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
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
|
/*****************************************************************************
Copyright (c) 1997, 2017, Oracle and/or its affiliates. All Rights Reserved.
Copyright (c) 2008, Google Inc.
Copyright (c) 2015, 2019, MariaDB Corporation.
Portions of this file contain modifications contributed and copyrighted by
Google, Inc. Those modifications are gratefully acknowledged and are described
briefly in the InnoDB documentation. The contributions by Google are
incorporated with their permission, and subject to the conditions contained in
the file COPYING.Google.
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 02110-1335 USA
*****************************************************************************/
/***************************************************//**
@file row/row0sel.cc
Select
Created 12/19/1997 Heikki Tuuri
*******************************************************/
#include "row0sel.h"
#include "dict0dict.h"
#include "dict0boot.h"
#include "trx0undo.h"
#include "trx0trx.h"
#include "btr0btr.h"
#include "btr0cur.h"
#include "btr0sea.h"
#include "gis0rtree.h"
#include "mach0data.h"
#include "que0que.h"
#include "row0upd.h"
#include "row0row.h"
#include "row0vers.h"
#include "rem0cmp.h"
#include "lock0lock.h"
#include "eval0eval.h"
#include "pars0sym.h"
#include "pars0pars.h"
#include "row0mysql.h"
#include "buf0lru.h"
#include "srv0srv.h"
#include "srv0mon.h"
#ifdef WITH_WSREP
#include "mysql/service_wsrep.h" /* For wsrep_thd_skip_locking */
#endif
/* Maximum number of rows to prefetch; MySQL interface has another parameter */
#define SEL_MAX_N_PREFETCH 16
/* Number of rows fetched, after which to start prefetching; MySQL interface
has another parameter */
#define SEL_PREFETCH_LIMIT 1
/* When a select has accessed about this many pages, it returns control back
to que_run_threads: this is to allow canceling runaway queries */
#define SEL_COST_LIMIT 100
/* Flags for search shortcut */
#define SEL_FOUND 0
#define SEL_EXHAUSTED 1
#define SEL_RETRY 2
/********************************************************************//**
Returns TRUE if the user-defined column in a secondary index record
is alphabetically the same as the corresponding BLOB column in the clustered
index record.
NOTE: the comparison is NOT done as a binary comparison, but character
fields are compared with collation!
@return TRUE if the columns are equal */
static
ibool
row_sel_sec_rec_is_for_blob(
/*========================*/
ulint mtype, /*!< in: main type */
ulint prtype, /*!< in: precise type */
ulint mbminlen, /*!< in: minimum length of
a character, in bytes */
ulint mbmaxlen, /*!< in: maximum length of
a character, in bytes */
const byte* clust_field, /*!< in: the locally stored part of
the clustered index column, including
the BLOB pointer; the clustered
index record must be covered by
a lock or a page latch to protect it
against deletion (rollback or purge) */
ulint clust_len, /*!< in: length of clust_field */
const byte* sec_field, /*!< in: column in secondary index */
ulint sec_len, /*!< in: length of sec_field */
ulint prefix_len, /*!< in: index column prefix length
in bytes */
dict_table_t* table) /*!< in: table */
{
ulint len;
byte buf[REC_VERSION_56_MAX_INDEX_COL_LEN];
/* This function should never be invoked on tables in
ROW_FORMAT=REDUNDANT or ROW_FORMAT=COMPACT, because they
should always contain enough prefix in the clustered index record. */
ut_ad(dict_table_has_atomic_blobs(table));
ut_a(clust_len >= BTR_EXTERN_FIELD_REF_SIZE);
ut_ad(prefix_len >= sec_len);
ut_ad(prefix_len > 0);
ut_a(prefix_len <= sizeof buf);
if (!memcmp(clust_field + clust_len - BTR_EXTERN_FIELD_REF_SIZE,
field_ref_zero, BTR_EXTERN_FIELD_REF_SIZE)) {
/* The externally stored field was not written yet.
This record should only be seen by
recv_recovery_rollback_active() or any
TRX_ISO_READ_UNCOMMITTED transactions. */
return(FALSE);
}
len = btr_copy_externally_stored_field_prefix(
buf, prefix_len, table->space->zip_size(),
clust_field, clust_len);
if (len == 0) {
/* The BLOB was being deleted as the server crashed.
There should not be any secondary index records
referring to this clustered index record, because
btr_free_externally_stored_field() is called after all
secondary index entries of the row have been purged. */
return(FALSE);
}
len = dtype_get_at_most_n_mbchars(prtype, mbminlen, mbmaxlen,
prefix_len, len, (const char*) buf);
return(!cmp_data_data(mtype, prtype, buf, len, sec_field, sec_len));
}
/** Returns TRUE if the user-defined column values in a secondary index record
are alphabetically the same as the corresponding columns in the clustered
index record.
NOTE: the comparison is NOT done as a binary comparison, but character
fields are compared with collation!
