ha_archive.cc 29 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030
/* Copyright (C) 2003 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 */

#ifdef __GNUC__
#pragma implementation        // gcc: Class implementation
#endif

#include <mysql_priv.h>

#ifdef HAVE_ARCHIVE_DB
#include "ha_archive.h"
#include <my_dir.h>

/*
  First, if you want to understand storage engines you should look at 
  ha_example.cc and ha_example.h. 
  This example was written as a test case for a customer who needed
  a storage engine without indexes that could compress data very well.
  So, welcome to a completely compressed storage engine. This storage
  engine only does inserts. No replace, deletes, or updates. All reads are 
  complete table scans. Compression is done through gzip (bzip compresses
  better, but only marginally, if someone asks I could add support for
  it too, but beaware that it costs a lot more in CPU time then gzip).
  
  We keep a file pointer open for each instance of ha_archive for each read
  but for writes we keep one open file handle just for that. We flush it
  only if we have a read occur. gzip handles compressing lots of records
  at once much better then doing lots of little records between writes.
  It is possible to not lock on writes but this would then mean we couldn't
  handle bulk inserts as well (that is if someone was trying to read at
  the same time since we would want to flush).

  A "meta" file is kept alongside the data file. This file serves two purpose.
  The first purpose is to track the number of rows in the table. The second 
  purpose is to determine if the table was closed properly or not. When the 
  meta file is first opened it is marked as dirty. It is opened when the table 
  itself is opened for writing. When the table is closed the new count for rows 
  is written to the meta file and the file is marked as clean. If the meta file 
  is opened and it is marked as dirty, it is assumed that a crash occured. At 
  this point an error occurs and the user is told to rebuild the file.
  A rebuild scans the rows and rewrites the meta file. If corruption is found
  in the data file then the meta file is not repaired.

  At some point a recovery method for such a drastic case needs to be divised.

  Locks are row level, and you will get a consistant read. 

  For performance as far as table scans go it is quite fast. I don't have
  good numbers but locally it has out performed both Innodb and MyISAM. For
  Innodb the question will be if the table can be fit into the buffer
  pool. For MyISAM its a question of how much the file system caches the
  MyISAM file. With enough free memory MyISAM is faster. Its only when the OS
  doesn't have enough memory to cache entire table that archive turns out 
  to be any faster. For writes it is always a bit slower then MyISAM. It has no
  internal limits though for row length.

  Examples between MyISAM (packed) and Archive.

  Table with 76695844 identical rows:
  29680807 a_archive.ARZ
  920350317 a.MYD


  Table with 8991478 rows (all of Slashdot's comments):
  1922964506 comment_archive.ARZ
  2944970297 comment_text.MYD


  TODO:
   Add bzip optional support.
   Allow users to set compression level.
   Add truncate table command.
   Implement versioning, should be easy.
   Allow for errors, find a way to mark bad rows.
   Talk to the gzip guys, come up with a writable format so that updates are doable
     without switching to a block method.
   Add optional feature so that rows can be flushed at interval (which will cause less
     compression but may speed up ordered searches).
   Checkpoint the meta file to allow for faster rebuilds.
   Dirty open (right now the meta file is repaired if a crash occured).
   Option to allow for dirty reads, this would lower the sync calls, which would make
     inserts a lot faster, but would mean highly arbitrary reads.

    -Brian
*/
/*
  Notes on file formats.
  The Meta file is layed out as:
  check - Just an int of 254 to make sure that the the file we are opening was
          never corrupted.
  version - The current version of the file format.
  rows - This is an unsigned long long which is the number of rows in the data
         file.
  check point - Reserved for future use
  dirty - Status of the file, whether or not its values are the latest. This
          flag is what causes a repair to occur

  The data file:
  check - Just an int of 254 to make sure that the the file we are opening was
          never corrupted.
  version - The current version of the file format.
  data - The data is stored in a "row +blobs" format.
*/

/* If the archive storage engine has been inited */
static bool archive_inited= 0;
/* Variables for archive share methods */
pthread_mutex_t archive_mutex;
static HASH archive_open_tables;

