summaryrefslogtreecommitdiff
path: root/sql/multi_range_read.cc
blob: ec3c85b34c4c30215d318ea64d0d0ac27ae8ed2a (plain)
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
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
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
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
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
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
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
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
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
1557
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
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
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
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
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
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
/* Copyright (C) 2010, 2011 Monty Program 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; version 2 of the License.

   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., 51 Franklin Street, Fifth Floor, Boston, MA 02111-1301 USA */

#include "mariadb.h"
#include "sql_parse.h"
#include <my_bit.h>
#include "sql_select.h"
#include "key.h"

/****************************************************************************
 * Default MRR implementation (MRR to non-MRR converter)
 ***************************************************************************/

/**
  Get cost and other information about MRR scan over a known list of ranges

  Calculate estimated cost and other information about an MRR scan for given
  sequence of ranges.

  @param keyno           Index number
  @param seq             Range sequence to be traversed
  @param seq_init_param  First parameter for seq->init()
  @param n_ranges_arg    Number of ranges in the sequence, or 0 if the caller
                         can't efficiently determine it
  @param bufsz    INOUT  IN:  Size of the buffer available for use
                         OUT: Size of the buffer that is expected to be actually
                              used, or 0 if buffer is not needed.
  @param flags    INOUT  A combination of HA_MRR_* flags
  @param cost     OUT    Estimated cost of MRR access

  @note
    This method (or an overriding one in a derived class) must check for
    thd->killed and return HA_POS_ERROR if it is not zero. This is required
    for a user to be able to interrupt the calculation by killing the
    connection/query.

  @retval
    HA_POS_ERROR  Error or the engine is unable to perform the requested
                  scan. Values of OUT parameters are undefined.
  @retval
    other         OK, *cost contains cost of the scan, *bufsz and *flags
                  contain scan parameters.
*/

ha_rows 
handler::multi_range_read_info_const(uint keyno, RANGE_SEQ_IF *seq,
                                     void *seq_init_param, uint n_ranges_arg,
                                     uint *bufsz, uint *flags, Cost_estimate *cost)
{
  KEY_MULTI_RANGE range;
  range_seq_t seq_it;
  ha_rows rows, total_rows= 0;
  uint n_ranges=0;
  THD *thd= table->in_use;
  
  /* Default MRR implementation doesn't need buffer */
  *bufsz= 0;

  seq_it= seq->init(seq_init_param, n_ranges, *flags);
  while (!seq->next(seq_it, &range))
  {
    if (unlikely(thd->killed != 0))
      return HA_POS_ERROR;
    
    n_ranges++;
    key_range *min_endp, *max_endp;
    if (range.range_flag & GEOM_FLAG)
    {
      /* In this case tmp_min_flag contains the handler-read-function */
      range.start_key.flag= (ha_rkey_function) (range.range_flag ^ GEOM_FLAG);
      min_endp= &range.start_key;
      max_endp= NULL;
    }
    else
    {
      min_endp= range.start_key.length? &range.start_key : NULL;
      max_endp= range.end_key.length? &range.end_key : NULL;
    }
    if ((range.range_flag & UNIQUE_RANGE) && !(range.range_flag & NULL_RANGE))
      rows= 1; /* there can be at most one row */
    else
    {
      if (HA_POS_ERROR == (rows= this->records_in_range(keyno, min_endp, 
                                                        max_endp)))
      {
        /* Can't scan one range => can't do MRR scan at all */
        total_rows= HA_POS_ERROR;
        break;
      }
    }
    total_rows += rows;
  }
  
  if (total_rows != HA_POS_ERROR)
  {
    /* The following calculation is the same as in multi_range_read_info(): */
    *flags |= HA_MRR_USE_DEFAULT_IMPL;
    cost->reset();
    cost->avg_io_cost= 1; /* assume random seeks */
    if ((*flags & HA_MRR_INDEX_ONLY) && total_rows > 2)
      cost->io_count= keyread_time(keyno, n_ranges, (uint)total_rows);
    else
      cost->io_count= read_time(keyno, n_ranges, total_rows);
    cost->cpu_cost= (double) total_rows / TIME_FOR_COMPARE + 0.01;
  }
  return total_rows;
}


/**
  Get cost and other information about MRR scan over some sequence of ranges

  Calculate estimated cost and other information about an MRR scan for some
  sequence of ranges.

  The ranges themselves will be known only at execution phase. When this
  function is called we only know number of ranges and a (rough) E(#records)
  within those ranges.

  Currently this function is only called for "n-keypart singlepoint" ranges,
  i.e. each range is "keypart1=someconst1 AND ... AND keypartN=someconstN"

  The flags parameter is a combination of those flags: HA_MRR_SORTED,
  HA_MRR_INDEX_ONLY, HA_MRR_NO_ASSOCIATION, HA_MRR_LIMITS.

  @param keyno           Index number
  @param n_ranges        Estimated number of ranges (i.e. intervals) in the
                         range sequence.
  @param n_rows          Estimated total number of records contained within all
                         of the ranges
  @param bufsz    INOUT  IN:  Size of the buffer available for use
                         OUT: Size of the buffer that will be actually used, or
                              0 if buffer is not needed.
  @param flags    INOUT  A combination of HA_MRR_* flags
  @param cost     OUT    Estimated cost of MRR access

  @retval
    0     OK, *cost contains cost of the scan, *bufsz and *flags contain scan
          parameters.
  @retval
    other Error or can't perform the requested scan
*/

ha_rows handler::multi_range_read_info(uint keyno, uint n_ranges, uint n_rows,
                                       uint key_parts, uint *bufsz, 
                                       uint *flags, Cost_estimate *cost)
{
  /* 
    Currently we expect this function to be called only in preparation of scan
    with HA_MRR_SINGLE_POINT property.
  */
  DBUG_ASSERT(*flags | HA_MRR_SINGLE_POINT);

  *bufsz= 0; /* Default implementation doesn't need a buffer */
  *flags |= HA_MRR_USE_DEFAULT_IMPL;

  cost->reset();
  cost->avg_io_cost= 1; /* assume random seeks */

  /* Produce the same cost as non-MRR code does */
  if (*flags & HA_MRR_INDEX_ONLY)
    cost->io_count= keyread_time(keyno, n_ranges, n_rows);
  else
    cost->io_count= read_time(keyno, n_ranges, n_rows);
  return 0;
}


/**
  Initialize the MRR scan

  Initialize the MRR scan. This function may do heavyweight scan 
  initialization like row prefetching/sorting/etc (NOTE: but better not do
  it here as we may not need it, e.g. if we never satisfy WHERE clause on
  previous tables. For many implementations it would be natural to do such
  initializations in the first multi_read_range_next() call)

  mode is a combination of the following flags: HA_MRR_SORTED,
  HA_MRR_INDEX_ONLY, HA_MRR_NO_ASSOCIATION 

  @param seq             Range sequence to be traversed
  @param seq_init_param  First parameter for seq->init()
  @param n_ranges        Number of ranges in the sequence
  @param mode            Flags, see the description section for the details
  @param buf             INOUT: memory buffer to be used

  @note
    One must have called index_init() before calling this function. Several
    multi_range_read_init() calls may be made in course of one query.

    Buffer memory management is done according to the following scenario:
    The caller allocates the buffer and provides it to the callee by filling
    the members of HANDLER_BUFFER structure.
    The callee consumes all or some fraction of the provided buffer space, and
    sets the HANDLER_BUFFER members accordingly.
    The callee may use the buffer memory until the next multi_range_read_init()
    call is made, all records have been read, or until index_end() call is
    made, whichever comes first.