@param[in] sec_rec secondary index record
@param[in] sec_index secondary index
@param[in] clust_rec clustered index record;
must be protected by a page s-latch
@param[in] clust_index clustered index
@param[in] thr query thread
@return TRUE if the secondary record is equal to the corresponding
fields in the clustered record, when compared with collation;
FALSE if not equal or if the clustered record has been marked for deletion */
static
ibool
row_sel_sec_rec_is_for_clust_rec(
const rec_t* sec_rec,
dict_index_t* sec_index,
const rec_t* clust_rec,
dict_index_t* clust_index,
que_thr_t* thr)
{
const byte* sec_field;
ulint sec_len;
const byte* clust_field;
ulint n;
ulint i;
mem_heap_t* heap = NULL;
ulint clust_offsets_[REC_OFFS_NORMAL_SIZE];
ulint sec_offsets_[REC_OFFS_SMALL_SIZE];
ulint* clust_offs = clust_offsets_;
ulint* sec_offs = sec_offsets_;
ibool is_equal = TRUE;
VCOL_STORAGE* vcol_storage= 0;
byte* record;
rec_offs_init(clust_offsets_);
rec_offs_init(sec_offsets_);
if (rec_get_deleted_flag(clust_rec,
dict_table_is_comp(clust_index->table))) {
/* In delete-marked records, DB_TRX_ID must
always refer to an existing undo log record. */
ut_ad(rec_get_trx_id(clust_rec, clust_index));
/* The clustered index record is delete-marked;
it is not visible in the read view. Besides,
if there are any externally stored columns,
some of them may have already been purged. */
return(FALSE);
}
heap = mem_heap_create(256);
clust_offs = rec_get_offsets(clust_rec, clust_index, clust_offs,
true, ULINT_UNDEFINED, &heap);
sec_offs = rec_get_offsets(sec_rec, sec_index, sec_offs,
true, ULINT_UNDEFINED, &heap);
n = dict_index_get_n_ordering_defined_by_user(sec_index);
for (i = 0; i < n; i++) {
const dict_field_t* ifield;
const dict_col_t* col;
ulint clust_pos = 0;
ulint clust_len = 0;
ulint len;
bool is_virtual;
ifield = dict_index_get_nth_field(sec_index, i);
col = dict_field_get_col(ifield);
is_virtual = col->is_virtual();
/* For virtual column, its value will need to be
reconstructed from base column in cluster index */
if (is_virtual) {
const dict_v_col_t* v_col;
dfield_t* vfield;
row_ext_t* ext;
if (!vcol_storage)
{
TABLE *mysql_table= thr->prebuilt->m_mysql_table;
innobase_allocate_row_for_vcol(thr_get_trx(thr)->mysql_thd,
clust_index,
&heap,
&mysql_table,
&record,
&vcol_storage);
}
v_col = reinterpret_cast<const dict_v_col_t*>(col);
dtuple_t* row = row_build(
ROW_COPY_POINTERS,
clust_index, clust_rec,
clust_offs,
NULL, NULL, NULL, &ext, heap);
vfield = innobase_get_computed_value(
row, v_col, clust_index,
&heap, NULL, NULL,
thr_get_trx(thr)->mysql_thd,
thr->prebuilt->m_mysql_table,
record, NULL, NULL, NULL);
clust_len = vfield->len;
clust_field = static_cast<byte*>(vfield->data);
} else {
clust_pos = dict_col_get_clust_pos(col, clust_index);
clust_field = rec_get_nth_cfield(
clust_rec, clust_index, clust_offs,
clust_pos, &clust_len);
}
sec_field = rec_get_nth_field(sec_rec, sec_offs, i, &sec_len);
len = clust_len;
if (ifield->prefix_len > 0 && len != UNIV_SQL_NULL
&& sec_len != UNIV_SQL_NULL && !is_virtual) {
if (rec_offs_nth_extern(clust_offs, clust_pos)) {
len -= BTR_EXTERN_FIELD_REF_SIZE;
}
len = dtype_get_at_most_n_mbchars(
col->prtype, col->mbminlen, col->mbmaxlen,
ifield->prefix_len, len, (char*) clust_field);
if (rec_offs_nth_extern(clust_offs, clust_pos)
&& len < sec_len) {
if (!row_sel_sec_rec_is_for_blob(
col->mtype, col->prtype,
col->mbminlen, col->mbmaxlen,
clust_field, clust_len,
sec_field, sec_len,
ifield->prefix_len,
clust_index->table)) {
goto inequal;
}
continue;
}
}
/* For spatial index, the first field is MBR, we check
if the MBR is equal or not. */
if (dict_index_is_spatial(sec_index) && i == 0) {
rtr_mbr_t tmp_mbr;
rtr_mbr_t sec_mbr;
byte* dptr =
const_cast<byte*>(clust_field);
ut_ad(clust_len != UNIV_SQL_NULL);
/* For externally stored field, we need to get full
geo data to generate the MBR for comparing. */
if (rec_offs_nth_extern(clust_offs, clust_pos)) {
dptr = btr_copy_externally_stored_field(
&clust_len, dptr,
clust_index->table->space->zip_size(),
len, heap);
}
rtree_mbr_from_wkb(dptr + GEO_DATA_HEADER_SIZE,
static_cast<uint>(clust_len
- GEO_DATA_HEADER_SIZE),
SPDIMS,
reinterpret_cast<double*>(
&tmp_mbr));
rtr_read_mbr(sec_field, &sec_mbr);
if (!MBR_EQUAL_CMP(&sec_mbr, &tmp_mbr)) {
is_equal = FALSE;
goto func_exit;
}
} else {
if (0 != cmp_data_data(col->mtype, col->prtype,
clust_field, len,
sec_field, sec_len)) {
inequal:
is_equal = FALSE;
goto func_exit;
}
}
}
func_exit:
if (UNIV_LIKELY_NULL(heap)) {
if (UNIV_LIKELY_NULL(vcol_storage))
innobase_free_row_for_vcol(vcol_storage);
mem_heap_free(heap);
}
return(is_equal);
}
/*********************************************************************//**
Creates a select node struct.