/* The file extension */
#define ARZ ".ARZ"               // The data file
#define ARN ".ARN"               // Files used during an optimize call
#define ARM ".ARM"               // Meta file
/*
  uchar + uchar + ulonglong + ulonglong + uchar
*/
#define META_BUFFER_SIZE 19      // Size of the data used in the meta file
/*
  uchar + uchar
*/
#define DATA_BUFFER_SIZE 2       // Size of the data used in the data file
#define ARCHIVE_CHECK_HEADER 254 // The number we use to determine corruption

/* dummy handlerton - only to have something to return from archive_db_init */
static handlerton archive_hton = {
  "archive",
  0,       /* slot */
  0,       /* savepoint size. */
  0,       /* close_connection */
  0,       /* savepoint */
  0,       /* rollback to savepoint */
  0,       /* releas savepoint */
  0,       /* commit */
  0,       /* rollback */
  0,       /* prepare */
  0,       /* recover */
  0,       /* commit_by_xid */
  0        /* rollback_by_xid */
};


/*
  Used for hash table that tracks open tables.
*/
static byte* archive_get_key(ARCHIVE_SHARE *share,uint *length,
                             my_bool not_used __attribute__((unused)))
{
  *length=share->table_name_length;
  return (byte*) share->table_name;
}


/*
  Initialize the archive handler.

  SYNOPSIS
    archive_db_init()
    void

  RETURN
    &archive_hton OK
    0             Error
*/

handlerton *archive_db_init()
{
  archive_inited= 1;
  VOID(pthread_mutex_init(&archive_mutex, MY_MUTEX_INIT_FAST));
  if (hash_init(&archive_open_tables, system_charset_info, 32, 0, 0,
                (hash_get_key) archive_get_key, 0, 0))
    return 0;
  return &archive_hton;
}

/*
  Release the archive handler.

  SYNOPSIS
    archive_db_end()
    void

  RETURN
    FALSE       OK
*/

bool archive_db_end()
{
  if (archive_inited)
  {
    hash_free(&archive_open_tables);
    VOID(pthread_mutex_destroy(&archive_mutex));
  }
  archive_inited= 0;
  return FALSE;
}


/*
  This method reads the header of a datafile and returns whether or not it was successful.
*/
int ha_archive::read_data_header(gzFile file_to_read)
{
  uchar data_buffer[DATA_BUFFER_SIZE];
  DBUG_ENTER("ha_archive::read_data_header");

  if (gzrewind(file_to_read) == -1)
    DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);

  if (gzread(file_to_read, data_buffer, DATA_BUFFER_SIZE) != DATA_BUFFER_SIZE)
    DBUG_RETURN(errno ? errno : -1);
  
  DBUG_PRINT("ha_archive::read_data_header", ("Check %u", data_buffer[0]));
  DBUG_PRINT("ha_archive::read_data_header", ("Version %u", data_buffer[1]));
  
  if ((data_buffer[0] != (uchar)ARCHIVE_CHECK_HEADER) &&  
      (data_buffer[1] != (uchar)ARCHIVE_VERSION))
    DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);

  DBUG_RETURN(0);
}

/*
  This method writes out the header of a datafile and returns whether or not it was successful.
*/
int ha_archive::write_data_header(gzFile file_to_write)
{
  uchar data_buffer[DATA_BUFFER_SIZE];
  DBUG_ENTER("ha_archive::write_data_header");

  data_buffer[0]= (uchar)ARCHIVE_CHECK_HEADER;
  data_buffer[1]= (uchar)ARCHIVE_VERSION;

  if (gzwrite(file_to_write, &data_buffer, DATA_BUFFER_SIZE) != 
      sizeof(DATA_BUFFER_SIZE))
    goto error;
  DBUG_PRINT("ha_archive::write_data_header", ("Check %u", (uint)data_buffer[0]));
  DBUG_PRINT("ha_archive::write_data_header", ("Version %u", (uint)data_buffer[1]));