  @retval 0  OK
  @retval 1  Error
*/

int
handler::multi_range_read_init(RANGE_SEQ_IF *seq_funcs, void *seq_init_param,
                               uint n_ranges, uint mode, HANDLER_BUFFER *buf)
{
  DBUG_ENTER("handler::multi_range_read_init");
  mrr_iter= seq_funcs->init(seq_init_param, n_ranges, mode);
  mrr_funcs= *seq_funcs;
  mrr_is_output_sorted= MY_TEST(mode & HA_MRR_SORTED);
  mrr_have_range= FALSE;
  DBUG_RETURN(0);
}

/**
  Get next record in MRR scan

  Default MRR implementation: read the next record

  @param range_info  OUT  Undefined if HA_MRR_NO_ASSOCIATION flag is in effect
                          Otherwise, the opaque value associated with the range
                          that contains the returned record.

  @retval 0      OK
  @retval other  Error code
*/

int handler::multi_range_read_next(range_id_t *range_info)
{
  int result= HA_ERR_END_OF_FILE;
  bool range_res;
  DBUG_ENTER("handler::multi_range_read_next");

  if (!mrr_have_range)
  {
    mrr_have_range= TRUE;
    goto start;
  }

  do
  {
    /* Save a call if there can be only one row in range. */
    if (mrr_cur_range.range_flag != (UNIQUE_RANGE | EQ_RANGE))
    {
      result= read_range_next();
      /* On success or non-EOF errors jump to the end. */
      if (result != HA_ERR_END_OF_FILE)
        break;
    }
    else
    {
      if (ha_was_semi_consistent_read())
      {
        /*
          The following assignment is redundant, but for extra safety and to
          remove the compiler warning:
        */
        range_res= FALSE;
        goto scan_it_again;
      }
      /*
        We need to set this for the last range only, but checking this
        condition is more expensive than just setting the result code.
      */
      result= HA_ERR_END_OF_FILE;
    }

start:
    /* Try the next range(s) until one matches a record. */
    while (!(range_res= mrr_funcs.next(mrr_iter, &mrr_cur_range)))
    {
scan_it_again:
      result= read_range_first(mrr_cur_range.start_key.keypart_map ?
                                 &mrr_cur_range.start_key : 0,
                               mrr_cur_range.end_key.keypart_map ?
                                 &mrr_cur_range.end_key : 0,
                               MY_TEST(mrr_cur_range.range_flag & EQ_RANGE),
                               mrr_is_output_sorted);
      if (result != HA_ERR_END_OF_FILE)
        break;
    }
  }
  while ((result == HA_ERR_END_OF_FILE) && !range_res);

  *range_info= mrr_cur_range.ptr;
  DBUG_PRINT("exit",("handler::multi_range_read_next result %d", result));
  DBUG_RETURN(result);
}

/****************************************************************************
 * Mrr_*_reader classes (building blocks for DS-MRR)
 ***************************************************************************/

int Mrr_simple_index_reader::init(handler *h_arg, RANGE_SEQ_IF *seq_funcs, 
                                  void *seq_init_param, uint n_ranges,
                                  uint mode,  Key_parameters *key_par_arg,
                                  Lifo_buffer *key_buffer_arg,
                                  Buffer_manager *buf_manager_arg)
{
  HANDLER_BUFFER no_buffer = {NULL, NULL, NULL};
  file= h_arg;
  return file->handler::multi_range_read_init(seq_funcs, seq_init_param,
                                              n_ranges, mode, &no_buffer);
}


int Mrr_simple_index_reader::get_next(range_id_t *range_info)
{
  int res;
  while (!(res= file->handler::multi_range_read_next(range_info)))
  {
    KEY_MULTI_RANGE *curr_range= &file->handler::mrr_cur_range;
    if (!file->mrr_funcs.skip_index_tuple ||
        !file->mrr_funcs.skip_index_tuple(file->mrr_iter, curr_range->ptr))
      break;
  }
  if (res && res != HA_ERR_END_OF_FILE && res != HA_ERR_KEY_NOT_FOUND)
    file->print_error(res, MYF(0));             // Fatal error
  return res;
}


/**
  @brief Get next index record

  @param range_info  OUT identifier of range that the returned record belongs to
  
  @note
    We actually iterate over nested sequences:
    - an ordered sequence of groups of identical keys
      - each key group has key value, which has multiple matching records 
        - thus, each record matches all members of the key group

  @retval 0                   OK, next record was successfully read
  @retval HA_ERR_END_OF_FILE  End of records
  @retval Other               Some other error; Error is printed
*/

int Mrr_ordered_index_reader::get_next(range_id_t *range_info)
{
  int res;
  DBUG_ENTER("Mrr_ordered_index_reader::get_next");
  
  for(;;)
  {
    if (!scanning_key_val_iter)
    {
      while ((res= kv_it.init(this)))
      {
        if ((res != HA_ERR_KEY_NOT_FOUND && res != HA_ERR_END_OF_FILE))
          DBUG_RETURN(res); /* Some fatal error */

        if (key_buffer->is_empty())
        {
          DBUG_RETURN(HA_ERR_END_OF_FILE);
        }
      }
      scanning_key_val_iter= TRUE;
    }

    if ((res= kv_it.get_next(range_info)))
    {
      scanning_key_val_iter= FALSE;
      if ((res != HA_ERR_KEY_NOT_FOUND && res != HA_ERR_END_OF_FILE))
        DBUG_RETURN(res);
      kv_it.move_to_next_key_value();
      continue;
    }
    if (!skip_index_tuple(*range_info) &&
        !skip_record(*range_info, NULL))
    {
      break;
    }
    /* Go get another (record, range_id) combination */
  } /* while */

  DBUG_RETURN(0);
}


/*
  Supply index reader with the O(1)space it needs for scan interrupt/restore
  operation
*/

bool Mrr_ordered_index_reader::set_interruption_temp_buffer(uint rowid_length,
                                                            uint key_len, 
                                                            uint saved_pk_len,
                                                            uchar **space_start,
                                                            uchar *space_end)
{
  if (space_end - *space_start <= (ptrdiff_t)(rowid_length + key_len + saved_pk_len))
    return TRUE;
  support_scan_interruptions= TRUE; 
  
  saved_rowid= *space_start;
  *space_start += rowid_length;
  
  if (saved_pk_len)
  {
    saved_primary_key= *space_start;
    *space_start += saved_pk_len;
  }
  else
    saved_primary_key= NULL;

  saved_key_tuple= *space_start;
  *space_start += key_len;

  have_saved_rowid= FALSE;
  read_was_interrupted= FALSE;
  return FALSE;
}

void Mrr_ordered_index_reader::set_no_interruption_temp_buffer()
{
  support_scan_interruptions= FALSE;
  saved_key_tuple= saved_rowid= saved_primary_key= NULL; /* safety */
  have_saved_rowid= FALSE;
  read_was_interrupted= FALSE;
}

void Mrr_ordered_index_reader::interrupt_read()
{
  DBUG_ASSERT(support_scan_interruptions);
  TABLE *table= file->get_table();
  KEY *used_index= &table->key_info[file->active_index];
  /* Save the current key value */
  key_copy(saved_key_tuple, table->record[0],
           used_index, used_index->key_length);
  
  if (saved_primary_key)
  {
    key_copy(saved_primary_key, table->record[0], 
             &table->key_info[table->s->primary_key],
             table->key_info[table->s->primary_key].key_length);
  }
  read_was_interrupted= TRUE;

  /* Save the last rowid */
  memcpy(saved_rowid, file->ref, file->ref_length);
  have_saved_rowid= TRUE;
}

void Mrr_ordered_index_reader::position()
{
  if (have_saved_rowid)
    memcpy(file->ref, saved_rowid, file->ref_length);
  else
    Mrr_index_reader::position();
}

void Mrr_ordered_index_reader::resume_read()
{
  TABLE *table= file->get_table();

  if (!read_was_interrupted)
    return;

