sql_partition.cc 193 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
/* Copyright (C) 2005 MySQL AB

   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2 of the License, or
   (at your option) any later version.

   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software
   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */

/*
18
  This file is a container for general functionality related
19
  to partitioning introduced in MySQL version 5.1. It contains functionality
20 21
  used by all handlers that support partitioning, such as
  the partitioning handler itself and the NDB handler.
22

patg@krsna.patg.net's avatar
patg@krsna.patg.net committed
23
  The first version was written by Mikael Ronstrom.
24 25 26 27 28

  This version supports RANGE partitioning, LIST partitioning, HASH
  partitioning and composite partitioning (hereafter called subpartitioning)
  where each RANGE/LIST partitioning is HASH partitioned. The hash function
  can either be supplied by the user or by only a list of fields (also
29
  called KEY partitioning), where the MySQL server will use an internal
30 31 32 33 34 35
  hash function.
  There are quite a few defaults that can be used as well.
*/

/* Some general useful functions */

36
#define MYSQL_LEX 1
37 38 39 40 41
#include "mysql_priv.h"
#include <errno.h>
#include <m_ctype.h>
#include "md5.h"

42
#ifdef WITH_PARTITION_STORAGE_ENGINE
kent@mysql.com's avatar
kent@mysql.com committed
43
#include "ha_partition.h"
44 45 46
/*
  Partition related functions declarations and some static constants;
*/
47 48 49 50 51 52 53 54 55
const LEX_STRING partition_keywords[]=
{
  { (char *) STRING_WITH_LEN("HASH") },
  { (char *) STRING_WITH_LEN("RANGE") },
  { (char *) STRING_WITH_LEN("LIST") }, 
  { (char *) STRING_WITH_LEN("KEY") },
  { (char *) STRING_WITH_LEN("MAXVALUE") },
  { (char *) STRING_WITH_LEN("LINEAR ") }
};
56 57 58 59 60 61 62 63
static const char *part_str= "PARTITION";
static const char *sub_str= "SUB";
static const char *by_str= "BY";
static const char *space_str= " ";
static const char *equal_str= "=";
static const char *end_paren_str= ")";
static const char *begin_paren_str= "(";
static const char *comma_str= ",";
64 65
static char buff[22];

66
int get_partition_id_list(partition_info *part_info,
67 68
                           uint32 *part_id,
                           longlong *func_value);
69
int get_partition_id_range(partition_info *part_info,
70 71
                            uint32 *part_id,
                            longlong *func_value);
72
int get_partition_id_hash_nosub(partition_info *part_info,
73 74
                                 uint32 *part_id,
                                 longlong *func_value);
75
int get_partition_id_key_nosub(partition_info *part_info,
76 77
                                uint32 *part_id,
                                longlong *func_value);
78
int get_partition_id_linear_hash_nosub(partition_info *part_info,
79 80
                                        uint32 *part_id,
                                        longlong *func_value);
81
int get_partition_id_linear_key_nosub(partition_info *part_info,
82 83
                                       uint32 *part_id,
                                       longlong *func_value);
84
int get_partition_id_range_sub_hash(partition_info *part_info,
85 86
                                     uint32 *part_id,
                                     longlong *func_value);
87
int get_partition_id_range_sub_key(partition_info *part_info,
88 89
                                    uint32 *part_id,
                                    longlong *func_value);
90
int get_partition_id_range_sub_linear_hash(partition_info *part_info,
91 92
                                            uint32 *part_id,
                                            longlong *func_value);
93
int get_partition_id_range_sub_linear_key(partition_info *part_info,
94 95
                                           uint32 *part_id,
                                           longlong *func_value);
96
int get_partition_id_list_sub_hash(partition_info *part_info,
97 98
                                    uint32 *part_id,
                                    longlong *func_value);
99
int get_partition_id_list_sub_key(partition_info *part_info,
100 101
                                   uint32 *part_id,
                                   longlong *func_value);
102
int get_partition_id_list_sub_linear_hash(partition_info *part_info,
103 104
                                           uint32 *part_id,
                                           longlong *func_value);
105
int get_partition_id_list_sub_linear_key(partition_info *part_info,
106 107
                                          uint32 *part_id,
                                          longlong *func_value);
108 109 110 111
uint32 get_partition_id_hash_sub(partition_info *part_info); 
uint32 get_partition_id_key_sub(partition_info *part_info); 
uint32 get_partition_id_linear_hash_sub(partition_info *part_info); 
uint32 get_partition_id_linear_key_sub(partition_info *part_info); 
112 113
#endif

114 115 116 117 118 119
static uint32 get_next_partition_via_walking(PARTITION_ITERATOR*);
static uint32 get_next_subpartition_via_walking(PARTITION_ITERATOR*);
uint32 get_next_partition_id_range(PARTITION_ITERATOR* part_iter);
uint32 get_next_partition_id_list(PARTITION_ITERATOR* part_iter);
int get_part_iter_for_interval_via_mapping(partition_info *part_info,
                                           bool is_subpart,
120
                                           char *min_value, char *max_value,
121 122 123 124
                                           uint flags,
                                           PARTITION_ITERATOR *part_iter);
int get_part_iter_for_interval_via_walking(partition_info *part_info,
                                           bool is_subpart,
125
                                           char *min_value, char *max_value,
126 127 128
                                           uint flags,
                                           PARTITION_ITERATOR *part_iter);
static void set_up_range_analysis_info(partition_info *part_info);
129 130 131 132

/*
  A routine used by the parser to decide whether we are specifying a full
  partitioning or if only partitions to add or to split.
133

134 135 136
  SYNOPSIS
    is_partition_management()
    lex                    Reference to the lex object
137

138 139 140
  RETURN VALUE
    TRUE                   Yes, it is part of a management partition command
    FALSE                  No, not a management partition command
141

142
  DESCRIPTION
143 144
    This needs to be outside of WITH_PARTITION_STORAGE_ENGINE since it is
    used from the sql parser that doesn't have any #ifdef's
145 146 147 148 149 150
*/

my_bool is_partition_management(LEX *lex)
{
  return (lex->sql_command == SQLCOM_ALTER_TABLE &&
          (lex->alter_info.flags == ALTER_ADD_PARTITION ||
151
           lex->alter_info.flags == ALTER_REORGANIZE_PARTITION));
152 153
}

154
#ifdef WITH_PARTITION_STORAGE_ENGINE
155
/*
156 157
  A support function to check if a name is in a list of strings

158
  SYNOPSIS
159 160 161 162
    is_name_in_list()
    name               String searched for
    list_names         A list of names searched in

163 164 165 166 167
  RETURN VALUES
    TRUE               String found
    FALSE              String not found
*/

168 169
bool is_name_in_list(char *name,
                          List<char> list_names)
170
{
171 172
  List_iterator<char> names_it(list_names);
  uint no_names= list_names.elements;
173
  uint i= 0;
174

175 176
  do
  {
177 178
    char *list_name= names_it++;
    if (!(my_strcasecmp(system_charset_info, name, list_name)))
179 180 181 182 183 184
      return TRUE;
  } while (++i < no_names);
  return FALSE;
}


185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200

/*
  Set-up defaults for partitions. 

  SYNOPSIS
    partition_default_handling()
    table                         Table object
    table_name                    Table name to use when getting no_parts
    db_name                       Database name to use when getting no_parts
    part_info                     Partition info to set up

  RETURN VALUES
    TRUE                          Error
    FALSE                         Success
*/

201 202
bool partition_default_handling(TABLE *table, partition_info *part_info,
                                const char *normalized_path)
203 204 205 206 207
{
  DBUG_ENTER("partition_default_handling");

  if (part_info->use_default_no_partitions)
  {
208
    if (table->file->get_no_parts(normalized_path, &part_info->no_parts))
209 210 211 212
    {
      DBUG_RETURN(TRUE);
    }
  }
213
  else if (part_info->is_sub_partitioned() &&
214 215 216
           part_info->use_default_no_subpartitions)
  {
    uint no_parts;
217
    if (table->file->get_no_parts(normalized_path, &no_parts))
218 219 220 221 222 223 224
    {
      DBUG_RETURN(TRUE);
    }
    DBUG_ASSERT(part_info->no_parts > 0);
    part_info->no_subparts= no_parts / part_info->no_parts;
    DBUG_ASSERT((no_parts % part_info->no_parts) == 0);
  }
225 226
  part_info->set_up_defaults_for_partitioning(table->file,
                                              (ulonglong)0, (uint)0);
227 228 229 230
  DBUG_RETURN(FALSE);
}


231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
/*
  Check that the reorganized table will not have duplicate partitions.

  SYNOPSIS
    check_reorganise_list()
    new_part_info      New partition info
    old_part_info      Old partition info
    list_part_names    The list of partition names that will go away and can be reused in the
                       new table.

  RETURN VALUES
    TRUE               Inacceptable name conflict detected.
    FALSE              New names are OK.

  DESCRIPTION
    Can handle that the 'new_part_info' and 'old_part_info' the same
    in which case it checks that the list of names in the partitions
    doesn't contain any duplicated names.
*/

bool check_reorganise_list(partition_info *new_part_info,
                           partition_info *old_part_info,
                           List<char> list_part_names)
{
  uint new_count, old_count;
  uint no_new_parts= new_part_info->partitions.elements;
  uint no_old_parts= old_part_info->partitions.elements;
  List_iterator<partition_element> new_parts_it(new_part_info->partitions);
  bool same_part_info= (new_part_info == old_part_info);
  DBUG_ENTER("check_reorganise_list");

  new_count= 0;
  do
  {
    List_iterator<partition_element> old_parts_it(old_part_info->partitions);
    char *new_name= (new_parts_it++)->partition_name;
    new_count++;
    old_count= 0;
    do
    {
      char *old_name= (old_parts_it++)->partition_name;
      old_count++;
      if (same_part_info && old_count == new_count)
        break;
      if (!(my_strcasecmp(system_charset_info, old_name, new_name)))
      {
277
        if (!is_name_in_list(old_name, list_part_names))
278 279 280 281 282 283 284 285
          DBUG_RETURN(TRUE);
      }
    } while (old_count < no_old_parts);
  } while (new_count < no_new_parts);
  DBUG_RETURN(FALSE);
}


286 287 288
/*
  A useful routine used by update_row for partition handlers to calculate
  the partition ids of the old and the new record.
289

290 291 292 293 294 295
  SYNOPSIS
    get_part_for_update()
    old_data                Buffer of old record
    new_data                Buffer of new record
    rec0                    Reference to table->record[0]
    part_info               Reference to partition information
296 297 298
    out:old_part_id         The returned partition id of old record 
    out:new_part_id         The returned partition id of new record

299 300 301 302 303 304 305
  RETURN VALUE
    0                       Success
    > 0                     Error code
*/

int get_parts_for_update(const byte *old_data, byte *new_data,
                         const byte *rec0, partition_info *part_info,
306 307
                         uint32 *old_part_id, uint32 *new_part_id,
                         longlong *new_func_value)
308 309 310
{
  Field **part_field_array= part_info->full_part_field_array;
  int error;
311
  longlong old_func_value;
312 313
  DBUG_ENTER("get_parts_for_update");

314
  DBUG_ASSERT(new_data == rec0);
315
  set_field_ptr(part_field_array, old_data, rec0);
316 317
  error= part_info->get_partition_id(part_info, old_part_id,
                                     &old_func_value);
318 319 320 321 322 323 324 325 326 327
  set_field_ptr(part_field_array, rec0, old_data);
  if (unlikely(error))                             // Should never happen
  {
    DBUG_ASSERT(0);
    DBUG_RETURN(error);
  }
#ifdef NOT_NEEDED
  if (new_data == rec0)
#endif
  {
328 329 330
    if (unlikely(error= part_info->get_partition_id(part_info,
                                                    new_part_id,
                                                    new_func_value)))
331 332 333 334 335 336 337 338 339 340 341 342 343
    {
      DBUG_RETURN(error);
    }
  }
#ifdef NOT_NEEDED
  else
  {
    /*
      This branch should never execute but it is written anyways for
      future use. It will be tested by ensuring that the above
      condition is false in one test situation before pushing the code.
    */
    set_field_ptr(part_field_array, new_data, rec0);
344 345
    error= part_info->get_partition_id(part_info, new_part_id,
                                       new_func_value);
346 347 348 349 350 351 352 353 354 355 356 357 358 359
    set_field_ptr(part_field_array, rec0, new_data);
    if (unlikely(error))
    {
      DBUG_RETURN(error);
    }
  }
#endif
  DBUG_RETURN(0);
}


/*
  A useful routine used by delete_row for partition handlers to calculate
  the partition id.
360

361 362 363 364 365
  SYNOPSIS
    get_part_for_delete()
    buf                     Buffer of old record
    rec0                    Reference to table->record[0]
    part_info               Reference to partition information
366 367
    out:part_id             The returned partition id to delete from

368 369 370
  RETURN VALUE
    0                       Success
    > 0                     Error code
371

372 373 374 375 376 377 378 379 380 381
  DESCRIPTION
    Dependent on whether buf is not record[0] we need to prepare the
    fields. Then we call the function pointer get_partition_id to
    calculate the partition id.
*/

int get_part_for_delete(const byte *buf, const byte *rec0,
                        partition_info *part_info, uint32 *part_id)
{
  int error;
382
  longlong func_value;
383 384 385 386
  DBUG_ENTER("get_part_for_delete");

  if (likely(buf == rec0))
  {
387 388
    if (unlikely((error= part_info->get_partition_id(part_info, part_id,
                                                     &func_value))))
389 390 391 392 393 394 395 396 397
    {
      DBUG_RETURN(error);
    }
    DBUG_PRINT("info", ("Delete from partition %d", *part_id));
  }
  else
  {
    Field **part_field_array= part_info->full_part_field_array;
    set_field_ptr(part_field_array, buf, rec0);
398
    error= part_info->get_partition_id(part_info, part_id, &func_value);
399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414
    set_field_ptr(part_field_array, rec0, buf);
    if (unlikely(error))
    {
      DBUG_RETURN(error);
    }
    DBUG_PRINT("info", ("Delete from partition %d (path2)", *part_id));
  }
  DBUG_RETURN(0);
}


/*
  This routine allocates an array for all range constants to achieve a fast
  check what partition a certain value belongs to. At the same time it does
  also check that the range constants are defined in increasing order and
  that the expressions are constant integer expressions.
415

416 417
  SYNOPSIS
    check_range_constants()
418 419
    part_info             Partition info

420 421 422
  RETURN VALUE
    TRUE                An error occurred during creation of range constants
    FALSE               Successful creation of range constant mapping
423

424 425 426 427 428 429 430 431 432 433
  DESCRIPTION
    This routine is called from check_partition_info to get a quick error
    before we came too far into the CREATE TABLE process. It is also called
    from fix_partition_func every time we open the .frm file. It is only
    called for RANGE PARTITIONed tables.
*/

static bool check_range_constants(partition_info *part_info)
{
  partition_element* part_def;
434 435 436 437
  longlong current_largest_int= LONGLONG_MIN;
  longlong part_range_value_int;
  uint no_parts= part_info->no_parts;
  uint i;
438 439 440 441 442 443 444 445 446 447
  List_iterator<partition_element> it(part_info->partitions);
  bool result= TRUE;
  DBUG_ENTER("check_range_constants");
  DBUG_PRINT("enter", ("INT_RESULT with %d parts", no_parts));

  part_info->part_result_type= INT_RESULT;
  part_info->range_int_array= 
                      (longlong*)sql_alloc(no_parts * sizeof(longlong));
  if (unlikely(part_info->range_int_array == NULL))
  {
448
    mem_alloc_error(no_parts * sizeof(longlong));
449 450 451 452 453 454 455
    goto end;
  }
  i= 0;
  do
  {
    part_def= it++;
    if ((i != (no_parts - 1)) || !part_info->defined_max_value)
456
      part_range_value_int= part_def->range_value; 
457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478
    else
      part_range_value_int= LONGLONG_MAX;
    if (likely(current_largest_int < part_range_value_int))
    {
      current_largest_int= part_range_value_int;
      part_info->range_int_array[i]= part_range_value_int;
    }
    else
    {
      my_error(ER_RANGE_NOT_INCREASING_ERROR, MYF(0));
      goto end;
    }
  } while (++i < no_parts);
  result= FALSE;
end:
  DBUG_RETURN(result);
}


/*
  A support routine for check_list_constants used by qsort to sort the
  constant list expressions.
479

480 481 482 483
  SYNOPSIS
    list_part_cmp()
      a                First list constant to compare with
      b                Second list constant to compare with
484

485 486 487 488 489 490 491 492
  RETURN VALUE
    +1                 a > b
    0                  a  == b
    -1                 a < b
*/

static int list_part_cmp(const void* a, const void* b)
{
493 494
  longlong a1= ((LIST_PART_ENTRY*)a)->list_value;
  longlong b1= ((LIST_PART_ENTRY*)b)->list_value;
495 496 497 498 499 500 501 502 503 504 505 506 507 508
  if (a1 < b1)
    return -1;
  else if (a1 > b1)
    return +1;
  else
    return 0;
}


/*
  This routine allocates an array for all list constants to achieve a fast
  check what partition a certain value belongs to. At the same time it does
  also check that there are no duplicates among the list constants and that
  that the list expressions are constant integer expressions.
509

510 511
  SYNOPSIS
    check_list_constants()
512 513
    part_info             Partition info

514 515 516
  RETURN VALUE
    TRUE                  An error occurred during creation of list constants
    FALSE                 Successful creation of list constant mapping
517

518 519 520 521 522 523 524 525 526
  DESCRIPTION
    This routine is called from check_partition_info to get a quick error
    before we came too far into the CREATE TABLE process. It is also called
    from fix_partition_func every time we open the .frm file. It is only
    called for LIST PARTITIONed tables.
*/

static bool check_list_constants(partition_info *part_info)
{
527 528 529
  uint i, no_parts;
  uint no_list_values= 0;
  uint list_index= 0;
530
  longlong *list_value;
531 532
  bool not_first;
  bool result= TRUE;
533 534
  longlong curr_value, prev_value;
  partition_element* part_def;
535
  bool found_null= FALSE;
536 537 538 539 540 541 542 543 544 545 546
  List_iterator<partition_element> list_func_it(part_info->partitions);
  DBUG_ENTER("check_list_constants");

  part_info->part_result_type= INT_RESULT;

  /*
    We begin by calculating the number of list values that have been
    defined in the first step.

    We use this number to allocate a properly sized array of structs
    to keep the partition id and the value to use in that partition.
547
    In the second traversal we assign them values in the struct array.
548 549 550 551 552 553 554 555 556 557 558 559 560

    Finally we sort the array of structs in order of values to enable
    a quick binary search for the proper value to discover the
    partition id.
    After sorting the array we check that there are no duplicates in the
    list.
  */

  no_parts= part_info->no_parts;
  i= 0;
  do
  {
    part_def= list_func_it++;
561 562 563 564 565 566 567 568 569 570 571
    if (part_def->has_null_value)
    {
      if (found_null)
      {
        my_error(ER_MULTIPLE_DEF_CONST_IN_LIST_PART_ERROR, MYF(0));
        goto end;
      }
      part_info->has_null_value= TRUE;
      part_info->has_null_part_id= i;
      found_null= TRUE;
    }
572
    List_iterator<longlong> list_val_it1(part_def->list_val_list);
573 574 575 576 577 578 579 580 581
    while (list_val_it1++)
      no_list_values++;
  } while (++i < no_parts);
  list_func_it.rewind();
  part_info->no_list_values= no_list_values;
  part_info->list_array=
      (LIST_PART_ENTRY*)sql_alloc(no_list_values*sizeof(LIST_PART_ENTRY));
  if (unlikely(part_info->list_array == NULL))
  {
582
    mem_alloc_error(no_list_values * sizeof(LIST_PART_ENTRY));
583 584 585 586 587 588 589
    goto end;
  }

  i= 0;
  do
  {
    part_def= list_func_it++;
590 591
    List_iterator<longlong> list_val_it2(part_def->list_val_list);
    while ((list_value= list_val_it2++))
592
    {
593 594
      part_info->list_array[list_index].list_value= *list_value;
      part_info->list_array[list_index++].partition_id= i;
595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622
    }
  } while (++i < no_parts);

  qsort((void*)part_info->list_array, no_list_values,
        sizeof(LIST_PART_ENTRY), &list_part_cmp);

  not_first= FALSE;
  i= prev_value= 0; //prev_value initialised to quiet compiler
  do
  {
    curr_value= part_info->list_array[i].list_value;
    if (likely(!not_first || prev_value != curr_value))
    {
      prev_value= curr_value;
      not_first= TRUE;
    }
    else
    {
      my_error(ER_MULTIPLE_DEF_CONST_IN_LIST_PART_ERROR, MYF(0));
      goto end;
    }
  } while (++i < no_list_values);
  result= FALSE;
end:
  DBUG_RETURN(result);
}


623

624 625 626 627 628 629 630




/*
  Check that all partitions use the same storage engine.
  This is currently a limitation in this version.
631

632 633 634 635
  SYNOPSIS
    check_engine_mix()
    engine_array           An array of engine identifiers
    no_parts               Total number of partitions
636

637 638 639
  RETURN VALUE
    TRUE                   Error, mixed engines
    FALSE                  Ok, no mixed engines
640 641 642
  DESCRIPTION
    Current check verifies only that all handlers are the same.
    Later this check will be more sophisticated.
643 644
*/

645
static bool check_engine_mix(handlerton **engine_array, uint no_parts)
646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663
{
  uint i= 0;
  bool result= FALSE;
  DBUG_ENTER("check_engine_mix");

  do
  {
    if (engine_array[i] != engine_array[0])
    {
      result= TRUE;
      break;
    }
  } while (++i < no_parts);
  DBUG_RETURN(result);
}


/*
664 665
  This code is used early in the CREATE TABLE and ALTER TABLE process.

666 667 668 669 670
  SYNOPSIS
    check_partition_info()
    part_info           The reference to all partition information
    file                A reference to a handler of the table
    max_rows            Maximum number of rows stored in the table
671 672
    engine_type         Return value for used engine in partitions

673 674 675
  RETURN VALUE
    TRUE                 Error, something went wrong
    FALSE                Ok, full partition data structures are now generated
676

677
  DESCRIPTION
678 679 680 681 682
    We will check that the partition info requested is possible to set-up in
    this version. This routine is an extension of the parser one could say.
    If defaults were used we will generate default data structures for all
    partitions.