  DBUG_RETURN(0);
error:
  DBUG_RETURN(errno);
}

/*
  This method reads the header of a meta file and returns whether or not it was successful.
  *rows will contain the current number of rows in the data file upon success.
*/
int ha_archive::read_meta_file(File meta_file, ulonglong *rows)
{
  uchar meta_buffer[META_BUFFER_SIZE];
  ulonglong check_point;

  DBUG_ENTER("ha_archive::read_meta_file");

  VOID(my_seek(meta_file, 0, MY_SEEK_SET, MYF(0)));
  if (my_read(meta_file, (byte*)meta_buffer, META_BUFFER_SIZE, 0) != META_BUFFER_SIZE)
    DBUG_RETURN(-1);
  
  /*
    Parse out the meta data, we ignore version at the moment
  */
  *rows= uint8korr(meta_buffer + 2);
  check_point= uint8korr(meta_buffer + 10);

  DBUG_PRINT("ha_archive::read_meta_file", ("Check %d", (uint)meta_buffer[0]));
  DBUG_PRINT("ha_archive::read_meta_file", ("Version %d", (uint)meta_buffer[1]));
  DBUG_PRINT("ha_archive::read_meta_file", ("Rows %lld", *rows));
  DBUG_PRINT("ha_archive::read_meta_file", ("Checkpoint %lld", check_point));
  DBUG_PRINT("ha_archive::read_meta_file", ("Dirty %d", (int)meta_buffer[18]));

  if ((meta_buffer[0] != (uchar)ARCHIVE_CHECK_HEADER) || 
      ((bool)meta_buffer[18] == TRUE))
    DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);

  my_sync(meta_file, MYF(MY_WME));

  DBUG_RETURN(0);
}

/*
  This method writes out the header of a meta file and returns whether or not it was successful.
  By setting dirty you say whether or not the file represents the actual state of the data file.
  Upon ::open() we set to dirty, and upon ::close() we set to clean.
*/
int ha_archive::write_meta_file(File meta_file, ulonglong rows, bool dirty)
{
  uchar meta_buffer[META_BUFFER_SIZE];
  ulonglong check_point= 0; //Reserved for the future

  DBUG_ENTER("ha_archive::write_meta_file");

  meta_buffer[0]= (uchar)ARCHIVE_CHECK_HEADER;
  meta_buffer[1]= (uchar)ARCHIVE_VERSION;
  int8store(meta_buffer + 2, rows); 
  int8store(meta_buffer + 10, check_point); 
  *(meta_buffer + 18)= (uchar)dirty;
  DBUG_PRINT("ha_archive::write_meta_file", ("Check %d", (uint)ARCHIVE_CHECK_HEADER));
  DBUG_PRINT("ha_archive::write_meta_file", ("Version %d", (uint)ARCHIVE_VERSION));
  DBUG_PRINT("ha_archive::write_meta_file", ("Rows %llu", rows));
  DBUG_PRINT("ha_archive::write_meta_file", ("Checkpoint %llu", check_point));
  DBUG_PRINT("ha_archive::write_meta_file", ("Dirty %d", (uint)dirty));

  VOID(my_seek(meta_file, 0, MY_SEEK_SET, MYF(0)));
  if (my_write(meta_file, (byte *)meta_buffer, META_BUFFER_SIZE, 0) != META_BUFFER_SIZE)
    DBUG_RETURN(-1);
  
  my_sync(meta_file, MYF(MY_WME));

  DBUG_RETURN(0);
}


/*
  We create the shared memory space that we will use for the open table. 
  See ha_example.cc for a longer description.
*/
ARCHIVE_SHARE *ha_archive::get_share(const char *table_name, TABLE *table)
{
  ARCHIVE_SHARE *share;
  char meta_file_name[FN_REFLEN];
  uint length;
  char *tmp_name;

  pthread_mutex_lock(&archive_mutex);
  length=(uint) strlen(table_name);

  if (!(share=(ARCHIVE_SHARE*) hash_search(&archive_open_tables,
                                           (byte*) table_name,
                                           length)))
  {
    if (!my_multi_malloc(MYF(MY_WME | MY_ZEROFILL),
                          &share, sizeof(*share),
                          &tmp_name, length+1,
                          NullS)) 
    {
      pthread_mutex_unlock(&archive_mutex);
      return NULL;
    }