  KEY *used_index= &table->key_info[file->active_index];
  key_restore(table->record[0], saved_key_tuple, 
              used_index, used_index->key_length);
  if (saved_primary_key)
  {
    key_restore(table->record[0], saved_primary_key, 
                &table->key_info[table->s->primary_key],
                table->key_info[table->s->primary_key].key_length);
  }
}


/**
  Fill the buffer with (lookup_tuple, range_id) pairs and sort

  @return 
    0                   OK, the buffer is non-empty and sorted
    HA_ERR_END_OF_FILE  Source exhausted, the buffer is empty.
*/

int Mrr_ordered_index_reader::refill_buffer(bool initial)
{
  KEY_MULTI_RANGE cur_range;
  DBUG_ENTER("Mrr_ordered_index_reader::refill_buffer");

  DBUG_ASSERT(key_buffer->is_empty());

  if (source_exhausted)
    DBUG_RETURN(HA_ERR_END_OF_FILE);

  buf_manager->reset_buffer_sizes(buf_manager->arg);
  key_buffer->reset();
  key_buffer->setup_writing(keypar.key_size_in_keybuf,
                            is_mrr_assoc? sizeof(range_id_t) : 0);

  while (key_buffer->can_write() && 
         !(source_exhausted= mrr_funcs.next(mrr_iter, &cur_range)))
  {
    DBUG_ASSERT(cur_range.range_flag & EQ_RANGE);

    /* Put key, or {key, range_id} pair into the buffer */
    key_buffer->write_ptr1= keypar.use_key_pointers ?
                              (uchar*)&cur_range.start_key.key : 
                              (uchar*)cur_range.start_key.key;
    key_buffer->write_ptr2= (uchar*)&cur_range.ptr;
    key_buffer->write();
  }
  
  /* Force get_next() to start with kv_it.init() call: */
  scanning_key_val_iter= FALSE;

  if (source_exhausted && key_buffer->is_empty())
    DBUG_RETURN(HA_ERR_END_OF_FILE);

  if (!initial)
  {
    /* This is a non-initial buffer fill and we've got a non-empty buffer */
    THD *thd= current_thd;
    status_var_increment(thd->status_var.ha_mrr_key_refills_count);
  }

  key_buffer->sort((key_buffer->type() == Lifo_buffer::FORWARD)? 
                     (qsort2_cmp)Mrr_ordered_index_reader::compare_keys_reverse : 
                     (qsort2_cmp)Mrr_ordered_index_reader::compare_keys, 
                   this);
  DBUG_RETURN(0);
}


int Mrr_ordered_index_reader::init(handler *h_arg, RANGE_SEQ_IF *seq_funcs,
                                   void *seq_init_param, uint n_ranges,
                                   uint mode, Key_parameters *key_par_arg,
                                   Lifo_buffer *key_buffer_arg,
                                   Buffer_manager *buf_manager_arg)
{
  file= h_arg;
  key_buffer= key_buffer_arg;
  buf_manager= buf_manager_arg;
  keypar= *key_par_arg;

  KEY *key_info= &file->get_table()->key_info[file->active_index];
  keypar.index_ranges_unique= MY_TEST(key_info->flags & HA_NOSAME &&
                                      key_info->user_defined_key_parts ==
                                      my_count_bits(keypar.key_tuple_map));

  mrr_iter= seq_funcs->init(seq_init_param, n_ranges, mode);
  is_mrr_assoc= !MY_TEST(mode & HA_MRR_NO_ASSOCIATION);
  mrr_funcs= *seq_funcs;
  source_exhausted= FALSE;
  read_was_interrupted= false;
  have_saved_rowid= FALSE;
  return 0;
}


static int rowid_cmp_reverse(void *file, uchar *a, uchar *b)
{
  return - ((handler*)file)->cmp_ref(a, b);
}


int Mrr_ordered_rndpos_reader::init(handler *h_arg, 
                                    Mrr_index_reader *index_reader_arg,
                                    uint mode,
                                    Lifo_buffer *buf)
{
  file= h_arg;
  index_reader= index_reader_arg;
  rowid_buffer= buf;
  is_mrr_assoc= !MY_TEST(mode & HA_MRR_NO_ASSOCIATION);
  index_reader_exhausted= FALSE;
  index_reader_needs_refill= TRUE;
  return 0;
}


/**
  DS-MRR: Fill and sort the rowid buffer

  Scan the MRR ranges and collect ROWIDs (or {ROWID, range_id} pairs) into 
  buffer. When the buffer is full or scan is completed, sort the buffer by 
  rowid and return.

  When this function returns, either rowid buffer is not empty, or the source
  of lookup keys (i.e. ranges) is exhaused.
  
  @retval 0      OK, the next portion of rowids is in the buffer,
                 properly ordered
  @retval other  Error
*/

int Mrr_ordered_rndpos_reader::refill_buffer(bool initial)
{
  int res;
  bool first_call= initial;
  DBUG_ENTER("Mrr_ordered_rndpos_reader::refill_buffer");

  if (index_reader_exhausted)
    DBUG_RETURN(HA_ERR_END_OF_FILE);

  while (initial || index_reader_needs_refill || 
         (res= refill_from_index_reader()) == HA_ERR_END_OF_FILE)
  {
    if ((res= index_reader->refill_buffer(initial)))
    {
      if (res == HA_ERR_END_OF_FILE)
        index_reader_exhausted= TRUE;
      break;
    }
    initial= FALSE;
    index_reader_needs_refill= FALSE;
  }

  if (!first_call && !index_reader_exhausted)
  {
    /* Ok, this was a successful buffer refill operation */
    THD *thd= current_thd;
    status_var_increment(thd->status_var.ha_mrr_rowid_refills_count);
  }

  DBUG_RETURN(res);
}


void Mrr_index_reader::position()
{
  file->position(file->get_table()->record[0]);
}


/* 
  @brief Try to refill the rowid buffer without calling
  index_reader->refill_buffer(). 
*/

int Mrr_ordered_rndpos_reader::refill_from_index_reader()
{
  range_id_t range_info;
  int res;
  DBUG_ENTER("Mrr_ordered_rndpos_reader::refill_from_index_reader");

  DBUG_ASSERT(rowid_buffer->is_empty());
  index_rowid= index_reader->get_rowid_ptr();
  rowid_buffer->reset();
  rowid_buffer->setup_writing(file->ref_length,
                              is_mrr_assoc? sizeof(range_id_t) : 0);

  last_identical_rowid= NULL;

  index_reader->resume_read();
  while (rowid_buffer->can_write())
  {
    res= index_reader->get_next(&range_info);

    if (res)
    {
      if (res != HA_ERR_END_OF_FILE)
        DBUG_RETURN(res);
      index_reader_needs_refill=TRUE;
      break;
    }

    index_reader->position();

    /* Put rowid, or {rowid, range_id} pair into the buffer */
    rowid_buffer->write_ptr1= index_rowid;
    rowid_buffer->write_ptr2= (uchar*)&range_info;
    rowid_buffer->write();
  }
  
  /*
    When index_reader_needs_refill=TRUE, this means we've got all of index
    tuples for lookups keys that index_reader had. We are not in the middle
    of an index read, so there is no need to call interrupt_read.