@return own: select node struct */
sel_node_t*
sel_node_create(
/*============*/
mem_heap_t* heap) /*!< in: memory heap where created */
{
sel_node_t* node;
node = static_cast<sel_node_t*>(
mem_heap_alloc(heap, sizeof(sel_node_t)));
node->common.type = QUE_NODE_SELECT;
node->state = SEL_NODE_OPEN;
node->plans = NULL;
return(node);
}
/*********************************************************************//**
Frees the memory private to a select node when a query graph is freed,
does not free the heap where the node was originally created. */
void
sel_node_free_private(
/*==================*/
sel_node_t* node) /*!< in: select node struct */
{
ulint i;
plan_t* plan;
if (node->plans != NULL) {
for (i = 0; i < node->n_tables; i++) {
plan = sel_node_get_nth_plan(node, i);
btr_pcur_close(&(plan->pcur));
btr_pcur_close(&(plan->clust_pcur));
if (plan->old_vers_heap) {
mem_heap_free(plan->old_vers_heap);
}
}
}
}
/*********************************************************************//**
Evaluates the values in a select list. If there are aggregate functions,
their argument value is added to the aggregate total. */
UNIV_INLINE
void
sel_eval_select_list(
/*=================*/
sel_node_t* node) /*!< in: select node */
{
que_node_t* exp;
exp = node->select_list;
while (exp) {
eval_exp(exp);
exp = que_node_get_next(exp);
}
}
/*********************************************************************//**
Assigns the values in the select list to the possible into-variables in
SELECT ... INTO ... */
UNIV_INLINE
void
sel_assign_into_var_values(
/*=======================*/
sym_node_t* var, /*!< in: first variable in a list of
variables */
sel_node_t* node) /*!< in: select node */
{
que_node_t* exp;
if (var == NULL) {
return;
}
for (exp = node->select_list;
var != 0;
var = static_cast<sym_node_t*>(que_node_get_next(var))) {
ut_ad(exp);
eval_node_copy_val(var->alias, exp);
exp = que_node_get_next(exp);
}
}
/*********************************************************************//**
Resets the aggregate value totals in the select list of an aggregate type
query. */
UNIV_INLINE
void
sel_reset_aggregate_vals(
/*=====================*/
sel_node_t* node) /*!< in: select node */
{
func_node_t* func_node;
ut_ad(node->is_aggregate);
for (func_node = static_cast<func_node_t*>(node->select_list);
func_node != 0;
func_node = static_cast<func_node_t*>(
que_node_get_next(func_node))) {
eval_node_set_int_val(func_node, 0);
}
node->aggregate_already_fetched = FALSE;
}
/*********************************************************************//**
Copies the input variable values when an explicit cursor is opened. */
UNIV_INLINE
void
row_sel_copy_input_variable_vals(
/*=============================*/
sel_node_t* node) /*!< in: select node */
{
sym_node_t* var;
var = UT_LIST_GET_FIRST(node->copy_variables);
while (var) {
eval_node_copy_val(var, var->alias);
var->indirection = NULL;
var = UT_LIST_GET_NEXT(col_var_list, var);
}
}
/*********************************************************************//**
Fetches the column values from a record. */
static
void
row_sel_fetch_columns(
/*==================*/
dict_index_t* index, /*!< in: record index */
const rec_t* rec, /*!< in: record in a clustered or non-clustered
index; must be protected by a page latch */
const ulint* offsets,/*!< in: rec_get_offsets(rec, index) */
sym_node_t* column) /*!< in: first column in a column list, or
NULL */
{
dfield_t* val;
ulint index_type;
ulint field_no;
const byte* data;
ulint len;
ut_ad(rec_offs_validate(rec, index, offsets));
if (dict_index_is_clust(index)) {
index_type = SYM_CLUST_FIELD_NO;
} else {
index_type = SYM_SEC_FIELD_NO;
}
while (column) {
mem_heap_t* heap = NULL;
ibool needs_copy;
field_no = column->field_nos[index_type];
if (field_no != ULINT_UNDEFINED) {
if (UNIV_UNLIKELY(rec_offs_nth_extern(
offsets, field_no) != 0)) {
/* Copy an externally stored field to the
temporary heap, if possible. */
heap = mem_heap_create(1);
data = btr_rec_copy_externally_stored_field(
rec, offsets,
index->table->space->zip_size(),
field_no, &len, heap);
/* data == NULL means that the
externally stored field was not
written yet. This record
should only be seen by
recv_recovery_rollback_active() or any
TRX_ISO_READ_UNCOMMITTED
transactions. The InnoDB SQL parser
(the sole caller of this function)