683 684
*/

685
bool check_partition_info(partition_info *part_info,handlerton **eng_type,
686 687
                          handler *file, ulonglong max_rows)
{
688
  handlerton **engine_array= NULL;
689 690
  uint part_count= 0;
  uint i, no_parts, tot_partitions;
691
  bool result= TRUE;
692
  char *same_name;
693 694
  DBUG_ENTER("check_partition_info");

695 696 697 698 699 700 701
  if (unlikely(!part_info->is_sub_partitioned() &&
               !(part_info->use_default_subpartitions &&
                 part_info->use_default_no_subpartitions)))
  {
    my_error(ER_SUBPARTITION_ERROR, MYF(0));
    goto end;
  }
702
  if (unlikely(part_info->is_sub_partitioned() &&
703 704 705 706
              (!(part_info->part_type == RANGE_PARTITION ||
                 part_info->part_type == LIST_PARTITION))))
  {
    /* Only RANGE and LIST partitioning can be subpartitioned */
707
    my_error(ER_PARTITION_SUBPART_MIX_ERROR, MYF(0));
708 709
    goto end;
  }
710 711 712
  if (unlikely(part_info->set_up_defaults_for_partitioning(file,
                                                           max_rows, 
                                                           (uint)0)))
713
    goto end;
714
  tot_partitions= part_info->get_tot_partitions();
715 716 717 718 719
  if (unlikely(tot_partitions > MAX_PARTITIONS))
  {
    my_error(ER_TOO_MANY_PARTITIONS_ERROR, MYF(0));
    goto end;
  }
720
  if ((same_name= part_info->has_unique_names()))
721
  {
722
    my_error(ER_SAME_NAME_PARTITION, MYF(0), same_name);
723 724
    goto end;
  }
725 726
  engine_array= (handlerton**)my_malloc(tot_partitions * sizeof(handlerton *), 
                                        MYF(MY_WME));
727 728 729 730 731
  if (unlikely(!engine_array))
    goto end;
  i= 0;
  no_parts= part_info->no_parts;
  {
732 733
    List_iterator<partition_element> part_it(part_info->partitions);
    do
734
    {
735
      partition_element *part_elem= part_it++;
736
      if (!part_info->is_sub_partitioned())
737
      {
738
        if (part_elem->engine_type == NULL)
739 740 741
          part_elem->engine_type= part_info->default_engine_type;
        DBUG_PRINT("info", ("engine = %d",
                   ha_legacy_type(part_elem->engine_type)));
742
        engine_array[part_count++]= part_elem->engine_type;
743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759
      }
      else
      {
        uint j= 0, no_subparts= part_info->no_subparts;;
        List_iterator<partition_element> sub_it(part_elem->subpartitions);
        do
        {
          part_elem= sub_it++;
          if (part_elem->engine_type == NULL)
            part_elem->engine_type= part_info->default_engine_type;
          DBUG_PRINT("info", ("engine = %u",
                     ha_legacy_type(part_elem->engine_type)));
          engine_array[part_count++]= part_elem->engine_type;
        } while (++j < no_subparts);
      }
    } while (++i < part_info->no_parts);
  }
760 761 762 763 764 765
  if (unlikely(check_engine_mix(engine_array, part_count)))
  {
    my_error(ER_MIX_HANDLER_ERROR, MYF(0));
    goto end;
  }

766 767 768
  if (eng_type)
    *eng_type= (handlerton*)engine_array[0];

769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787
  /*
    We need to check all constant expressions that they are of the correct
    type and that they are increasing for ranges and not overlapping for
    list constants.
  */

  if (unlikely((part_info->part_type == RANGE_PARTITION &&
                check_range_constants(part_info)) ||
               (part_info->part_type == LIST_PARTITION &&
                check_list_constants(part_info))))
    goto end;
  result= FALSE;
end:
  my_free((char*)engine_array,MYF(MY_ALLOW_ZERO_PTR));
  DBUG_RETURN(result);
}


/*
788 789 790
  This method is used to set-up both partition and subpartitioning
  field array and used for all types of partitioning.
  It is part of the logic around fix_partition_func.
791 792 793 794 795

  SYNOPSIS
    set_up_field_array()
    table                TABLE object for which partition fields are set-up
    sub_part             Is the table subpartitioned as well
796

797 798 799
  RETURN VALUE
    TRUE                 Error, some field didn't meet requirements
    FALSE                Ok, partition field array set-up
800

801
  DESCRIPTION
802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826

    A great number of functions below here is part of the fix_partition_func
    method. It is used to set up the partition structures for execution from
    openfrm. It is called at the end of the openfrm when the table struct has
    been set-up apart from the partition information.
    It involves:
    1) Setting arrays of fields for the partition functions.
    2) Setting up binary search array for LIST partitioning
    3) Setting up array for binary search for RANGE partitioning
    4) Setting up key_map's to assist in quick evaluation whether one
       can deduce anything from a given index of what partition to use
    5) Checking whether a set of partitions can be derived from a range on
       a field in the partition function.
    As part of doing this there is also a great number of error controls.
    This is actually the place where most of the things are checked for
    partition information when creating a table.
    Things that are checked includes
    1) All fields of partition function in Primary keys and unique indexes
       (if not supported)


    Create an array of partition fields (NULL terminated). Before this method
    is called fix_fields or find_table_in_sef has been called to set
    GET_FIXED_FIELDS_FLAG on all fields that are part of the partition
    function.
827
*/
828

829
static bool set_up_field_array(TABLE *table,
830
                              bool is_sub_part)
831 832
{
  Field **ptr, *field, **field_array;
833 834 835
  uint no_fields= 0;
  uint size_field_array;
  uint i= 0;
836
  partition_info *part_info= table->part_info;
837 838 839 840 841 842 843 844 845
  int result= FALSE;
  DBUG_ENTER("set_up_field_array");

  ptr= table->field;
  while ((field= *(ptr++))) 
  {
    if (field->flags & GET_FIXED_FIELDS_FLAG)
      no_fields++;
  }
846 847 848 849 850 851 852 853
  if (no_fields == 0)
  {
    /*
      We are using hidden key as partitioning field
    */
    DBUG_ASSERT(!is_sub_part);
    DBUG_RETURN(result);
  }
854 855 856 857
  size_field_array= (no_fields+1)*sizeof(Field*);
  field_array= (Field**)sql_alloc(size_field_array);
  if (unlikely(!field_array))
  {
858
    mem_alloc_error(size_field_array);
859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888
    result= TRUE;
  }
  ptr= table->field;
  while ((field= *(ptr++))) 
  {
    if (field->flags & GET_FIXED_FIELDS_FLAG)
    {
      field->flags&= ~GET_FIXED_FIELDS_FLAG;
      field->flags|= FIELD_IN_PART_FUNC_FLAG;
      if (likely(!result))
      {
        field_array[i++]= field;

        /*
          We check that the fields are proper. It is required for each
          field in a partition function to:
          1) Not be a BLOB of any type
            A BLOB takes too long time to evaluate so we don't want it for
            performance reasons.
        */

        if (unlikely(field->flags & BLOB_FLAG))
        {
          my_error(ER_BLOB_FIELD_IN_PART_FUNC_ERROR, MYF(0));
          result= TRUE;
        }
      }
    }
  }
  field_array[no_fields]= 0;
889
  if (!is_sub_part)
890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905
  {
    part_info->part_field_array= field_array;
    part_info->no_part_fields= no_fields;
  }
  else
  {
    part_info->subpart_field_array= field_array;
    part_info->no_subpart_fields= no_fields;
  }
  DBUG_RETURN(result);
}


/*
  Create a field array including all fields of both the partitioning and the
  subpartitioning functions.
906

907 908 909 910
  SYNOPSIS
    create_full_part_field_array()
    table                TABLE object for which partition fields are set-up
    part_info            Reference to partitioning data structure
911

912 913 914
  RETURN VALUE
    TRUE                 Memory allocation of field array failed
    FALSE                Ok
915

916 917 918 919 920 921 922 923 924 925 926 927 928
  DESCRIPTION
    If there is no subpartitioning then the same array is used as for the
    partitioning. Otherwise a new array is built up using the flag
    FIELD_IN_PART_FUNC in the field object.
    This function is called from fix_partition_func
*/

static bool create_full_part_field_array(TABLE *table,
                                         partition_info *part_info)
{
  bool result= FALSE;
  DBUG_ENTER("create_full_part_field_array");

929
  if (!part_info->is_sub_partitioned())
930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947
  {
    part_info->full_part_field_array= part_info->part_field_array;
    part_info->no_full_part_fields= part_info->no_part_fields;
  }
  else
  {
    Field **ptr, *field, **field_array;
    uint no_part_fields=0, size_field_array;
    ptr= table->field;
    while ((field= *(ptr++)))
    {
      if (field->flags & FIELD_IN_PART_FUNC_FLAG)
        no_part_fields++;
    }
    size_field_array= (no_part_fields+1)*sizeof(Field*);
    field_array= (Field**)sql_alloc(size_field_array);
    if (unlikely(!field_array))
    {
948
      mem_alloc_error(size_field_array);
949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971
      result= TRUE;
      goto end;
    }
    no_part_fields= 0;
    ptr= table->field;
    while ((field= *(ptr++)))
    {
      if (field->flags & FIELD_IN_PART_FUNC_FLAG)
        field_array[no_part_fields++]= field;
    }
    field_array[no_part_fields]=0;
    part_info->full_part_field_array= field_array;
    part_info->no_full_part_fields= no_part_fields;
  }
end:
  DBUG_RETURN(result);
}


/*

  Clear flag GET_FIXED_FIELDS_FLAG in all fields of a key previously set by
  set_indicator_in_key_fields (always used in pairs).
972

973 974 975
  SYNOPSIS
    clear_indicator_in_key_fields()
    key_info                  Reference to find the key fields
976 977 978 979 980 981 982 983 984 985 986 987

  RETURN VALUE
    NONE

  DESCRIPTION
    These support routines is used to set/reset an indicator of all fields
    in a certain key. It is used in conjunction with another support routine
    that traverse all fields in the PF to find if all or some fields in the
    PF is part of the key. This is used to check primary keys and unique
    keys involve all fields in PF (unless supported) and to derive the
    key_map's used to quickly decide whether the index can be used to
    derive which partitions are needed to scan.
988 989 990 991 992 993 994 995 996 997 998 999 1000
*/

static void clear_indicator_in_key_fields(KEY *key_info)
{
  KEY_PART_INFO *key_part;
  uint key_parts= key_info->key_parts, i;
  for (i= 0, key_part=key_info->key_part; i < key_parts; i++, key_part++)
    key_part->field->flags&= (~GET_FIXED_FIELDS_FLAG);
}


/*
  Set flag GET_FIXED_FIELDS_FLAG in all fields of a key.
1001

1002 1003 1004
  SYNOPSIS
    set_indicator_in_key_fields
    key_info                  Reference to find the key fields
1005 1006 1007

  RETURN VALUE
    NONE
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
*/

static void set_indicator_in_key_fields(KEY *key_info)
{
  KEY_PART_INFO *key_part;
  uint key_parts= key_info->key_parts, i;
  for (i= 0, key_part=key_info->key_part; i < key_parts; i++, key_part++)
    key_part->field->flags|= GET_FIXED_FIELDS_FLAG;
}


/*
  Check if all or some fields in partition field array is part of a key
  previously used to tag key fields.
1022

1023 1024 1025
  SYNOPSIS
    check_fields_in_PF()
    ptr                  Partition field array
1026 1027 1028
    out:all_fields       Is all fields of partition field array used in key
    out:some_fields      Is some fields of partition field array used in key

1029 1030 1031 1032 1033 1034 1035 1036
  RETURN VALUE
    all_fields, some_fields
*/

static void check_fields_in_PF(Field **ptr, bool *all_fields,
                               bool *some_fields)
{
  DBUG_ENTER("check_fields_in_PF");
1037

1038 1039
  *all_fields= TRUE;
  *some_fields= FALSE;
1040 1041 1042 1043 1044
  if ((!ptr) || !(*ptr))
  {
    *all_fields= FALSE;
    DBUG_VOID_RETURN;
  }
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
  do
  {
  /* Check if the field of the PF is part of the current key investigated */
    if ((*ptr)->flags & GET_FIXED_FIELDS_FLAG)
      *some_fields= TRUE; 
    else
      *all_fields= FALSE;
  } while (*(++ptr));
  DBUG_VOID_RETURN;
}


/*
  Clear flag GET_FIXED_FIELDS_FLAG in all fields of the table.
  This routine is used for error handling purposes.
1060

1061 1062 1063
  SYNOPSIS
    clear_field_flag()
    table                TABLE object for which partition fields are set-up
1064 1065 1066

  RETURN VALUE
    NONE
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
*/

static void clear_field_flag(TABLE *table)
{
  Field **ptr;
  DBUG_ENTER("clear_field_flag");

  for (ptr= table->field; *ptr; ptr++)
    (*ptr)->flags&= (~GET_FIXED_FIELDS_FLAG);
  DBUG_VOID_RETURN;
}


/*
1081 1082 1083
  find_field_in_table_sef finds the field given its name. All fields get
  GET_FIXED_FIELDS_FLAG set.

1084 1085 1086 1087 1088 1089
  SYNOPSIS
    handle_list_of_fields()
    it                   A list of field names for the partition function
    table                TABLE object for which partition fields are set-up
    part_info            Reference to partitioning data structure
    sub_part             Is the table subpartitioned as well
1090

1091 1092 1093
  RETURN VALUE
    TRUE                 Fields in list of fields not part of table
    FALSE                All fields ok and array created
1094

1095
  DESCRIPTION
1096 1097 1098 1099
    This routine sets-up the partition field array for KEY partitioning, it
    also verifies that all fields in the list of fields is actually a part of
    the table.

1100 1101
*/

1102

1103 1104 1105
static bool handle_list_of_fields(List_iterator<char> it,
                                  TABLE *table,
                                  partition_info *part_info,
1106
                                  bool is_sub_part)
1107 1108 1109 1110
{
  Field *field;
  bool result;
  char *field_name;
1111
  bool is_list_empty= TRUE;
1112 1113 1114 1115
  DBUG_ENTER("handle_list_of_fields");

  while ((field_name= it++))
  {
1116
    is_list_empty= FALSE;
1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
    field= find_field_in_table_sef(table, field_name);
    if (likely(field != 0))
      field->flags|= GET_FIXED_FIELDS_FLAG;
    else
    {
      my_error(ER_FIELD_NOT_FOUND_PART_ERROR, MYF(0));
      clear_field_flag(table);
      result= TRUE;
      goto end;
    }
  }
1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164
  if (is_list_empty)
  {
    uint primary_key= table->s->primary_key;
    if (primary_key != MAX_KEY)
    {
      uint no_key_parts= table->key_info[primary_key].key_parts, i;
      /*
        In the case of an empty list we use primary key as partition key.
      */
      for (i= 0; i < no_key_parts; i++)
      {
        Field *field= table->key_info[primary_key].key_part[i].field;
        field->flags|= GET_FIXED_FIELDS_FLAG;
      }
    }
    else
    {
      if (table->s->db_type->partition_flags &&
          (table->s->db_type->partition_flags() & HA_USE_AUTO_PARTITION) &&
          (table->s->db_type->partition_flags() & HA_CAN_PARTITION))
      {
        /*
          This engine can handle automatic partitioning and there is no
          primary key. In this case we rely on that the engine handles
          partitioning based on a hidden key. Thus we allocate no
          array for partitioning fields.
        */
        DBUG_RETURN(FALSE);
      }
      else
      {
        my_error(ER_FIELD_NOT_FOUND_PART_ERROR, MYF(0));
        DBUG_RETURN(TRUE);
      }
    }
  }
  result= set_up_field_array(table, is_sub_part);
1165 1166 1167 1168 1169 1170
end:
  DBUG_RETURN(result);
}


/*
1171 1172 1173 1174 1175
  The function uses a new feature in fix_fields where the flag 
  GET_FIXED_FIELDS_FLAG is set for all fields in the item tree.
  This field must always be reset before returning from the function
  since it is used for other purposes as well.

1176 1177 1178 1179 1180 1181 1182
  SYNOPSIS
    fix_fields_part_func()
    thd                  The thread object
    tables               A list of one table, the partitioned table
    func_expr            The item tree reference of the partition function
    part_info            Reference to partitioning data structure
    sub_part             Is the table subpartitioned as well
1183

1184 1185 1186 1187
  RETURN VALUE
    TRUE                 An error occurred, something was wrong with the
                         partition function.
    FALSE                Ok, a partition field array was created
1188

1189
  DESCRIPTION
1190 1191 1192 1193 1194 1195
    This function is used to build an array of partition fields for the
    partitioning function and subpartitioning function. The partitioning
    function is an item tree that must reference at least one field in the
    table. This is checked first in the parser that the function doesn't
    contain non-cacheable parts (like a random function) and by checking
    here that the function isn't a constant function.
1196 1197 1198 1199 1200 1201 1202

    Calculate the number of fields in the partition function.
    Use it allocate memory for array of Field pointers.
    Initialise array of field pointers. Use information set when
    calling fix_fields and reset it immediately after.
    The get_fields_in_item_tree activates setting of bit in flags
    on the field object.
1203
*/
1204

1205 1206 1207 1208
static bool fix_fields_part_func(THD *thd, TABLE_LIST *tables,
                                 Item* func_expr, partition_info *part_info,
                                 bool is_sub_part)
{
1209 1210
  bool result= TRUE;
  TABLE *table= tables->table;
1211
  TABLE_LIST *save_table_list, *save_first_table, *save_last_table;
1212
  int error;
1213
  Name_resolution_context *context;
1214
  const char *save_where;
1215 1216
  DBUG_ENTER("fix_fields_part_func");

1217
  context= thd->lex->current_context();
1218 1219
  table->map= 1; //To ensure correct calculation of const item
  table->get_fields_in_item_tree= TRUE;
1220 1221 1222
  save_table_list= context->table_list;
  save_first_table= context->first_name_resolution_table;
  save_last_table= context->last_name_resolution_table;
1223
  context->table_list= tables;
1224 1225 1226
  context->first_name_resolution_table= tables;
  context->last_name_resolution_table= NULL;
  func_expr->walk(&Item::change_context_processor, (byte*) context);
1227
  save_where= thd->where;
1228 1229
  thd->where= "partition function";
  error= func_expr->fix_fields(thd, (Item**)0);
1230 1231 1232
  context->table_list= save_table_list;
  context->first_name_resolution_table= save_first_table;
  context->last_name_resolution_table= save_last_table;
1233 1234 1235 1236 1237 1238
  if (unlikely(error))
  {
    DBUG_PRINT("info", ("Field in partition function not part of table"));
    clear_field_flag(table);
    goto end;
  }
1239
  thd->where= save_where;
1240 1241 1242 1243 1244 1245
  if (unlikely(func_expr->const_item()))
  {
    my_error(ER_CONST_EXPR_IN_PARTITION_FUNC_ERROR, MYF(0));
    clear_field_flag(table);
    goto end;
  }
1246
  result= set_up_field_array(table, is_sub_part);
1247 1248 1249 1250 1251 1252 1253 1254
end:
  table->get_fields_in_item_tree= FALSE;
  table->map= 0; //Restore old value
  DBUG_RETURN(result);
}


/*
1255 1256
  Check that the primary key contains all partition fields if defined

1257 1258 1259
  SYNOPSIS
    check_primary_key()
    table                TABLE object for which partition fields are set-up
1260

1261 1262 1263 1264 1265
  RETURN VALUES
    TRUE                 Not all fields in partitioning function was part
                         of primary key
    FALSE                Ok, all fields of partitioning function were part
                         of primary key
1266 1267 1268 1269 1270 1271

  DESCRIPTION
    This function verifies that if there is a primary key that it contains
    all the fields of the partition function.
    This is a temporary limitation that will hopefully be removed after a
    while.
1272 1273 1274 1275 1276
*/

static bool check_primary_key(TABLE *table)
{
  uint primary_key= table->s->primary_key;
1277 1278
  bool all_fields, some_fields;
  bool result= FALSE;
1279 1280 1281 1282 1283
  DBUG_ENTER("check_primary_key");

  if (primary_key < MAX_KEY)
  {
    set_indicator_in_key_fields(table->key_info+primary_key);
1284
    check_fields_in_PF(table->part_info->full_part_field_array,
1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
                        &all_fields, &some_fields);
    clear_indicator_in_key_fields(table->key_info+primary_key);
    if (unlikely(!all_fields))
    {
      my_error(ER_UNIQUE_KEY_NEED_ALL_FIELDS_IN_PF,MYF(0),"PRIMARY KEY");
      result= TRUE;
    }
  }
  DBUG_RETURN(result);
}


/*
1298 1299
  Check that unique keys contains all partition fields

1300 1301 1302
  SYNOPSIS
    check_unique_keys()
    table                TABLE object for which partition fields are set-up
1303

1304 1305 1306 1307 1308
  RETURN VALUES
    TRUE                 Not all fields in partitioning function was part
                         of all unique keys
    FALSE                Ok, all fields of partitioning function were part
                         of unique keys
1309 1310 1311 1312 1313 1314

  DESCRIPTION
    This function verifies that if there is a unique index that it contains
    all the fields of the partition function.
    This is a temporary limitation that will hopefully be removed after a
    while.
1315 1316 1317 1318
*/

static bool check_unique_keys(TABLE *table)
{
1319 1320 1321 1322
  bool all_fields, some_fields;
  bool result= FALSE;
  uint keys= table->s->keys;
  uint i;
1323
  DBUG_ENTER("check_unique_keys");
1324

1325 1326 1327 1328 1329
  for (i= 0; i < keys; i++)
  {
    if (table->key_info[i].flags & HA_NOSAME) //Unique index
    {
      set_indicator_in_key_fields(table->key_info+i);
1330
      check_fields_in_PF(table->part_info->full_part_field_array,
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387
                         &all_fields, &some_fields);
      clear_indicator_in_key_fields(table->key_info+i);
      if (unlikely(!all_fields))
      {
        my_error(ER_UNIQUE_KEY_NEED_ALL_FIELDS_IN_PF,MYF(0),"UNIQUE INDEX");
        result= TRUE;
        break;
      }
    }
  }
  DBUG_RETURN(result);
}


/*
  An important optimisation is whether a range on a field can select a subset
  of the partitions.
  A prerequisite for this to happen is that the PF is a growing function OR
  a shrinking function.
  This can never happen for a multi-dimensional PF. Thus this can only happen
  with PF with at most one field involved in the PF.
  The idea is that if the function is a growing function and you know that
  the field of the PF is 4 <= A <= 6 then we can convert this to a range
  in the PF instead by setting the range to PF(4) <= PF(A) <= PF(6). In the
  case of RANGE PARTITIONING and LIST PARTITIONING this can be used to
  calculate a set of partitions rather than scanning all of them.
  Thus the following prerequisites are there to check if sets of partitions
  can be found.
  1) Only possible for RANGE and LIST partitioning (not for subpartitioning)
  2) Only possible if PF only contains 1 field
  3) Possible if PF is a growing function of the field
  4) Possible if PF is a shrinking function of the field
  OBSERVATION:
  1) IF f1(A) is a growing function AND f2(A) is a growing function THEN
     f1(A) + f2(A) is a growing function
     f1(A) * f2(A) is a growing function if f1(A) >= 0 and f2(A) >= 0
  2) IF f1(A) is a growing function and f2(A) is a shrinking function THEN
     f1(A) / f2(A) is a growing function if f1(A) >= 0 and f2(A) > 0
  3) IF A is a growing function then a function f(A) that removes the
     least significant portion of A is a growing function
     E.g. DATE(datetime) is a growing function
     MONTH(datetime) is not a growing/shrinking function
  4) IF f1(A) is a growing function and f2(A) is a growing function THEN
     f1(f2(A)) and f2(f1(A)) are also growing functions
  5) IF f1(A) is a shrinking function and f2(A) is a growing function THEN
     f1(f2(A)) is a shrinking function and f2(f1(A)) is a shrinking function
  6) f1(A) = A is a growing function
  7) f1(A) = A*a + b (where a and b are constants) is a growing function

  By analysing the item tree of the PF we can use these deducements and
  derive whether the PF is a growing function or a shrinking function or
  neither of it.

  If the PF is range capable then a flag is set on the table object
  indicating this to notify that we can use also ranges on the field
  of the PF to deduce a set of partitions if the fields of the PF were
  not all fully bound.
1388

1389 1390 1391
  SYNOPSIS
    check_range_capable_PF()
    table                TABLE object for which partition fields are set-up
1392

1393 1394 1395 1396 1397 1398 1399
  DESCRIPTION
    Support for this is not implemented yet.
*/

void check_range_capable_PF(TABLE *table)
{
  DBUG_ENTER("check_range_capable_PF");
1400

1401 1402 1403 1404
  DBUG_VOID_RETURN;
}


1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
/*
  Set up partition bitmap

  SYNOPSIS
    set_up_partition_bitmap()
    thd                  Thread object
    part_info            Reference to partitioning data structure

  RETURN VALUE
    TRUE                 Memory allocation failure
    FALSE                Success

  DESCRIPTION
    Allocate memory for bitmap of the partitioned table
    and initialise it.
*/

static bool set_up_partition_bitmap(THD *thd, partition_info *part_info)
{
  uint32 *bitmap_buf;
  uint bitmap_bits= part_info->no_subparts? 
                     (part_info->no_subparts* part_info->no_parts):
                      part_info->no_parts;
  uint bitmap_bytes= bitmap_buffer_size(bitmap_bits);
  DBUG_ENTER("set_up_partition_bitmap");

  if (!(bitmap_buf= (uint32*)thd->alloc(bitmap_bytes)))
  {
    mem_alloc_error(bitmap_bytes);
    DBUG_RETURN(TRUE);
  }
  bitmap_init(&part_info->used_partitions, bitmap_buf, bitmap_bytes*8, FALSE);
1437
  bitmap_set_all(&part_info->used_partitions);
1438 1439 1440 1441
  DBUG_RETURN(FALSE);
}


1442 1443
/*
  Set up partition key maps
1444

1445 1446 1447 1448
  SYNOPSIS
    set_up_partition_key_maps()
    table                TABLE object for which partition fields are set-up
    part_info            Reference to partitioning data structure
1449

1450 1451
  RETURN VALUES
    None
1452

1453
  DESCRIPTION
1454 1455 1456 1457 1458 1459 1460 1461 1462 1463
    This function sets up a couple of key maps to be able to quickly check
    if an index ever can be used to deduce the partition fields or even
    a part of the fields of the  partition function.
    We set up the following key_map's.
    PF = Partition Function
    1) All fields of the PF is set even by equal on the first fields in the
       key
    2) All fields of the PF is set if all fields of the key is set
    3) At least one field in the PF is set if all fields is set
    4) At least one field in the PF is part of the key
1464 1465 1466 1467 1468
*/

static void set_up_partition_key_maps(TABLE *table,
                                      partition_info *part_info)
{
1469 1470
  uint keys= table->s->keys;
  uint i;
1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486
  bool all_fields, some_fields;
  DBUG_ENTER("set_up_partition_key_maps");

  part_info->all_fields_in_PF.clear_all();
  part_info->all_fields_in_PPF.clear_all();
  part_info->all_fields_in_SPF.clear_all();
  part_info->some_fields_in_PF.clear_all();
  for (i= 0; i < keys; i++)
  {
    set_indicator_in_key_fields(table->key_info+i);
    check_fields_in_PF(part_info->full_part_field_array,
                       &all_fields, &some_fields);
    if (all_fields)
      part_info->all_fields_in_PF.set_bit(i);
    if (some_fields)
      part_info->some_fields_in_PF.set_bit(i);
1487
    if (part_info->is_sub_partitioned())
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
    {
      check_fields_in_PF(part_info->part_field_array,
                         &all_fields, &some_fields);
      if (all_fields)
        part_info->all_fields_in_PPF.set_bit(i);
      check_fields_in_PF(part_info->subpart_field_array,
                         &all_fields, &some_fields);
      if (all_fields)
        part_info->all_fields_in_SPF.set_bit(i);
    }
    clear_indicator_in_key_fields(table->key_info+i);
  }
  DBUG_VOID_RETURN;
}


/*
1505 1506
  Set up function pointers for partition function

1507
  SYNOPSIS
1508
    set_up_partition_func_pointers()
1509
    part_info            Reference to partitioning data structure
1510 1511 1512 1513 1514 1515 1516 1517 1518

  RETURN VALUE
    NONE

  DESCRIPTION
    Set-up all function pointers for calculation of partition id,
    subpartition id and the upper part in subpartitioning. This is to speed up
    execution of get_partition_id which is executed once every record to be
    written and deleted and twice for updates.
1519 1520 1521 1522
*/

static void set_up_partition_func_pointers(partition_info *part_info)
{
1523 1524
  DBUG_ENTER("set_up_partition_func_pointers");