    share->use_count= 0;
    share->table_name_length= length;
    share->table_name= tmp_name;
    share->crashed= FALSE;
    fn_format(share->data_file_name,table_name,"",ARZ,MY_REPLACE_EXT|MY_UNPACK_FILENAME);
    fn_format(meta_file_name,table_name,"",ARM,MY_REPLACE_EXT|MY_UNPACK_FILENAME);
    strmov(share->table_name,table_name);
    /*
      We will use this lock for rows.
    */
    VOID(pthread_mutex_init(&share->mutex,MY_MUTEX_INIT_FAST));
    if ((share->meta_file= my_open(meta_file_name, O_RDWR, MYF(0))) == -1)
      goto error;
    
    /*
      After we read, we set the file to dirty. When we close, we will do the 
      opposite. If the meta file will not open we assume it is crashed and
      leave it up to the user to fix.
    */
    if (read_meta_file(share->meta_file, &share->rows_recorded))
      share->crashed= TRUE;
    else
      (void)write_meta_file(share->meta_file, share->rows_recorded, TRUE);

    /* 
      It is expensive to open and close the data files and since you can't have
      a gzip file that can be both read and written we keep a writer open
      that is shared amoung all open tables.
    */
    if ((share->archive_write= gzopen(share->data_file_name, "ab")) == NULL)
      goto error2;
    if (my_hash_insert(&archive_open_tables, (byte*) share))
      goto error3;
    thr_lock_init(&share->lock);
  }
  share->use_count++;
  pthread_mutex_unlock(&archive_mutex);

  return share;

error3:
  /* We close, but ignore errors since we already have errors */
  (void)gzclose(share->archive_write);
error2:
  my_close(share->meta_file,MYF(0));
error:
  pthread_mutex_unlock(&archive_mutex);
  VOID(pthread_mutex_destroy(&share->mutex));
  my_free((gptr) share, MYF(0));

  return NULL;
}


/* 
  Free the share.
  See ha_example.cc for a description.
*/
int ha_archive::free_share(ARCHIVE_SHARE *share)
{
  int rc= 0;
  pthread_mutex_lock(&archive_mutex);
  if (!--share->use_count)
  {
    hash_delete(&archive_open_tables, (byte*) share);
    thr_lock_delete(&share->lock);
    VOID(pthread_mutex_destroy(&share->mutex));
    (void)write_meta_file(share->meta_file, share->rows_recorded, FALSE);
    if (gzclose(share->archive_write) == Z_ERRNO)
      rc= 1;
    if (my_close(share->meta_file, MYF(0)))
      rc= 1;
    my_free((gptr) share, MYF(0));
  }
  pthread_mutex_unlock(&archive_mutex);

  return rc;
}


/*
  We just implement one additional file extension.
*/
static const char *ha_archive_exts[] = {
  ARZ,
  ARN,
  ARM,
  NullS
};

const char **ha_archive::bas_ext() const
{
  return ha_archive_exts;
}


/*
  Rename all files that this handler defines in bas_ext list

  NOTE Don't care if the .arn file is missing
*/
int ha_archive::rename_table(const char * from, const char * to)
{
  DBUG_ENTER("ha_archive::rename_table");
  for (const char **ext=bas_ext(); *ext ; ext++)
  {
    if (rename_file_ext(from,to,*ext))
    {
      if (my_errno == ENOENT &&
          !my_strcasecmp(system_charset_info, *ext, ARN))
        continue;

      DBUG_RETURN(my_errno);
    }

  }
  DBUG_RETURN(0);
}


/* 
  When opening a file we:
  Create/get our shared structure.
  Init out lock.
  We open the file we will read from.
*/
int ha_archive::open(const char *name, int mode, uint test_if_locked)
{
  DBUG_ENTER("ha_archive::open");

  if (!(share= get_share(name, table)))
    DBUG_RETURN(-1);
  thr_lock_data_init(&share->lock,&lock,NULL);

  if ((archive= gzopen(share->data_file_name, "rb")) == NULL)
  {
    (void)free_share(share); //We void since we already have an error
    DBUG_RETURN(errno ? errno : -1);
  }

  DBUG_RETURN(0);
}


/*
  Closes the file.