does not implement READ UNCOMMITTED,
and it is not involved during rollback. */
ut_a(data);
ut_a(len != UNIV_SQL_NULL);
needs_copy = TRUE;
} else {
data = rec_get_nth_cfield(rec, index, offsets,
field_no, &len);
needs_copy = column->copy_val;
}
if (needs_copy) {
eval_node_copy_and_alloc_val(column, data,
len);
} else {
val = que_node_get_val(column);
dfield_set_data(val, data, len);
}
if (UNIV_LIKELY_NULL(heap)) {
mem_heap_free(heap);
}
}
column = UT_LIST_GET_NEXT(col_var_list, column);
}
}
/*********************************************************************//**
Allocates a prefetch buffer for a column when prefetch is first time done. */
static
void
sel_col_prefetch_buf_alloc(
/*=======================*/
sym_node_t* column) /*!< in: symbol table node for a column */
{
sel_buf_t* sel_buf;
ulint i;
ut_ad(que_node_get_type(column) == QUE_NODE_SYMBOL);
column->prefetch_buf = static_cast<sel_buf_t*>(
ut_malloc_nokey(SEL_MAX_N_PREFETCH * sizeof(sel_buf_t)));
for (i = 0; i < SEL_MAX_N_PREFETCH; i++) {
sel_buf = column->prefetch_buf + i;
sel_buf->data = NULL;
sel_buf->len = 0;
sel_buf->val_buf_size = 0;
}
}
/*********************************************************************//**
Frees a prefetch buffer for a column, including the dynamically allocated
memory for data stored there. */
void
sel_col_prefetch_buf_free(
/*======================*/
sel_buf_t* prefetch_buf) /*!< in, own: prefetch buffer */
{
sel_buf_t* sel_buf;
ulint i;
for (i = 0; i < SEL_MAX_N_PREFETCH; i++) {
sel_buf = prefetch_buf + i;
if (sel_buf->val_buf_size > 0) {
ut_free(sel_buf->data);
}
}
ut_free(prefetch_buf);
}
/*********************************************************************//**
Pops the column values for a prefetched, cached row from the column prefetch
buffers and places them to the val fields in the column nodes. */
static
void
sel_dequeue_prefetched_row(
/*=======================*/
plan_t* plan) /*!< in: plan node for a table */
{
sym_node_t* column;
sel_buf_t* sel_buf;
dfield_t* val;
byte* data;
ulint len;
ulint val_buf_size;
ut_ad(plan->n_rows_prefetched > 0);
column = UT_LIST_GET_FIRST(plan->columns);
while (column) {
val = que_node_get_val(column);
if (!column->copy_val) {
/* We did not really push any value for the
column */
ut_ad(!column->prefetch_buf);
ut_ad(que_node_get_val_buf_size(column) == 0);
ut_d(dfield_set_null(val));
goto next_col;
}
ut_ad(column->prefetch_buf);
ut_ad(!dfield_is_ext(val));
sel_buf = column->prefetch_buf + plan->first_prefetched;
data = sel_buf->data;
len = sel_buf->len;
val_buf_size = sel_buf->val_buf_size;
/* We must keep track of the allocated memory for
column values to be able to free it later: therefore
we swap the values for sel_buf and val */
sel_buf->data = static_cast<byte*>(dfield_get_data(val));
sel_buf->len = dfield_get_len(val);
sel_buf->val_buf_size = que_node_get_val_buf_size(column);
dfield_set_data(val, data, len);
que_node_set_val_buf_size(column, val_buf_size);
next_col:
column = UT_LIST_GET_NEXT(col_var_list, column);
}
plan->n_rows_prefetched--;
plan->first_prefetched++;
}
/*********************************************************************//**
Pushes the column values for a prefetched, cached row to the column prefetch
buffers from the val fields in the column nodes. */
UNIV_INLINE
void
sel_enqueue_prefetched_row(
/*=======================*/
plan_t* plan) /*!< in: plan node for a table */
{
sym_node_t* column;
sel_buf_t* sel_buf;
dfield_t* val;
byte* data;
ulint len;
ulint pos;
ulint val_buf_size;
if (plan->n_rows_prefetched == 0) {
pos = 0;
plan->first_prefetched = 0;
} else {
pos = plan->n_rows_prefetched;
/* We have the convention that pushing new rows starts only
after the prefetch stack has been emptied: */
ut_ad(plan->first_prefetched == 0);
}
plan->n_rows_prefetched++;
ut_ad(pos < SEL_MAX_N_PREFETCH);
for (column = UT_LIST_GET_FIRST(plan->columns);
column != 0;
column = UT_LIST_GET_NEXT(col_var_list, column)) {
if (!column->copy_val) {
/* There is no sense to push pointers to database
page fields when we do not keep latch on the page! */
continue;
}
if (!column->prefetch_buf) {
/* Allocate a new prefetch buffer */
sel_col_prefetch_buf_alloc(column);
}
sel_buf = column->prefetch_buf + pos;
val = que_node_get_val(column);