1525
  if (part_info->is_sub_partitioned())
1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556
  {
    if (part_info->part_type == RANGE_PARTITION)
    {
      part_info->get_part_partition_id= get_partition_id_range;
      if (part_info->list_of_subpart_fields)
      {
        if (part_info->linear_hash_ind)
        {
          part_info->get_partition_id= get_partition_id_range_sub_linear_key;
          part_info->get_subpartition_id= get_partition_id_linear_key_sub;
        }
        else
        {
          part_info->get_partition_id= get_partition_id_range_sub_key;
          part_info->get_subpartition_id= get_partition_id_key_sub;
        }
      }
      else
      {
        if (part_info->linear_hash_ind)
        {
          part_info->get_partition_id= get_partition_id_range_sub_linear_hash;
          part_info->get_subpartition_id= get_partition_id_linear_hash_sub;
        }
        else
        {
          part_info->get_partition_id= get_partition_id_range_sub_hash;
          part_info->get_subpartition_id= get_partition_id_hash_sub;
        }
      }
    }
1557
    else /* LIST Partitioning */
1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
    {
      part_info->get_part_partition_id= get_partition_id_list;
      if (part_info->list_of_subpart_fields)
      {
        if (part_info->linear_hash_ind)
        {
          part_info->get_partition_id= get_partition_id_list_sub_linear_key;
          part_info->get_subpartition_id= get_partition_id_linear_key_sub;
        }
        else
        {
          part_info->get_partition_id= get_partition_id_list_sub_key;
          part_info->get_subpartition_id= get_partition_id_key_sub;
        }
      }
      else
      {
        if (part_info->linear_hash_ind)
        {
          part_info->get_partition_id= get_partition_id_list_sub_linear_hash;
          part_info->get_subpartition_id= get_partition_id_linear_hash_sub;
        }
        else
        {
          part_info->get_partition_id= get_partition_id_list_sub_hash;
          part_info->get_subpartition_id= get_partition_id_hash_sub;
        }
      }
    }
  }
1588
  else /* No subpartitioning */
1589 1590 1591 1592 1593 1594 1595
  {
    part_info->get_part_partition_id= NULL;
    part_info->get_subpartition_id= NULL;
    if (part_info->part_type == RANGE_PARTITION)
      part_info->get_partition_id= get_partition_id_range;
    else if (part_info->part_type == LIST_PARTITION)
      part_info->get_partition_id= get_partition_id_list;
1596
    else /* HASH partitioning */
1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
    {
      if (part_info->list_of_part_fields)
      {
        if (part_info->linear_hash_ind)
          part_info->get_partition_id= get_partition_id_linear_key_nosub;
        else
          part_info->get_partition_id= get_partition_id_key_nosub;
      }
      else
      {
        if (part_info->linear_hash_ind)
          part_info->get_partition_id= get_partition_id_linear_hash_nosub;
        else
          part_info->get_partition_id= get_partition_id_hash_nosub;
      }
    }
  }
1614
  DBUG_VOID_RETURN;
1615
}
1616 1617


1618 1619 1620
/*
  For linear hashing we need a mask which is on the form 2**n - 1 where
  2**n >= no_parts. Thus if no_parts is 6 then mask is 2**3 - 1 = 8 - 1 = 7.
1621

1622 1623 1624 1625
  SYNOPSIS
    set_linear_hash_mask()
    part_info            Reference to partitioning data structure
    no_parts             Number of parts in linear hash partitioning
1626 1627 1628

  RETURN VALUE
    NONE
1629 1630 1631 1632 1633
*/

static void set_linear_hash_mask(partition_info *part_info, uint no_parts)
{
  uint mask;
1634

1635 1636 1637 1638 1639 1640 1641 1642 1643
  for (mask= 1; mask < no_parts; mask<<=1)
    ;
  part_info->linear_hash_mask= mask - 1;
}


/*
  This function calculates the partition id provided the result of the hash
  function using linear hashing parameters, mask and number of partitions.
1644

1645 1646 1647 1648 1649
  SYNOPSIS
    get_part_id_from_linear_hash()
    hash_value          Hash value calculated by HASH function or KEY function
    mask                Mask calculated previously by set_linear_hash_mask
    no_parts            Number of partitions in HASH partitioned part
1650

1651 1652
  RETURN VALUE
    part_id             The calculated partition identity (starting at 0)
1653

1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664
  DESCRIPTION
    The partition is calculated according to the theory of linear hashing.
    See e.g. Linear hashing: a new tool for file and table addressing,
    Reprinted from VLDB-80 in Readings Database Systems, 2nd ed, M. Stonebraker
    (ed.), Morgan Kaufmann 1994.
*/

static uint32 get_part_id_from_linear_hash(longlong hash_value, uint mask,
                                           uint no_parts)
{
  uint32 part_id= (uint32)(hash_value & mask);
1665

1666 1667 1668
  if (part_id >= no_parts)
  {
    uint new_mask= ((mask + 1) >> 1) - 1;
1669
    part_id= (uint32)(hash_value & new_mask);
1670 1671 1672 1673 1674
  }
  return part_id;
}

/*
1675 1676
  fix partition functions

1677 1678 1679 1680 1681
  SYNOPSIS
    fix_partition_func()
    thd                  The thread object
    name                 The name of the partitioned table
    table                TABLE object for which partition fields are set-up
1682 1683
    create_table_ind     Indicator of whether openfrm was called as part of
                         CREATE or ALTER TABLE
1684

1685
  RETURN VALUE
1686 1687
    TRUE                 Error
    FALSE                Success
1688

1689 1690 1691 1692
  DESCRIPTION
    The name parameter contains the full table name and is used to get the
    database name of the table which is used to set-up a correct
    TABLE_LIST object for use in fix_fields.
1693 1694 1695 1696 1697 1698 1699

NOTES
    This function is called as part of opening the table by opening the .frm
    file. It is a part of CREATE TABLE to do this so it is quite permissible
    that errors due to erroneus syntax isn't found until we come here.
    If the user has used a non-existing field in the table is one such example
    of an error that is not discovered until here.
1700 1701
*/

1702 1703
bool fix_partition_func(THD *thd, const char* name, TABLE *table,
                        bool is_create_table_ind)
1704 1705 1706 1707 1708 1709 1710
{
  bool result= TRUE;
  uint dir_length, home_dir_length;
  TABLE_LIST tables;
  TABLE_SHARE *share= table->s;
  char db_name_string[FN_REFLEN];
  char* db_name;
1711
  partition_info *part_info= table->part_info;
1712 1713 1714
  ulong save_set_query_id= thd->set_query_id;
  DBUG_ENTER("fix_partition_func");

1715 1716 1717 1718
  if (part_info->fixed)
  {
    DBUG_RETURN(FALSE);
  }
1719
  thd->set_query_id= 0;
1720
  DBUG_PRINT("info", ("thd->set_query_id: %d", thd->set_query_id));
1721
  /*
1722 1723 1724
    Set-up the TABLE_LIST object to be a list with a single table
    Set the object to zero to create NULL pointers and set alias
    and real name to table name and get database name from file name.
1725 1726 1727
  */

  bzero((void*)&tables, sizeof(TABLE_LIST));
1728
  tables.alias= tables.table_name= (char*) share->table_name.str;
1729
  tables.table= table;
1730 1731
  tables.next_local= 0;
  tables.next_name_resolution_table= 0;
1732 1733 1734 1735 1736 1737 1738
  strmov(db_name_string, name);
  dir_length= dirname_length(db_name_string);
  db_name_string[dir_length - 1]= 0;
  home_dir_length= dirname_length(db_name_string);
  db_name= &db_name_string[home_dir_length];
  tables.db= db_name;

1739 1740
  if (!is_create_table_ind)
  {
1741 1742
    if (partition_default_handling(table, part_info,
                                   table->s->normalized_path.str))
1743 1744 1745 1746
    {
      DBUG_RETURN(TRUE);
    }
  }
1747
  if (part_info->is_sub_partitioned())
1748 1749 1750
  {
    DBUG_ASSERT(part_info->subpart_type == HASH_PARTITION);
    /*
1751 1752
      Subpartition is defined. We need to verify that subpartitioning
      function is correct.
1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764
    */
    if (part_info->linear_hash_ind)
      set_linear_hash_mask(part_info, part_info->no_subparts);
    if (part_info->list_of_subpart_fields)
    {
      List_iterator<char> it(part_info->subpart_field_list);
      if (unlikely(handle_list_of_fields(it, table, part_info, TRUE)))
        goto end;
    }
    else
    {
      if (unlikely(fix_fields_part_func(thd, &tables,
1765 1766
                                        part_info->subpart_expr, part_info,
                                        TRUE)))
1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
        goto end;
      if (unlikely(part_info->subpart_expr->result_type() != INT_RESULT))
      {
        my_error(ER_PARTITION_FUNC_NOT_ALLOWED_ERROR, MYF(0),
                 "SUBPARTITION");
        goto end;
      }
    }
  }
  DBUG_ASSERT(part_info->part_type != NOT_A_PARTITION);
  /*
1778 1779
    Partition is defined. We need to verify that partitioning
    function is correct.
1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805
  */
  if (part_info->part_type == HASH_PARTITION)
  {
    if (part_info->linear_hash_ind)
      set_linear_hash_mask(part_info, part_info->no_parts);
    if (part_info->list_of_part_fields)
    {
      List_iterator<char> it(part_info->part_field_list);
      if (unlikely(handle_list_of_fields(it, table, part_info, FALSE)))
        goto end;
    }
    else
    {
      if (unlikely(fix_fields_part_func(thd, &tables, part_info->part_expr,
                                        part_info, FALSE)))
        goto end;
      if (unlikely(part_info->part_expr->result_type() != INT_RESULT))
      {
        my_error(ER_PARTITION_FUNC_NOT_ALLOWED_ERROR, MYF(0), part_str);
        goto end;
      }
      part_info->part_result_type= INT_RESULT;
    }
  }
  else
  {
1806
    const char *error_str;
1807 1808
    if (part_info->part_type == RANGE_PARTITION)
    {
1809
      error_str= partition_keywords[PKW_RANGE].str; 
1810 1811 1812 1813 1814
      if (unlikely(check_range_constants(part_info)))
        goto end;
    }
    else if (part_info->part_type == LIST_PARTITION)
    {
1815
      error_str= partition_keywords[PKW_LIST].str; 
1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842
      if (unlikely(check_list_constants(part_info)))
        goto end;
    }
    else
    {
      DBUG_ASSERT(0);
      my_error(ER_INCONSISTENT_PARTITION_INFO_ERROR, MYF(0));
      goto end;
    }
    if (unlikely(part_info->no_parts < 1))
    {
      my_error(ER_PARTITIONS_MUST_BE_DEFINED_ERROR, MYF(0), error_str);
      goto end;
    }
    if (unlikely(fix_fields_part_func(thd, &tables, part_info->part_expr,
                                      part_info, FALSE)))
      goto end;
    if (unlikely(part_info->part_expr->result_type() != INT_RESULT))
    {
      my_error(ER_PARTITION_FUNC_NOT_ALLOWED_ERROR, MYF(0), part_str);
      goto end;
    }
  }
  if (unlikely(create_full_part_field_array(table, part_info)))
    goto end;
  if (unlikely(check_primary_key(table)))
    goto end;
1843 1844
  if (unlikely((!(table->s->db_type->partition_flags &&
      (table->s->db_type->partition_flags() & HA_CAN_PARTITION_UNIQUE))) &&
1845 1846
               check_unique_keys(table)))
    goto end;
1847 1848
  if (unlikely(set_up_partition_bitmap(thd, part_info)))
    goto end;
1849 1850 1851
  check_range_capable_PF(table);
  set_up_partition_key_maps(table, part_info);
  set_up_partition_func_pointers(part_info);
1852
  part_info->fixed= TRUE;
1853
  set_up_range_analysis_info(part_info);
1854 1855 1856
  result= FALSE;
end:
  thd->set_query_id= save_set_query_id;
1857
  DBUG_PRINT("info", ("thd->set_query_id: %d", thd->set_query_id));
1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871
  DBUG_RETURN(result);
}


/*
  The code below is support routines for the reverse parsing of the 
  partitioning syntax. This feature is very useful to generate syntax for
  all default values to avoid all default checking when opening the frm
  file. It is also used when altering the partitioning by use of various
  ALTER TABLE commands. Finally it is used for SHOW CREATE TABLES.
*/

static int add_write(File fptr, const char *buf, uint len)
{
1872
  uint len_written= my_write(fptr, (const byte*)buf, len, MYF(0));
1873

1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917
  if (likely(len == len_written))
    return 0;
  else
    return 1;
}

static int add_string(File fptr, const char *string)
{
  return add_write(fptr, string, strlen(string));
}

static int add_string_len(File fptr, const char *string, uint len)
{
  return add_write(fptr, string, len);
}

static int add_space(File fptr)
{
  return add_string(fptr, space_str);
}

static int add_comma(File fptr)
{
  return add_string(fptr, comma_str);
}

static int add_equal(File fptr)
{
  return add_string(fptr, equal_str);
}

static int add_end_parenthesis(File fptr)
{
  return add_string(fptr, end_paren_str);
}

static int add_begin_parenthesis(File fptr)
{
  return add_string(fptr, begin_paren_str);
}

static int add_part_key_word(File fptr, const char *key_string)
{
  int err= add_string(fptr, key_string);
1918

1919 1920 1921 1922 1923 1924
  err+= add_space(fptr);
  return err + add_begin_parenthesis(fptr);
}

static int add_hash(File fptr)
{
1925
  return add_part_key_word(fptr, partition_keywords[PKW_HASH].str);
1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
}

static int add_partition(File fptr)
{
  strxmov(buff, part_str, space_str, NullS);
  return add_string(fptr, buff);
}

static int add_subpartition(File fptr)
{
  int err= add_string(fptr, sub_str);
1937

1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949
  return err + add_partition(fptr);
}

static int add_partition_by(File fptr)
{
  strxmov(buff, part_str, space_str, by_str, space_str, NullS);
  return add_string(fptr, buff);
}

static int add_subpartition_by(File fptr)
{
  int err= add_string(fptr, sub_str);
1950

1951 1952 1953 1954 1955 1956 1957
  return err + add_partition_by(fptr);
}

static int add_key_partition(File fptr, List<char> field_list)
{
  uint i, no_fields;
  int err;
1958

1959
  List_iterator<char> part_it(field_list);
1960
  err= add_part_key_word(fptr, partition_keywords[PKW_KEY].str);
1961 1962
  no_fields= field_list.elements;
  i= 0;
1963
  while (i < no_fields)
1964 1965 1966 1967 1968
  {
    const char *field_str= part_it++;
    err+= add_string(fptr, field_str);
    if (i != (no_fields-1))
      err+= add_comma(fptr);
1969 1970
    i++;
  }
1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
  return err;
}

static int add_int(File fptr, longlong number)
{
  llstr(number, buff);
  return add_string(fptr, buff);
}

static int add_keyword_string(File fptr, const char *keyword,
1981
                              bool should_use_quotes, 
1982 1983 1984
                              const char *keystr)
{
  int err= add_string(fptr, keyword);
1985

1986 1987 1988
  err+= add_space(fptr);
  err+= add_equal(fptr);
  err+= add_space(fptr);
1989 1990
  if (should_use_quotes)
    err+= add_string(fptr, "'");
1991
  err+= add_string(fptr, keystr);
1992 1993
  if (should_use_quotes)
    err+= add_string(fptr, "'");
1994 1995 1996 1997 1998 1999
  return err + add_space(fptr);
}

static int add_keyword_int(File fptr, const char *keyword, longlong num)
{
  int err= add_string(fptr, keyword);
2000

2001 2002 2003 2004 2005 2006 2007
  err+= add_space(fptr);
  err+= add_equal(fptr);
  err+= add_space(fptr);
  err+= add_int(fptr, num);
  return err + add_space(fptr);
}

2008
static int add_engine(File fptr, handlerton *engine_type)
2009
{
2010
  const char *engine_str= engine_type->name;
2011
  DBUG_PRINT("info", ("ENGINE = %s", engine_str));
2012 2013 2014 2015 2016 2017 2018
  int err= add_string(fptr, "ENGINE = ");
  return err + add_string(fptr, engine_str);
}

static int add_partition_options(File fptr, partition_element *p_elem)
{
  int err= 0;
2019

2020
  if (p_elem->tablespace_name)
2021 2022
    err+= add_keyword_string(fptr,"TABLESPACE", FALSE, 
                             p_elem->tablespace_name);
2023 2024 2025 2026 2027 2028 2029
  if (p_elem->nodegroup_id != UNDEF_NODEGROUP)
    err+= add_keyword_int(fptr,"NODEGROUP",(longlong)p_elem->nodegroup_id);
  if (p_elem->part_max_rows)
    err+= add_keyword_int(fptr,"MAX_ROWS",(longlong)p_elem->part_max_rows);
  if (p_elem->part_min_rows)
    err+= add_keyword_int(fptr,"MIN_ROWS",(longlong)p_elem->part_min_rows);
  if (p_elem->data_file_name)
2030 2031
    err+= add_keyword_string(fptr, "DATA DIRECTORY", TRUE, 
                             p_elem->data_file_name);
2032
  if (p_elem->index_file_name)
2033 2034
    err+= add_keyword_string(fptr, "INDEX DIRECTORY", TRUE, 
                             p_elem->index_file_name);
2035
  if (p_elem->part_comment)
2036
    err+= add_keyword_string(fptr, "COMMENT", FALSE, p_elem->part_comment);
2037 2038 2039 2040 2041 2042 2043
  return err + add_engine(fptr,p_elem->engine_type);
}

static int add_partition_values(File fptr, partition_info *part_info,
                         partition_element *p_elem)
{
  int err= 0;
2044

2045 2046 2047
  if (part_info->part_type == RANGE_PARTITION)
  {
    err+= add_string(fptr, "VALUES LESS THAN ");
2048
    if (p_elem->range_value != LONGLONG_MAX)
2049 2050
    {
      err+= add_begin_parenthesis(fptr);
2051
      err+= add_int(fptr, p_elem->range_value);
2052 2053 2054
      err+= add_end_parenthesis(fptr);
    }
    else
2055
      err+= add_string(fptr, partition_keywords[PKW_MAXVALUE].str);
2056 2057 2058 2059
  }
  else if (part_info->part_type == LIST_PARTITION)
  {
    uint i;
2060
    List_iterator<longlong> list_val_it(p_elem->list_val_list);
2061
    err+= add_string(fptr, "VALUES IN ");
2062
    uint no_items= p_elem->list_val_list.elements;
2063
    err+= add_begin_parenthesis(fptr);
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073
    if (p_elem->has_null_value)
    {
      err+= add_string(fptr, "NULL");
      if (no_items == 0)
      {
        err+= add_end_parenthesis(fptr);
        goto end;
      }
      err+= add_comma(fptr);
    }
2074 2075 2076
    i= 0;
    do
    {
2077 2078
      longlong *list_value= list_val_it++;
      err+= add_int(fptr, *list_value);
2079 2080 2081 2082 2083
      if (i != (no_items-1))
        err+= add_comma(fptr);
    } while (++i < no_items);
    err+= add_end_parenthesis(fptr);
  }
2084
end:
2085 2086 2087 2088 2089 2090 2091
  return err + add_space(fptr);
}

/*
  Generate the partition syntax from the partition data structure.
  Useful for support of generating defaults, SHOW CREATE TABLES
  and easy partition management.
2092

2093 2094 2095 2096 2097 2098
  SYNOPSIS
    generate_partition_syntax()
    part_info                  The partitioning data structure
    buf_length                 A pointer to the returned buffer length
    use_sql_alloc              Allocate buffer from sql_alloc if true
                               otherwise use my_malloc
2099 2100
    write_all                  Write everything, also default values

2101 2102 2103
  RETURN VALUES
    NULL error
    buf, buf_length            Buffer and its length
2104

2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126
  DESCRIPTION
  Here we will generate the full syntax for the given command where all
  defaults have been expanded. By so doing the it is also possible to
  make lots of checks of correctness while at it.
  This could will also be reused for SHOW CREATE TABLES and also for all
  type ALTER TABLE commands focusing on changing the PARTITION structure
  in any fashion.

  The implementation writes the syntax to a temporary file (essentially
  an abstraction of a dynamic array) and if all writes goes well it
  allocates a buffer and writes the syntax into this one and returns it.

  As a security precaution the file is deleted before writing into it. This
  means that no other processes on the machine can open and read the file
  while this processing is ongoing.

  The code is optimised for minimal code size since it is not used in any
  common queries.
*/

char *generate_partition_syntax(partition_info *part_info,
                                uint *buf_length,
2127
                                bool use_sql_alloc,
2128
                                bool write_all)
2129
{
2130
  uint i,j, tot_no_parts, no_subparts, no_parts;
2131
  partition_element *part_elem;
2132
  partition_element *save_part_elem= NULL;
2133 2134 2135
  ulonglong buffer_length;
  char path[FN_REFLEN];
  int err= 0;
2136 2137
  List_iterator<partition_element> part_it(part_info->partitions);
  List_iterator<partition_element> temp_it(part_info->temp_partitions);
2138 2139
  File fptr;
  char *buf= NULL; //Return buffer
2140 2141 2142 2143 2144 2145
  uint use_temp= 0;
  uint no_temp_parts= part_info->temp_partitions.elements;
  bool write_part_state;
  DBUG_ENTER("generate_partition_syntax");

  write_part_state= (part_info->part_state && !part_info->part_state_len);
2146 2147 2148
  if (unlikely(((fptr= create_temp_file(path,mysql_tmpdir,"psy", 
                                        O_RDWR | O_BINARY | O_TRUNC |  
                                        O_TEMPORARY, MYF(MY_WME)))) < 0))
2149
    DBUG_RETURN(NULL);
2150 2151
#ifndef __WIN__
  unlink(path);
2152 2153 2154 2155 2156 2157
#endif
  err+= add_space(fptr);
  err+= add_partition_by(fptr);
  switch (part_info->part_type)
  {
    case RANGE_PARTITION:
2158
      err+= add_part_key_word(fptr, partition_keywords[PKW_RANGE].str);
2159 2160
      break;
    case LIST_PARTITION:
2161
      err+= add_part_key_word(fptr, partition_keywords[PKW_LIST].str);
2162 2163 2164
      break;
    case HASH_PARTITION:
      if (part_info->linear_hash_ind)
2165
        err+= add_string(fptr, partition_keywords[PKW_LINEAR].str);
2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181
      if (part_info->list_of_part_fields)
        err+= add_key_partition(fptr, part_info->part_field_list);
      else
        err+= add_hash(fptr);
      break;
    default:
      DBUG_ASSERT(0);
      /* We really shouldn't get here, no use in continuing from here */
      current_thd->fatal_error();
      DBUG_RETURN(NULL);
  }
  if (part_info->part_expr)
    err+= add_string_len(fptr, part_info->part_func_string,
                         part_info->part_func_len);
  err+= add_end_parenthesis(fptr);
  err+= add_space(fptr);
2182 2183 2184 2185 2186 2187 2188
  if ((!part_info->use_default_no_partitions) &&
       part_info->use_default_partitions)
  {
    err+= add_string(fptr, "PARTITIONS ");
    err+= add_int(fptr, part_info->no_parts);
    err+= add_space(fptr);
  }
2189
  if (part_info->is_sub_partitioned())
2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201
  {
    err+= add_subpartition_by(fptr);
    /* Must be hash partitioning for subpartitioning */
    if (part_info->list_of_subpart_fields)
      err+= add_key_partition(fptr, part_info->subpart_field_list);
    else
      err+= add_hash(fptr);
    if (part_info->subpart_expr)
      err+= add_string_len(fptr, part_info->subpart_func_string,
                           part_info->subpart_func_len);
    err+= add_end_parenthesis(fptr);
    err+= add_space(fptr);
2202 2203 2204 2205 2206 2207 2208 2209
    if ((!part_info->use_default_no_subpartitions) && 
          part_info->use_default_subpartitions)
    {
      err+= add_string(fptr, "SUBPARTITIONS ");
      err+= add_int(fptr, part_info->no_subparts);
      err+= add_space(fptr);
    }
  }
2210
  no_parts= part_info->no_parts;
2211
  tot_no_parts= no_parts + no_temp_parts;
2212
  no_subparts= part_info->no_subparts;
2213 2214

  if (write_all || (!part_info->use_default_partitions))
2215
  {
2216 2217 2218
    err+= add_begin_parenthesis(fptr);
    i= 0;
    do
2219
    {
2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263
      /*
        We need to do some clever list manipulation here since we have two
        different needs for our list processing and here we take some of the
        cost of using a simpler list processing for the other parts of the
        code.