    Actually, we must not call interrupt_read(), because it could be that we
    haven't read a single row (because all index lookups returned
    HA_ERR_KEY_NOT_FOUND). In this case, interrupt_read() will cause [harmless]
    valgrind warnings when trying to save garbage from table->record[0].
  */
  if (!index_reader_needs_refill)
    index_reader->interrupt_read();
  /* Sort the buffer contents by rowid */
  rowid_buffer->sort((qsort2_cmp)rowid_cmp_reverse, (void*)file);

  rowid_buffer->setup_reading(file->ref_length,
                              is_mrr_assoc ? sizeof(range_id_t) : 0);
  DBUG_RETURN(rowid_buffer->is_empty()? HA_ERR_END_OF_FILE : 0);
}


/*
  Get the next {record, range_id} using ordered array of rowid+range_id pairs

  @note
    Since we have sorted rowids, we try not to make multiple rnd_pos() calls
    with the same rowid value.
*/

int Mrr_ordered_rndpos_reader::get_next(range_id_t *range_info)
{
  int res;
  
  /* 
    First, check if rowid buffer has elements with the same rowid value as
    the previous.
  */
  while (last_identical_rowid)
  {
    /*
      Current record (the one we've returned in previous call) was obtained
      from a rowid that matched multiple range_ids. Return this record again,
      with next matching range_id.
    */
    (void)rowid_buffer->read();

    if (rowid_buffer->read_ptr1 == last_identical_rowid)
      last_identical_rowid= NULL; /* reached the last of identical rowids */

    if (!is_mrr_assoc)
      return 0;

    memcpy(range_info, rowid_buffer->read_ptr2, sizeof(range_id_t));
    if (!index_reader->skip_record(*range_info, rowid_buffer->read_ptr1))
      return 0;
  }
  
  /* 
     Ok, last_identical_rowid==NULL, it's time to read next different rowid
     value and get record for it.
  */
  for(;;)
  {
    /* Return eof if there are no rowids in the buffer after re-fill attempt */
    if (rowid_buffer->read())
      return HA_ERR_END_OF_FILE;

    if (is_mrr_assoc)
    {
      memcpy(range_info, rowid_buffer->read_ptr2, sizeof(range_id_t));
      if (index_reader->skip_record(*range_info, rowid_buffer->read_ptr1))
        continue;
    }

    res= file->ha_rnd_pos(file->get_table()->record[0], 
                          rowid_buffer->read_ptr1);

    if (res == HA_ERR_RECORD_DELETED)
    {
      /* not likely to get this code with current storage engines, but still */
      continue;
    }

    if (res)
      return res; /* Some fatal error */

    break; /* Got another record */
  }

  /* 
    Check if subsequent buffer elements have the same rowid value as this
    one. If yes, remember this fact so that we don't make any more rnd_pos()
    calls with this value.

    Note: this implies that SQL layer doesn't touch table->record[0]
    between calls.
  */
  Lifo_buffer_iterator it;
  it.init(rowid_buffer);
  while (!it.read())
  {
    if (file->cmp_ref(it.read_ptr1, rowid_buffer->read_ptr1))
      break;
    last_identical_rowid= it.read_ptr1;
  }
  return 0;
}


/****************************************************************************
 * Top-level DS-MRR implementation functions (the ones called by storage engine)
 ***************************************************************************/

/**
  DS-MRR: Initialize and start MRR scan

  Initialize and start the MRR scan. Depending on the mode parameter, this
  may use default or DS-MRR implementation.

  @param h_arg           Table handler to be used
  @param key             Index to be used
  @param seq_funcs       Interval sequence enumeration functions
  @param seq_init_param  Interval sequence enumeration parameter
  @param n_ranges        Number of ranges in the sequence.
  @param mode            HA_MRR_* modes to use
  @param buf             INOUT Buffer to use

  @retval 0     Ok, Scan started.
  @retval other Error
*/

int DsMrr_impl::dsmrr_init(handler *h_arg, RANGE_SEQ_IF *seq_funcs, 
                           void *seq_init_param, uint n_ranges, uint mode,
                           HANDLER_BUFFER *buf)
{
  THD *thd= h_arg->get_table()->in_use;
  int res;
  Key_parameters keypar;
  uint UNINIT_VAR(key_buff_elem_size); /* set/used when do_sort_keys==TRUE */
  handler *h_idx;
  Mrr_ordered_rndpos_reader *disk_strategy= NULL;
  bool do_sort_keys= FALSE;
  DBUG_ENTER("DsMrr_impl::dsmrr_init");
  /*
    index_merge may invoke a scan on an object for which dsmrr_info[_const]
    has not been called, so set the owner handler here as well.
  */
  primary_file= h_arg;
  is_mrr_assoc= !MY_TEST(mode & HA_MRR_NO_ASSOCIATION);

  strategy_exhausted= FALSE;
  
  /* By default, have do-nothing buffer manager */
  buf_manager.arg= this;
  buf_manager.reset_buffer_sizes= do_nothing;
  buf_manager.redistribute_buffer_space= do_nothing;

  if (mode & (HA_MRR_USE_DEFAULT_IMPL | HA_MRR_SORTED))
    goto use_default_impl;
  
  /*
    Determine whether we'll need to do key sorting and/or rnd_pos() scan
  */
  index_strategy= NULL;
  if ((mode & HA_MRR_SINGLE_POINT) &&
      optimizer_flag(thd, OPTIMIZER_SWITCH_MRR_SORT_KEYS))
  {
    do_sort_keys= TRUE;
    index_strategy= &reader_factory.ordered_index_reader;
  }
  else
    index_strategy= &reader_factory.simple_index_reader;

  strategy= index_strategy;
  /*
    We don't need a rowid-to-rndpos step if
     - We're doing a scan on clustered primary key
     - [In the future] We're doing an index_only read
  */
  DBUG_ASSERT(primary_file->inited == handler::INDEX || 
              (primary_file->inited == handler::RND && 
               secondary_file && 
               secondary_file->inited == handler::INDEX));

  h_idx= (primary_file->inited == handler::INDEX)? primary_file: secondary_file;
  keyno= h_idx->active_index;

  if (!(keyno == table->s->primary_key && h_idx->primary_key_is_clustered()))
  {
    strategy= disk_strategy= &reader_factory.ordered_rndpos_reader;
  }

  full_buf= buf->buffer;
  full_buf_end= buf->buffer_end;

  if (do_sort_keys)
  {
    /* Pre-calculate some parameters of key sorting */
    keypar.use_key_pointers= MY_TEST(mode & HA_MRR_MATERIALIZED_KEYS);
    seq_funcs->get_key_info(seq_init_param, &keypar.key_tuple_length, 
                            &keypar.key_tuple_map);
    keypar.key_size_in_keybuf= keypar.use_key_pointers? 
                                 sizeof(char*) : keypar.key_tuple_length;
    key_buff_elem_size= keypar.key_size_in_keybuf + (int)is_mrr_assoc * sizeof(void*);
    
    /* Ordered index reader needs some space to store an index tuple */
    if (strategy != index_strategy)
    {
      uint saved_pk_length=0;
      if (h_idx->primary_key_is_clustered())
      {
        uint pk= h_idx->get_table()->s->primary_key;
        if (pk != MAX_KEY)
          saved_pk_length= h_idx->get_table()->key_info[pk].key_length;
      }
      
      KEY *used_index= &h_idx->get_table()->key_info[h_idx->active_index];
      if (reader_factory.ordered_index_reader.
            set_interruption_temp_buffer(primary_file->ref_length,
                                         used_index->key_length,
                                         saved_pk_length,
                                         &full_buf, full_buf_end))
        goto use_default_impl;
    }
    else
      reader_factory.ordered_index_reader.set_no_interruption_temp_buffer();
  }

  if (strategy == index_strategy)
  {
    /* 
      Index strategy alone handles the record retrieval. Give all buffer space
      to it. Key buffer should have forward orientation so we can return the
      end of it.
    */
    key_buffer= &forward_key_buf;
    key_buffer->set_buffer_space(full_buf, full_buf_end);
    
