table t1(a blob unique) partition by hash(a);
 ERROR HY000: A BLOB field is not allowed in partition function
 #key length > 2^16 -1
 create table t1(a blob, unique(a(65536)));
-ERROR HY000: Max key segment length is 65535
+ERROR 42000: Specified key part was too long; max key part length is 65535 bytes
 create table t1(a blob, unique(a(65535)));
 show create table t1;
 Table	Create Table
 t1	CREATE TABLE `t1` (
   `a` blob DEFAULT NULL,
-  UNIQUE KEY `a` (`a`) USING HASH
+  UNIQUE KEY `a` (`a`(65535)) USING HASH
 ) ENGINE=MyISAM DEFAULT CHARSET=latin1
 drop table t1;
 #64 indexes

mysql-test/main/long_unique.test

@@ -1,6 +1,10 @@
 let datadir=`select @@datadir`;
 --source include/have_partition.inc
 
+#
+# MDEV-371 Unique indexes for blobs
+#
+
 --echo #Structure of tests
 --echo #First we will check all option for
 --echo #table containing single unique column
@@ -475,7 +479,7 @@ drop table t1;
 --error ER_BLOB_FIELD_IN_PART_FUNC_ERROR
 create table t1(a blob unique) partition by hash(a);
 --echo #key length > 2^16 -1
---error ER_TOO_LONG_HASH_KEYSEG
+--error ER_TOO_LONG_KEYPART
 create table t1(a blob, unique(a(65536)));
 create table t1(a blob, unique(a(65535)));
 show create table t1;

mysql-test/main/long_unique_debug.test

@@ -1,5 +1,9 @@
 --source include/have_debug.inc
---source include/have_innodb.inc
+
+#
+# MDEV-371 Unique indexes for blobs
+#
+
 --echo #In this test case we will check what will happen in the case of hash collision
 
 SET debug_dbug="d,same_long_unique_hash";

mysql-test/main/long_unique_innodb.result

@@ -3,6 +3,16 @@ insert into t1 values('RUC');
 insert into t1 values ('RUC');
 ERROR 23000: Duplicate entry 'RUC' for key 'a'
 drop table t1;
+create table t1 (a blob unique , c int unique) engine=innodb;
+show create table t1;
+Table	Create Table
+t1	CREATE TABLE `t1` (
+  `a` blob DEFAULT NULL,
+  `c` int(11) DEFAULT NULL,
+  UNIQUE KEY `a` (`a`) USING HASH,
+  UNIQUE KEY `c` (`c`)
+) ENGINE=InnoDB DEFAULT CHARSET=latin1
+drop table t1;
 #test for concurrent insert of long unique in innodb
 create table t1(a blob unique) engine= InnoDB;
 show create table t1;

mysql-test/main/long_unique_innodb.test

@@ -1,11 +1,19 @@
 --source include/have_innodb.inc
 
+#
+# MDEV-371 Unique indexes for blobs
+#
+
 create table t1(a blob unique) engine= InnoDB;
 insert into t1 values('RUC');
 --error ER_DUP_ENTRY
 insert into t1 values ('RUC');
 drop table t1;
 
+create table t1 (a blob unique , c int unique) engine=innodb;
+show create table t1;
+drop table t1;
+
 --echo #test for concurrent insert of long unique in innodb
 create table t1(a blob unique) engine= InnoDB;
 show create table t1;
@@ -33,7 +41,6 @@ insert into t1 values('RC');
 commit;
 set  transaction isolation level READ COMMITTED;
 start transaction;
---error ER_DUP_ENTRY
 --error ER_LOCK_WAIT_TIMEOUT
 insert into t1 values ('RC');
 commit;
@@ -47,7 +54,6 @@ insert into t1 values('RR');
 commit;
 set  transaction isolation level REPEATABLE READ;
 start transaction;
---error ER_DUP_ENTRY
 --error ER_LOCK_WAIT_TIMEOUT
 insert into t1 values ('RR');
 