  SYNOPSIS
    close();
  
  IMPLEMENTATION:

  We first close this storage engines file handle to the archive and
  then remove our reference count to the table (and possibly free it
  as well).

  RETURN
    0  ok
    1  Error
*/

int ha_archive::close(void)
{
  int rc= 0;
  DBUG_ENTER("ha_archive::close");

  /* First close stream */
  if (gzclose(archive) == Z_ERRNO)
    rc= 1;
  /* then also close share */
  rc|= free_share(share);

  DBUG_RETURN(rc);
}


/*
  We create our data file here. The format is pretty simple. 
  You can read about the format of the data file above.
  Unlike other storage engines we do not "pack" our data. Since we 
  are about to do a general compression, packing would just be a waste of 
  CPU time. If the table has blobs they are written after the row in the order 
  of creation.
*/

int ha_archive::create(const char *name, TABLE *table_arg,
                       HA_CREATE_INFO *create_info)
{
  File create_file;  // We use to create the datafile and the metafile
  char name_buff[FN_REFLEN];
  int error;
  DBUG_ENTER("ha_archive::create");

  if ((create_file= my_create(fn_format(name_buff,name,"",ARM,
                                        MY_REPLACE_EXT|MY_UNPACK_FILENAME),0,
                              O_RDWR | O_TRUNC,MYF(MY_WME))) < 0)
  {
    error= my_errno;
    goto error;
  }
  write_meta_file(create_file, 0, FALSE);
  my_close(create_file,MYF(0));

  /* 
    We reuse name_buff since it is available.
  */
  if ((create_file= my_create(fn_format(name_buff,name,"",ARZ,
                                        MY_REPLACE_EXT|MY_UNPACK_FILENAME),0,
                              O_RDWR | O_TRUNC,MYF(MY_WME))) < 0)
  {
    error= my_errno;
    goto error;
  }
  if ((archive= gzdopen(create_file, "ab")) == NULL)
  {
    error= errno;
    goto error2;
  }
  if (write_data_header(archive))
  {
    error= errno;
    goto error3;
  }

  if (gzclose(archive))
  {
    error= errno;
    goto error2;
  }

  my_close(create_file, MYF(0));

  DBUG_RETURN(0);

error3:
  /* We already have an error, so ignore results of gzclose. */
  (void)gzclose(archive);
error2:
  my_close(create_file, MYF(0));
  delete_table(name);
error:
  /* Return error number, if we got one */
  DBUG_RETURN(error ? error : -1);
}


/* 
  Look at ha_archive::open() for an explanation of the row format.
  Here we just write out the row.

  Wondering about start_bulk_insert()? We don't implement it for
  archive since it optimizes for lots of writes. The only save
  for implementing start_bulk_insert() is that we could skip 
  setting dirty to true each time.
*/
int ha_archive::write_row(byte * buf)
{
  z_off_t written;
  uint *ptr, *end;
  DBUG_ENTER("ha_archive::write_row");

  if (share->crashed)
      DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);

  statistic_increment(table->in_use->status_var.ha_write_count, &LOCK_status);
  if (table->timestamp_field_type & TIMESTAMP_AUTO_SET_ON_INSERT)
    table->timestamp_field->set_time();
  pthread_mutex_lock(&share->mutex);
  written= gzwrite(share->archive_write, buf, table->s->reclength);
  DBUG_PRINT("ha_archive::write_row", ("Wrote %d bytes expected %d", written, table->s->reclength));
  if (!delayed_insert || !bulk_insert)
    share->dirty= TRUE;

  if (written != table->s->reclength)
    goto error;
  /*
    We should probably mark the table as damagaged if the record is written
    but the blob fails.
  */
  for (ptr= table->s->blob_field, end= ptr + table->s->blob_fields ;
       ptr != end ;
       ptr++)
  {
    char *data_ptr;
    uint32 size= ((Field_blob*) table->field[*ptr])->get_length();

    if (size)
    {
      ((Field_blob*) table->field[*ptr])->get_ptr(&data_ptr);
      written= gzwrite(share->archive_write, data_ptr, (unsigned)size);
      if (written != size)
        goto error;
    }
  }
  share->rows_recorded++;
  pthread_mutex_unlock(&share->mutex);