    /* Safety: specify that rowid buffer has zero size: */
    rowid_buffer.set_buffer_space(full_buf_end, full_buf_end);

    if (do_sort_keys && !key_buffer->have_space_for(key_buff_elem_size))
      goto use_default_impl;

    if ((res= index_strategy->init(primary_file, seq_funcs, seq_init_param, n_ranges,
                                   mode, &keypar, key_buffer, &buf_manager)))
      goto error;
  }
  else
  {
    /* We'll have both index and rndpos strategies working together */
    if (do_sort_keys)
    {
      /* Both strategies will need buffer space, share the buffer */
      if (setup_buffer_sharing(keypar.key_size_in_keybuf, keypar.key_tuple_map))
        goto use_default_impl;

      buf_manager.reset_buffer_sizes= reset_buffer_sizes;
      buf_manager.redistribute_buffer_space= redistribute_buffer_space;
    }
    else
    {
      /* index strategy doesn't need buffer, give all space to rowids*/
      rowid_buffer.set_buffer_space(full_buf, full_buf_end);
      if (!rowid_buffer.have_space_for(primary_file->ref_length + 
                                       (int)is_mrr_assoc * sizeof(range_id_t)))
        goto use_default_impl;
    }

    if ((res= setup_two_handlers()))
      goto error;

    if ((res= index_strategy->init(secondary_file, seq_funcs, seq_init_param,
                                   n_ranges, mode, &keypar, key_buffer, 
                                   &buf_manager)) || 
        (res= disk_strategy->init(primary_file, index_strategy, mode, 
                                  &rowid_buffer)))
    {
      goto error;
    }
  }
  
  /* 
    At this point, we're sure that we're running a native MRR scan (i.e. we
    didnt fall back to default implementation for some reason).
  */
  status_var_increment(thd->status_var.ha_mrr_init_count);

  res= strategy->refill_buffer(TRUE);
  if (res)
  {
    if (res != HA_ERR_END_OF_FILE)
      goto error;
    strategy_exhausted= TRUE;
  }

  /*
    If we have scanned through all intervals in *seq, then adjust *buf to 
    indicate that the remaining buffer space will not be used.
  */
//  if (dsmrr_eof) 
//    buf->end_of_used_area= rowid_buffer.end_of_space();

  
  DBUG_RETURN(0);
error:
  close_second_handler();
   /* Safety, not really needed but: */
  strategy= NULL;
  DBUG_RETURN(res);

use_default_impl:
  if (primary_file->inited != handler::INDEX)
  {
    /* We can get here when 
       - we've previously successfully done a DS-MRR scan (and so have 
         secondary_file!= NULL, secondary_file->inited= INDEX, 
         primary_file->inited=RND)
       - for this invocation, we haven't got enough buffer space, and so we
         have to use the default MRR implementation.

      note: primary_file->ha_index_end() will call dsmrr_close() which will
      close/destroy the secondary_file, this is intentional. 
      (Yes this is slow, but one can't expect performance with join buffer 
       so small that it can accomodate one rowid and one index tuple)
    */
    if ((res= primary_file->ha_rnd_end()) || 
        (res= primary_file->ha_index_init(keyno, MY_TEST(mode & HA_MRR_SORTED))))
    {
      DBUG_RETURN(res);
    }
  }
  /* Call correct init function and assign to top level object */
  Mrr_simple_index_reader *s= &reader_factory.simple_index_reader;
  res= s->init(primary_file, seq_funcs, seq_init_param, n_ranges, mode, NULL, 
               NULL, NULL);
  strategy= s;
  DBUG_RETURN(res);
}


/*
  Whatever the current state is, make it so that we have two handler objects:
  - primary_file       -  initialized for rnd_pos() scan
  - secondary_file     -  initialized for scanning the index specified in
                          this->keyno
  RETURN 
    0        OK
    HA_XXX   Error code
*/

int DsMrr_impl::setup_two_handlers()
{
  int res;
  THD *thd= primary_file->get_table()->in_use;
  DBUG_ENTER("DsMrr_impl::setup_two_handlers");
  if (!secondary_file)
  {
    handler *new_h2;
    Item *pushed_cond= NULL;
    DBUG_ASSERT(primary_file->inited == handler::INDEX);
    /* Create a separate handler object to do rnd_pos() calls. */
    /*
      ::clone() takes up a lot of stack, especially on 64 bit platforms.
      The constant 5 is an empiric result.
    */
    if (check_stack_overrun(thd, 5*STACK_MIN_SIZE, (uchar*) &new_h2))
      DBUG_RETURN(1);

    /* Create a separate handler object to do rnd_pos() calls. */
    if (!(new_h2= primary_file->clone(primary_file->get_table()->s->
                                      normalized_path.str,
                                      thd->mem_root)) || 
        new_h2->ha_external_lock(thd, F_RDLCK))
    {
      delete new_h2;
      DBUG_RETURN(1);
    }

    if (keyno == primary_file->pushed_idx_cond_keyno)
      pushed_cond= primary_file->pushed_idx_cond;
    
    Mrr_reader *save_strategy= strategy;
    strategy= NULL;
    /*
      Caution: this call will invoke this->dsmrr_close(). Do not put the
      created secondary table handler new_h2 into this->secondary_file or it 
      will delete it. Also, save the picked strategy
    */
    res= primary_file->ha_index_end();

    strategy= save_strategy;
    secondary_file= new_h2;

    if (res || (res= (primary_file->ha_rnd_init(FALSE))))
      goto error;

    table->prepare_for_position();
    secondary_file->extra(HA_EXTRA_KEYREAD);
    secondary_file->mrr_iter= primary_file->mrr_iter;

    if ((res= secondary_file->ha_index_init(keyno, FALSE)))
      goto error;

    if (pushed_cond)
      secondary_file->idx_cond_push(keyno, pushed_cond);
  }
  else
  {
    DBUG_ASSERT(secondary_file && secondary_file->inited==handler::INDEX);
    /* 
      We get here when the access alternates betwen MRR scan(s) and non-MRR
      scans.

      Calling primary_file->index_end() will invoke dsmrr_close() for this object,
      which will delete secondary_file. We need to keep it, so put it away and dont
      let it be deleted:
    */
    if (primary_file->inited == handler::INDEX)
    {
      handler *save_h2= secondary_file;
      Mrr_reader *save_strategy= strategy;
      secondary_file= NULL;
      strategy= NULL;
      res= primary_file->ha_index_end();
      secondary_file= save_h2;
      strategy= save_strategy;
      if (res)
        goto error;
    }
    if ((primary_file->inited != handler::RND) && 
        (res= primary_file->ha_rnd_init(FALSE)))
      goto error;
  }
  DBUG_RETURN(0);

error:
  DBUG_RETURN(res);
}


void DsMrr_impl::close_second_handler()
{
  if (secondary_file)
  {
    secondary_file->extra(HA_EXTRA_NO_KEYREAD);
    secondary_file->ha_index_or_rnd_end();
    secondary_file->ha_external_lock(current_thd, F_UNLCK);
    secondary_file->ha_close();
    delete secondary_file;
    secondary_file= NULL;
  }
}


void DsMrr_impl::dsmrr_close()
{
  DBUG_ENTER("DsMrr_impl::dsmrr_close");
  close_second_handler();
  strategy= NULL;
  DBUG_VOID_RETURN;
}


/* 
  my_qsort2-compatible static member function to compare key tuples 
*/

int Mrr_ordered_index_reader::compare_keys(void* arg, uchar* key1_arg, 
                                           uchar* key2_arg)
{
  Mrr_ordered_index_reader *reader= (Mrr_ordered_index_reader*)arg;
  TABLE *table= reader->file->get_table();
  KEY_PART_INFO *part= table->key_info[reader->file->active_index].key_part;
  uchar *key1, *key2;
   
  if (reader->keypar.use_key_pointers)
  {
    /* the buffer stores pointers to keys, get to the keys */
    memcpy(&key1, key1_arg, sizeof(char*));
    memcpy(&key2, key2_arg, sizeof(char*));
  }
  else
  {
    key1= key1_arg;
    key2= key2_arg;
  }

  return key_tuple_cmp(part, key1, key2, reader->keypar.key_tuple_length);
}


int Mrr_ordered_index_reader::compare_keys_reverse(void* arg, uchar* key1, 
                                                   uchar* key2)
{
  return -compare_keys(arg, key1, key2);
}