        ALTER TABLE REORGANIZE PARTITIONS has the list of partitions to be
        the final list as the main list and the reorganised partitions is in
        the temporary partition list. Thus when finding the first part added
        we insert the temporary list if there is such a list. If there is no
        temporary list we are performing an ADD PARTITION.
      */
      if (use_temp && use_temp <= no_temp_parts)
      {
        part_elem= temp_it++;
        DBUG_ASSERT(no_temp_parts);
        no_temp_parts--;
      }
      else if (use_temp)
      {
        DBUG_ASSERT(no_parts);
        part_elem= save_part_elem;
        use_temp= 0;
        no_parts--;
      }
      else
      {
        part_elem= part_it++;
        if ((part_elem->part_state == PART_TO_BE_ADDED ||
             part_elem->part_state == PART_IS_ADDED) && no_temp_parts)
        {
          save_part_elem= part_elem;
          part_elem= temp_it++;
          no_temp_parts--;
          use_temp= 1;
        }
        else
        {
          DBUG_ASSERT(no_parts);
          no_parts--;
        }
      }

      if (part_elem->part_state != PART_IS_DROPPED)
2264
      {
2265 2266 2267 2268 2269 2270 2271
        if (write_part_state)
        {
          uint32 part_state_id= part_info->part_state_len;
          part_info->part_state[part_state_id]= (uchar)part_elem->part_state;
          part_info->part_state_len= part_state_id+1;
        }
        err+= add_partition(fptr);
2272 2273
        err+= add_string(fptr, part_elem->partition_name);
        err+= add_space(fptr);
2274
        err+= add_partition_values(fptr, part_info, part_elem);
2275
        if (!part_info->is_sub_partitioned())
2276
          err+= add_partition_options(fptr, part_elem);
2277
        if (part_info->is_sub_partitioned() &&
2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300
            (write_all || (!part_info->use_default_subpartitions)))
        {
          err+= add_space(fptr);
          err+= add_begin_parenthesis(fptr);
          List_iterator<partition_element> sub_it(part_elem->subpartitions);
          j= 0;
          do
          {
            part_elem= sub_it++;
            err+= add_subpartition(fptr);
            err+= add_string(fptr, part_elem->partition_name);
            err+= add_space(fptr);
            err+= add_partition_options(fptr, part_elem);
            if (j != (no_subparts-1))
            {
              err+= add_comma(fptr);
              err+= add_space(fptr);
            }
            else
              err+= add_end_parenthesis(fptr);
          } while (++j < no_subparts);
        }
        if (i != (tot_no_parts-1))
2301 2302 2303 2304
        {
          err+= add_comma(fptr);
          err+= add_space(fptr);
        }
2305 2306 2307 2308 2309
      }
      if (i == (tot_no_parts-1))
        err+= add_end_parenthesis(fptr);
    } while (++i < tot_no_parts);
    DBUG_ASSERT(!no_parts && !no_temp_parts);
2310
  }
2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325
  if (err)
    goto close_file;
  buffer_length= my_seek(fptr, 0L,MY_SEEK_END,MYF(0));
  if (unlikely(buffer_length == MY_FILEPOS_ERROR))
    goto close_file;
  if (unlikely(my_seek(fptr, 0L, MY_SEEK_SET, MYF(0)) == MY_FILEPOS_ERROR))
    goto close_file;
  *buf_length= (uint)buffer_length;
  if (use_sql_alloc)
    buf= sql_alloc(*buf_length+1);
  else
    buf= my_malloc(*buf_length+1, MYF(MY_WME));
  if (!buf)
    goto close_file;

2326
  if (unlikely(my_read(fptr, (byte*)buf, *buf_length, MYF(MY_FNABP))))
2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344
  {
    if (!use_sql_alloc)
      my_free(buf, MYF(0));
    else
      buf= NULL;
  }
  else
    buf[*buf_length]= 0;

close_file:
  my_close(fptr, MYF(0));
  DBUG_RETURN(buf);
}


/*
  Check if partition key fields are modified and if it can be handled by the
  underlying storage engine.
2345

2346 2347 2348 2349
  SYNOPSIS
    partition_key_modified
    table                TABLE object for which partition fields are set-up
    fields               A list of the to be modifed
2350

2351 2352 2353 2354 2355 2356 2357 2358
  RETURN VALUES
    TRUE                 Need special handling of UPDATE
    FALSE                Normal UPDATE handling is ok
*/

bool partition_key_modified(TABLE *table, List<Item> &fields)
{
  List_iterator_fast<Item> f(fields);
2359
  partition_info *part_info= table->part_info;
2360 2361
  Item_field *item_field;
  DBUG_ENTER("partition_key_modified");
2362

2363 2364
  if (!part_info)
    DBUG_RETURN(FALSE);
2365 2366
  if (table->s->db_type->partition_flags &&
      (table->s->db_type->partition_flags() & HA_CAN_UPDATE_PARTITION_KEY))
2367 2368 2369 2370 2371 2372 2373 2374 2375
    DBUG_RETURN(FALSE);
  f.rewind();
  while ((item_field=(Item_field*) f++))
    if (item_field->field->flags & FIELD_IN_PART_FUNC_FLAG)
      DBUG_RETURN(TRUE);
  DBUG_RETURN(FALSE);
}


2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398
/*
  A function to handle correct handling of NULL values in partition
  functions.
  SYNOPSIS
    part_val_int()
    item_expr                 The item expression to evaluate
  RETURN VALUES
    The value of the partition function, LONGLONG_MIN if any null value
    in function
*/

static
inline
longlong
part_val_int(Item *item_expr)
{
  longlong value= item_expr->val_int();
  if (item_expr->null_value)
    value= LONGLONG_MIN;
  return value;
}


2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418
/*
  The next set of functions are used to calculate the partition identity.
  A handler sets up a variable that corresponds to one of these functions
  to be able to quickly call it whenever the partition id needs to calculated
  based on the record in table->record[0] (or set up to fake that).
  There are 4 functions for hash partitioning and 2 for RANGE/LIST partitions.
  In addition there are 4 variants for RANGE subpartitioning and 4 variants
  for LIST subpartitioning thus in total there are 14 variants of this
  function.

  We have a set of support functions for these 14 variants. There are 4
  variants of hash functions and there is a function for each. The KEY
  partitioning uses the function calculate_key_value to calculate the hash
  value based on an array of fields. The linear hash variants uses the
  method get_part_id_from_linear_hash to get the partition id using the
  hash value and some parameters calculated from the number of partitions.
*/

/*
  Calculate hash value for KEY partitioning using an array of fields.
2419

2420 2421 2422
  SYNOPSIS
    calculate_key_value()
    field_array             An array of the fields in KEY partitioning
2423

2424 2425
  RETURN VALUE
    hash_value calculated
2426

2427 2428 2429 2430 2431 2432 2433 2434 2435
  DESCRIPTION
    Uses the hash function on the character set of the field. Integer and
    floating point fields use the binary character set by default.
*/

static uint32 calculate_key_value(Field **field_array)
{
  uint32 hashnr= 0;
  ulong nr2= 4;
2436

2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459
  do
  {
    Field *field= *field_array;
    if (field->is_null())
    {
      hashnr^= (hashnr << 1) | 1;
    }
    else
    {
      uint len= field->pack_length();
      ulong nr1= 1;
      CHARSET_INFO *cs= field->charset();
      cs->coll->hash_sort(cs, (uchar*)field->ptr, len, &nr1, &nr2);
      hashnr^= (uint32)nr1;
    }
  } while (*(++field_array));
  return hashnr;
}


/*
  A simple support function to calculate part_id given local part and
  sub part.
2460

2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477
  SYNOPSIS
    get_part_id_for_sub()
    loc_part_id             Local partition id
    sub_part_id             Subpartition id
    no_subparts             Number of subparts
*/

inline
static uint32 get_part_id_for_sub(uint32 loc_part_id, uint32 sub_part_id,
                                  uint no_subparts)
{
  return (uint32)((loc_part_id * no_subparts) + sub_part_id);
}


/*
  Calculate part_id for (SUB)PARTITION BY HASH
2478

2479 2480 2481 2482
  SYNOPSIS
    get_part_id_hash()
    no_parts                 Number of hash partitions
    part_expr                Item tree of hash function
2483
    out:func_value      Value of hash function
2484

2485 2486 2487 2488 2489 2490
  RETURN VALUE
    Calculated partition id
*/

inline
static uint32 get_part_id_hash(uint no_parts,
2491 2492
                               Item *part_expr,
                               longlong *func_value)
2493 2494
{
  DBUG_ENTER("get_part_id_hash");
2495
  *func_value= part_val_int(part_expr);
2496
  longlong int_hash_id= *func_value % no_parts;
2497
  DBUG_RETURN(int_hash_id < 0 ? -int_hash_id : int_hash_id);
2498 2499 2500 2501 2502
}


/*
  Calculate part_id for (SUB)PARTITION BY LINEAR HASH
2503

2504 2505 2506 2507 2508 2509
  SYNOPSIS
    get_part_id_linear_hash()
    part_info           A reference to the partition_info struct where all the
                        desired information is given
    no_parts            Number of hash partitions
    part_expr           Item tree of hash function
2510
    out:func_value      Value of hash function
2511

2512 2513 2514 2515 2516 2517 2518
  RETURN VALUE
    Calculated partition id
*/

inline
static uint32 get_part_id_linear_hash(partition_info *part_info,
                                      uint no_parts,
2519 2520
                                      Item *part_expr,
                                      longlong *func_value)
2521 2522
{
  DBUG_ENTER("get_part_id_linear_hash");
2523

2524
  *func_value= part_val_int(part_expr);
2525
  DBUG_RETURN(get_part_id_from_linear_hash(*func_value,
2526 2527 2528 2529 2530 2531 2532
                                           part_info->linear_hash_mask,
                                           no_parts));
}


/*
  Calculate part_id for (SUB)PARTITION BY KEY
2533

2534 2535 2536 2537
  SYNOPSIS
    get_part_id_key()
    field_array         Array of fields for PARTTION KEY
    no_parts            Number of KEY partitions
2538

2539 2540 2541 2542 2543 2544
  RETURN VALUE
    Calculated partition id
*/

inline
static uint32 get_part_id_key(Field **field_array,
2545 2546
                              uint no_parts,
                              longlong *func_value)
2547 2548
{
  DBUG_ENTER("get_part_id_key");
2549 2550
  *func_value= calculate_key_value(field_array);
  DBUG_RETURN(*func_value % no_parts);
2551 2552 2553 2554 2555
}


/*
  Calculate part_id for (SUB)PARTITION BY LINEAR KEY
2556

2557 2558 2559 2560 2561 2562
  SYNOPSIS
    get_part_id_linear_key()
    part_info           A reference to the partition_info struct where all the
                        desired information is given
    field_array         Array of fields for PARTTION KEY
    no_parts            Number of KEY partitions
2563

2564 2565 2566 2567 2568 2569 2570
  RETURN VALUE
    Calculated partition id
*/

inline
static uint32 get_part_id_linear_key(partition_info *part_info,
                                     Field **field_array,
2571 2572
                                     uint no_parts,
                                     longlong *func_value)
2573 2574
{
  DBUG_ENTER("get_partition_id_linear_key");
2575

2576 2577
  *func_value= calculate_key_value(field_array);
  DBUG_RETURN(get_part_id_from_linear_hash(*func_value,
2578 2579 2580 2581 2582 2583 2584 2585
                                           part_info->linear_hash_mask,
                                           no_parts));
}

/*
  This function is used to calculate the partition id where all partition
  fields have been prepared to point to a record where the partition field
  values are bound.
2586

2587 2588 2589 2590
  SYNOPSIS
    get_partition_id()
    part_info           A reference to the partition_info struct where all the
                        desired information is given
2591 2592
    out:part_id         The partition id is returned through this pointer

2593 2594 2595 2596
  RETURN VALUE
    part_id
    return TRUE means that the fields of the partition function didn't fit
    into any partition and thus the values of the PF-fields are not allowed.
2597

2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626
  DESCRIPTION
    A routine used from write_row, update_row and delete_row from any
    handler supporting partitioning. It is also a support routine for
    get_partition_set used to find the set of partitions needed to scan
    for a certain index scan or full table scan.
    
    It is actually 14 different variants of this function which are called
    through a function pointer.

    get_partition_id_list
    get_partition_id_range
    get_partition_id_hash_nosub
    get_partition_id_key_nosub
    get_partition_id_linear_hash_nosub
    get_partition_id_linear_key_nosub
    get_partition_id_range_sub_hash
    get_partition_id_range_sub_key
    get_partition_id_range_sub_linear_hash
    get_partition_id_range_sub_linear_key
    get_partition_id_list_sub_hash
    get_partition_id_list_sub_key
    get_partition_id_list_sub_linear_hash
    get_partition_id_list_sub_linear_key
*/

/*
  This function is used to calculate the main partition to use in the case of
  subpartitioning and we don't know enough to get the partition identity in
  total.
2627

2628 2629 2630 2631
  SYNOPSIS
    get_part_partition_id()
    part_info           A reference to the partition_info struct where all the
                        desired information is given
2632 2633
    out:part_id         The partition id is returned through this pointer

2634 2635 2636 2637
  RETURN VALUE
    part_id
    return TRUE means that the fields of the partition function didn't fit
    into any partition and thus the values of the PF-fields are not allowed.
2638

2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652
  DESCRIPTION
    
    It is actually 6 different variants of this function which are called
    through a function pointer.

    get_partition_id_list
    get_partition_id_range
    get_partition_id_hash_nosub
    get_partition_id_key_nosub
    get_partition_id_linear_hash_nosub
    get_partition_id_linear_key_nosub
*/


2653
int get_partition_id_list(partition_info *part_info,
2654 2655
                           uint32 *part_id,
                           longlong *func_value)
2656 2657
{
  LIST_PART_ENTRY *list_array= part_info->list_array;
2658
  int list_index;
2659
  longlong list_value;
2660 2661
  int min_list_index= 0;
  int max_list_index= part_info->no_list_values - 1;
2662
  longlong part_func_value= part_val_int(part_info->part_expr);
2663 2664
  DBUG_ENTER("get_partition_id_list");

2665 2666 2667 2668 2669 2670 2671 2672 2673
  if (part_info->part_expr->null_value)
  {
    if (part_info->has_null_value)
    {
      *part_id= part_info->has_null_part_id;
      DBUG_RETURN(0);
    }
    goto notfound;
  }
2674
  *func_value= part_func_value;
2675 2676 2677 2678 2679 2680 2681
  while (max_list_index >= min_list_index)
  {
    list_index= (max_list_index + min_list_index) >> 1;
    list_value= list_array[list_index].list_value;
    if (list_value < part_func_value)
      min_list_index= list_index + 1;
    else if (list_value > part_func_value)
2682 2683 2684
    {
      if (!list_index)
        goto notfound;
2685
      max_list_index= list_index - 1;
2686 2687 2688
    }
    else
    {
2689
      *part_id= (uint32)list_array[list_index].partition_id;
2690
      DBUG_RETURN(0);
2691 2692
    }
  }
2693
notfound:
2694
  *part_id= 0;
2695
  DBUG_RETURN(HA_ERR_NO_PARTITION_FOUND);
2696 2697 2698
}


2699
/*
2700 2701
  Find the sub-array part_info->list_array that corresponds to given interval

2702 2703 2704 2705 2706 2707 2708 2709
  SYNOPSIS 
    get_list_array_idx_for_endpoint()
      part_info         Partitioning info (partitioning type must be LIST)
      left_endpoint     TRUE  - the interval is [a; +inf) or (a; +inf)
                        FALSE - the interval is (-inf; a] or (-inf; a)
      include_endpoint  TRUE iff the interval includes the endpoint

  DESCRIPTION
2710
    This function finds the sub-array of part_info->list_array where values of
2711 2712 2713
    list_array[idx].list_value are contained within the specifed interval.
    list_array is ordered by list_value, so
    1. For [a; +inf) or (a; +inf)-type intervals (left_endpoint==TRUE), the 
2714
       sought sub-array starts at some index idx and continues till array end.
2715 2716 2717 2718
       The function returns first number idx, such that 
       list_array[idx].list_value is contained within the passed interval.
       
    2. For (-inf; a] or (-inf; a)-type intervals (left_endpoint==FALSE), the
2719
       sought sub-array starts at array start and continues till some last 
2720 2721 2722 2723 2724 2725 2726
       index idx.
       The function returns first number idx, such that 
       list_array[idx].list_value is NOT contained within the passed interval.
       If all array elements are contained, part_info->no_list_values is
       returned.

  NOTE
2727
    The caller will call this function and then will run along the sub-array of
2728 2729 2730 2731 2732 2733
    list_array to collect partition ids. If the number of list values is 
    significantly higher then number of partitions, this could be slow and
    we could invent some other approach. The "run over list array" part is
    already wrapped in a get_next()-like function.

  RETURN
2734
    The edge of corresponding sub-array of part_info->list_array
2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745
*/

uint32 get_list_array_idx_for_endpoint(partition_info *part_info,
                                       bool left_endpoint,
                                       bool include_endpoint)
{
  DBUG_ENTER("get_list_array_idx_for_endpoint");
  LIST_PART_ENTRY *list_array= part_info->list_array;
  uint list_index;
  longlong list_value;
  uint min_list_index= 0, max_list_index= part_info->no_list_values - 1;
2746
  /* Get the partitioning function value for the endpoint */
2747
  longlong part_func_value= part_val_int(part_info->part_expr);
2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770
  while (max_list_index >= min_list_index)
  {
    list_index= (max_list_index + min_list_index) >> 1;
    list_value= list_array[list_index].list_value;
    if (list_value < part_func_value)
      min_list_index= list_index + 1;
    else if (list_value > part_func_value)
    {
      if (!list_index)
        goto notfound;
      max_list_index= list_index - 1;
    }
    else 
    {
      DBUG_RETURN(list_index + test(left_endpoint ^ include_endpoint));
    }
  }
notfound:
  if (list_value < part_func_value)
    list_index++;
  DBUG_RETURN(list_index);
}

2771

2772
int get_partition_id_range(partition_info *part_info,
2773 2774
                            uint32 *part_id,
                            longlong *func_value)
2775 2776 2777
{
  longlong *range_array= part_info->range_int_array;
  uint max_partition= part_info->no_parts - 1;
2778 2779 2780
  uint min_part_id= 0;
  uint max_part_id= max_partition;
  uint loc_part_id;
2781
  longlong part_func_value= part_val_int(part_info->part_expr);
2782 2783
  DBUG_ENTER("get_partition_id_int_range");

2784 2785 2786 2787 2788
  if (part_info->part_expr->null_value)
  {
    *part_id= 0;
    DBUG_RETURN(0);
  }
2789 2790 2791
  while (max_part_id > min_part_id)
  {
    loc_part_id= (max_part_id + min_part_id + 1) >> 1;
patg@govinda.patg.net's avatar
patg@govinda.patg.net committed
2792
    if (range_array[loc_part_id] <= part_func_value)
2793 2794 2795 2796 2797 2798 2799 2800 2801
      min_part_id= loc_part_id + 1;
    else
      max_part_id= loc_part_id - 1;
  }
  loc_part_id= max_part_id;
  if (part_func_value >= range_array[loc_part_id])
    if (loc_part_id != max_partition)
      loc_part_id++;
  *part_id= (uint32)loc_part_id;
2802
  *func_value= part_func_value;
2803 2804 2805
  if (loc_part_id == max_partition)
    if (range_array[loc_part_id] != LONGLONG_MAX)
      if (part_func_value >= range_array[loc_part_id])
2806 2807
        DBUG_RETURN(HA_ERR_NO_PARTITION_FOUND);
  DBUG_RETURN(0);
2808 2809
}

2810 2811

/*
2812 2813
  Find the sub-array of part_info->range_int_array that covers given interval
 
2814 2815 2816 2817 2818 2819 2820 2821 2822
  SYNOPSIS 
    get_partition_id_range_for_endpoint()
      part_info         Partitioning info (partitioning type must be RANGE)
      left_endpoint     TRUE  - the interval is [a; +inf) or (a; +inf)
                        FALSE - the interval is (-inf; a] or (-inf; a).
      include_endpoint  TRUE <=> the endpoint itself is included in the
                        interval

  DESCRIPTION
2823
    This function finds the sub-array of part_info->range_int_array where the
2824
    elements have non-empty intersections with the given interval.
2825
 
2826 2827 2828 2829 2830 2831 2832
    A range_int_array element at index idx represents the interval
      
      [range_int_array[idx-1], range_int_array[idx]),

    intervals are disjoint and ordered by their right bound, so
    
    1. For [a; +inf) or (a; +inf)-type intervals (left_endpoint==TRUE), the
2833
       sought sub-array starts at some index idx and continues till array end.
2834 2835 2836 2837 2838
       The function returns first number idx, such that the interval
       represented by range_int_array[idx] has non empty intersection with 
       the passed interval.
       
    2. For (-inf; a] or (-inf; a)-type intervals (left_endpoint==FALSE), the
2839
       sought sub-array starts at array start and continues till some last
2840 2841 2842 2843 2844 2845 2846 2847 2848
       index idx.
       The function returns first number idx, such that the interval
       represented by range_int_array[idx] has EMPTY intersection with the
       passed interval.
       If the interval represented by the last array element has non-empty 
       intersection with the passed interval, part_info->no_parts is
       returned.
       
  RETURN
2849
    The edge of corresponding part_info->range_int_array sub-array.
2850 2851 2852 2853 2854 2855 2856 2857 2858 2859
*/

uint32 get_partition_id_range_for_endpoint(partition_info *part_info,
                                           bool left_endpoint,
                                           bool include_endpoint)
{
  DBUG_ENTER("get_partition_id_range_for_endpoint");
  longlong *range_array= part_info->range_int_array;
  uint max_partition= part_info->no_parts - 1;
  uint min_part_id= 0, max_part_id= max_partition, loc_part_id;
2860
  /* Get the partitioning function value for the endpoint */
2861
  longlong part_func_value= part_val_int(part_info->part_expr);
2862 2863 2864 2865

  if (part_info->part_expr->null_value)
    DBUG_RETURN(0);

2866 2867 2868
  while (max_part_id > min_part_id)
  {
    loc_part_id= (max_part_id + min_part_id + 1) >> 1;
patg@govinda.patg.net's avatar
patg@govinda.patg.net committed
2869
    if (range_array[loc_part_id] <= part_func_value)
2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896
      min_part_id= loc_part_id + 1;
    else
      max_part_id= loc_part_id - 1;
  }
  loc_part_id= max_part_id;
  if (loc_part_id < max_partition && 
      part_func_value >= range_array[loc_part_id+1])
  {
     loc_part_id++;
  }
  if (left_endpoint)
  {
    if (part_func_value >= range_array[loc_part_id])
      loc_part_id++;
  }
  else 
  {
    if (part_func_value == range_array[loc_part_id])
      loc_part_id += test(include_endpoint);
    else if (part_func_value > range_array[loc_part_id])
      loc_part_id++;
    loc_part_id++;
  }
  DBUG_RETURN(loc_part_id);
}


2897
int get_partition_id_hash_nosub(partition_info *part_info,
2898 2899
                                 uint32 *part_id,
                                 longlong *func_value)
2900
{
2901 2902
  *part_id= get_part_id_hash(part_info->no_parts, part_info->part_expr,
                             func_value);
2903
  return 0;
2904 2905 2906
}


2907
int get_partition_id_linear_hash_nosub(partition_info *part_info,
2908 2909
                                        uint32 *part_id,
                                        longlong *func_value)
2910 2911
{
  *part_id= get_part_id_linear_hash(part_info, part_info->no_parts,
2912
                                    part_info->part_expr, func_value);
2913
  return 0;
2914 2915 2916
}


2917
int get_partition_id_key_nosub(partition_info *part_info,
2918 2919
                                uint32 *part_id,
                                longlong *func_value)
2920
{
2921 2922
  *part_id= get_part_id_key(part_info->part_field_array,
                            part_info->no_parts, func_value);
2923
  return 0;
2924 2925 2926
}


2927
int get_partition_id_linear_key_nosub(partition_info *part_info,
2928 2929
                                       uint32 *part_id,
                                       longlong *func_value)
2930 2931 2932
{
  *part_id= get_part_id_linear_key(part_info,
                                   part_info->part_field_array,
2933
                                   part_info->no_parts, func_value);
2934
  return 0;
2935 2936 2937
}


2938
int get_partition_id_range_sub_hash(partition_info *part_info,
2939 2940
                                     uint32 *part_id,
                                     longlong *func_value)
2941 2942 2943
{
  uint32 loc_part_id, sub_part_id;
  uint no_subparts;
2944
  longlong local_func_value;
2945
  int error;
2946
  DBUG_ENTER("get_partition_id_range_sub_hash");
2947

2948 2949
  if (unlikely((error= get_partition_id_range(part_info, &loc_part_id,
                                              func_value))))
2950
  {
2951
    DBUG_RETURN(error);
2952 2953
  }
  no_subparts= part_info->no_subparts;
2954 2955
  sub_part_id= get_part_id_hash(no_subparts, part_info->subpart_expr,
                                &local_func_value);
2956
  *part_id= get_part_id_for_sub(loc_part_id, sub_part_id, no_subparts);
2957
  DBUG_RETURN(0);
2958 2959 2960
}


2961
int get_partition_id_range_sub_linear_hash(partition_info *part_info,
2962 2963
                                            uint32 *part_id,
                                            longlong *func_value)
2964 2965 2966
{
  uint32 loc_part_id, sub_part_id;
  uint no_subparts;
2967
  longlong local_func_value;
2968
  int error;
2969
  DBUG_ENTER("get_partition_id_range_sub_linear_hash");
2970

2971 2972
  if (unlikely((error= get_partition_id_range(part_info, &loc_part_id,
                                              func_value))))
2973
  {
2974
    DBUG_RETURN(error);
2975 2976 2977
  }
  no_subparts= part_info->no_subparts;
  sub_part_id= get_part_id_linear_hash(part_info, no_subparts,
2978 2979
                                       part_info->subpart_expr,
                                       &local_func_value);
2980
  *part_id= get_part_id_for_sub(loc_part_id, sub_part_id, no_subparts);
2981
  DBUG_RETURN(0);
2982 2983 2984
}


2985
int get_partition_id_range_sub_key(partition_info *part_info,
2986 2987
                                    uint32 *part_id,
                                    longlong *func_value)
2988 2989 2990
{
  uint32 loc_part_id, sub_part_id;
  uint no_subparts;
2991
  longlong local_func_value;
2992
  int error;
2993
  DBUG_ENTER("get_partition_id_range_sub_key");
2994