@@ -60,7 +66,6 @@ insert into t1 values('S');
 commit;
 set  transaction isolation level SERIALIZABLE;
 start transaction;
---error ER_DUP_ENTRY
 --error ER_LOCK_WAIT_TIMEOUT
 insert into t1 values ('S');
 commit;

mysql-test/main/long_unique_update.test

@@ -1,3 +1,6 @@
+#
+# MDEV-371 Unique indexes for blobs
+#
 --echo #structure of tests;
 --echo #1 test of table containing single unique blob column;
 --echo #2 test of table containing another unique int/ varchar etc column; 

mysql-test/main/long_unique_using_hash.test

@@ -1,3 +1,6 @@
+#
+# MDEV-371 Unique indexes for blobs
+#
 
 create table t1(a blob , unique(a) using hash);
 --query_vertical show keys from t1;

sql/share/errmsg-utf8.txt

@@ -7950,8 +7950,7 @@ ER_PERIOD_NOT_FOUND
         eng "Period %`s is not found in table"
 ER_PERIOD_COLUMNS_UPDATED
         eng "Column %`s used in period %`s specified in update SET list"
-
 ER_PERIOD_CONSTRAINT_DROP
         eng "Can't DROP CONSTRAINT `%s`. Use DROP PERIOD `%s` for this"
-ER_TOO_LONG_HASH_KEYSEG
+ER_TOO_LONG_KEYPART 42000 S1009
-        eng "Max key segment length is 65535"
+        eng "Specified key part was too long; max key part length is %u bytes"

sql/sql_show.cc

@@ -2352,9 +2352,7 @@ int show_create_table(THD *thd, TABLE_LIST *table_list, String *packet,
       if (key_part->field &&
           (key_part->length !=
            table->field[key_part->fieldnr-1]->key_length() &&
-           !(key_info->flags & (HA_FULLTEXT | HA_SPATIAL))) &&
+           !(key_info->flags & (HA_FULLTEXT | HA_SPATIAL))))
-          (key_info->algorithm != HA_KEY_ALG_LONG_HASH ||
-           key_info->algorithm == HA_KEY_ALG_LONG_HASH && key_part->length))
       {
         packet->append_parenthesized((long) key_part->length /
                                       key_part->field->charset()->mbmaxlen);
@@ -6644,9 +6642,7 @@ static int get_schema_stat_record(THD *thd, TABLE_LIST *tables,
         if (!(key_info->flags & HA_FULLTEXT) &&
             (key_part->field &&
              key_part->length !=
-             show_table->s->field[key_part->fieldnr-1]->key_length()) &&
+             show_table->s->field[key_part->fieldnr-1]->key_length()))
-          (key_info->algorithm != HA_KEY_ALG_LONG_HASH ||
-           key_info->algorithm == HA_KEY_ALG_LONG_HASH && key_part->length))
         {
           table->field[10]->store((longlong) key_part->length /
                                   key_part->field->charset()->mbmaxlen, TRUE);

sql/sql_table.cc

@@ -2778,23 +2778,26 @@ bool quick_rm_table(THD *thd, handlerton *base, const LEX_CSTRING *db,
 
   This will make checking for duplicated keys faster and ensure that
   PRIMARY keys are prioritized.
-  This will not reorder LONG_HASH indexes, because they must match the
-  order of their LONG_UNIQUE_HASH_FIELD's.
 */
 
 static int sort_keys(KEY *a, KEY *b)
 {
   ulong a_flags= a->flags, b_flags= b->flags;
   
+  /*
+    Do not reorder LONG_HASH indexes, because they must match the order
+    of their LONG_UNIQUE_HASH_FIELD's.
+  */
+  if (a->algorithm == HA_KEY_ALG_LONG_HASH &&
+      b->algorithm == HA_KEY_ALG_LONG_HASH)
+    return a->usable_key_parts - b->usable_key_parts;
+
   if (a_flags & HA_NOSAME)
   {
     if (!(b_flags & HA_NOSAME))
       return -1;
     if ((a_flags ^ b_flags) & HA_NULL_PART_KEY)
     {
-      if (a->algorithm == HA_KEY_ALG_LONG_HASH &&
-              b->algorithm == HA_KEY_ALG_LONG_HASH)
-        return a->usable_key_parts - b->usable_key_parts;
       /* Sort NOT NULL keys before other keys */
       return (a_flags & HA_NULL_PART_KEY) ? 1 : -1;
     }
@@ -2817,9 +2820,7 @@ static int sort_keys(KEY *a, KEY *b)
     Prefer original key order.	usable_key_parts contains here
     the original key position.
   */
-  return ((a->usable_key_parts < b->usable_key_parts) ? -1 :
+  return a->usable_key_parts - b->usable_key_parts;
-	  (a->usable_key_parts > b->usable_key_parts) ? 1 :
-	  0);
 }
 