  DBUG_RETURN(0);
error:
  pthread_mutex_unlock(&share->mutex);
  DBUG_RETURN(errno ? errno : -1);
}


/*
  All calls that need to scan the table start with this method. If we are told
  that it is a table scan we rewind the file to the beginning, otherwise
  we assume the position will be set.
*/

int ha_archive::rnd_init(bool scan)
{
  DBUG_ENTER("ha_archive::rnd_init");
  
  if (share->crashed)
      DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);

  /* We rewind the file so that we can read from the beginning if scan */
  if (scan)
  {
    scan_rows= share->rows_recorded;
    records= 0;

    /* 
      If dirty, we lock, and then reset/flush the data.
      I found that just calling gzflush() doesn't always work.
    */
    if (share->dirty == TRUE)
    {
      pthread_mutex_lock(&share->mutex);
      if (share->dirty == TRUE)
      {
        gzflush(share->archive_write, Z_SYNC_FLUSH);
        share->dirty= FALSE;
      }
      pthread_mutex_unlock(&share->mutex);
    }

    if (read_data_header(archive))
      DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
  }

  DBUG_RETURN(0);
}


/*
  This is the method that is used to read a row. It assumes that the row is 
  positioned where you want it.
*/
int ha_archive::get_row(gzFile file_to_read, byte *buf)
{
  int read; // Bytes read, gzread() returns int
  uint *ptr, *end;
  char *last;
  size_t total_blob_length= 0;
  DBUG_ENTER("ha_archive::get_row");

  read= gzread(file_to_read, buf, table->s->reclength);
  DBUG_PRINT("ha_archive::get_row", ("Read %d bytes expected %d", read, table->s->reclength));

  if (read == Z_STREAM_ERROR)
    DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);

  /* If we read nothing we are at the end of the file */
  if (read == 0)
    DBUG_RETURN(HA_ERR_END_OF_FILE);

  /* 
    If the record is the wrong size, the file is probably damaged, unless 
    we are dealing with a delayed insert or a bulk insert.
  */
  if ((ulong) read != table->s->reclength)
    DBUG_RETURN(HA_ERR_END_OF_FILE);

  /* Calculate blob length, we use this for our buffer */
  for (ptr= table->s->blob_field, end=ptr + table->s->blob_fields ;
       ptr != end ;
       ptr++)
    total_blob_length += ((Field_blob*) table->field[*ptr])->get_length();

  /* Adjust our row buffer if we need be */
  buffer.alloc(total_blob_length);
  last= (char *)buffer.ptr();

  /* Loop through our blobs and read them */
  for (ptr= table->s->blob_field, end=ptr + table->s->blob_fields ;
       ptr != end ;
       ptr++)
  {
    size_t size= ((Field_blob*) table->field[*ptr])->get_length();
    if (size)
    {
      read= gzread(file_to_read, last, size);
      if ((size_t) read != size)
        DBUG_RETURN(HA_ERR_END_OF_FILE);
      ((Field_blob*) table->field[*ptr])->set_ptr(size, last);
      last += size;
    }
  }
  DBUG_RETURN(0);
}


/* 
  Called during ORDER BY. Its position is either from being called sequentially
  or by having had ha_archive::rnd_pos() called before it is called.
*/

int ha_archive::rnd_next(byte *buf)
{
  int rc;
  DBUG_ENTER("ha_archive::rnd_next");

  if (share->crashed)
      DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);

  if (!scan_rows)
    DBUG_RETURN(HA_ERR_END_OF_FILE);
  scan_rows--;

  statistic_increment(table->in_use->status_var.ha_read_rnd_next_count,
		      &LOCK_status);
  current_position= gztell(archive);
  rc= get_row(archive, buf);


  if (rc != HA_ERR_END_OF_FILE)
    records++;