/**
  Set the buffer space to be shared between rowid and key buffer

  @return FALSE  ok 
  @return TRUE   There is so little buffer space that we won't be able to use
                 the strategy. 
                 This happens when we don't have enough space for one rowid 
                 element and one key element so this is mainly targeted at
                 testing.
*/

bool DsMrr_impl::setup_buffer_sharing(uint key_size_in_keybuf, 
                                      key_part_map key_tuple_map)
{
  long key_buff_elem_size= key_size_in_keybuf + 
                           (int)is_mrr_assoc * sizeof(range_id_t);
  
  KEY *key_info= &primary_file->get_table()->key_info[keyno];
  /* 
    Ok if we got here we need to allocate one part of the buffer 
    for keys and another part for rowids.
  */
  ulonglong rowid_buf_elem_size= primary_file->ref_length + 
                                 (int)is_mrr_assoc * sizeof(range_id_t);
  
  /*
    Use rec_per_key statistics as a basis to find out how many rowids 
    we'll get for each key value.
     TODO: what should be the default value to use when there is no 
           statistics?
  */
  uint parts= my_count_bits(key_tuple_map);
  ha_rows rpc;
  ulonglong rowids_size= rowid_buf_elem_size;
  if ((rpc= (ha_rows) key_info->actual_rec_per_key(parts - 1)))
    rowids_size= rowid_buf_elem_size * rpc;

  double fraction_for_rowids=
    (ulonglong2double(rowids_size) / 
     (ulonglong2double(rowids_size) + key_buff_elem_size));

  ptrdiff_t bytes_for_rowids= 
    (ptrdiff_t)floor(0.5 + fraction_for_rowids * (full_buf_end - full_buf));
  
  ptrdiff_t bytes_for_keys= (full_buf_end - full_buf) - bytes_for_rowids;

  if (bytes_for_keys < key_buff_elem_size + 1 ||
      bytes_for_rowids < (ptrdiff_t)rowid_buf_elem_size + 1)
    return TRUE; /* Failed to provide minimum space for one of the buffers */

  rowid_buffer_end= full_buf + bytes_for_rowids;
  rowid_buffer.set_buffer_space(full_buf, rowid_buffer_end);
  key_buffer= &backward_key_buf;
  key_buffer->set_buffer_space(rowid_buffer_end, full_buf_end); 

  /* The above code guarantees that the buffers are big enough */
  DBUG_ASSERT(key_buffer->have_space_for(key_buff_elem_size) &&
              rowid_buffer.have_space_for((size_t)rowid_buf_elem_size));

  return FALSE;
}


void DsMrr_impl::do_nothing(void *dsmrr_arg)
{
  /* Do nothing */
}


void DsMrr_impl::reset_buffer_sizes(void *dsmrr_arg)
{
  DsMrr_impl *dsmrr= (DsMrr_impl*)dsmrr_arg;
  dsmrr->rowid_buffer.set_buffer_space(dsmrr->full_buf, 
                                       dsmrr->rowid_buffer_end);
  dsmrr->key_buffer->set_buffer_space(dsmrr->rowid_buffer_end, 
                                      dsmrr->full_buf_end);
}


/*
  Take unused space from the key buffer and give it to the rowid buffer
*/

void DsMrr_impl::redistribute_buffer_space(void *dsmrr_arg)
{
  DsMrr_impl *dsmrr= (DsMrr_impl*)dsmrr_arg;
  uchar *unused_start, *unused_end;
  dsmrr->key_buffer->remove_unused_space(&unused_start, &unused_end);
  dsmrr->rowid_buffer.grow(unused_start, unused_end);
}


/*
  @brief Initialize the iterator
  
  @note
  Initialize the iterator to produce matches for the key of the first element 
  in owner_arg->key_buffer

  @retval  0                    OK
  @retval  HA_ERR_END_OF_FILE   Either the owner->key_buffer is empty or 
                                no matches for the key we've tried (check
                                key_buffer->is_empty() to tell these apart)
  @retval  other code           Fatal error
*/

int Key_value_records_iterator::init(Mrr_ordered_index_reader *owner_arg)
{
  int res;
  owner= owner_arg;

  identical_key_it.init(owner->key_buffer);
  owner->key_buffer->setup_reading(owner->keypar.key_size_in_keybuf,
                                   owner->is_mrr_assoc ? sizeof(void*) : 0);

  if (identical_key_it.read())
    return HA_ERR_END_OF_FILE;

  uchar *key_in_buf= last_identical_key_ptr= identical_key_it.read_ptr1;

  uchar *index_tuple= key_in_buf;
  if (owner->keypar.use_key_pointers)
    memcpy(&index_tuple, key_in_buf, sizeof(char*));
  
  /* Check out how many more identical keys are following */
  while (!identical_key_it.read())
  {
    if (Mrr_ordered_index_reader::compare_keys(owner, key_in_buf, 
                                               identical_key_it.read_ptr1))
      break;
    last_identical_key_ptr= identical_key_it.read_ptr1;
  }
  identical_key_it.init(owner->key_buffer);
  res= owner->file->ha_index_read_map(owner->file->get_table()->record[0], 
                                      index_tuple, 
                                      owner->keypar.key_tuple_map, 
                                      HA_READ_KEY_EXACT);

  if (res)
  {
    /* Failed to find any matching records */
    move_to_next_key_value();
    return res;
  }
  owner->have_saved_rowid= FALSE;
  get_next_row= FALSE;
  return 0;
}


int Key_value_records_iterator::get_next(range_id_t *range_info)
{
  int res;

  if (get_next_row)
  {
    if (owner->keypar.index_ranges_unique)
    {
      /* We're using a full unique key, no point to call index_next_same */
      return HA_ERR_END_OF_FILE;
    }
    
    handler *h= owner->file;
    uchar *lookup_key;
    if (owner->keypar.use_key_pointers)
      memcpy(&lookup_key, identical_key_it.read_ptr1, sizeof(void*));
    else
      lookup_key= identical_key_it.read_ptr1;

    if ((res= h->ha_index_next_same(h->get_table()->record[0], 
                                    lookup_key,
                                    owner->keypar.key_tuple_length)))
    {
      /* It's either HA_ERR_END_OF_FILE or some other error */
      return res; 
    }
    identical_key_it.init(owner->key_buffer);
    owner->have_saved_rowid= FALSE;
    get_next_row= FALSE;
  }

  identical_key_it.read(); /* This gets us next range_id */
  memcpy(range_info, identical_key_it.read_ptr2, sizeof(range_id_t));

  if (!last_identical_key_ptr || 
      (identical_key_it.read_ptr1 == last_identical_key_ptr))
  {
    /* 
      We've reached the last of the identical keys that current record is a
      match for.  Set get_next_row=TRUE so that we read the next index record
      on the next call to this function.
    */
    get_next_row= TRUE;
  }
  return 0;
}


void Key_value_records_iterator::move_to_next_key_value()
{
  while (!owner->key_buffer->read() && 
         (owner->key_buffer->read_ptr1 != last_identical_key_ptr)) {}
}


/**
  DS-MRR implementation: multi_range_read_next() function.