data = static_cast<byte*>(dfield_get_data(val));
len = dfield_get_len(val);
val_buf_size = que_node_get_val_buf_size(column);
/* We must keep track of the allocated memory for
column values to be able to free it later: therefore
we swap the values for sel_buf and val */
dfield_set_data(val, sel_buf->data, sel_buf->len);
que_node_set_val_buf_size(column, sel_buf->val_buf_size);
sel_buf->data = data;
sel_buf->len = len;
sel_buf->val_buf_size = val_buf_size;
}
}
/*********************************************************************//**
Builds a previous version of a clustered index record for a consistent read
@return DB_SUCCESS or error code */
static MY_ATTRIBUTE((nonnull, warn_unused_result))
dberr_t
row_sel_build_prev_vers(
/*====================*/
ReadView* read_view, /*!< in: read view */
dict_index_t* index, /*!< in: plan node for table */
rec_t* rec, /*!< in: record in a clustered index */
ulint** offsets, /*!< in/out: offsets returned by
rec_get_offsets(rec, plan->index) */
mem_heap_t** offset_heap, /*!< in/out: memory heap from which
the offsets are allocated */
mem_heap_t** old_vers_heap, /*!< out: old version heap to use */
rec_t** old_vers, /*!< out: old version, or NULL if the
record does not exist in the view:
i.e., it was freshly inserted
afterwards */
mtr_t* mtr) /*!< in: mtr */
{
dberr_t err;
if (*old_vers_heap) {
mem_heap_empty(*old_vers_heap);
} else {
*old_vers_heap = mem_heap_create(512);
}
err = row_vers_build_for_consistent_read(
rec, mtr, index, offsets, read_view, offset_heap,
*old_vers_heap, old_vers, NULL);
return(err);
}
/*********************************************************************//**
Builds the last committed version of a clustered index record for a
semi-consistent read. */
static
void
row_sel_build_committed_vers_for_mysql(
/*===================================*/
dict_index_t* clust_index, /*!< in: clustered index */
row_prebuilt_t* prebuilt, /*!< in: prebuilt struct */
const rec_t* rec, /*!< in: record in a clustered index */
ulint** offsets, /*!< in/out: offsets returned by
rec_get_offsets(rec, clust_index) */
mem_heap_t** offset_heap, /*!< in/out: memory heap from which
the offsets are allocated */
const rec_t** old_vers, /*!< out: old version, or NULL if the
record does not exist in the view:
i.e., it was freshly inserted
afterwards */
dtuple_t** vrow, /*!< out: to be filled with old virtual
column version if any */
mtr_t* mtr) /*!< in: mtr */
{
if (prebuilt->old_vers_heap) {
mem_heap_empty(prebuilt->old_vers_heap);
} else {
prebuilt->old_vers_heap = mem_heap_create(
rec_offs_size(*offsets));
}
row_vers_build_for_semi_consistent_read(prebuilt->trx,
rec, mtr, clust_index, offsets, offset_heap,
prebuilt->old_vers_heap, old_vers, vrow);
}
/*********************************************************************//**
Tests the conditions which determine when the index segment we are searching
through has been exhausted.
@return TRUE if row passed the tests */
UNIV_INLINE
ibool
row_sel_test_end_conds(
/*===================*/
plan_t* plan) /*!< in: plan for the table; the column values must
already have been retrieved and the right sides of
comparisons evaluated */
{
func_node_t* cond;
/* All conditions in end_conds are comparisons of a column to an
expression */
for (cond = UT_LIST_GET_FIRST(plan->end_conds);
cond != 0;
cond = UT_LIST_GET_NEXT(cond_list, cond)) {
/* Evaluate the left side of the comparison, i.e., get the
column value if there is an indirection */
eval_sym(static_cast<sym_node_t*>(cond->args));
/* Do the comparison */
if (!eval_cmp(cond)) {
return(FALSE);
}
}
return(TRUE);
}
/*********************************************************************//**
Tests the other conditions.
@return TRUE if row passed the tests */
UNIV_INLINE
ibool
row_sel_test_other_conds(
/*=====================*/
plan_t* plan) /*!< in: plan for the table; the column values must
already have been retrieved */
{
func_node_t* cond;
cond = UT_LIST_GET_FIRST(plan->other_conds);
while (cond) {
eval_exp(cond);
if (!eval_node_get_ibool_val(cond)) {
return(FALSE);
}
cond = UT_LIST_GET_NEXT(cond_list, cond);
}
return(TRUE);
}
/*********************************************************************//**
Retrieves the clustered index record corresponding to a record in a
non-clustered index. Does the necessary locking.