  DBUG_RETURN(rc);
}


/*
  Thanks to the table flag HA_REC_NOT_IN_SEQ this will be called after
  each call to ha_archive::rnd_next() if an ordering of the rows is
  needed.
*/

void ha_archive::position(const byte *record)
{
  DBUG_ENTER("ha_archive::position");
  my_store_ptr(ref, ref_length, current_position);
  DBUG_VOID_RETURN;
}


/*
  This is called after a table scan for each row if the results of the
  scan need to be ordered. It will take *pos and use it to move the
  cursor in the file so that the next row that is called is the
  correctly ordered row.
*/

int ha_archive::rnd_pos(byte * buf, byte *pos)
{
  DBUG_ENTER("ha_archive::rnd_pos");
  statistic_increment(table->in_use->status_var.ha_read_rnd_next_count,
		      &LOCK_status);
  current_position= my_get_ptr(pos, ref_length);
  z_off_t seek= gzseek(archive, current_position, SEEK_SET);

  DBUG_RETURN(get_row(archive, buf));
}

/*
  This method repairs the meta file. It does this by walking the datafile and 
  rewriting the meta file.
*/
int ha_archive::repair(THD* thd, HA_CHECK_OPT* check_opt)
{
  int rc;
  byte *buf; 
  ulonglong rows_recorded= 0;
  gzFile rebuild_file; // Archive file we are working with 
  File meta_file; // Meta file we use 
  char data_file_name[FN_REFLEN];
  DBUG_ENTER("ha_archive::repair");

  /*
    Open up the meta file to recreate it.
  */
  fn_format(data_file_name, share->table_name, "", ARZ,
            MY_REPLACE_EXT|MY_UNPACK_FILENAME);
  if ((rebuild_file= gzopen(data_file_name, "rb")) == NULL)
    DBUG_RETURN(errno ? errno : -1);

  if ((rc= read_data_header(rebuild_file)))
    goto error;

  /*
    We malloc up the buffer we will use for counting the rows. 
    I know, this malloc'ing memory but this should be a very 
    rare event.
  */
  if (!(buf= (byte*) my_malloc(table->s->rec_buff_length > sizeof(ulonglong) +1 ? 
                               table->s->rec_buff_length : sizeof(ulonglong) +1 ,
                               MYF(MY_WME))))
  {
    rc= HA_ERR_CRASHED_ON_USAGE;
    goto error;
  }

  while (!(rc= get_row(rebuild_file, buf)))
    rows_recorded++;

  /* 
    Only if we reach the end of the file do we assume we can rewrite.
    At this point we reset rc to a non-message state.
  */
  if (rc == HA_ERR_END_OF_FILE)
  {
    fn_format(data_file_name,share->table_name,"",ARM,MY_REPLACE_EXT|MY_UNPACK_FILENAME);
    if ((meta_file= my_open(data_file_name, O_RDWR, MYF(0))) == -1)
    {
      rc= HA_ERR_CRASHED_ON_USAGE;
      goto error;
    }
    (void)write_meta_file(meta_file, rows_recorded, TRUE);
    my_close(meta_file,MYF(0));
    rc= 0;
  }

  my_free((gptr) buf, MYF(0));
  share->crashed= FALSE;
error:
  gzclose(rebuild_file);

  DBUG_RETURN(rc);
}

/*
  The table can become fragmented if data was inserted, read, and then
  inserted again. What we do is open up the file and recompress it completely. 
*/
int ha_archive::optimize(THD* thd, HA_CHECK_OPT* check_opt)
{
  DBUG_ENTER("ha_archive::optimize");
  int read; // Bytes read, gzread() returns int
  gzFile reader, writer;
  char block[IO_SIZE];
  char writer_filename[FN_REFLEN];

  /* Closing will cause all data waiting to be flushed */
  gzclose(share->archive_write);
  share->archive_write= NULL; 