  Calling convention is like multi_range_read_next() has.
*/

int DsMrr_impl::dsmrr_next(range_id_t *range_info)
{
  int res;
  if (strategy_exhausted)
    return HA_ERR_END_OF_FILE;

  while ((res= strategy->get_next(range_info)) == HA_ERR_END_OF_FILE)
  {
    if ((res= strategy->refill_buffer(FALSE)))
      break; /* EOF or error */
  }
  return res;
}


/**
  DS-MRR implementation: multi_range_read_info() function
*/
ha_rows DsMrr_impl::dsmrr_info(uint keyno, uint n_ranges, uint rows, 
                               uint key_parts,
                               uint *bufsz, uint *flags, Cost_estimate *cost)
{  
  ha_rows res __attribute__((unused));
  uint def_flags= *flags;
  uint def_bufsz= *bufsz;

  /* Get cost/flags/mem_usage of default MRR implementation */
  res= primary_file->handler::multi_range_read_info(keyno, n_ranges, rows,
                                                    key_parts, &def_bufsz, 
                                                    &def_flags, cost);
  DBUG_ASSERT(!res);

  if ((*flags & HA_MRR_USE_DEFAULT_IMPL) || 
      choose_mrr_impl(keyno, rows, flags, bufsz, cost))
  {
    /* Default implementation is choosen */
    DBUG_PRINT("info", ("Default MRR implementation choosen"));
    *flags= def_flags;
    *bufsz= def_bufsz;
  }
  else
  {
    /* *flags and *bufsz were set by choose_mrr_impl */
    DBUG_PRINT("info", ("DS-MRR implementation choosen"));
  }
  return 0;
}


/**
  DS-MRR Implementation: multi_range_read_info_const() function
*/

ha_rows DsMrr_impl::dsmrr_info_const(uint keyno, RANGE_SEQ_IF *seq,
                                 void *seq_init_param, uint n_ranges, 
                                 uint *bufsz, uint *flags, Cost_estimate *cost)
{
  ha_rows rows;
  uint def_flags= *flags;
  uint def_bufsz= *bufsz;
  /* Get cost/flags/mem_usage of default MRR implementation */
  rows= primary_file->handler::multi_range_read_info_const(keyno, seq, 
                                                           seq_init_param,
                                                           n_ranges, 
                                                           &def_bufsz, 
                                                           &def_flags, cost);
  if (rows == HA_POS_ERROR)
  {
    /* Default implementation can't perform MRR scan => we can't either */
    return rows;
  }

  /*
    If HA_MRR_USE_DEFAULT_IMPL has been passed to us, that is an order to
    use the default MRR implementation (we need it for UPDATE/DELETE).
    Otherwise, make a choice based on cost and @@optimizer_switch settings
  */
  if ((*flags & HA_MRR_USE_DEFAULT_IMPL) ||
      choose_mrr_impl(keyno, rows, flags, bufsz, cost))
  {
    DBUG_PRINT("info", ("Default MRR implementation choosen"));
    *flags= def_flags;
    *bufsz= def_bufsz;
  }
  else
  {
    /* *flags and *bufsz were set by choose_mrr_impl */
    DBUG_PRINT("info", ("DS-MRR implementation choosen"));
  }
  return rows;
}


/**
  Check if key has partially-covered columns

  We can't use DS-MRR to perform range scans when the ranges are over
  partially-covered keys, because we'll not have full key part values
  (we'll have their prefixes from the index) and will not be able to check
  if we've reached the end the range.

  @param keyno  Key to check

  @todo
    Allow use of DS-MRR in cases where the index has partially-covered
    components but they are not used for scanning.

  @retval TRUE   Yes
  @retval FALSE  No
*/

bool key_uses_partial_cols(TABLE_SHARE *share, uint keyno)
{
  KEY_PART_INFO *kp= share->key_info[keyno].key_part;
  KEY_PART_INFO *kp_end= kp + share->key_info[keyno].user_defined_key_parts;
  for (; kp != kp_end; kp++)
  {
    if (!kp->field->part_of_key.is_set(keyno))
      return TRUE;
  }
  return FALSE;
}


/*
  Check if key/flags allow DS-MRR/CPK strategy to be used
  
  @param thd
  @param keyno      Index that will be used
  @param  mrr_flags  
  
  @retval TRUE   DS-MRR/CPK should be used
  @retval FALSE  Otherwise
*/

bool DsMrr_impl::check_cpk_scan(THD *thd, TABLE_SHARE *share, uint keyno, 
                                uint mrr_flags)
{
  return MY_TEST((mrr_flags & HA_MRR_SINGLE_POINT) &&
                 keyno == share->primary_key &&
                 primary_file->primary_key_is_clustered() &&
                 optimizer_flag(thd, OPTIMIZER_SWITCH_MRR_SORT_KEYS));
}


/*
  DS-MRR Internals: Choose between Default MRR implementation and DS-MRR

  Make the choice between using Default MRR implementation and DS-MRR.
  This function contains common functionality factored out of dsmrr_info()
  and dsmrr_info_const(). The function assumes that the default MRR
  implementation's applicability requirements are satisfied.

  @param keyno       Index number
  @param rows        E(full rows to be retrieved)
  @param flags  IN   MRR flags provided by the MRR user
                OUT  If DS-MRR is choosen, flags of DS-MRR implementation
                     else the value is not modified
  @param bufsz  IN   If DS-MRR is choosen, buffer use of DS-MRR implementation
                     else the value is not modified
  @param cost   IN   Cost of default MRR implementation
                OUT  If DS-MRR is choosen, cost of DS-MRR scan
                     else the value is not modified

  @retval TRUE   Default MRR implementation should be used
  @retval FALSE  DS-MRR implementation should be used
*/


bool DsMrr_impl::choose_mrr_impl(uint keyno, ha_rows rows, uint *flags,
                                 uint *bufsz, Cost_estimate *cost)
{
  Cost_estimate dsmrr_cost;
  bool res;
  THD *thd= primary_file->get_table()->in_use;
  TABLE_SHARE *share= primary_file->get_table_share();

  bool doing_cpk_scan= check_cpk_scan(thd, share, keyno, *flags); 
  bool using_cpk= MY_TEST(keyno == share->primary_key &&
                          primary_file->primary_key_is_clustered());
  *flags &= ~HA_MRR_IMPLEMENTATION_FLAGS;
  if (!optimizer_flag(thd, OPTIMIZER_SWITCH_MRR) ||
      *flags & HA_MRR_INDEX_ONLY ||
      (using_cpk && !doing_cpk_scan) || key_uses_partial_cols(share, keyno))
  {
    /* Use the default implementation */
    *flags |= HA_MRR_USE_DEFAULT_IMPL;
    *flags &= ~HA_MRR_IMPLEMENTATION_FLAGS;
    return TRUE;
  }

  uint add_len= share->key_info[keyno].key_length + primary_file->ref_length; 
  *bufsz -= add_len;
  if (get_disk_sweep_mrr_cost(keyno, rows, *flags, bufsz, &dsmrr_cost))
    return TRUE;
  *bufsz += add_len;
  
  bool force_dsmrr;
  /* 
    If mrr_cost_based flag is not set, then set cost of DS-MRR to be minimum of
    DS-MRR and Default implementations cost. This allows one to force use of
    DS-MRR whenever it is applicable without affecting other cost-based
    choices.
  */
  if ((force_dsmrr= !optimizer_flag(thd, OPTIMIZER_SWITCH_MRR_COST_BASED)) &&
      dsmrr_cost.total_cost() > cost->total_cost())
    dsmrr_cost= *cost;

  if (force_dsmrr || dsmrr_cost.total_cost() <= cost->total_cost())
  {
    *flags &= ~HA_MRR_USE_DEFAULT_IMPL;  /* Use the DS-MRR implementation */
    *flags &= ~HA_MRR_SORTED;          /* We will return unordered output */
    *cost= dsmrr_cost;
    res= FALSE;


    if ((using_cpk && doing_cpk_scan) ||
        (optimizer_flag(thd, OPTIMIZER_SWITCH_MRR_SORT_KEYS) &&
         *flags & HA_MRR_SINGLE_POINT))
    {
      *flags |= DSMRR_IMPL_SORT_KEYS;
    }
    
    if (!(using_cpk && doing_cpk_scan) &&
        !(*flags & HA_MRR_INDEX_ONLY))
    {
      *flags |= DSMRR_IMPL_SORT_ROWIDS;
    }
    /*
    if ((*flags & HA_MRR_SINGLE_POINT) && 
         optimizer_flag(thd, OPTIMIZER_SWITCH_MRR_SORT_KEYS))
      *flags |= HA_MRR_MATERIALIZED_KEYS;
    */
  }
  else
  {
    /* Use the default MRR implementation */
    res= TRUE;
  }
  return res;
}