@return DB_SUCCESS or error code */
static MY_ATTRIBUTE((nonnull, warn_unused_result))
dberr_t
row_sel_get_clust_rec(
/*==================*/
sel_node_t* node, /*!< in: select_node */
plan_t* plan, /*!< in: plan node for table */
rec_t* rec, /*!< in: record in a non-clustered index */
que_thr_t* thr, /*!< in: query thread */
rec_t** out_rec,/*!< out: clustered record or an old version of
it, NULL if the old version did not exist
in the read view, i.e., it was a fresh
inserted version */
mtr_t* mtr) /*!< in: mtr used to get access to the
non-clustered record; the same mtr is used to
access the clustered index */
{
dict_index_t* index;
rec_t* clust_rec;
rec_t* old_vers;
dberr_t err;
mem_heap_t* heap = NULL;
ulint offsets_[REC_OFFS_NORMAL_SIZE];
ulint* offsets = offsets_;
rec_offs_init(offsets_);
*out_rec = NULL;
offsets = rec_get_offsets(rec,
btr_pcur_get_btr_cur(&plan->pcur)->index,
offsets, true, ULINT_UNDEFINED, &heap);
row_build_row_ref_fast(plan->clust_ref, plan->clust_map, rec, offsets);
index = dict_table_get_first_index(plan->table);
btr_pcur_open_with_no_init(index, plan->clust_ref, PAGE_CUR_LE,
BTR_SEARCH_LEAF, &plan->clust_pcur,
0, mtr);
clust_rec = btr_pcur_get_rec(&(plan->clust_pcur));
/* Note: only if the search ends up on a non-infimum record is the
low_match value the real match to the search tuple */
if (!page_rec_is_user_rec(clust_rec)
|| btr_pcur_get_low_match(&(plan->clust_pcur))
< dict_index_get_n_unique(index)) {
ut_a(rec_get_deleted_flag(rec,
dict_table_is_comp(plan->table)));
ut_a(node->read_view);
/* In a rare case it is possible that no clust rec is found
for a delete-marked secondary index record: if in row0umod.cc
in row_undo_mod_remove_clust_low() we have already removed
the clust rec, while purge is still cleaning and removing
secondary index records associated with earlier versions of
the clustered index record. In that case we know that the
clustered index record did not exist in the read view of
trx. */
goto func_exit;
}
offsets = rec_get_offsets(clust_rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
if (!node->read_view) {
/* Try to place a lock on the index record */
ulint lock_type;
trx_t* trx;
trx = thr_get_trx(thr);
/* At READ UNCOMMITTED or READ COMMITTED isolation level
we lock only the record, i.e., next-key locking is
not used. */
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED) {
lock_type = LOCK_REC_NOT_GAP;
} else {
lock_type = LOCK_ORDINARY;
}
err = lock_clust_rec_read_check_and_lock(
0, btr_pcur_get_block(&plan->clust_pcur),
clust_rec, index, offsets,
static_cast<lock_mode>(node->row_lock_mode),
lock_type,
thr);
switch (err) {
case DB_SUCCESS:
case DB_SUCCESS_LOCKED_REC:
/* Declare the variable uninitialized in Valgrind.
It should be set to DB_SUCCESS at func_exit. */
UNIV_MEM_INVALID(&err, sizeof err);
break;
default:
goto err_exit;
}
} else {
/* This is a non-locking consistent read: if necessary, fetch
a previous version of the record */
old_vers = NULL;
if (!lock_clust_rec_cons_read_sees(clust_rec, index, offsets,
node->read_view)) {
err = row_sel_build_prev_vers(
node->read_view, index, clust_rec,
&offsets, &heap, &plan->old_vers_heap,
&old_vers, mtr);
if (err != DB_SUCCESS) {
goto err_exit;
}
clust_rec = old_vers;
if (clust_rec == NULL) {
goto func_exit;
}
}
/* If we had to go to an earlier version of row or the
secondary index record is delete marked, then it may be that
the secondary index record corresponding to clust_rec
(or old_vers) is not rec; in that case we must ignore
such row because in our snapshot rec would not have existed.
Remember that from rec we cannot see directly which transaction
id corresponds to it: we have to go to the clustered index
record. A query where we want to fetch all rows where
the secondary index value is in some interval would return
a wrong result if we would not drop rows which we come to
visit through secondary index records that would not really
exist in our snapshot. */
if ((old_vers
|| rec_get_deleted_flag(rec, dict_table_is_comp(
plan->table)))
&& !row_sel_sec_rec_is_for_clust_rec(rec, plan->index,
clust_rec, index,
thr)) {
goto func_exit;
}
}
/* Fetch the columns needed in test conditions. The clustered
index record is protected by a page latch that was acquired
when plan->clust_pcur was positioned. The latch will not be
released until mtr->commit(). */
ut_ad(!rec_get_deleted_flag(clust_rec, rec_offs_comp(offsets)));
row_sel_fetch_columns(index, clust_rec, offsets,
UT_LIST_GET_FIRST(plan->columns));
*out_rec = clust_rec;
func_exit:
err = DB_SUCCESS;
err_exit:
if (UNIV_LIKELY_NULL(heap)) {
mem_heap_free(heap);
}
return(err);
}
/*********************************************************************//**
Sets a lock on a page of R-Tree record. This is all or none action,
mostly due to we cannot reposition a record in R-Tree (with the
nature of splitting)
@return DB_SUCCESS, DB_SUCCESS_LOCKED_REC, or error code */
UNIV_INLINE
dberr_t
sel_set_rtr_rec_lock(
/*=================*/
btr_pcur_t* pcur, /*!< in: cursor */
const rec_t* first_rec,/*!< in: record */
dict_index_t* index, /*!< in: index */
const ulint* offsets,/*!< in: rec_get_offsets(rec, index) */
ulint mode, /*!< in: lock mode */
ulint type, /*!< in: LOCK_ORDINARY, LOCK_GAP, or
LOC_REC_NOT_GAP */
que_thr_t* thr, /*!< in: query thread */
mtr_t* mtr) /*!< in: mtr */
{
matched_rec_t* match = pcur->btr_cur.rtr_info->matches;
mem_heap_t* heap = NULL;
dberr_t err = DB_SUCCESS;
trx_t* trx = thr_get_trx(thr);
buf_block_t* cur_block = btr_pcur_get_block(pcur);
ulint offsets_[REC_OFFS_NORMAL_SIZE];
ulint* my_offsets = const_cast<ulint*>(offsets);
rec_t* rec = const_cast<rec_t*>(first_rec);
rtr_rec_vector* match_rec;
rtr_rec_vector::iterator end;
rec_offs_init(offsets_);
if (match->locked || page_rec_is_supremum(first_rec)) {
return(DB_SUCCESS_LOCKED_REC);
}
ut_ad(page_align(first_rec) == cur_block->frame);
ut_ad(match->valid);
rw_lock_x_lock(&(match->block.lock));
retry:
cur_block = btr_pcur_get_block(pcur);
ut_ad(rw_lock_own_flagged(&match->block.lock,
RW_LOCK_FLAG_X | RW_LOCK_FLAG_S));
ut_ad(page_is_leaf(buf_block_get_frame(cur_block)));
err = lock_sec_rec_read_check_and_lock(
0, cur_block, rec, index, my_offsets,
static_cast<lock_mode>(mode), type, thr);
if (err == DB_LOCK_WAIT) {
re_scan:
mtr->commit();
trx->error_state = err;
que_thr_stop_for_mysql(thr);
thr->lock_state = QUE_THR_LOCK_ROW;
if (row_mysql_handle_errors(
&err, trx, thr, NULL)) {
thr->lock_state = QUE_THR_LOCK_NOLOCK;
mtr->start();
mutex_enter(&match->rtr_match_mutex);
if (!match->valid && match->matched_recs->empty()) {
mutex_exit(&match->rtr_match_mutex);
err = DB_RECORD_NOT_FOUND;
goto func_end;
}
mutex_exit(&match->rtr_match_mutex);
/* MDEV-14059 FIXME: why re-latch the block?