2995 2996
  if (unlikely((error= get_partition_id_range(part_info, &loc_part_id,
                                              func_value))))
2997
  {
2998
    DBUG_RETURN(error);
2999 3000
  }
  no_subparts= part_info->no_subparts;
3001 3002
  sub_part_id= get_part_id_key(part_info->subpart_field_array,
                               no_subparts, &local_func_value);
3003
  *part_id= get_part_id_for_sub(loc_part_id, sub_part_id, no_subparts);
3004
  DBUG_RETURN(0);
3005 3006 3007
}


3008
int get_partition_id_range_sub_linear_key(partition_info *part_info,
3009 3010
                                           uint32 *part_id,
                                           longlong *func_value)
3011 3012 3013
{
  uint32 loc_part_id, sub_part_id;
  uint no_subparts;
3014
  longlong local_func_value;
3015
  int error;
3016
  DBUG_ENTER("get_partition_id_range_sub_linear_key");
3017

3018 3019
  if (unlikely((error= get_partition_id_range(part_info, &loc_part_id,
                                              func_value))))
3020
  {
3021
    DBUG_RETURN(error);
3022 3023 3024 3025
  }
  no_subparts= part_info->no_subparts;
  sub_part_id= get_part_id_linear_key(part_info,
                                      part_info->subpart_field_array,
3026
                                      no_subparts, &local_func_value);
3027
  *part_id= get_part_id_for_sub(loc_part_id, sub_part_id, no_subparts);
3028
  DBUG_RETURN(0);
3029 3030 3031
}


3032
int get_partition_id_list_sub_hash(partition_info *part_info,
3033 3034
                                    uint32 *part_id,
                                    longlong *func_value)
3035 3036 3037
{
  uint32 loc_part_id, sub_part_id;
  uint no_subparts;
3038
  longlong local_func_value;
3039
  int error;
3040
  DBUG_ENTER("get_partition_id_list_sub_hash");
3041

3042 3043
  if (unlikely((error= get_partition_id_list(part_info, &loc_part_id,
                                             func_value))))
3044
  {
3045
    DBUG_RETURN(error);
3046 3047
  }
  no_subparts= part_info->no_subparts;
3048 3049
  sub_part_id= get_part_id_hash(no_subparts, part_info->subpart_expr,
                                &local_func_value);
3050
  *part_id= get_part_id_for_sub(loc_part_id, sub_part_id, no_subparts);
3051
  DBUG_RETURN(0);
3052 3053 3054
}


3055
int get_partition_id_list_sub_linear_hash(partition_info *part_info,
3056 3057
                                           uint32 *part_id,
                                           longlong *func_value)
3058 3059 3060
{
  uint32 loc_part_id, sub_part_id;
  uint no_subparts;
3061
  longlong local_func_value;
3062
  int error;
3063
  DBUG_ENTER("get_partition_id_list_sub_linear_hash");
3064

3065 3066
  if (unlikely((error= get_partition_id_list(part_info, &loc_part_id,
                                             func_value))))
3067
  {
3068
    DBUG_RETURN(error);
3069 3070
  }
  no_subparts= part_info->no_subparts;
3071 3072 3073
  sub_part_id= get_part_id_linear_hash(part_info, no_subparts,
                                       part_info->subpart_expr,
                                       &local_func_value);
3074
  *part_id= get_part_id_for_sub(loc_part_id, sub_part_id, no_subparts);
3075
  DBUG_RETURN(0);
3076 3077 3078
}


3079
int get_partition_id_list_sub_key(partition_info *part_info,
3080 3081
                                   uint32 *part_id,
                                   longlong *func_value)
3082 3083 3084
{
  uint32 loc_part_id, sub_part_id;
  uint no_subparts;
3085
  longlong local_func_value;
3086
  int error;
3087
  DBUG_ENTER("get_partition_id_range_sub_key");
3088

3089 3090
  if (unlikely((error= get_partition_id_list(part_info, &loc_part_id,
                                             func_value))))
3091
  {
3092
    DBUG_RETURN(error);
3093 3094
  }
  no_subparts= part_info->no_subparts;
3095 3096
  sub_part_id= get_part_id_key(part_info->subpart_field_array,
                               no_subparts, &local_func_value);
3097
  *part_id= get_part_id_for_sub(loc_part_id, sub_part_id, no_subparts);
3098
  DBUG_RETURN(0);
3099 3100 3101
}


3102
int get_partition_id_list_sub_linear_key(partition_info *part_info,
3103 3104
                                          uint32 *part_id,
                                          longlong *func_value)
3105 3106 3107
{
  uint32 loc_part_id, sub_part_id;
  uint no_subparts;
3108
  longlong local_func_value;
3109
  int error;
3110
  DBUG_ENTER("get_partition_id_list_sub_linear_key");
3111

3112 3113
  if (unlikely((error= get_partition_id_list(part_info, &loc_part_id,
                                             func_value))))
3114
  {
3115
    DBUG_RETURN(error);
3116 3117 3118 3119
  }
  no_subparts= part_info->no_subparts;
  sub_part_id= get_part_id_linear_key(part_info,
                                      part_info->subpart_field_array,
3120
                                      no_subparts, &local_func_value);
3121
  *part_id= get_part_id_for_sub(loc_part_id, sub_part_id, no_subparts);
3122
  DBUG_RETURN(0);
3123 3124 3125 3126 3127
}


/*
  This function is used to calculate the subpartition id
3128

3129 3130 3131 3132
  SYNOPSIS
    get_subpartition_id()
    part_info           A reference to the partition_info struct where all the
                        desired information is given
3133

3134
  RETURN VALUE
3135 3136
    part_id             The subpartition identity

3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151
  DESCRIPTION
    A routine used in some SELECT's when only partial knowledge of the
    partitions is known.
    
    It is actually 4 different variants of this function which are called
    through a function pointer.

    get_partition_id_hash_sub
    get_partition_id_key_sub
    get_partition_id_linear_hash_sub
    get_partition_id_linear_key_sub
*/

uint32 get_partition_id_hash_sub(partition_info *part_info)
{
3152 3153 3154
  longlong func_value;
  return get_part_id_hash(part_info->no_subparts, part_info->subpart_expr,
                          &func_value);
3155 3156 3157 3158 3159
}


uint32 get_partition_id_linear_hash_sub(partition_info *part_info)
{
3160
  longlong func_value;
3161
  return get_part_id_linear_hash(part_info, part_info->no_subparts,
3162
                                 part_info->subpart_expr, &func_value);
3163 3164 3165 3166 3167
}


uint32 get_partition_id_key_sub(partition_info *part_info)
{
3168
  longlong func_value;
3169
  return get_part_id_key(part_info->subpart_field_array,
3170
                         part_info->no_subparts, &func_value);
3171 3172 3173 3174 3175
}


uint32 get_partition_id_linear_key_sub(partition_info *part_info)
{
3176
  longlong func_value;
3177 3178
  return get_part_id_linear_key(part_info,
                                part_info->subpart_field_array,
3179
                                part_info->no_subparts, &func_value);
3180 3181 3182 3183
}


/*
3184 3185
  Set an indicator on all partition fields that are set by the key

3186 3187 3188 3189
  SYNOPSIS
    set_PF_fields_in_key()
    key_info                   Information about the index
    key_length                 Length of key
3190

3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230
  RETURN VALUE
    TRUE                       Found partition field set by key
    FALSE                      No partition field set by key
*/

static bool set_PF_fields_in_key(KEY *key_info, uint key_length)
{
  KEY_PART_INFO *key_part;
  bool found_part_field= FALSE;
  DBUG_ENTER("set_PF_fields_in_key");

  for (key_part= key_info->key_part; (int)key_length > 0; key_part++)
  {
    if (key_part->null_bit)
      key_length--;
    if (key_part->type == HA_KEYTYPE_BIT)
    {
      if (((Field_bit*)key_part->field)->bit_len)
        key_length--;
    }
    if (key_part->key_part_flag & (HA_BLOB_PART + HA_VAR_LENGTH_PART))
    {
      key_length-= HA_KEY_BLOB_LENGTH;
    }
    if (key_length < key_part->length)
      break;
    key_length-= key_part->length;
    if (key_part->field->flags & FIELD_IN_PART_FUNC_FLAG)
    {
      found_part_field= TRUE;
      key_part->field->flags|= GET_FIXED_FIELDS_FLAG;
    }
  }
  DBUG_RETURN(found_part_field);
}


/*
  We have found that at least one partition field was set by a key, now
  check if a partition function has all its fields bound or not.
3231

3232 3233 3234
  SYNOPSIS
    check_part_func_bound()
    ptr                     Array of fields NULL terminated (partition fields)
3235

3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260
  RETURN VALUE
    TRUE                    All fields in partition function are set
    FALSE                   Not all fields in partition function are set
*/

static bool check_part_func_bound(Field **ptr)
{
  bool result= TRUE;
  DBUG_ENTER("check_part_func_bound");

  for (; *ptr; ptr++)
  {
    if (!((*ptr)->flags & GET_FIXED_FIELDS_FLAG))
    {
      result= FALSE;
      break;
    }
  }
  DBUG_RETURN(result);
}


/*
  Get the id of the subpartitioning part by using the key buffer of the
  index scan.
3261

3262 3263 3264 3265 3266 3267
  SYNOPSIS
    get_sub_part_id_from_key()
    table         The table object
    buf           A buffer that can be used to evaluate the partition function
    key_info      The index object
    key_spec      A key_range containing key and key length
3268

3269 3270
  RETURN VALUES
    part_id       Subpartition id to use
3271

3272 3273 3274 3275 3276 3277 3278 3279 3280 3281
  DESCRIPTION
    Use key buffer to set-up record in buf, move field pointers and
    get the partition identity and restore field pointers afterwards.
*/

static uint32 get_sub_part_id_from_key(const TABLE *table,byte *buf,
                                       KEY *key_info,
                                       const key_range *key_spec)
{
  byte *rec0= table->record[0];
3282
  partition_info *part_info= table->part_info;
3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301
  uint32 part_id;
  DBUG_ENTER("get_sub_part_id_from_key");

  key_restore(buf, (byte*)key_spec->key, key_info, key_spec->length);
  if (likely(rec0 == buf))
    part_id= part_info->get_subpartition_id(part_info);
  else
  {
    Field **part_field_array= part_info->subpart_field_array;
    set_field_ptr(part_field_array, buf, rec0);
    part_id= part_info->get_subpartition_id(part_info);
    set_field_ptr(part_field_array, rec0, buf);
  }
  DBUG_RETURN(part_id);
}

/*
  Get the id of the partitioning part by using the key buffer of the
  index scan.
3302

3303 3304 3305 3306 3307 3308
  SYNOPSIS
    get_part_id_from_key()
    table         The table object
    buf           A buffer that can be used to evaluate the partition function
    key_info      The index object
    key_spec      A key_range containing key and key length
3309 3310
    out:part_id   Partition to use

3311 3312 3313
  RETURN VALUES
    TRUE          Partition to use not found
    FALSE         Ok, part_id indicates partition to use
3314

3315 3316 3317 3318
  DESCRIPTION
    Use key buffer to set-up record in buf, move field pointers and
    get the partition identity and restore field pointers afterwards.
*/
3319

3320 3321 3322 3323 3324
bool get_part_id_from_key(const TABLE *table, byte *buf, KEY *key_info,
                          const key_range *key_spec, uint32 *part_id)
{
  bool result;
  byte *rec0= table->record[0];
3325
  partition_info *part_info= table->part_info;
3326
  longlong func_value;
3327 3328 3329 3330
  DBUG_ENTER("get_part_id_from_key");

  key_restore(buf, (byte*)key_spec->key, key_info, key_spec->length);
  if (likely(rec0 == buf))
3331 3332
    result= part_info->get_part_partition_id(part_info, part_id,
                                             &func_value);
3333 3334 3335 3336
  else
  {
    Field **part_field_array= part_info->part_field_array;
    set_field_ptr(part_field_array, buf, rec0);
3337 3338
    result= part_info->get_part_partition_id(part_info, part_id,
                                             &func_value);
3339 3340 3341 3342 3343 3344 3345 3346
    set_field_ptr(part_field_array, rec0, buf);
  }
  DBUG_RETURN(result);
}

/*
  Get the partitioning id of the full PF by using the key buffer of the
  index scan.
3347

3348 3349 3350 3351 3352 3353
  SYNOPSIS
    get_full_part_id_from_key()
    table         The table object
    buf           A buffer that is used to evaluate the partition function
    key_info      The index object
    key_spec      A key_range containing key and key length
3354 3355
    out:part_spec A partition id containing start part and end part

3356 3357 3358
  RETURN VALUES
    part_spec
    No partitions to scan is indicated by end_part > start_part when returning
3359

3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370
  DESCRIPTION
    Use key buffer to set-up record in buf, move field pointers if needed and
    get the partition identity and restore field pointers afterwards.
*/

void get_full_part_id_from_key(const TABLE *table, byte *buf,
                               KEY *key_info,
                               const key_range *key_spec,
                               part_id_range *part_spec)
{
  bool result;
3371
  partition_info *part_info= table->part_info;
3372
  byte *rec0= table->record[0];
3373
  longlong func_value;
3374 3375 3376 3377
  DBUG_ENTER("get_full_part_id_from_key");

  key_restore(buf, (byte*)key_spec->key, key_info, key_spec->length);
  if (likely(rec0 == buf))
3378 3379
    result= part_info->get_partition_id(part_info, &part_spec->start_part,
                                        &func_value);
3380 3381 3382 3383
  else
  {
    Field **part_field_array= part_info->full_part_field_array;
    set_field_ptr(part_field_array, buf, rec0);
3384 3385
    result= part_info->get_partition_id(part_info, &part_spec->start_part,
                                        &func_value);
3386 3387 3388 3389 3390 3391 3392
    set_field_ptr(part_field_array, rec0, buf);
  }
  part_spec->end_part= part_spec->start_part;
  if (unlikely(result))
    part_spec->start_part++;
  DBUG_VOID_RETURN;
}
3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429

/*
  Prune the set of partitions to use in query 

  SYNOPSIS
    prune_partition_set()
    table         The table object
    out:part_spec Contains start part, end part 

  DESCRIPTION
    This function is called to prune the range of partitions to scan by
    checking the used_partitions bitmap.
    If start_part > end_part at return it means no partition needs to be
    scanned. If start_part == end_part it always means a single partition
    needs to be scanned.

  RETURN VALUE
    part_spec
*/
void prune_partition_set(const TABLE *table, part_id_range *part_spec)
{
  int last_partition= -1;
  uint i;
  partition_info *part_info= table->part_info;

  DBUG_ENTER("prune_partition_set");
  for (i= part_spec->start_part; i <= part_spec->end_part; i++)
  {
    if (bitmap_is_set(&(part_info->used_partitions), i))
    {
      DBUG_PRINT("info", ("Partition %d is set", i));
      if (last_partition == -1)
        /* First partition found in set and pruned bitmap */
        part_spec->start_part= i;
      last_partition= i;
    }
  }
mskold@mysql.com's avatar
mskold@mysql.com committed
3430 3431 3432 3433
  if (last_partition == -1)
    /* No partition found in pruned bitmap */
    part_spec->start_part= part_spec->end_part + 1;  
  else //if (last_partition != -1)
3434 3435 3436 3437 3438
    part_spec->end_part= last_partition;

  DBUG_VOID_RETURN;
}

3439 3440
/*
  Get the set of partitions to use in query.
3441

3442 3443 3444 3445 3446 3447
  SYNOPSIS
    get_partition_set()
    table         The table object
    buf           A buffer that can be used to evaluate the partition function
    index         The index of the key used, if MAX_KEY no index used
    key_spec      A key_range containing key and key length
3448
    out:part_spec Contains start part, end part and indicator if bitmap is
3449
                  used for which partitions to scan
3450

3451 3452 3453 3454 3455 3456 3457 3458 3459
  DESCRIPTION
    This function is called to discover which partitions to use in an index
    scan or a full table scan.
    It returns a range of partitions to scan. If there are holes in this
    range with partitions that are not needed to scan a bit array is used
    to signal which partitions to use and which not to use.
    If start_part > end_part at return it means no partition needs to be
    scanned. If start_part == end_part it always means a single partition
    needs to be scanned.
3460

3461 3462 3463 3464 3465 3466
  RETURN VALUE
    part_spec
*/
void get_partition_set(const TABLE *table, byte *buf, const uint index,
                       const key_range *key_spec, part_id_range *part_spec)
{
3467
  partition_info *part_info= table->part_info;
3468
  uint no_parts= part_info->get_tot_partitions();
3469
  uint i, part_id;
3470 3471
  uint sub_part= no_parts;
  uint32 part_part= no_parts;
3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501
  KEY *key_info= NULL;
  bool found_part_field= FALSE;
  DBUG_ENTER("get_partition_set");

  part_spec->start_part= 0;
  part_spec->end_part= no_parts - 1;
  if ((index < MAX_KEY) && 
       key_spec->flag == (uint)HA_READ_KEY_EXACT &&
       part_info->some_fields_in_PF.is_set(index))
  {
    key_info= table->key_info+index;
    /*
      The index can potentially provide at least one PF-field (field in the
      partition function). Thus it is interesting to continue our probe.
    */
    if (key_spec->length == key_info->key_length)
    {
      /*
        The entire key is set so we can check whether we can immediately
        derive either the complete PF or if we can derive either
        the top PF or the subpartitioning PF. This can be established by
        checking precalculated bits on each index.
      */
      if (part_info->all_fields_in_PF.is_set(index))
      {
        /*
          We can derive the exact partition to use, no more than this one
          is needed.
        */
        get_full_part_id_from_key(table,buf,key_info,key_spec,part_spec);
3502 3503 3504 3505
        /*
          Check if range can be adjusted by looking in used_partitions
        */
        prune_partition_set(table, part_spec);
3506 3507
        DBUG_VOID_RETURN;
      }
3508
      else if (part_info->is_sub_partitioned())
3509 3510 3511 3512 3513
      {
        if (part_info->all_fields_in_SPF.is_set(index))
          sub_part= get_sub_part_id_from_key(table, buf, key_info, key_spec);
        else if (part_info->all_fields_in_PPF.is_set(index))
        {
3514 3515
          if (get_part_id_from_key(table,buf,key_info,
                                   key_spec,(uint32*)&part_part))
3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547
          {
            /*
              The value of the RANGE or LIST partitioning was outside of
              allowed values. Thus it is certain that the result of this
              scan will be empty.
            */
            part_spec->start_part= no_parts;
            DBUG_VOID_RETURN;
          }
        }
      }
    }
    else
    {
      /*
        Set an indicator on all partition fields that are bound.
        If at least one PF-field was bound it pays off to check whether
        the PF or PPF or SPF has been bound.
        (PF = Partition Function, SPF = Subpartition Function and
         PPF = Partition Function part of subpartitioning)
      */
      if ((found_part_field= set_PF_fields_in_key(key_info,
                                                  key_spec->length)))
      {
        if (check_part_func_bound(part_info->full_part_field_array))
        {
          /*
            We were able to bind all fields in the partition function even
            by using only a part of the key. Calculate the partition to use.
          */
          get_full_part_id_from_key(table,buf,key_info,key_spec,part_spec);
          clear_indicator_in_key_fields(key_info);
3548 3549 3550 3551
          /*
            Check if range can be adjusted by looking in used_partitions
          */
          prune_partition_set(table, part_spec);
3552 3553
          DBUG_VOID_RETURN; 
        }
3554
        else if (part_info->is_sub_partitioned())
3555
        {
3556 3557 3558
          if (check_part_func_bound(part_info->subpart_field_array))
            sub_part= get_sub_part_id_from_key(table, buf, key_info, key_spec);
          else if (check_part_func_bound(part_info->part_field_array))
3559
          {
3560 3561 3562 3563 3564 3565
            if (get_part_id_from_key(table,buf,key_info,key_spec,&part_part))
            {
              part_spec->start_part= no_parts;
              clear_indicator_in_key_fields(key_info);
              DBUG_VOID_RETURN;
            }
3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594
          }
        }
      }
    }
  }
  {
    /*
      The next step is to analyse the table condition to see whether any
      information about which partitions to scan can be derived from there.
      Currently not implemented.
    */
  }
  /*
    If we come here we have found a range of sorts we have either discovered
    nothing or we have discovered a range of partitions with possible holes
    in it. We need a bitvector to further the work here.
  */
  if (!(part_part == no_parts && sub_part == no_parts))
  {
    /*
      We can only arrive here if we are using subpartitioning.
    */
    if (part_part != no_parts)
    {
      /*
        We know the top partition and need to scan all underlying
        subpartitions. This is a range without holes.
      */
      DBUG_ASSERT(sub_part == no_parts);
3595
      part_spec->start_part= part_part * part_info->no_subparts;
3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610
      part_spec->end_part= part_spec->start_part+part_info->no_subparts - 1;
    }
    else
    {
      DBUG_ASSERT(sub_part != no_parts);
      part_spec->start_part= sub_part;
      part_spec->end_part=sub_part+
                           (part_info->no_subparts*(part_info->no_parts-1));
      for (i= 0, part_id= sub_part; i < part_info->no_parts;
           i++, part_id+= part_info->no_subparts)
        ; //Set bit part_id in bit array
    }
  }
  if (found_part_field)
    clear_indicator_in_key_fields(key_info);
3611 3612 3613 3614
  /*
    Check if range can be adjusted by looking in used_partitions
  */
  prune_partition_set(table, part_spec);
3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635
  DBUG_VOID_RETURN;
}

/*
   If the table is partitioned we will read the partition info into the
   .frm file here.
   -------------------------------
   |  Fileinfo     64 bytes      |
   -------------------------------
   | Formnames     7 bytes       |
   -------------------------------
   | Not used    4021 bytes      |
   -------------------------------
   | Keyinfo + record            |
   -------------------------------
   | Padded to next multiple     |
   | of IO_SIZE                  |
   -------------------------------
   | Forminfo     288 bytes      |
   -------------------------------
   | Screen buffer, to make      |
3636
   | field names readable        |
3637 3638
   -------------------------------
   | Packed field info           |
3639
   | 17 + 1 + strlen(field_name) |
3640 3641 3642 3643 3644 3645 3646 3647
   | + 1 end of file character   |
   -------------------------------
   | Partition info              |
   -------------------------------
   We provide the length of partition length in Fileinfo[55-58].

   Read the partition syntax from the frm file and parse it to get the
   data structures of the partitioning.
3648

3649 3650 3651
   SYNOPSIS
     mysql_unpack_partition()
     thd                           Thread object
3652
     part_buf                      Partition info from frm file
3653 3654
     part_info_len                 Length of partition syntax
     table                         Table object of partitioned table
3655 3656 3657
     create_table_ind              Is it called from CREATE TABLE
     default_db_type               What is the default engine of the table

3658 3659 3660
   RETURN VALUE
     TRUE                          Error
     FALSE                         Sucess
3661

3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672
   DESCRIPTION
     Read the partition syntax from the current position in the frm file.
     Initiate a LEX object, save the list of item tree objects to free after
     the query is done. Set-up partition info object such that parser knows
     it is called from internally. Call parser to create data structures
     (best possible recreation of item trees and so forth since there is no
     serialisation of these objects other than in parseable text format).
     We need to save the text of the partition functions since it is not
     possible to retrace this given an item tree.
*/

3673
bool mysql_unpack_partition(THD *thd, const uchar *part_buf,
3674 3675 3676
                            uint part_info_len,
                            uchar *part_state, uint part_state_len,
                            TABLE* table, bool is_create_table_ind,
3677
                            handlerton *default_db_type)
3678 3679 3680 3681
{
  Item *thd_free_list= thd->free_list;
  bool result= TRUE;
  partition_info *part_info;
3682
  CHARSET_INFO *old_character_set_client= thd->variables.character_set_client;
3683 3684
  LEX *old_lex= thd->lex;
  LEX lex;
3685
  DBUG_ENTER("mysql_unpack_partition");
3686

3687
  thd->lex= &lex;
3688
  thd->variables.character_set_client= system_charset_info;
3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707
  lex_start(thd, part_buf, part_info_len);
  /*
    We need to use the current SELECT_LEX since I need to keep the
    Name_resolution_context object which is referenced from the
    Item_field objects.
    This is not a nice solution since if the parser uses current_select
    for anything else it will corrupt the current LEX object.
  */
  thd->lex->current_select= old_lex->current_select; 
  /*
    All Items created is put into a free list on the THD object. This list
    is used to free all Item objects after completing a query. We don't
    want that to happen with the Item tree created as part of the partition
    info. This should be attached to the table object and remain so until
    the table object is released.
    Thus we move away the current list temporarily and start a new list that
    we then save in the partition info structure.
  */
  thd->free_list= NULL;
3708
  lex.part_info= new partition_info();/* Indicates MYSQLparse from this place */
3709 3710 3711 3712 3713 3714 3715 3716
  if (!lex.part_info)
  {
    mem_alloc_error(sizeof(partition_info));
    goto end;
  }
  lex.part_info->part_state= part_state;
  lex.part_info->part_state_len= part_state_len;
  DBUG_PRINT("info", ("Parse: %s", part_buf));
3717
  if (MYSQLparse((void*)thd) || thd->is_fatal_error)
3718 3719 3720 3721
  {
    free_items(thd->free_list);
    goto end;
  }
3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737
  /*
    The parsed syntax residing in the frm file can still contain defaults.
    The reason is that the frm file is sometimes saved outside of this
    MySQL Server and used in backup and restore of clusters or partitioned
    tables. It is not certain that the restore will restore exactly the
    same default partitioning.
    