 /*
@@ -3302,6 +3303,7 @@ static inline void make_long_hash_field_name(LEX_CSTRING *buf, uint num)
   buf->length= my_snprintf((char *)buf->str,
           LONG_HASH_FIELD_NAME_LENGTH, "DB_ROW_HASH_%u", num);
 }
+
 /**
   Add fully invisible hash field to table in case of long
   unique column
@@ -3313,7 +3315,6 @@ static Create_field * add_hash_field(THD * thd, List<Create_field> *create_list,
                                       KEY *key_info)
 {
   List_iterator<Create_field> it(*create_list);
-//  CHARSET_INFO *field_cs;
   Create_field *dup_field, *cf= new (thd->mem_root) Create_field();
   cf->flags|= UNSIGNED_FLAG | LONG_UNIQUE_HASH_FIELD;
   cf->decimals= 0;
@@ -3336,18 +3337,6 @@ static Create_field * add_hash_field(THD * thd, List<Create_field> *create_list,
       it.rewind();
     }
   }
-  /*  for (uint i= 0; i < key_info->user_defined_key_parts; i++)
-  {
-    dup_field= create_list->elem(key_info->key_part[i].fieldnr);
-    if (!i)
-      field_cs= dup_field->charset;
-    else if(field_cs != dup_field->charset)
-    {
-	  my_error(ER_MULTIPLE_CS_HASH_KEY, MYF(0));
-      return NULL;
-    }
-  }
-  cf->charset= field_cs;*/
   cf->field_name= field_name;
   cf->set_handler(&type_handler_longlong);
   key_info->algorithm= HA_KEY_ALG_LONG_HASH;
@@ -4107,19 +4096,14 @@ mysql_prepare_create_table(THD *thd, HA_CREATE_INFO *create_info,
           }
         }
       }
-      /* We can not store key_part_length more then 2^16 - 1 in frm
+      /* We can not store key_part_length more then 2^16 - 1 in frm */
-         So we will simply make it zero */
+      if (is_hash_field_needed && column->length > UINT16_MAX)
-      if (is_hash_field_needed && column->length > (1<<16) - 1)
       {
-	    my_error(ER_TOO_LONG_HASH_KEYSEG, MYF(0));
+        my_error(ER_TOO_LONG_KEYPART, MYF(0),  UINT16_MAX);
         DBUG_RETURN(TRUE);
       }
       else
         key_part_info->length= (uint16) key_part_length;
-      if (is_hash_field_needed &&
-           (key_part_info->length == sql_field->char_length * sql_field->charset->mbmaxlen ||
-            key_part_info->length == (1<<16) -1))
-        key_part_info->length= 0;
       /* Use packed keys for long strings on the first column */
       if (!((*db_options) & HA_OPTION_NO_PACK_KEYS) &&
           !((create_info->table_options & HA_OPTION_NO_PACK_KEYS)) &&
@@ -8385,13 +8369,6 @@ mysql_prepare_alter_table(THD *thd, TABLE *table,
       if (cfield->field)			// Not new field
       {
         /*
-        if (key_info->algorithm == HA_KEY_ALG_LONG_HASH)
-        {
-          Field *fld= cfield->field;
-          if (fld->max_display_length() == cfield->length*fld->charset()->mbmaxlen
-                      && fld->max_data_length() != key_part->length)
-            cfield->length= cfield->char_length= key_part->length;
-        }
           If the field can't have only a part used in a key according to its
           new type, or should not be used partially according to its
           previous type, or the field length is less than the key part