pcur is already positioned on it! */
ulint page_no = page_get_page_no(
btr_pcur_get_page(pcur));
cur_block = buf_page_get_gen(
page_id_t(index->table->space_id, page_no),
index->table->space->zip_size(),
RW_X_LATCH, NULL, BUF_GET,
__FILE__, __LINE__, mtr, &err);
} else {
mtr->start();
goto func_end;
}
DEBUG_SYNC_C("rtr_set_lock_wait");
if (!match->valid) {
/* Page got deleted */
mtr->commit();
mtr->start();
err = DB_RECORD_NOT_FOUND;
goto func_end;
}
match->matched_recs->clear();
rtr_cur_search_with_match(
cur_block, index,
pcur->btr_cur.rtr_info->search_tuple,
pcur->btr_cur.rtr_info->search_mode,
&pcur->btr_cur.page_cur,
pcur->btr_cur.rtr_info);
if (!page_is_leaf(buf_block_get_frame(cur_block))) {
/* Page got splitted and promoted (only for
root page it is possible). Release the
page and ask for a re-search */
mtr->commit();
mtr->start();
err = DB_RECORD_NOT_FOUND;
goto func_end;
}
rec = btr_pcur_get_rec(pcur);
my_offsets = offsets_;
my_offsets = rec_get_offsets(rec, index, my_offsets, true,
ULINT_UNDEFINED, &heap);
/* No match record */
if (page_rec_is_supremum(rec) || !match->valid) {
mtr->commit();
mtr->start();
err = DB_RECORD_NOT_FOUND;
goto func_end;
}
goto retry;
}
my_offsets = offsets_;
match_rec = match->matched_recs;
end = match_rec->end();
for (rtr_rec_vector::iterator it = match_rec->begin();
it != end; ++it) {
rtr_rec_t* rtr_rec = &(*it);
my_offsets = rec_get_offsets(
rtr_rec->r_rec, index, my_offsets, true,
ULINT_UNDEFINED, &heap);
err = lock_sec_rec_read_check_and_lock(
0, &match->block, rtr_rec->r_rec, index,
my_offsets, static_cast<lock_mode>(mode),
type, thr);
if (err == DB_SUCCESS || err == DB_SUCCESS_LOCKED_REC) {
rtr_rec->locked = true;
} else if (err == DB_LOCK_WAIT) {
goto re_scan;
} else {
goto func_end;
}
}
match->locked = true;
func_end:
rw_lock_x_unlock(&(match->block.lock));
if (heap != NULL) {
mem_heap_free(heap);
}
ut_ad(err != DB_LOCK_WAIT);
return(err);
}
/*********************************************************************//**
Sets a lock on a record.