    The easiest manner of handling this is to simply continue using the
    part_info we already built up during mysql_create_table if we are
    in the process of creating a table. If the table already exists we
    need to discover the number of partitions for the default parts. Since
    the handler object hasn't been created here yet we need to postpone this
    to the fix_partition_func method.
  */

  DBUG_PRINT("info", ("Successful parse"));
3738
  part_info= lex.part_info;
3739 3740 3741
  DBUG_PRINT("info", ("default engine = %d, default_db_type = %d",
             ha_legacy_type(part_info->default_engine_type),
             ha_legacy_type(default_db_type)));
3742 3743 3744 3745 3746 3747 3748 3749 3750
  if (is_create_table_ind)
  {
    if (old_lex->name)
    {
      /*
        This code is executed when we do a CREATE TABLE t1 LIKE t2
        old_lex->name contains the t2 and the table we are opening has 
        name t1.
      */
3751 3752 3753 3754 3755 3756
      Table_ident *table_ident= (Table_ident *)old_lex->name;
      char *src_db= table_ident->db.str ? table_ident->db.str : thd->db;
      char *src_table= table_ident->table.str;
      char buf[FN_REFLEN];
      build_table_filename(buf, sizeof(buf), src_db, src_table, "");
      if (partition_default_handling(table, part_info, buf))
3757
      {
3758 3759
        result= TRUE;
        goto end;
3760 3761 3762 3763 3764
      }
    }
    else
      part_info= old_lex->part_info;
  }
3765
  table->part_info= part_info;
3766
  table->file->set_part_info(part_info);
3767
  if (part_info->default_engine_type == NULL)
3768
  {
3769
    part_info->default_engine_type= default_db_type;
3770
  }
3771 3772 3773 3774
  else
  {
    DBUG_ASSERT(part_info->default_engine_type == default_db_type);
  }
3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788
  part_info->item_free_list= thd->free_list;

  {
  /*
    This code part allocates memory for the serialised item information for
    the partition functions. In most cases this is not needed but if the
    table is used for SHOW CREATE TABLES or ALTER TABLE that modifies
    partition information it is needed and the info is lost if we don't
    save it here so unfortunately we have to do it here even if in most
    cases it is not needed. This is a consequence of that item trees are
    not serialisable.
  */
    uint part_func_len= part_info->part_func_len;
    uint subpart_func_len= part_info->subpart_func_len; 
3789 3790 3791 3792
    char *part_func_string= NULL;
    char *subpart_func_string= NULL;
    if ((part_func_len &&
        !((part_func_string= thd->alloc(part_func_len)))) ||
3793
        (subpart_func_len &&
3794
        !((subpart_func_string= thd->alloc(subpart_func_len)))))
3795
    {
3796
      mem_alloc_error(part_func_len);
3797 3798 3799 3800
      free_items(thd->free_list);
      part_info->item_free_list= 0;
      goto end;
    }
3801 3802
    if (part_func_len)
      memcpy(part_func_string, part_info->part_func_string, part_func_len);
3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813
    if (subpart_func_len)
      memcpy(subpart_func_string, part_info->subpart_func_string,
             subpart_func_len);
    part_info->part_func_string= part_func_string;
    part_info->subpart_func_string= subpart_func_string;
  }

  result= FALSE;
end:
  thd->free_list= thd_free_list;
  thd->lex= old_lex;
3814
  thd->variables.character_set_client= old_character_set_client;
3815 3816
  DBUG_RETURN(result);
}
3817

3818 3819 3820

/*
  SYNOPSIS
3821 3822 3823 3824 3825 3826
    fast_alter_partition_error_handler()
    lpt                           Container for parameters

  RETURN VALUES
    None

3827
  DESCRIPTION
3828 3829
    Support routine to clean up after failures of on-line ALTER TABLE
    for partition management.
3830 3831
*/

3832
static void fast_alter_partition_error_handler(ALTER_PARTITION_PARAM_TYPE *lpt)
3833
{
3834 3835
  DBUG_ENTER("fast_alter_partition_error_handler");
  /* TODO: WL 2826 Error handling */
3836 3837 3838 3839 3840 3841
  DBUG_VOID_RETURN;
}


/*
  SYNOPSIS
3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853
    fast_end_partition()
    thd                           Thread object
    out:copied                    Number of records copied
    out:deleted                   Number of records deleted
    table_list                    Table list with the one table in it
    empty                         Has nothing been done
    lpt                           Struct to be used by error handler

  RETURN VALUES
    FALSE                         Success
    TRUE                          Failure

3854
  DESCRIPTION
3855 3856
    Support routine to handle the successful cases for partition
    management.
3857 3858
*/

3859 3860 3861 3862 3863
static int fast_end_partition(THD *thd, ulonglong copied,
                              ulonglong deleted,
                              TABLE_LIST *table_list, bool is_empty,
                              ALTER_PARTITION_PARAM_TYPE *lpt,
                              bool written_bin_log)
3864
{
3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892
  int error;
  DBUG_ENTER("fast_end_partition");

  thd->proc_info="end";
  if (!is_empty)
    query_cache_invalidate3(thd, table_list, 0);
  error= ha_commit_stmt(thd);
  if (ha_commit(thd))
    error= 1;
  if (!error || is_empty)
  {
    char tmp_name[80];
    if ((!is_empty) && (!written_bin_log) &&
        (!thd->lex->no_write_to_binlog))
      write_bin_log(thd, FALSE, thd->query, thd->query_length);
    close_thread_tables(thd);
    my_snprintf(tmp_name, sizeof(tmp_name), ER(ER_INSERT_INFO),
                (ulong) (copied + deleted),
                (ulong) deleted,
                (ulong) 0);
    send_ok(thd,copied+deleted,0L,tmp_name);
    DBUG_RETURN(FALSE);
  }
  fast_alter_partition_error_handler(lpt);
  DBUG_RETURN(TRUE);
}


3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933
/*
  Check engine mix that it is correct
  SYNOPSIS
    check_engine_condition()
    p_elem                   Partition element
    default_engine           Have user specified engine on table level
    inout::engine_type       Current engine used
    inout::first             Is it first partition
  RETURN VALUE
    TRUE                     Failed check
    FALSE                    Ok
  DESCRIPTION
    (specified partition handler ) specified table handler
    (NDB, NDB) NDB           OK
    (MYISAM, MYISAM) -       OK
    (MYISAM, -)      -       NOT OK
    (MYISAM, -)    MYISAM    OK
    (- , MYISAM)   -         NOT OK
    (- , -)        MYISAM    OK
    (-,-)          -         OK
    (NDB, MYISAM) *          NOT OK
*/

static bool check_engine_condition(partition_element *p_elem,
                                   bool default_engine,
                                   handlerton **engine_type,
                                   bool *first)
{
  if (*first && default_engine)
    *engine_type= p_elem->engine_type;
  *first= FALSE;
  if ((!default_engine &&
      (p_elem->engine_type != *engine_type &&
       !p_elem->engine_type)) ||
      (default_engine &&
       p_elem->engine_type != *engine_type))
    return TRUE;
  else
    return FALSE;
}

3934 3935 3936
/*
  We need to check if engine used by all partitions can handle
  partitioning natively.
3937

3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961
  SYNOPSIS
    check_native_partitioned()
    create_info            Create info in CREATE TABLE
    out:ret_val            Return value
    part_info              Partition info
    thd                    Thread object

  RETURN VALUES
  Value returned in bool ret_value
    TRUE                   Native partitioning supported by engine
    FALSE                  Need to use partition handler

  Return value from function
    TRUE                   Error
    FALSE                  Success
*/

static bool check_native_partitioned(HA_CREATE_INFO *create_info,bool *ret_val,
                                     partition_info *part_info, THD *thd)
{
  List_iterator<partition_element> part_it(part_info->partitions);
  bool first= TRUE;
  bool default_engine;
  handlerton *engine_type= create_info->db_type;
3962
  handlerton *old_engine_type= engine_type;
3963 3964
  uint i= 0;
  handler *file;
3965
  uint no_parts= part_info->partitions.elements;
3966 3967 3968 3969 3970 3971 3972
  DBUG_ENTER("check_native_partitioned");

  default_engine= (create_info->used_fields | HA_CREATE_USED_ENGINE) ?
                   TRUE : FALSE;
  DBUG_PRINT("info", ("engine_type = %u, default = %u",
                       ha_legacy_type(engine_type),
                       default_engine));
3973
  if (no_parts)
3974
  {
3975
    do
3976
    {
3977
      partition_element *part_elem= part_it++;
3978
      if (part_info->is_sub_partitioned() &&
3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005
          part_elem->subpartitions.elements)
      {
        uint no_subparts= part_elem->subpartitions.elements;
        uint j= 0;
        List_iterator<partition_element> sub_it(part_elem->subpartitions);
        do
        {
          partition_element *sub_elem= sub_it++;
          if (check_engine_condition(sub_elem, default_engine,
                                     &engine_type, &first))
            goto error;
        } while (++j < no_subparts);
        /*
          In case of subpartitioning and defaults we allow that only
          subparts have specified engines, as long as the parts haven't
          specified the wrong engine it's ok.
        */
        if (check_engine_condition(part_elem, FALSE,
                                   &engine_type, &first))
          goto error;
      }
      else if (check_engine_condition(part_elem, default_engine,
                                      &engine_type, &first))
        goto error;
    } while (++i < no_parts);
  }

4006 4007 4008 4009
  /*
    All engines are of the same type. Check if this engine supports
    native partitioning.
  */
4010 4011 4012 4013 4014

  if (!engine_type)
    engine_type= old_engine_type;
  DBUG_PRINT("info", ("engine_type = %s",
              ha_resolve_storage_engine_name(engine_type)));
4015 4016 4017 4018 4019 4020 4021 4022
  if (engine_type->partition_flags &&
      (engine_type->partition_flags() & HA_CAN_PARTITION))
  {
    create_info->db_type= engine_type;
    DBUG_PRINT("info", ("Changed to native partitioning"));
    *ret_val= TRUE;
  }
  DBUG_RETURN(FALSE);
4023 4024 4025 4026 4027 4028 4029
error:
  /*
    Mixed engines not yet supported but when supported it will need
    the partition handler
  */
  *ret_val= FALSE;
  DBUG_RETURN(TRUE);
4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070
}


/*
  Prepare for ALTER TABLE of partition structure

  SYNOPSIS
    prep_alter_part_table()
    thd                        Thread object
    table                      Table object
    inout:alter_info           Alter information
    inout:create_info          Create info for CREATE TABLE
    old_db_type                Old engine type
    out:partition_changed      Boolean indicating whether partition changed
    out:fast_alter_partition   Boolean indicating whether fast partition
                               change is requested

  RETURN VALUES
    TRUE                       Error
    FALSE                      Success
    partition_changed
    fast_alter_partition

  DESCRIPTION
    This method handles all preparations for ALTER TABLE for partitioned
    tables
    We need to handle both partition management command such as Add Partition
    and others here as well as an ALTER TABLE that completely changes the
    partitioning and yet others that don't change anything at all. We start
    by checking the partition management variants and then check the general
    change patterns.
*/

uint prep_alter_part_table(THD *thd, TABLE *table, ALTER_INFO *alter_info,
                           HA_CREATE_INFO *create_info,
                           handlerton *old_db_type,
                           bool *partition_changed,
                           uint *fast_alter_partition)
{
  DBUG_ENTER("prep_alter_part_table");

4071 4072 4073 4074 4075 4076 4077 4078 4079
  /*
    We are going to manipulate the partition info on the table object
    so we need to ensure that the data structure of the table object
    is freed by setting version to 0. table->s->version= 0 forces a
    flush of the table object in close_thread_tables().
  */
  if (table->part_info)
    table->s->version= 0L;

4080 4081 4082 4083 4084 4085 4086 4087
  if (alter_info->flags &
      (ALTER_ADD_PARTITION | ALTER_DROP_PARTITION |
       ALTER_COALESCE_PARTITION | ALTER_REORGANIZE_PARTITION |
       ALTER_TABLE_REORG | ALTER_OPTIMIZE_PARTITION |
       ALTER_CHECK_PARTITION | ALTER_ANALYZE_PARTITION |
       ALTER_REPAIR_PARTITION | ALTER_REBUILD_PARTITION))
  {
    partition_info *tab_part_info= table->part_info;
4088
    partition_info *alt_part_info= thd->lex->part_info;
4089
    uint flags= 0;
4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141
    if (!tab_part_info)
    {
      my_error(ER_PARTITION_MGMT_ON_NONPARTITIONED, MYF(0));
      DBUG_RETURN(TRUE);
    }
    if (alter_info->flags == ALTER_TABLE_REORG)
    {
      uint new_part_no, curr_part_no;
      ulonglong max_rows= table->s->max_rows;
      if (tab_part_info->part_type != HASH_PARTITION ||
          tab_part_info->use_default_no_partitions)
      {
        my_error(ER_REORG_NO_PARAM_ERROR, MYF(0));
        DBUG_RETURN(TRUE);
      }
      new_part_no= table->file->get_default_no_partitions(max_rows);
      curr_part_no= tab_part_info->no_parts;
      if (new_part_no == curr_part_no)
      {
        /*
          No change is needed, we will have the same number of partitions
          after the change as before. Thus we can reply ok immediately
          without any changes at all.
        */
        DBUG_RETURN(fast_end_partition(thd, ULL(0), ULL(0), NULL,
                                       TRUE, NULL, FALSE));
      }
      else if (new_part_no > curr_part_no)
      {
        /*
          We will add more partitions, we use the ADD PARTITION without
          setting the flag for no default number of partitions
        */
        alter_info->flags|= ALTER_ADD_PARTITION;
        thd->lex->part_info->no_parts= new_part_no - curr_part_no;
      }
      else
      {
        /*
          We will remove hash partitions, we use the COALESCE PARTITION
          without setting the flag for no default number of partitions
        */
        alter_info->flags|= ALTER_COALESCE_PARTITION;
        alter_info->no_parts= curr_part_no - new_part_no;
      }
    }
    if (table->s->db_type->alter_table_flags &&
        (!(flags= table->s->db_type->alter_table_flags(alter_info->flags))))
    {
      my_error(ER_PARTITION_FUNCTION_FAILURE, MYF(0));
      DBUG_RETURN(1);
    }
4142 4143 4144 4145
    *fast_alter_partition=
      ((flags & (HA_FAST_CHANGE_PARTITION | HA_PARTITION_ONE_PHASE)) != 0);
    DBUG_PRINT("info", ("*fast_alter_partition: %d  flags: 0x%x",
                        *fast_alter_partition, flags));
4146 4147
    if (((alter_info->flags & ALTER_ADD_PARTITION) ||
         (alter_info->flags & ALTER_REORGANIZE_PARTITION)) &&
mikael@zim.(none)'s avatar
mikael@zim.(none) committed
4148 4149
         (thd->lex->part_info->part_type != tab_part_info->part_type) &&
         (thd->lex->part_info->part_type != NOT_A_PARTITION))
4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174
    {
      if (thd->lex->part_info->part_type == RANGE_PARTITION)
      {
        my_error(ER_PARTITION_WRONG_VALUES_ERROR, MYF(0),
                 "RANGE", "LESS THAN");
      }
      else if (thd->lex->part_info->part_type == LIST_PARTITION)
      {
        DBUG_ASSERT(thd->lex->part_info->part_type == LIST_PARTITION);
        my_error(ER_PARTITION_WRONG_VALUES_ERROR, MYF(0),
                 "LIST", "IN");
      }
      else if (tab_part_info->part_type == RANGE_PARTITION)
      {
        my_error(ER_PARTITION_REQUIRES_VALUES_ERROR, MYF(0),
                 "RANGE", "LESS THAN");
      }
      else
      {
        DBUG_ASSERT(tab_part_info->part_type == LIST_PARTITION);
        my_error(ER_PARTITION_REQUIRES_VALUES_ERROR, MYF(0),
                 "LIST", "IN");
      }
      DBUG_RETURN(TRUE);
    }
4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201
    if (alter_info->flags & ALTER_ADD_PARTITION)
    {
      /*
        We start by moving the new partitions to the list of temporary
        partitions. We will then check that the new partitions fit in the
        partitioning scheme as currently set-up.
        Partitions are always added at the end in ADD PARTITION.
      */
      uint no_new_partitions= alt_part_info->no_parts;
      uint no_orig_partitions= tab_part_info->no_parts;
      uint check_total_partitions= no_new_partitions + no_orig_partitions;
      uint new_total_partitions= check_total_partitions;
      /*
        We allow quite a lot of values to be supplied by defaults, however we
        must know the number of new partitions in this case.
      */
      if (thd->lex->no_write_to_binlog &&
          tab_part_info->part_type != HASH_PARTITION)
      {
        my_error(ER_NO_BINLOG_ERROR, MYF(0));
        DBUG_RETURN(TRUE);
      } 
      if (no_new_partitions == 0)
      {
        my_error(ER_ADD_PARTITION_NO_NEW_PARTITION, MYF(0));
        DBUG_RETURN(TRUE);
      }
4202
      if (tab_part_info->is_sub_partitioned())
4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219
      {
        if (alt_part_info->no_subparts == 0)
          alt_part_info->no_subparts= tab_part_info->no_subparts;
        else if (alt_part_info->no_subparts != tab_part_info->no_subparts)
        {
          my_error(ER_ADD_PARTITION_SUBPART_ERROR, MYF(0));
          DBUG_RETURN(TRUE);
        }
        check_total_partitions= new_total_partitions*
                                alt_part_info->no_subparts;
      }
      if (check_total_partitions > MAX_PARTITIONS)
      {
        my_error(ER_TOO_MANY_PARTITIONS_ERROR, MYF(0));
        DBUG_RETURN(TRUE);
      }
      alt_part_info->part_type= tab_part_info->part_type;
4220 4221 4222
      if (alt_part_info->set_up_defaults_for_partitioning(table->file,
                                                          ULL(0), 
                                                          tab_part_info->no_parts))
4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776
      {
        DBUG_RETURN(TRUE);
      }
/*
Handling of on-line cases:

ADD PARTITION for RANGE/LIST PARTITIONING:
------------------------------------------
For range and list partitions add partition is simply adding a
new empty partition to the table. If the handler support this we
will use the simple method of doing this. The figure below shows
an example of this and the states involved in making this change.
            
Existing partitions                                     New added partitions
------       ------        ------        ------      |  ------    ------
|    |       |    |        |    |        |    |      |  |    |    |    |
| p0 |       | p1 |        | p2 |        | p3 |      |  | p4 |    | p5 |
------       ------        ------        ------      |  ------    ------
PART_NORMAL  PART_NORMAL   PART_NORMAL   PART_NORMAL    PART_TO_BE_ADDED*2
PART_NORMAL  PART_NORMAL   PART_NORMAL   PART_NORMAL    PART_IS_ADDED*2

The first line is the states before adding the new partitions and the 
second line is after the new partitions are added. All the partitions are
in the partitions list, no partitions are placed in the temp_partitions
list.

ADD PARTITION for HASH PARTITIONING
-----------------------------------
This little figure tries to show the various partitions involved when
adding two new partitions to a linear hash based partitioned table with
four partitions to start with, which lists are used and the states they
pass through. Adding partitions to a normal hash based is similar except
that it is always all the existing partitions that are reorganised not
only a subset of them.

Existing partitions                                     New added partitions
------       ------        ------        ------      |  ------    ------
|    |       |    |        |    |        |    |      |  |    |    |    |
| p0 |       | p1 |        | p2 |        | p3 |      |  | p4 |    | p5 |
------       ------        ------        ------      |  ------    ------
PART_CHANGED PART_CHANGED  PART_NORMAL   PART_NORMAL    PART_TO_BE_ADDED
PART_IS_CHANGED*2          PART_NORMAL   PART_NORMAL    PART_IS_ADDED
PART_NORMAL  PART_NORMAL   PART_NORMAL   PART_NORMAL    PART_IS_ADDED

Reorganised existing partitions
------      ------
|    |      |    |
| p0'|      | p1'|
------      ------

p0 - p5 will be in the partitions list of partitions.
p0' and p1' will actually not exist as separate objects, there presence can
be deduced from the state of the partition and also the names of those
partitions can be deduced this way.

After adding the partitions and copying the partition data to p0', p1',
p4 and p5 from p0 and p1 the states change to adapt for the new situation
where p0 and p1 is dropped and replaced by p0' and p1' and the new p4 and
p5 are in the table again.

The first line above shows the states of the partitions before we start
adding and copying partitions, the second after completing the adding
and copying and finally the third line after also dropping the partitions
that are reorganised.
*/
      if (*fast_alter_partition &&
          tab_part_info->part_type == HASH_PARTITION)
      {
        uint part_no= 0, start_part= 1, start_sec_part= 1;
        uint end_part= 0, end_sec_part= 0;
        uint upper_2n= tab_part_info->linear_hash_mask + 1;
        uint lower_2n= upper_2n >> 1;
        bool all_parts= TRUE;
        if (tab_part_info->linear_hash_ind &&
            no_new_partitions < upper_2n)
        {
          /*
            An analysis of which parts needs reorganisation shows that it is
            divided into two intervals. The first interval is those parts
            that are reorganised up until upper_2n - 1. From upper_2n and
            onwards it starts again from partition 0 and goes on until
            it reaches p(upper_2n - 1). If the last new partition reaches
            beyond upper_2n - 1 then the first interval will end with
            p(lower_2n - 1) and start with p(no_orig_partitions - lower_2n).
            If lower_2n partitions are added then p0 to p(lower_2n - 1) will
            be reorganised which means that the two interval becomes one
            interval at this point. Thus only when adding less than
            lower_2n partitions and going beyond a total of upper_2n we
            actually get two intervals.

            To exemplify this assume we have 6 partitions to start with and
            add 1, 2, 3, 5, 6, 7, 8, 9 partitions.
            The first to add after p5 is p6 = 110 in bit numbers. Thus we
            can see that 10 = p2 will be partition to reorganise if only one
            partition.
            If 2 partitions are added we reorganise [p2, p3]. Those two
            cases are covered by the second if part below.
            If 3 partitions are added we reorganise [p2, p3] U [p0,p0]. This
            part is covered by the else part below.
            If 5 partitions are added we get [p2,p3] U [p0, p2] = [p0, p3].
            This is covered by the first if part where we need the max check
            to here use lower_2n - 1.
            If 7 partitions are added we get [p2,p3] U [p0, p4] = [p0, p4].
            This is covered by the first if part but here we use the first
            calculated end_part.
            Finally with 9 new partitions we would also reorganise p6 if we
            used the method below but we cannot reorganise more partitions
            than what we had from the start and thus we simply set all_parts
            to TRUE. In this case we don't get into this if-part at all.
          */
          all_parts= FALSE;
          if (no_new_partitions >= lower_2n)
          {
            /*
              In this case there is only one interval since the two intervals
              overlap and this starts from zero to last_part_no - upper_2n
            */
            start_part= 0;
            end_part= new_total_partitions - (upper_2n + 1);
            end_part= max(lower_2n - 1, end_part);
          }
          else if (new_total_partitions <= upper_2n)
          {
            /*
              Also in this case there is only one interval since we are not
              going over a 2**n boundary
            */
            start_part= no_orig_partitions - lower_2n;
            end_part= start_part + (no_new_partitions - 1);
          }
          else
          {
            /* We have two non-overlapping intervals since we are not
               passing a 2**n border and we have not at least lower_2n
               new parts that would ensure that the intervals become
               overlapping.
            */
            start_part= no_orig_partitions - lower_2n;
            end_part= upper_2n - 1;
            start_sec_part= 0;
            end_sec_part= new_total_partitions - (upper_2n + 1);
          }
        }
        List_iterator<partition_element> tab_it(tab_part_info->partitions);
        part_no= 0;
        do
        {
          partition_element *p_elem= tab_it++;
          if (all_parts ||
              (part_no >= start_part && part_no <= end_part) ||
              (part_no >= start_sec_part && part_no <= end_sec_part))
          {
            p_elem->part_state= PART_CHANGED;
          }
        } while (++part_no < no_orig_partitions);
      }
      /*
        Need to concatenate the lists here to make it possible to check the
        partition info for correctness using check_partition_info.
        For on-line add partition we set the state of this partition to
        PART_TO_BE_ADDED to ensure that it is known that it is not yet
        usable (becomes usable when partition is created and the switch of
        partition configuration is made.
      */
      {
        List_iterator<partition_element> alt_it(alt_part_info->partitions);
        uint part_count= 0;
        do
        {
          partition_element *part_elem= alt_it++;
          if (*fast_alter_partition)
            part_elem->part_state= PART_TO_BE_ADDED;
          if (tab_part_info->partitions.push_back(part_elem))
          {
            mem_alloc_error(1);
            DBUG_RETURN(TRUE);
          }
        } while (++part_count < no_new_partitions);
        tab_part_info->no_parts+= no_new_partitions;
      }
      /*
        If we specify partitions explicitly we don't use defaults anymore.
        Using ADD PARTITION also means that we don't have the default number
        of partitions anymore. We use this code also for Table reorganisations
        and here we don't set any default flags to FALSE.
      */
      if (!(alter_info->flags & ALTER_TABLE_REORG))
      {
        if (!alt_part_info->use_default_partitions)
        {
          DBUG_PRINT("info", ("part_info= %x", tab_part_info));
          tab_part_info->use_default_partitions= FALSE;
        }
        tab_part_info->use_default_no_partitions= FALSE;
      }
    }
    else if (alter_info->flags == ALTER_DROP_PARTITION)
    {
      /*
        Drop a partition from a range partition and list partitioning is
        always safe and can be made more or less immediate. It is necessary
        however to ensure that the partition to be removed is safely removed
        and that REPAIR TABLE can remove the partition if for some reason the
        command to drop the partition failed in the middle.
      */
      uint part_count= 0;
      uint no_parts_dropped= alter_info->partition_names.elements;
      uint no_parts_found= 0;
      List_iterator<partition_element> part_it(tab_part_info->partitions);
      if (!(tab_part_info->part_type == RANGE_PARTITION ||
            tab_part_info->part_type == LIST_PARTITION))
      {
        my_error(ER_ONLY_ON_RANGE_LIST_PARTITION, MYF(0), "DROP");
        DBUG_RETURN(TRUE);
      }
      if (no_parts_dropped >= tab_part_info->no_parts)
      {
        my_error(ER_DROP_LAST_PARTITION, MYF(0));
        DBUG_RETURN(TRUE);
      }
      do
      {
        partition_element *part_elem= part_it++;
        if (is_name_in_list(part_elem->partition_name,
                            alter_info->partition_names))
        {
          /*
            Set state to indicate that the partition is to be dropped.
          */
          no_parts_found++;
          part_elem->part_state= PART_TO_BE_DROPPED;
        }
      } while (++part_count < tab_part_info->no_parts);
      if (no_parts_found != no_parts_dropped)
      {
        my_error(ER_DROP_PARTITION_NON_EXISTENT, MYF(0), "DROP");
        DBUG_RETURN(TRUE);
      }
      if (table->file->is_fk_defined_on_table_or_index(MAX_KEY))
      {
        my_error(ER_ROW_IS_REFERENCED, MYF(0));
        DBUG_RETURN(TRUE);
      }
    }
    else if ((alter_info->flags & ALTER_OPTIMIZE_PARTITION) ||
             (alter_info->flags & ALTER_ANALYZE_PARTITION) ||
             (alter_info->flags & ALTER_CHECK_PARTITION) ||
             (alter_info->flags & ALTER_REPAIR_PARTITION) ||
             (alter_info->flags & ALTER_REBUILD_PARTITION))
    {
      uint no_parts_opt= alter_info->partition_names.elements;
      uint part_count= 0;
      uint no_parts_found= 0;
      List_iterator<partition_element> part_it(tab_part_info->partitions);