sql/table.cc

@@ -2443,8 +2443,6 @@ int TABLE_SHARE::init_from_binary_frm_image(THD *thd, bool write,
       if (keyinfo->algorithm == HA_KEY_ALG_LONG_HASH)
       {
         share->long_unique_table= 1;
-        if (share->frm_version < FRM_VER_EXPRESSSIONS)
-          share->frm_version= FRM_VER_EXPRESSSIONS;
         hash_keypart= keyinfo->key_part + keyinfo->user_defined_key_parts;
         hash_keypart->length= HA_HASH_KEY_LENGTH_WITHOUT_NULL;
         hash_keypart->store_length= hash_keypart->length;
@@ -2453,8 +2451,6 @@ int TABLE_SHARE::init_from_binary_frm_image(THD *thd, bool write,
         hash_keypart->key_type= 32834;
         /* Last n fields are unique_index_hash fields*/
         hash_keypart->offset= offset;
-//        hash_keypart->offset= share->reclength
-//                       - HA_HASH_FIELD_LENGTH*(share->fields - hash_field_used_no);
         hash_keypart->fieldnr= hash_field_used_no + 1;
         hash_field= share->field[hash_field_used_no];
         hash_field->flags|= LONG_UNIQUE_HASH_FIELD;//Used in parse_vcol_defs
@@ -2566,7 +2562,7 @@ int TABLE_SHARE::init_from_binary_frm_image(THD *thd, bool write,
         for (i= 0; i < keyinfo->user_defined_key_parts; i++)
         {
           uint fieldnr= keyinfo->key_part[i].fieldnr;
-          field= share->field[keyinfo->key_part[i].fieldnr-1];
+          field= share->field[fieldnr-1];
 
           if (field->null_ptr)
             len_null_byte= HA_KEY_NULL_LENGTH;
@@ -2581,8 +2577,7 @@ int TABLE_SHARE::init_from_binary_frm_image(THD *thd, bool write,
 
           ext_key_length+= keyinfo->key_part[i].length + len_null_byte
                             + length_bytes;
-          if (share->field[fieldnr-1]->key_length() !=
+          if (field->key_length() != keyinfo->key_part[i].length)
-              keyinfo->key_part[i].length)
 	  {
             add_keyparts_for_this_key= 0;
             break;
@@ -4258,6 +4253,8 @@ void prepare_frm_header(THD *thd, uint reclength, uchar *fileinfo,
 {
   size_t key_comment_total_bytes= 0;
   uint i;
+  uchar frm_format= create_info->expression_length ? FRM_VER_EXPRESSSIONS
+                                                   : FRM_VER_TRUE_VARCHAR;
   DBUG_ENTER("prepare_frm_header");
 
   /* Fix this when we have new .frm files;  Current limit is 4G rows (TODO) */
@@ -4266,17 +4263,6 @@ void prepare_frm_header(THD *thd, uint reclength, uchar *fileinfo,
   if (create_info->min_rows > UINT_MAX32)
     create_info->min_rows= UINT_MAX32;
 
-  size_t key_length, tmp_key_length, tmp, csid;
-  bzero((char*) fileinfo, FRM_HEADER_SIZE);
-  /* header */
-  fileinfo[0]=(uchar) 254;
-  fileinfo[1]= 1;
-  fileinfo[2]= (create_info->expression_length == 0 ? FRM_VER_TRUE_VARCHAR :
-                FRM_VER_EXPRESSSIONS);
-
-  DBUG_ASSERT(ha_storage_engine_is_enabled(create_info->db_type));
-  fileinfo[3]= (uchar) ha_legacy_type(create_info->db_type);
-
   /*
     Keep in sync with pack_keys() in unireg.cc
     For each key:
@@ -4295,8 +4281,20 @@ void prepare_frm_header(THD *thd, uint reclength, uchar *fileinfo,
                 (key_info[i].comment.length > 0));
     if (key_info[i].flags & HA_USES_COMMENT)
       key_comment_total_bytes += 2 + key_info[i].comment.length;
+    if (key_info[i].algorithm == HA_KEY_ALG_LONG_HASH)
+      frm_format= FRM_VER_EXPRESSSIONS;
   }
 
+  size_t key_length, tmp_key_length, tmp, csid;
+  bzero((char*) fileinfo, FRM_HEADER_SIZE);
+  /* header */
+  fileinfo[0]=(uchar) 254;
+  fileinfo[1]= 1;
+  fileinfo[2]= frm_format;
+
+  DBUG_ASSERT(ha_storage_engine_is_enabled(create_info->db_type));
+  fileinfo[3]= (uchar) ha_legacy_type(create_info->db_type);
+
   key_length= keys * (8 + MAX_REF_PARTS * 9 + NAME_LEN + 1) + 16
               + key_comment_total_bytes;