  /* Lets create a file to contain the new data */
  fn_format(writer_filename, share->table_name, "", ARN, 
            MY_REPLACE_EXT|MY_UNPACK_FILENAME);

  if ((reader= gzopen(share->data_file_name, "rb")) == NULL)
    DBUG_RETURN(-1); 

  if ((writer= gzopen(writer_filename, "wb")) == NULL)
  {
    gzclose(reader);
    DBUG_RETURN(-1); 
  }

  while ((read= gzread(reader, block, IO_SIZE)))
    gzwrite(writer, block, read);

  gzclose(reader);
  gzclose(writer);

  my_rename(writer_filename,share->data_file_name,MYF(0));

  DBUG_RETURN(0); 
}


/* 
  Below is an example of how to setup row level locking.
*/
THR_LOCK_DATA **ha_archive::store_lock(THD *thd,
                                       THR_LOCK_DATA **to,
                                       enum thr_lock_type lock_type)
{
  if (lock_type == TL_WRITE_DELAYED)
    delayed_insert= TRUE;
  else
    delayed_insert= FALSE;

  if (lock_type != TL_IGNORE && lock.type == TL_UNLOCK) 
  {
    /* 
      Here is where we get into the guts of a row level lock.
      If TL_UNLOCK is set 
      If we are not doing a LOCK TABLE or DISCARD/IMPORT
      TABLESPACE, then allow multiple writers 
    */

    if ((lock_type >= TL_WRITE_CONCURRENT_INSERT &&
         lock_type <= TL_WRITE) && !thd->in_lock_tables
        && !thd->tablespace_op)
      lock_type = TL_WRITE_ALLOW_WRITE;

    /* 
      In queries of type INSERT INTO t1 SELECT ... FROM t2 ...
      MySQL would use the lock TL_READ_NO_INSERT on t2, and that
      would conflict with TL_WRITE_ALLOW_WRITE, blocking all inserts
      to t2. Convert the lock to a normal read lock to allow
      concurrent inserts to t2. 
    */

    if (lock_type == TL_READ_NO_INSERT && !thd->in_lock_tables) 
      lock_type = TL_READ;

    lock.type=lock_type;
  }

  *to++= &lock;

  return to;
}


/*
  Hints for optimizer, see ha_tina for more information
*/
void ha_archive::info(uint flag)
{
  DBUG_ENTER("ha_archive::info");
  /* 
    This should be an accurate number now, though bulk and delayed inserts can
    cause the number to be inaccurate.
  */
  records= share->rows_recorded;
  deleted= 0;
  /* Costs quite a bit more to get all information */
  if (flag & HA_STATUS_TIME)
  {
    MY_STAT file_stat;  // Stat information for the data file

    VOID(my_stat(share->data_file_name, &file_stat, MYF(MY_WME)));

    mean_rec_length= table->s->reclength + buffer.alloced_length();
    data_file_length= file_stat.st_size;
    create_time= file_stat.st_ctime;
    update_time= file_stat.st_mtime;
    max_data_file_length= share->rows_recorded * mean_rec_length;
  }
  delete_length= 0;
  index_file_length=0;

  DBUG_VOID_RETURN;
}


/*
  This method tells us that a bulk insert operation is about to occur. We set
  a flag which will keep write_row from saying that its data is dirty. This in
  turn will keep selects from causing a sync to occur.
  Basically, yet another optimizations to keep compression working well.
*/
void ha_archive::start_bulk_insert(ha_rows rows)
{
  DBUG_ENTER("ha_archive::start_bulk_insert");
  bulk_insert= TRUE;
  DBUG_VOID_RETURN;
}


/* 
  Other side of start_bulk_insert, is end_bulk_insert. Here we turn off the bulk insert
  flag, and set the share dirty so that the next select will call sync for us.
*/
int ha_archive::end_bulk_insert()
{
  DBUG_ENTER("ha_archive::end_bulk_insert");
  bulk_insert= FALSE;
  share->dirty= TRUE;
  DBUG_RETURN(0);
}
#endif /* HAVE_ARCHIVE_DB */