      do
      {
        partition_element *part_elem= part_it++;
        if ((alter_info->flags & ALTER_ALL_PARTITION) ||
            (is_name_in_list(part_elem->partition_name,
                             alter_info->partition_names)))
        {
          /*
            Mark the partition as a partition to be "changed" by
            analyzing/optimizing/rebuilding/checking/repairing
          */
          no_parts_found++;
          part_elem->part_state= PART_CHANGED;
        }
      } while (++part_count < tab_part_info->no_parts);
      if (no_parts_found != no_parts_opt &&
          (!(alter_info->flags & ALTER_ALL_PARTITION)))
      {
        const char *ptr;
        if (alter_info->flags & ALTER_OPTIMIZE_PARTITION)
          ptr= "OPTIMIZE";
        else if (alter_info->flags & ALTER_ANALYZE_PARTITION)
          ptr= "ANALYZE";
        else if (alter_info->flags & ALTER_CHECK_PARTITION)
          ptr= "CHECK";
        else if (alter_info->flags & ALTER_REPAIR_PARTITION)
          ptr= "REPAIR";
        else
          ptr= "REBUILD";
        my_error(ER_DROP_PARTITION_NON_EXISTENT, MYF(0), ptr);
        DBUG_RETURN(TRUE);
      }
    }
    else if (alter_info->flags & ALTER_COALESCE_PARTITION)
    {
      uint no_parts_coalesced= alter_info->no_parts;
      uint no_parts_remain= tab_part_info->no_parts - no_parts_coalesced;
      List_iterator<partition_element> part_it(tab_part_info->partitions);
      if (tab_part_info->part_type != HASH_PARTITION)
      {
        my_error(ER_COALESCE_ONLY_ON_HASH_PARTITION, MYF(0));
        DBUG_RETURN(TRUE);
      }
      if (no_parts_coalesced == 0)
      {
        my_error(ER_COALESCE_PARTITION_NO_PARTITION, MYF(0));
        DBUG_RETURN(TRUE);
      }
      if (no_parts_coalesced >= tab_part_info->no_parts)
      {
        my_error(ER_DROP_LAST_PARTITION, MYF(0));
        DBUG_RETURN(TRUE);
      }
/*
Online handling:
COALESCE PARTITION:
-------------------
The figure below shows the manner in which partitions are handled when
performing an on-line coalesce partition and which states they go through
at start, after adding and copying partitions and finally after dropping
the partitions to drop. The figure shows an example using four partitions
to start with, using linear hash and coalescing one partition (always the
last partition).

Using linear hash then all remaining partitions will have a new reorganised
part.

Existing partitions                     Coalesced partition 
------       ------              ------   |      ------
|    |       |    |              |    |   |      |    |
| p0 |       | p1 |              | p2 |   |      | p3 |
------       ------              ------   |      ------
PART_NORMAL  PART_CHANGED        PART_NORMAL     PART_REORGED_DROPPED
PART_NORMAL  PART_IS_CHANGED     PART_NORMAL     PART_TO_BE_DROPPED
PART_NORMAL  PART_NORMAL         PART_NORMAL     PART_IS_DROPPED

Reorganised existing partitions
            ------
            |    |
            | p1'|
            ------

p0 - p3 is in the partitions list.
The p1' partition will actually not be in any list it is deduced from the
state of p1.
*/
      {
        uint part_count= 0, start_part= 1, start_sec_part= 1;
        uint end_part= 0, end_sec_part= 0;
        bool all_parts= TRUE;
        if (*fast_alter_partition &&
            tab_part_info->linear_hash_ind)
        {
          uint upper_2n= tab_part_info->linear_hash_mask + 1;
          uint lower_2n= upper_2n >> 1;
          all_parts= FALSE;
          if (no_parts_coalesced >= lower_2n)
          {
            all_parts= TRUE;
          }
          else if (no_parts_remain >= lower_2n)
          {
            end_part= tab_part_info->no_parts - (lower_2n + 1);
            start_part= no_parts_remain - lower_2n;
          }
          else
          {
            start_part= 0;
            end_part= tab_part_info->no_parts - (lower_2n + 1);
            end_sec_part= (lower_2n >> 1) - 1;
            start_sec_part= end_sec_part - (lower_2n - (no_parts_remain + 1));
          }
        }
        do
        {
          partition_element *p_elem= part_it++;
          if (*fast_alter_partition &&
              (all_parts ||
              (part_count >= start_part && part_count <= end_part) ||
              (part_count >= start_sec_part && part_count <= end_sec_part)))
            p_elem->part_state= PART_CHANGED;
          if (++part_count > no_parts_remain)
          {
            if (*fast_alter_partition)
              p_elem->part_state= PART_REORGED_DROPPED;
            else
              part_it.remove();
          }
        } while (part_count < tab_part_info->no_parts);
        tab_part_info->no_parts= no_parts_remain;
      }
      if (!(alter_info->flags & ALTER_TABLE_REORG))
        tab_part_info->use_default_no_partitions= FALSE;
    }
    else if (alter_info->flags == ALTER_REORGANIZE_PARTITION)
    {
      /*
        Reorganise partitions takes a number of partitions that are next
        to each other (at least for RANGE PARTITIONS) and then uses those
        to create a set of new partitions. So data is copied from those
        partitions into the new set of partitions. Those new partitions
        can have more values in the LIST value specifications or less both
        are allowed. The ranges can be different but since they are 
        changing a set of consecutive partitions they must cover the same
        range as those changed from.
        This command can be used on RANGE and LIST partitions.
      */
      uint no_parts_reorged= alter_info->partition_names.elements;
      uint no_parts_new= thd->lex->part_info->partitions.elements;
      partition_info *alt_part_info= thd->lex->part_info;
      uint check_total_partitions;
      if (no_parts_reorged > tab_part_info->no_parts)
      {
        my_error(ER_REORG_PARTITION_NOT_EXIST, MYF(0));
        DBUG_RETURN(TRUE);
      }
      if (!(tab_part_info->part_type == RANGE_PARTITION ||
            tab_part_info->part_type == LIST_PARTITION) &&
           (no_parts_new != no_parts_reorged))
      {
        my_error(ER_REORG_HASH_ONLY_ON_SAME_NO, MYF(0));
        DBUG_RETURN(TRUE);
      }
      check_total_partitions= tab_part_info->no_parts + no_parts_new;
      check_total_partitions-= no_parts_reorged;
      if (check_total_partitions > MAX_PARTITIONS)
      {
        my_error(ER_TOO_MANY_PARTITIONS_ERROR, MYF(0));
        DBUG_RETURN(TRUE);
      }
/*
Online handling:
REORGANIZE PARTITION:
---------------------
The figure exemplifies the handling of partitions, their state changes and
how they are organised. It exemplifies four partitions where two of the
partitions are reorganised (p1 and p2) into two new partitions (p4 and p5).
The reason of this change could be to change range limits, change list
values or for hash partitions simply reorganise the partition which could
also involve moving them to new disks or new node groups (MySQL Cluster).

Existing partitions                                  
------       ------        ------        ------
|    |       |    |        |    |        |    |
| p0 |       | p1 |        | p2 |        | p3 |
------       ------        ------        ------
PART_NORMAL  PART_TO_BE_REORGED          PART_NORMAL
PART_NORMAL  PART_TO_BE_DROPPED          PART_NORMAL
PART_NORMAL  PART_IS_DROPPED             PART_NORMAL

Reorganised new partitions (replacing p1 and p2)
------      ------
|    |      |    |
| p4 |      | p5 |
------      ------
PART_TO_BE_ADDED
PART_IS_ADDED
PART_IS_ADDED

All unchanged partitions and the new partitions are in the partitions list
in the order they will have when the change is completed. The reorganised
partitions are placed in the temp_partitions list. PART_IS_ADDED is only a
temporary state not written in the frm file. It is used to ensure we write
the generated partition syntax in a correct manner.
*/
      {
        List_iterator<partition_element> tab_it(tab_part_info->partitions);
        uint part_count= 0;
        bool found_first= FALSE;
        bool found_last= FALSE;
        bool is_last_partition_reorged;
        uint drop_count= 0;
        longlong tab_max_range= 0, alt_max_range= 0;
        do
        {
          partition_element *part_elem= tab_it++;
          is_last_partition_reorged= FALSE;
          if (is_name_in_list(part_elem->partition_name,
                              alter_info->partition_names))
          {
            is_last_partition_reorged= TRUE;
            drop_count++;
            tab_max_range= part_elem->range_value;
            if (*fast_alter_partition &&
                tab_part_info->temp_partitions.push_back(part_elem))
            {
              mem_alloc_error(1);
              DBUG_RETURN(TRUE);
            }
            if (*fast_alter_partition)
              part_elem->part_state= PART_TO_BE_REORGED;
            if (!found_first)
            {
              uint alt_part_count= 0;
              found_first= TRUE;
              List_iterator<partition_element>
                                 alt_it(alt_part_info->partitions);
              do
              {
                partition_element *alt_part_elem= alt_it++;
                alt_max_range= alt_part_elem->range_value;
                if (*fast_alter_partition)
                  alt_part_elem->part_state= PART_TO_BE_ADDED;
                if (alt_part_count == 0)
                  tab_it.replace(alt_part_elem);
                else
                  tab_it.after(alt_part_elem);
              } while (++alt_part_count < no_parts_new);
            }
            else if (found_last)
            {
              my_error(ER_CONSECUTIVE_REORG_PARTITIONS, MYF(0));
              DBUG_RETURN(TRUE);
            }
            else
              tab_it.remove();
          }
          else
          {
            if (found_first)
              found_last= TRUE;
          }
        } while (++part_count < tab_part_info->no_parts);
        if (drop_count != no_parts_reorged)
        {
          my_error(ER_DROP_PARTITION_NON_EXISTENT, MYF(0), "REORGANIZE");
          DBUG_RETURN(TRUE);
        }
        if (tab_part_info->part_type == RANGE_PARTITION &&
            ((is_last_partition_reorged &&
               alt_max_range < tab_max_range) ||
              (!is_last_partition_reorged &&
               alt_max_range != tab_max_range)))
        {
          /*
            For range partitioning the total resulting range before and
            after the change must be the same except in one case. This is
            when the last partition is reorganised, in this case it is
            acceptable to increase the total range.
            The reason is that it is not allowed to have "holes" in the
            middle of the ranges and thus we should not allow to reorganise
            to create "holes". Also we should not allow using REORGANIZE
            to drop data.
          */
          my_error(ER_REORG_OUTSIDE_RANGE, MYF(0));
          DBUG_RETURN(TRUE);
        }
        tab_part_info->no_parts= check_total_partitions;
      }
    }
    else
    {
      DBUG_ASSERT(FALSE);
    }
    *partition_changed= TRUE;
    thd->lex->part_info= tab_part_info;
    if (alter_info->flags == ALTER_ADD_PARTITION ||
        alter_info->flags == ALTER_REORGANIZE_PARTITION)
    {
4777
      if (tab_part_info->use_default_subpartitions &&
4778 4779 4780 4781 4782
          !alt_part_info->use_default_subpartitions)
      {
        tab_part_info->use_default_subpartitions= FALSE;
        tab_part_info->use_default_no_subpartitions= FALSE;
      }
4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869
      if (check_partition_info(tab_part_info, (handlerton**)NULL,
                               table->file, ULL(0)))
      {
        DBUG_RETURN(TRUE);
      }
    }
  }
  else
  {
    /*
     When thd->lex->part_info has a reference to a partition_info the
     ALTER TABLE contained a definition of a partitioning.

     Case I:
       If there was a partition before and there is a new one defined.
       We use the new partitioning. The new partitioning is already
       defined in the correct variable so no work is needed to
       accomplish this.
       We do however need to update partition_changed to ensure that not
       only the frm file is changed in the ALTER TABLE command.

     Case IIa:
       There was a partitioning before and there is no new one defined.
       Also the user has not specified an explicit engine to use.

       We use the old partitioning also for the new table. We do this
       by assigning the partition_info from the table loaded in
       open_ltable to the partition_info struct used by mysql_create_table
       later in this method.

     Case IIb:
       There was a partitioning before and there is no new one defined.
       The user has specified an explicit engine to use.

       Since the user has specified an explicit engine to use we override
       the old partitioning info and create a new table using the specified
       engine. This is the reason for the extra check if old and new engine
       is equal.
       In this case the partition also is changed.

     Case III:
       There was no partitioning before altering the table, there is
       partitioning defined in the altered table. Use the new partitioning.
       No work needed since the partitioning info is already in the
       correct variable.

       In this case we discover one case where the new partitioning is using
       the same partition function as the default (PARTITION BY KEY or
       PARTITION BY LINEAR KEY with the list of fields equal to the primary
       key fields OR PARTITION BY [LINEAR] KEY() for tables without primary
       key)
       Also here partition has changed and thus a new table must be
       created.

     Case IV:
       There was no partitioning before and no partitioning defined.
       Obviously no work needed.
    */
    if (table->part_info)
    {
      if (!thd->lex->part_info &&
          create_info->db_type == old_db_type)
        thd->lex->part_info= table->part_info;
    }
    if (thd->lex->part_info)
    {
      /*
        Need to cater for engine types that can handle partition without
        using the partition handler.
      */
      if (thd->lex->part_info != table->part_info)
        *partition_changed= TRUE;
      if (create_info->db_type == &partition_hton)
      {
        if (table->part_info)
        {
          thd->lex->part_info->default_engine_type=
                               table->part_info->default_engine_type;
        }
        else
        {
          thd->lex->part_info->default_engine_type= 
                           ha_checktype(thd, DB_TYPE_DEFAULT, FALSE, FALSE);
        }
      }
      else
      {
4870
        bool is_native_partitioned= FALSE;
4871 4872 4873 4874 4875 4876 4877 4878 4879
        partition_info *part_info= thd->lex->part_info;
        part_info->default_engine_type= create_info->db_type;
        if (check_native_partitioned(create_info, &is_native_partitioned,
                                     part_info, thd))
        {
          DBUG_RETURN(TRUE);
        }
        if (!is_native_partitioned)
        {
4880
          DBUG_ASSERT(create_info->db_type != &default_hton);
4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408
          create_info->db_type= &partition_hton;
        }
      }
      DBUG_PRINT("info", ("default_db_type = %s",
                 thd->lex->part_info->default_engine_type->name));
    }
  }
  DBUG_RETURN(FALSE);
}


/*
  Change partitions, used to implement ALTER TABLE ADD/REORGANIZE/COALESCE
  partitions. This method is used to implement both single-phase and multi-
  phase implementations of ADD/REORGANIZE/COALESCE partitions.

  SYNOPSIS
    mysql_change_partitions()
    lpt                        Struct containing parameters

  RETURN VALUES
    TRUE                          Failure
    FALSE                         Success

  DESCRIPTION
    Request handler to add partitions as set in states of the partition

    Elements of the lpt parameters used:
    create_info                Create information used to create partitions
    db                         Database name
    table_name                 Table name
    copied                     Output parameter where number of copied
                               records are added
    deleted                    Output parameter where number of deleted
                               records are added
*/

static bool mysql_change_partitions(ALTER_PARTITION_PARAM_TYPE *lpt)
{
  char path[FN_REFLEN+1];
  DBUG_ENTER("mysql_change_partitions");

  build_table_filename(path, sizeof(path), lpt->db, lpt->table_name, "");
  DBUG_RETURN(lpt->table->file->change_partitions(lpt->create_info, path,
                                                  &lpt->copied,
                                                  &lpt->deleted,
                                                  lpt->pack_frm_data,
                                                  lpt->pack_frm_len));
}


/*
  Rename partitions in an ALTER TABLE of partitions

  SYNOPSIS
    mysql_rename_partitions()
    lpt                        Struct containing parameters

  RETURN VALUES
    TRUE                          Failure
    FALSE                         Success

  DESCRIPTION
    Request handler to rename partitions as set in states of the partition

    Parameters used:
    db                         Database name
    table_name                 Table name
*/

static bool mysql_rename_partitions(ALTER_PARTITION_PARAM_TYPE *lpt)
{
  char path[FN_REFLEN+1];
  DBUG_ENTER("mysql_rename_partitions");

  build_table_filename(path, sizeof(path), lpt->db, lpt->table_name, "");
  DBUG_RETURN(lpt->table->file->rename_partitions(path));
}


/*
  Drop partitions in an ALTER TABLE of partitions

  SYNOPSIS
    mysql_drop_partitions()
    lpt                        Struct containing parameters

  RETURN VALUES
    TRUE                          Failure
    FALSE                         Success
  DESCRIPTION
    Drop the partitions marked with PART_TO_BE_DROPPED state and remove
    those partitions from the list.

    Parameters used:
    table                       Table object
    db                          Database name
    table_name                  Table name
*/

static bool mysql_drop_partitions(ALTER_PARTITION_PARAM_TYPE *lpt)
{
  char path[FN_REFLEN+1];
  partition_info *part_info= lpt->table->part_info;
  List_iterator<partition_element> part_it(part_info->partitions);
  uint i= 0;
  uint remove_count= 0;
  DBUG_ENTER("mysql_drop_partitions");

  build_table_filename(path, sizeof(path), lpt->db, lpt->table_name, "");
  if (lpt->table->file->drop_partitions(path))
  {
    DBUG_RETURN(TRUE);
  }
  do
  {
    partition_element *part_elem= part_it++;
    if (part_elem->part_state == PART_IS_DROPPED)
    {
      part_it.remove();
      remove_count++;
    }
  } while (++i < part_info->no_parts);
  part_info->no_parts-= remove_count;
  DBUG_RETURN(FALSE);
}


/*
  Actually perform the change requested by ALTER TABLE of partitions
  previously prepared.

  SYNOPSIS
    fast_alter_partition_table()
    thd                           Thread object
    table                         Table object
    alter_info                    ALTER TABLE info
    create_info                   Create info for CREATE TABLE
    table_list                    List of the table involved
    create_list                   The fields in the resulting table
    key_list                      The keys in the resulting table
    db                            Database name of new table
    table_name                    Table name of new table

  RETURN VALUES
    TRUE                          Error
    FALSE                         Success

  DESCRIPTION
    Perform all ALTER TABLE operations for partitioned tables that can be
    performed fast without a full copy of the original table.
*/

uint fast_alter_partition_table(THD *thd, TABLE *table,
                                ALTER_INFO *alter_info,
                                HA_CREATE_INFO *create_info,
                                TABLE_LIST *table_list,
                                List<create_field> *create_list,
                                List<Key> *key_list, const char *db,
                                const char *table_name,
                                uint fast_alter_partition)
{
  /* Set-up struct used to write frm files */
  ulonglong copied= 0;
  ulonglong deleted= 0;
  partition_info *part_info= table->part_info;
  ALTER_PARTITION_PARAM_TYPE lpt_obj;
  ALTER_PARTITION_PARAM_TYPE *lpt= &lpt_obj;
  bool written_bin_log= TRUE;
  DBUG_ENTER("fast_alter_partition_table");

  lpt->thd= thd;
  lpt->create_info= create_info;
  lpt->create_list= create_list;
  lpt->key_list= key_list;
  lpt->db_options= create_info->table_options;
  if (create_info->row_type == ROW_TYPE_DYNAMIC)
    lpt->db_options|= HA_OPTION_PACK_RECORD;
  lpt->table= table;
  lpt->key_info_buffer= 0;
  lpt->key_count= 0;
  lpt->db= db;
  lpt->table_name= table_name;
  lpt->copied= 0;
  lpt->deleted= 0;
  lpt->pack_frm_data= NULL;
  lpt->pack_frm_len= 0;
  thd->lex->part_info= part_info;

  if (alter_info->flags & ALTER_OPTIMIZE_PARTITION ||
      alter_info->flags & ALTER_ANALYZE_PARTITION ||
      alter_info->flags & ALTER_CHECK_PARTITION ||
      alter_info->flags & ALTER_REPAIR_PARTITION)
  {
    /*
      In this case the user has specified that he wants a set of partitions
      to be optimised and the partition engine can handle optimising
      partitions natively without requiring a full rebuild of the
      partitions.

      In this case it is enough to call optimise_partitions, there is no
      need to change frm files or anything else.
    */
    written_bin_log= FALSE;
    if (((alter_info->flags & ALTER_OPTIMIZE_PARTITION) &&
         (table->file->optimize_partitions(thd))) ||
        ((alter_info->flags & ALTER_ANALYZE_PARTITION) &&
         (table->file->analyze_partitions(thd))) ||
        ((alter_info->flags & ALTER_CHECK_PARTITION) &&
         (table->file->check_partitions(thd))) ||
        ((alter_info->flags & ALTER_REPAIR_PARTITION) &&
         (table->file->repair_partitions(thd))))
    {
      fast_alter_partition_error_handler(lpt);
      DBUG_RETURN(TRUE);
    }
  }
  else if (fast_alter_partition & HA_PARTITION_ONE_PHASE)
  {
    /*
      In the case where the engine supports one phase online partition
      changes it is not necessary to have any exclusive locks. The
      correctness is upheld instead by transactions being aborted if they
      access the table after its partition definition has changed (if they
      are still using the old partition definition).

      The handler is in this case responsible to ensure that all users
      start using the new frm file after it has changed. To implement
      one phase it is necessary for the handler to have the master copy
      of the frm file and use discovery mechanisms to renew it. Thus
      write frm will write the frm, pack the new frm and finally
      the frm is deleted and the discovery mechanisms will either restore
      back to the old or installing the new after the change is activated.

      Thus all open tables will be discovered that they are old, if not
      earlier as soon as they try an operation using the old table. One
      should ensure that this is checked already when opening a table,
      even if it is found in the cache of open tables.

      change_partitions will perform all operations and it is the duty of
      the handler to ensure that the frm files in the system gets updated
      in synch with the changes made and if an error occurs that a proper
      error handling is done.

      If the MySQL Server crashes at this moment but the handler succeeds
      in performing the change then the binlog is not written for the
      change. There is no way to solve this as long as the binlog is not
      transactional and even then it is hard to solve it completely.
 
      The first approach here was to downgrade locks. Now a different approach
      is decided upon. The idea is that the handler will have access to the
      ALTER_INFO when store_lock arrives with TL_WRITE_ALLOW_READ. So if the
      handler knows that this functionality can be handled with a lower lock
      level it will set the lock level to TL_WRITE_ALLOW_WRITE immediately.
      Thus the need to downgrade the lock disappears.
      1) Write the new frm, pack it and then delete it
      2) Perform the change within the handler
    */
    if ((mysql_write_frm(lpt, WFRM_INITIAL_WRITE | WFRM_PACK_FRM)) ||
        (mysql_change_partitions(lpt)))
    {
      fast_alter_partition_error_handler(lpt);
      DBUG_RETURN(TRUE);
    }
  }
  else if (alter_info->flags == ALTER_DROP_PARTITION)
  {
    /*
      Now after all checks and setting state on dropped partitions we can
      start the actual dropping of the partitions.

      Drop partition is actually two things happening. The first is that
      a lot of records are deleted. The second is that the behaviour of
      subsequent updates and writes and deletes will change. The delete
      part can be handled without any particular high lock level by
      transactional engines whereas non-transactional engines need to
      ensure that this change is done with an exclusive lock on the table.
      The second part, the change of partitioning does however require
      an exclusive lock to install the new partitioning as one atomic
      operation. If this is not the case, it is possible for two
      transactions to see the change in a different order than their
      serialisation order. Thus we need an exclusive lock for both
      transactional and non-transactional engines.

      For LIST partitions it could be possible to avoid the exclusive lock
      (and for RANGE partitions if they didn't rearrange range definitions
      after a DROP PARTITION) if one ensured that failed accesses to the
      dropped partitions was aborted for sure (thus only possible for
      transactional engines).
      