/*
  Take the flags we've returned previously and print one of
  - Key-ordered scan
  - Rowid-ordered scan
  - Key-ordered Rowid-ordered scan
*/

int DsMrr_impl::dsmrr_explain_info(uint mrr_mode, char *str, size_t size)
{
  const char *key_ordered=   "Key-ordered scan";
  const char *rowid_ordered= "Rowid-ordered scan";
  const char *both_ordered=  "Key-ordered Rowid-ordered scan";
  const char *used_str="";
  const uint BOTH_FLAGS= (DSMRR_IMPL_SORT_KEYS | DSMRR_IMPL_SORT_ROWIDS);

  if (!(mrr_mode & HA_MRR_USE_DEFAULT_IMPL))
  {
    if ((mrr_mode & BOTH_FLAGS) == BOTH_FLAGS)
      used_str= both_ordered;
    else if (mrr_mode & DSMRR_IMPL_SORT_KEYS)
      used_str= key_ordered;
    else if (mrr_mode & DSMRR_IMPL_SORT_ROWIDS)
      used_str= rowid_ordered;

    size_t used_str_len= strlen(used_str);
    size_t copy_len= MY_MIN(used_str_len, size);
    memcpy(str, used_str, copy_len);
    return (int)copy_len;
  }
  return 0;
}


static void get_sort_and_sweep_cost(TABLE *table, ha_rows nrows, Cost_estimate *cost);


/**
  Get cost of DS-MRR scan

  @param keynr              Index to be used
  @param rows               E(Number of rows to be scanned)
  @param flags              Scan parameters (HA_MRR_* flags)
  @param buffer_size INOUT  Buffer size
  @param cost        OUT    The cost

  @retval FALSE  OK
  @retval TRUE   Error, DS-MRR cannot be used (the buffer is too small
                 for even 1 rowid)
*/

bool DsMrr_impl::get_disk_sweep_mrr_cost(uint keynr, ha_rows rows, uint flags,
                                         uint *buffer_size, Cost_estimate *cost)
{
  ulong max_buff_entries, elem_size;
  ha_rows rows_in_full_step;
  ha_rows rows_in_last_step;
  uint n_full_steps;
  double index_read_cost;

  elem_size= primary_file->ref_length + 
             sizeof(void*) * (!MY_TEST(flags & HA_MRR_NO_ASSOCIATION));
  max_buff_entries = *buffer_size / elem_size;

  if (!max_buff_entries)
    return TRUE; /* Buffer has not enough space for even 1 rowid */

  /* Number of iterations we'll make with full buffer */
  n_full_steps= (uint)floor(rows2double(rows) / max_buff_entries);
  
  /* 
    Get numbers of rows we'll be processing in 
     - non-last sweep, with full buffer 
     - last iteration, with non-full buffer
  */
  rows_in_full_step= max_buff_entries;
  rows_in_last_step= rows % max_buff_entries;
  
  /* Adjust buffer size if we expect to use only part of the buffer */
  if (n_full_steps)
  {
    get_sort_and_sweep_cost(table, rows_in_full_step, cost);
    cost->multiply(n_full_steps);
  }
  else
  {
    cost->reset();
    *buffer_size= (uint)MY_MAX(*buffer_size, 
                      (size_t)(1.2*rows_in_last_step) * elem_size + 
                      primary_file->ref_length + table->key_info[keynr].key_length);
  }
  
  Cost_estimate last_step_cost;
  get_sort_and_sweep_cost(table, rows_in_last_step, &last_step_cost);
  cost->add(&last_step_cost);
 
  if (n_full_steps != 0)
    cost->mem_cost= *buffer_size;
  else
    cost->mem_cost= (double)rows_in_last_step * elem_size;
  
  /* Total cost of all index accesses */
  index_read_cost= primary_file->keyread_time(keynr, 1, rows);
  cost->add_io(index_read_cost, 1 /* Random seeks */);
  return FALSE;
}


/* 
  Get cost of one sort-and-sweep step
  
  It consists of two parts:
   - sort an array of #nrows ROWIDs using qsort
   - read #nrows records from table in a sweep.

  @param table       Table being accessed
  @param nrows       Number of rows to be sorted and retrieved
  @param cost   OUT  The cost of scan
*/

static 
void get_sort_and_sweep_cost(TABLE *table, ha_rows nrows, Cost_estimate *cost)
{
  if (nrows)
  {
    get_sweep_read_cost(table, nrows, FALSE, cost);
    /* Add cost of qsort call: n * log2(n) * cost(rowid_comparison) */
    double cmp_op= rows2double(nrows) * (1.0 / TIME_FOR_COMPARE_ROWID);
    if (cmp_op < 3)
      cmp_op= 3;
    cost->cpu_cost += cmp_op * log2(cmp_op);
  }
  else
    cost->reset();
}


/**
  Get cost of reading nrows table records in a "disk sweep"

  A disk sweep read is a sequence of handler->rnd_pos(rowid) calls that made
  for an ordered sequence of rowids.

  We assume hard disk IO. The read is performed as follows:

   1. The disk head is moved to the needed cylinder
   2. The controller waits for the plate to rotate
   3. The data is transferred

  Time to do #3 is insignificant compared to #2+#1.

  Time to move the disk head is proportional to head travel distance.

  Time to wait for the plate to rotate depends on whether the disk head
  was moved or not. 

  If disk head wasn't moved, the wait time is proportional to distance
  between the previous block and the block we're reading.

  If the head was moved, we don't know how much we'll need to wait for the
  plate to rotate. We assume the wait time to be a variate with a mean of
  0.5 of full rotation time.

  Our cost units are "random disk seeks". The cost of random disk seek is
  actually not a constant, it depends one range of cylinders we're going
  to access. We make it constant by introducing a fuzzy concept of "typical 
  datafile length" (it's fuzzy as it's hard to tell whether it should
  include index file, temp.tables etc). Then random seek cost is:

    1 = half_rotation_cost + move_cost * 1/3 * typical_data_file_length

  We define half_rotation_cost as DISK_SEEK_BASE_COST=0.9.

  @param table             Table to be accessed
  @param nrows             Number of rows to retrieve
  @param interrupted       TRUE <=> Assume that the disk sweep will be
                           interrupted by other disk IO. FALSE - otherwise.
  @param cost         OUT  The cost.
*/

void get_sweep_read_cost(TABLE *table, ha_rows nrows, bool interrupted, 
                         Cost_estimate *cost)
{
  DBUG_ENTER("get_sweep_read_cost");

  cost->reset();
  if (table->file->primary_key_is_clustered())
  {
    cost->io_count= table->file->read_time(table->s->primary_key,
                                           (uint) nrows, nrows);
  }
  else
  {
    double n_blocks=
      ceil(ulonglong2double(table->file->stats.data_file_length) / IO_SIZE);
    double busy_blocks=
      n_blocks * (1.0 - pow(1.0 - 1.0/n_blocks, rows2double(nrows)));
    if (busy_blocks < 1.0)
      busy_blocks= 1.0;

    DBUG_PRINT("info",("sweep: nblocks=%g, busy_blocks=%g", n_blocks,
                       busy_blocks));
    cost->io_count= busy_blocks;

    if (!interrupted)
    {
      /* Assume reading is done in one 'sweep' */
      cost->avg_io_cost= (DISK_SEEK_BASE_COST +
                          DISK_SEEK_PROP_COST*n_blocks/busy_blocks);
    }
  }
  DBUG_PRINT("info",("returning cost=%g", cost->total_cost()));
  DBUG_VOID_RETURN;
}


/* **************************************************************************
 * DS-MRR implementation ends
 ***************************************************************************/