      1) Lock the table in TL_WRITE_ONLY to ensure all other accesses to
         the table have completed
      2) Write the new frm file where the partitions have changed but are
         still remaining with the state PART_TO_BE_DROPPED
      3) Write the bin log
      4) Prepare MyISAM handlers for drop of partitions
      5) Ensure that any users that has opened the table but not yet
         reached the abort lock do that before downgrading the lock.
      6) Drop the partitions
      7) Write the frm file that the partition has been dropped
      8) Wait until all accesses using the old frm file has completed
      9) Complete query
    */
    if ((abort_and_upgrade_lock(lpt)) ||
        (mysql_write_frm(lpt, WFRM_INITIAL_WRITE)) ||
        ((!thd->lex->no_write_to_binlog) &&
         (write_bin_log(thd, FALSE,
                       thd->query, thd->query_length), FALSE)) ||
        (table->file->extra(HA_EXTRA_PREPARE_FOR_DELETE)) ||
        (close_open_tables_and_downgrade(lpt), FALSE) || 
        (mysql_drop_partitions(lpt)) ||
        (mysql_write_frm(lpt, WFRM_CREATE_HANDLER_FILES)) ||
        (mysql_wait_completed_table(lpt, table), FALSE))
    {
      fast_alter_partition_error_handler(lpt);
      DBUG_RETURN(TRUE);
    }
  }
  else if ((alter_info->flags & ALTER_ADD_PARTITION) &&
           (part_info->part_type == RANGE_PARTITION ||
            part_info->part_type == LIST_PARTITION))
  {
    /*
      ADD RANGE/LIST PARTITIONS
      In this case there are no tuples removed and no tuples are added.
      Thus the operation is merely adding a new partition. Thus it is
      necessary to perform the change as an atomic operation. Otherwise
      someone reading without seeing the new partition could potentially
      miss updates made by a transaction serialised before it that are
      inserted into the new partition.

      1) Write the new frm file where state of added partitions is
         changed to PART_TO_BE_ADDED
      2) Add the new partitions
      3) Lock all partitions in TL_WRITE_ONLY to ensure that no users
         are still using the old partitioning scheme. Wait until all
         ongoing users have completed before progressing.
      4) Write a new frm file of the table where the partitions are added
         to the table.
      5) Write binlog
      6) Wait until all accesses using the old frm file has completed
      7) Complete query
    */
    if ((mysql_write_frm(lpt, WFRM_INITIAL_WRITE)) ||
        (mysql_change_partitions(lpt)) ||
        (abort_and_upgrade_lock(lpt)) ||
        (mysql_write_frm(lpt, WFRM_CREATE_HANDLER_FILES)) ||
        ((!thd->lex->no_write_to_binlog) &&
         (write_bin_log(thd, FALSE,
                        thd->query, thd->query_length), FALSE)) ||
        (close_open_tables_and_downgrade(lpt), FALSE))
    {
      fast_alter_partition_error_handler(lpt);
      DBUG_RETURN(TRUE);
    }
  }
  else
  {
    /*
      ADD HASH PARTITION/
      COALESCE PARTITION/
      REBUILD PARTITION/
      REORGANIZE PARTITION
 
      In this case all records are still around after the change although
      possibly organised into new partitions, thus by ensuring that all
      updates go to both the old and the new partitioning scheme we can
      actually perform this operation lock-free. The only exception to
      this is when REORGANIZE PARTITION adds/drops ranges. In this case
      there needs to be an exclusive lock during the time when the range
      changes occur.
      This is only possible if the handler can ensure double-write for a
      period. The double write will ensure that it doesn't matter where the
      data is read from since both places are updated for writes. If such
      double writing is not performed then it is necessary to perform the
      change with the usual exclusive lock. With double writes it is even
      possible to perform writes in parallel with the reorganisation of
      partitions.

      Without double write procedure we get the following procedure.
      The only difference with using double write is that we can downgrade
      the lock to TL_WRITE_ALLOW_WRITE. Double write in this case only
      double writes from old to new. If we had double writing in both
      directions we could perform the change completely without exclusive
      lock for HASH partitions.
      Handlers that perform double writing during the copy phase can actually
      use a lower lock level. This can be handled inside store_lock in the
      respective handler.

      1) Write the new frm file where state of added partitions is
         changed to PART_TO_BE_ADDED and the reorganised partitions
         are set in state PART_TO_BE_REORGED.
      2) Add the new partitions
         Copy from the reorganised partitions to the new partitions
      3) Lock all partitions in TL_WRITE_ONLY to ensure that no users
         are still using the old partitioning scheme. Wait until all
         ongoing users have completed before progressing.
      4) Prepare MyISAM handlers for rename and delete of partitions
      5) Write a new frm file of the table where the partitions are
         reorganised.
      6) Rename the reorged partitions such that they are no longer
         used and rename those added to their real new names.
      7) Write bin log
      8) Wait until all accesses using the old frm file has completed
      9) Drop the reorganised partitions
      10)Write a new frm file of the table where the partitions are
         reorganised.
      11)Wait until all accesses using the old frm file has completed
      12)Complete query
    */

    if ((mysql_write_frm(lpt, WFRM_INITIAL_WRITE)) ||
        (mysql_change_partitions(lpt)) ||
        (abort_and_upgrade_lock(lpt)) ||
        (mysql_write_frm(lpt, WFRM_CREATE_HANDLER_FILES)) ||
        (table->file->extra(HA_EXTRA_PREPARE_FOR_DELETE)) ||
        (mysql_rename_partitions(lpt)) ||
        ((!thd->lex->no_write_to_binlog) &&
         (write_bin_log(thd, FALSE,
                        thd->query, thd->query_length), FALSE)) ||
        (close_open_tables_and_downgrade(lpt), FALSE) ||
        (mysql_drop_partitions(lpt)) ||
        (mysql_write_frm(lpt, 0UL)) ||
        (mysql_wait_completed_table(lpt, table), FALSE))
    {
        fast_alter_partition_error_handler(lpt);
        DBUG_RETURN(TRUE);
    }
  }
  /*
    A final step is to write the query to the binlog and send ok to the
    user
  */
  DBUG_RETURN(fast_end_partition(thd, lpt->copied, lpt->deleted,
                                 table_list, FALSE, lpt,
                                 written_bin_log));
}
#endif


/*
  Prepare for calling val_int on partition function by setting fields to
  point to the record where the values of the PF-fields are stored.

  SYNOPSIS
    set_field_ptr()
    ptr                 Array of fields to change ptr
    new_buf             New record pointer
    old_buf             Old record pointer

  DESCRIPTION
    Set ptr in field objects of field array to refer to new_buf record
    instead of previously old_buf. Used before calling val_int and after
    it is used to restore pointers to table->record[0].
    This routine is placed outside of partition code since it can be useful
    also for other programs.
*/

void set_field_ptr(Field **ptr, const byte *new_buf,
                   const byte *old_buf)
{
  my_ptrdiff_t diff= (new_buf - old_buf);
  DBUG_ENTER("set_field_ptr");

  do
  {
    (*ptr)->move_field_offset(diff);
  } while (*(++ptr));
  DBUG_VOID_RETURN;
}


/*
  Prepare for calling val_int on partition function by setting fields to
  point to the record where the values of the PF-fields are stored.
  This variant works on a key_part reference.
  It is not required that all fields are NOT NULL fields.

  SYNOPSIS
    set_key_field_ptr()
    key_info            key info with a set of fields to change ptr
    new_buf             New record pointer
    old_buf             Old record pointer

  DESCRIPTION
    Set ptr in field objects of field array to refer to new_buf record
    instead of previously old_buf. Used before calling val_int and after
    it is used to restore pointers to table->record[0].
    This routine is placed outside of partition code since it can be useful
    also for other programs.
*/

void set_key_field_ptr(KEY *key_info, const byte *new_buf,
                       const byte *old_buf)
{
  KEY_PART_INFO *key_part= key_info->key_part;
  uint key_parts= key_info->key_parts;
  uint i= 0;
  my_ptrdiff_t diff= (new_buf - old_buf);
  DBUG_ENTER("set_key_field_ptr");

  do
  {
    key_part->field->move_field_offset(diff);
    key_part++;
  } while (++i < key_parts);
  DBUG_VOID_RETURN;
}


/*
  SYNOPSIS
    mem_alloc_error()
    size                Size of memory attempted to allocate
    None

  RETURN VALUES
    None

  DESCRIPTION
    A routine to use for all the many places in the code where memory
    allocation error can happen, a tremendous amount of them, needs
    simple routine that signals this error.
*/

void mem_alloc_error(size_t size)
{
  my_error(ER_OUTOFMEMORY, MYF(0), size);
5409
}
5410

5411
#ifdef WITH_PARTITION_STORAGE_ENGINE
5412
/*
5413 5414
  Return comma-separated list of used partitions in the provided given string

5415 5416 5417 5418
  SYNOPSIS
    make_used_partitions_str()
      part_info  IN  Partitioning info
      parts_str  OUT The string to fill
5419 5420 5421 5422 5423 5424 5425

  DESCRIPTION
    Generate a list of used partitions (from bits in part_info->used_partitions
    bitmap), asd store it into the provided String object.
    
  NOTE
    The produced string must not be longer then MAX_PARTITIONS * (1 + FN_LEN).
5426 5427 5428 5429 5430 5431 5432 5433 5434
*/

void make_used_partitions_str(partition_info *part_info, String *parts_str)
{
  parts_str->length(0);
  partition_element *pe;
  uint partition_id= 0;
  List_iterator<partition_element> it(part_info->partitions);
  
5435
  if (part_info->is_sub_partitioned())
5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473
  {
    partition_element *head_pe;
    while ((head_pe= it++))
    {
      List_iterator<partition_element> it2(head_pe->subpartitions);
      while ((pe= it2++))
      {
        if (bitmap_is_set(&part_info->used_partitions, partition_id))
        {
          if (parts_str->length())
            parts_str->append(',');
          parts_str->append(head_pe->partition_name,
                           strlen(head_pe->partition_name),
                           system_charset_info);
          parts_str->append('_');
          parts_str->append(pe->partition_name,
                           strlen(pe->partition_name),
                           system_charset_info);
        }
        partition_id++;
      }
    }
  }
  else
  {
    while ((pe= it++))
    {
      if (bitmap_is_set(&part_info->used_partitions, partition_id))
      {
        if (parts_str->length())
          parts_str->append(',');
        parts_str->append(pe->partition_name, strlen(pe->partition_name),
                         system_charset_info);
      }
      partition_id++;
    }
  }
}
5474
#endif
5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511

/****************************************************************************
 * Partition interval analysis support
 ***************************************************************************/

/*
  Setup partition_info::* members related to partitioning range analysis

  SYNOPSIS
    set_up_partition_func_pointers()
      part_info  Partitioning info structure

  DESCRIPTION
    Assuming that passed partition_info structure already has correct values
    for members that specify [sub]partitioning type, table fields, and
    functions, set up partition_info::* members that are related to
    Partitioning Interval Analysis (see get_partitions_in_range_iter for its
    definition)

  IMPLEMENTATION
    There are two available interval analyzer functions:
    (1) get_part_iter_for_interval_via_mapping 
    (2) get_part_iter_for_interval_via_walking

    They both have limited applicability:
    (1) is applicable for "PARTITION BY <RANGE|LIST>(func(t.field))", where
    func is a monotonic function.
    
    (2) is applicable for 
      "[SUB]PARTITION BY <any-partitioning-type>(any_func(t.integer_field))"
      
    If both are applicable, (1) is preferred over (2).
    
    This function sets part_info::get_part_iter_for_interval according to
    this criteria, and also sets some auxilary fields that the function
    uses.
*/
5512
#ifdef WITH_PARTITION_STORAGE_ENGINE
5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602
static void set_up_range_analysis_info(partition_info *part_info)
{
  enum_monotonicity_info minfo;

  /* Set the catch-all default */
  part_info->get_part_iter_for_interval= NULL;
  part_info->get_subpart_iter_for_interval= NULL;

  /* 
    Check if get_part_iter_for_interval_via_mapping() can be used for 
    partitioning
  */
  switch (part_info->part_type) {
  case RANGE_PARTITION:
  case LIST_PARTITION:
    minfo= part_info->part_expr->get_monotonicity_info();
    if (minfo != NON_MONOTONIC)
    {
      part_info->range_analysis_include_bounds=
        test(minfo == MONOTONIC_INCREASING);
      part_info->get_part_iter_for_interval=
        get_part_iter_for_interval_via_mapping;
      goto setup_subparts;
    }
  default:
    ;
  }
   
  /*
    Check get_part_iter_for_interval_via_walking() can be used for
    partitioning
  */
  if (part_info->no_part_fields == 1)
  {
    Field *field= part_info->part_field_array[0];
    switch (field->type()) {
    case MYSQL_TYPE_TINY:
    case MYSQL_TYPE_SHORT:
    case MYSQL_TYPE_LONG:
    case MYSQL_TYPE_LONGLONG:
      part_info->get_part_iter_for_interval=
        get_part_iter_for_interval_via_walking;
      break;
    default:
      ;
    }
  }

setup_subparts:
  /*
    Check get_part_iter_for_interval_via_walking() can be used for
    subpartitioning
  */
  if (part_info->no_subpart_fields == 1)
  {
    Field *field= part_info->subpart_field_array[0];
    switch (field->type()) {
    case MYSQL_TYPE_TINY:
    case MYSQL_TYPE_SHORT:
    case MYSQL_TYPE_LONG:
    case MYSQL_TYPE_LONGLONG:
      part_info->get_subpart_iter_for_interval=
        get_part_iter_for_interval_via_walking;
      break;
    default:
      ;
    }
  }
}


typedef uint32 (*get_endpoint_func)(partition_info*, bool left_endpoint,
                                    bool include_endpoint);

/*
  Partitioning Interval Analysis: Initialize the iterator for "mapping" case

  SYNOPSIS
    get_part_iter_for_interval_via_mapping()
      part_info   Partition info
      is_subpart  TRUE  - act for subpartitioning
                  FALSE - act for partitioning
      min_value   minimum field value, in opt_range key format.
      max_value   minimum field value, in opt_range key format.
      flags       Some combination of NEAR_MIN, NEAR_MAX, NO_MIN_RANGE,
                  NO_MAX_RANGE.
      part_iter   Iterator structure to be initialized

  DESCRIPTION
    Initialize partition set iterator to walk over the interval in
5603 5604
    ordered-array-of-partitions (for RANGE partitioning) or 
    ordered-array-of-list-constants (for LIST partitioning) space.
5605 5606

  IMPLEMENTATION
5607
    This function is used when partitioning is done by
5608 5609 5610 5611 5612 5613 5614 5615
    <RANGE|LIST>(ascending_func(t.field)), and we can map an interval in
    t.field space into a sub-array of partition_info::range_int_array or
    partition_info::list_array (see get_partition_id_range_for_endpoint,
    get_list_array_idx_for_endpoint for details).
    
    The function performs this interval mapping, and sets the iterator to
    traverse the sub-array and return appropriate partitions.
    
5616
  RETURN
5617 5618 5619 5620 5621 5622 5623
    0 - No matching partitions (iterator not initialized)
    1 - Ok, iterator intialized for traversal of matching partitions.
   -1 - All partitions would match (iterator not initialized)
*/

int get_part_iter_for_interval_via_mapping(partition_info *part_info,
                                           bool is_subpart,
5624
                                           char *min_value, char *max_value,
5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651
                                           uint flags,
                                           PARTITION_ITERATOR *part_iter)
{
  DBUG_ASSERT(!is_subpart);
  Field *field= part_info->part_field_array[0];
  uint32             max_endpoint_val;
  get_endpoint_func  get_endpoint;
  uint field_len= field->pack_length_in_rec();

  if (part_info->part_type == RANGE_PARTITION)
  {
    get_endpoint=        get_partition_id_range_for_endpoint;
    max_endpoint_val=    part_info->no_parts;
    part_iter->get_next= get_next_partition_id_range;
  }
  else if (part_info->part_type == LIST_PARTITION)
  {
    get_endpoint=        get_list_array_idx_for_endpoint;
    max_endpoint_val=    part_info->no_list_values;
    part_iter->get_next= get_next_partition_id_list;
    part_iter->part_info= part_info;
  }
  else
    DBUG_ASSERT(0);

  /* Find minimum */
  if (flags & NO_MIN_RANGE)
5652
    part_iter->part_nums.start= 0;
5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663
  else
  {
    /*
      Store the interval edge in the record buffer, and call the
      function that maps the edge in table-field space to an edge
      in ordered-set-of-partitions (for RANGE partitioning) or 
      index-in-ordered-array-of-list-constants (for LIST) space.
    */
    store_key_image_to_rec(field, min_value, field_len);
    bool include_endp= part_info->range_analysis_include_bounds ||
                       !test(flags & NEAR_MIN);
5664 5665
    part_iter->part_nums.start= get_endpoint(part_info, 1, include_endp);
    if (part_iter->part_nums.start == max_endpoint_val)
5666 5667 5668 5669 5670
      return 0; /* No partitions */
  }

  /* Find maximum, do the same as above but for right interval bound */
  if (flags & NO_MAX_RANGE)
5671
    part_iter->part_nums.end= max_endpoint_val;
5672 5673 5674 5675 5676
  else
  {
    store_key_image_to_rec(field, max_value, field_len);
    bool include_endp= part_info->range_analysis_include_bounds ||
                       !test(flags & NEAR_MAX);
5677 5678
    part_iter->part_nums.end= get_endpoint(part_info, 0, include_endp);
    if (part_iter->part_nums.start== part_iter->part_nums.end)
5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689
      return 0; /* No partitions */
  }
  return 1; /* Ok, iterator initialized */
}


/* See get_part_iter_for_interval_via_walking for definition of what this is */
#define MAX_RANGE_TO_WALK 10


/*
5690
  Partitioning Interval Analysis: Initialize iterator to walk field interval
5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705

  SYNOPSIS
    get_part_iter_for_interval_via_walking()
      part_info   Partition info
      is_subpart  TRUE  - act for subpartitioning
                  FALSE - act for partitioning
      min_value   minimum field value, in opt_range key format.
      max_value   minimum field value, in opt_range key format.
      flags       Some combination of NEAR_MIN, NEAR_MAX, NO_MIN_RANGE,
                  NO_MAX_RANGE.
      part_iter   Iterator structure to be initialized

  DESCRIPTION
    Initialize partition set iterator to walk over interval in integer field
    space. That is, for "const1 <=? t.field <=? const2" interval, initialize 
5706 5707
    the iterator to return a set of [sub]partitions obtained with the
    following procedure:
5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720
      get partition id for t.field = const1,   return it
      get partition id for t.field = const1+1, return it
       ...                 t.field = const1+2, ...
       ...                           ...       ...
       ...                 t.field = const2    ...

  IMPLEMENTATION
    See get_partitions_in_range_iter for general description of interval
    analysis. We support walking over the following intervals: 
      "t.field IS NULL" 
      "c1 <=? t.field <=? c2", where c1 and c2 are finite. 
    Intervals with +inf/-inf, and [NULL, c1] interval can be processed but
    that is more tricky and I don't have time to do it right now.
5721

5722 5723 5724 5725 5726 5727 5728 5729
    Additionally we have these requirements:
    * number of values in the interval must be less then number of
      [sub]partitions, and 
    * Number of values in the interval must be less then MAX_RANGE_TO_WALK.
    
    The rationale behind these requirements is that if they are not met
    we're likely to hit most of the partitions and traversing the interval
    will only add overhead. So it's better return "all partitions used" in
5730
    that case.
5731 5732 5733 5734 5735 5736 5737 5738 5739

  RETURN
    0 - No matching partitions, iterator not initialized
    1 - Some partitions would match, iterator intialized for traversing them
   -1 - All partitions would match, iterator not initialized
*/

int get_part_iter_for_interval_via_walking(partition_info *part_info,
                                           bool is_subpart,
5740
                                           char *min_value, char *max_value,
5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778
                                           uint flags,
                                           PARTITION_ITERATOR *part_iter)
{
  Field *field;
  uint total_parts;
  partition_iter_func get_next_func;
  if (is_subpart)
  {
    field= part_info->subpart_field_array[0];
    total_parts= part_info->no_subparts;
    get_next_func=  get_next_subpartition_via_walking;
  }
  else
  {
    field= part_info->part_field_array[0];
    total_parts= part_info->no_parts;
    get_next_func=  get_next_partition_via_walking;
  }

  /* Handle the "t.field IS NULL" interval, it is a special case */
  if (field->real_maybe_null() && !(flags & (NO_MIN_RANGE | NO_MAX_RANGE)) &&
      *min_value && *max_value)
  {
    /* 
      We don't have a part_iter->get_next() function that would find which
      partition "t.field IS NULL" belongs to, so find partition that contains 
      NULL right here, and return an iterator over singleton set.
    */
    uint32 part_id;
    field->set_null();
    if (is_subpart)
    {
      part_id= part_info->get_subpartition_id(part_info);
      init_single_partition_iterator(part_id, part_iter);
      return 1; /* Ok, iterator initialized */
    }
    else
    {
sergefp@mysql.com's avatar
sergefp@mysql.com committed
5779 5780
      longlong dummy;
      if (!part_info->get_partition_id(part_info, &part_id, &dummy))
5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807
      {
        init_single_partition_iterator(part_id, part_iter);
        return 1; /* Ok, iterator initialized */
      }
    }
    return 0; /* No partitions match */
  }

  if (flags & (NO_MIN_RANGE | NO_MAX_RANGE))
    return -1; /* Can't handle this interval, have to use all partitions */
  
  /* Get integers for left and right interval bound */
  longlong a, b;
  uint len= field->pack_length_in_rec();
  store_key_image_to_rec(field, min_value, len);
  a= field->val_int();
  
  store_key_image_to_rec(field, max_value, len);
  b= field->val_int();

  a += test(flags & NEAR_MIN);
  b += test(!(flags & NEAR_MAX));
  uint n_values= b - a;
  
  if (n_values > total_parts || n_values > MAX_RANGE_TO_WALK)
    return -1;

5808 5809
  part_iter->field_vals.start= a;
  part_iter->field_vals.end=   b;
5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825 5826 5827 5828 5829 5830 5831 5832 5833
  part_iter->part_info= part_info;
  part_iter->get_next=  get_next_func;
  return 1;
}


/*
  PARTITION_ITERATOR::get_next implementation: enumerate partitions in range

  SYNOPSIS
    get_next_partition_id_list()
      part_iter  Partition set iterator structure

  DESCRIPTION
    This is implementation of PARTITION_ITERATOR::get_next() that returns
    [sub]partition ids in [min_partition_id, max_partition_id] range.

  RETURN
    partition id
    NOT_A_PARTITION_ID if there are no more partitions
*/

uint32 get_next_partition_id_range(PARTITION_ITERATOR* part_iter)
{
5834
  if (part_iter->part_nums.start== part_iter->part_nums.end)
5835 5836
    return NOT_A_PARTITION_ID;
  else
5837
    return part_iter->part_nums.start++;
5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848
}


/*
  PARTITION_ITERATOR::get_next implementation for LIST partitioning

  SYNOPSIS
    get_next_partition_id_list()
      part_iter  Partition set iterator structure

  DESCRIPTION
5849
    This implementation of PARTITION_ITERATOR::get_next() is special for 
5850 5851 5852 5853 5854 5855 5856 5857 5858 5859
    LIST partitioning: it enumerates partition ids in 
    part_info->list_array[i] where i runs over [min_idx, max_idx] interval.

  RETURN 
    partition id
    NOT_A_PARTITION_ID if there are no more partitions
*/

uint32 get_next_partition_id_list(PARTITION_ITERATOR *part_iter)
{
5860
  if (part_iter->part_nums.start == part_iter->part_nums.end)
5861 5862 5863
    return NOT_A_PARTITION_ID;
  else
    return part_iter->part_info->list_array[part_iter->
5864
                                            part_nums.start++].partition_id;
5865 5866 5867 5868
}


/*
5869
  PARTITION_ITERATOR::get_next implementation: walk over field-space interval
5870 5871 5872 5873 5874 5875

  SYNOPSIS
    get_next_partition_via_walking()
      part_iter  Partitioning iterator

  DESCRIPTION
5876 5877 5878
    This implementation of PARTITION_ITERATOR::get_next() returns ids of
    partitions that contain records with partitioning field value within
    [start_val, end_val] interval.
5879 5880 5881 5882 5883 5884 5885 5886 5887 5888

  RETURN 
    partition id
    NOT_A_PARTITION_ID if there are no more partitioning.
*/

static uint32 get_next_partition_via_walking(PARTITION_ITERATOR *part_iter)
{
  uint32 part_id;
  Field *field= part_iter->part_info->part_field_array[0];
5889
  while (part_iter->field_vals.start != part_iter->field_vals.end)
5890
  {
5891 5892
    field->store(part_iter->field_vals.start, FALSE);
    part_iter->field_vals.start++;
sergefp@mysql.com's avatar
sergefp@mysql.com committed
5893
    longlong dummy;
5894
    if (part_iter->part_info->is_sub_partitioned() &&
5895 5896 5897
        !part_iter->part_info->get_part_partition_id(part_iter->part_info,
                                                     &part_id, &dummy) ||
        !part_iter->part_info->get_partition_id(part_iter->part_info,
sergefp@mysql.com's avatar
sergefp@mysql.com committed
5898
                                                &part_id, &dummy))
5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910
      return part_id;
  }
  return NOT_A_PARTITION_ID;
}


/* Same as get_next_partition_via_walking, but for subpartitions */

static uint32 get_next_subpartition_via_walking(PARTITION_ITERATOR *part_iter)
{
  uint32 part_id;
  Field *field= part_iter->part_info->subpart_field_array[0];
5911
  if (part_iter->field_vals.start == part_iter->field_vals.end)
5912
    return NOT_A_PARTITION_ID;
5913 5914
  field->store(part_iter->field_vals.start, FALSE);
  part_iter->field_vals.start++;
5915 5916
  return part_iter->part_info->get_subpartition_id(part_iter->part_info);
}
5917
#endif
5918