@return DB_SUCCESS, DB_SUCCESS_LOCKED_REC, or error code */
UNIV_INLINE
dberr_t
sel_set_rec_lock(
/*=============*/
btr_pcur_t* pcur, /*!< in: cursor */
const rec_t* rec, /*!< in: record */
dict_index_t* index, /*!< in: index */
const ulint* offsets,/*!< in: rec_get_offsets(rec, index) */
ulint mode, /*!< in: lock mode */
ulint type, /*!< in: LOCK_ORDINARY, LOCK_GAP, or
LOC_REC_NOT_GAP */
que_thr_t* thr, /*!< in: query thread */
mtr_t* mtr) /*!< in: mtr */
{
trx_t* trx;
dberr_t err = DB_SUCCESS;
const buf_block_t* block;
block = btr_pcur_get_block(pcur);
trx = thr_get_trx(thr);
if (UT_LIST_GET_LEN(trx->lock.trx_locks) > 10000) {
if (buf_LRU_buf_pool_running_out()) {
return(DB_LOCK_TABLE_FULL);
}
}
if (dict_index_is_clust(index)) {
err = lock_clust_rec_read_check_and_lock(
0, block, rec, index, offsets,
static_cast<lock_mode>(mode), type, thr);
} else {
if (dict_index_is_spatial(index)) {
if (type == LOCK_GAP || type == LOCK_ORDINARY) {
ut_ad(0);
ib::error() << "Incorrectly request GAP lock "
"on RTree";
return(DB_SUCCESS);
}
err = sel_set_rtr_rec_lock(pcur, rec, index, offsets,
mode, type, thr, mtr);
} else {
err = lock_sec_rec_read_check_and_lock(
0, block, rec, index, offsets,
static_cast<lock_mode>(mode), type, thr);
}
}
return(err);
}
/*********************************************************************//**
Opens a pcur to a table index. */
static
void
row_sel_open_pcur(
/*==============*/
plan_t* plan, /*!< in: table plan */
#ifdef BTR_CUR_HASH_ADAPT
rw_lock_t* ahi_latch,
/*!< in: the adaptive hash index latch */
#endif /* BTR_CUR_HASH_ADAPT */
mtr_t* mtr) /*!< in/out: mini-transaction */
{
dict_index_t* index;
func_node_t* cond;
que_node_t* exp;
ulint n_fields;
ulint i;
index = plan->index;
/* Calculate the value of the search tuple: the exact match columns
get their expressions evaluated when we evaluate the right sides of
end_conds */
cond = UT_LIST_GET_FIRST(plan->end_conds);
while (cond) {
eval_exp(que_node_get_next(cond->args));
cond = UT_LIST_GET_NEXT(cond_list, cond);
}
if (plan->tuple) {
n_fields = dtuple_get_n_fields(plan->tuple);
if (plan->n_exact_match < n_fields) {
/* There is a non-exact match field which must be
evaluated separately */
eval_exp(plan->tuple_exps[n_fields - 1]);
}
for (i = 0; i < n_fields; i++) {
exp = plan->tuple_exps[i];
dfield_copy_data(dtuple_get_nth_field(plan->tuple, i),
que_node_get_val(exp));
}
/* Open pcur to the index */
btr_pcur_open_with_no_init(index, plan->tuple, plan->mode,
BTR_SEARCH_LEAF, &plan->pcur,
ahi_latch, mtr);
} else {
/* Open the cursor to the start or the end of the index
(FALSE: no init) */
btr_pcur_open_at_index_side(plan->asc, index, BTR_SEARCH_LEAF,
&(plan->pcur), false, 0, mtr);
}
ut_ad(plan->n_rows_prefetched == 0);
ut_ad(plan->n_rows_fetched == 0);
ut_ad(plan->cursor_at_end == FALSE);
plan->pcur_is_open = TRUE;
}
/*********************************************************************//**
Restores a stored pcur position to a table index.
@return TRUE if the cursor should be moved to the next record after we
return from this function (moved to the previous, in the case of a
descending cursor) without processing again the current cursor
record */
static
ibool
row_sel_restore_pcur_pos(
/*=====================*/
plan_t* plan, /*!< in: table plan */
mtr_t* mtr) /*!< in: mtr */
{
ibool equal_position;
ulint relative_position;
ut_ad(!plan->cursor_at_end);
relative_position = btr_pcur_get_rel_pos(&(plan->pcur));
equal_position = btr_pcur_restore_position(BTR_SEARCH_LEAF,
&(plan->pcur), mtr);
/* If the cursor is traveling upwards, and relative_position is
(1) BTR_PCUR_BEFORE: this is not allowed, as we did not have a lock
yet on the successor of the page infimum;
(2) BTR_PCUR_AFTER: btr_pcur_restore_position placed the cursor on the
first record GREATER than the predecessor of a page supremum; we have
not yet processed the cursor record: no need to move the cursor to the
next record;
(3) BTR_PCUR_ON: btr_pcur_restore_position placed the cursor on the
last record LESS or EQUAL to the old stored user record; (a) if
equal_position is FALSE, this means that the cursor is now on a record
less than the old user record, and we must move to the next record;
(b) if equal_position is TRUE, then if
plan->stored_cursor_rec_processed is TRUE, we must move to the next
record, else there is no need to move the cursor. */
if (plan->asc) {
if (relative_position == BTR_PCUR_ON) {
if (equal_position) {
return(plan->stored_cursor_rec_processed);
}
return(TRUE);
}
ut_ad(relative_position == BTR_PCUR_AFTER
|| relative_position == BTR_PCUR_AFTER_LAST_IN_TREE);
return(FALSE);
}
/* If the cursor is traveling downwards, and relative_position is
(1) BTR_PCUR_BEFORE: btr_pcur_restore_position placed the cursor on
the last record LESS than the successor of a page infimum; we have not
processed the cursor record: no need to move the cursor;
(2) BTR_PCUR_AFTER: btr_pcur_restore_position placed the cursor on the
first record GREATER than the predecessor of a page supremum; we have
processed the cursor record: we should move the cursor to the previous
record;
(3) BTR_PCUR_ON: btr_pcur_restore_position placed the cursor on the
last record LESS or EQUAL to the old stored user record; (a) if
equal_position is FALSE, this means that the cursor is now on a record
less than the old user record, and we need not move to the previous
record; (b) if equal_position is TRUE, then if
plan->stored_cursor_rec_processed is TRUE, we must move to the previous
record, else there is no need to move the cursor. */
if (relative_position == BTR_PCUR_BEFORE
|| relative_position == BTR_PCUR_BEFORE_FIRST_IN_TREE) {
return(FALSE);
}
if (relative_position == BTR_PCUR_ON) {
if (equal_position) {
return(plan->stored_cursor_rec_processed);
}
return(FALSE);
}
ut_ad(relative_position == BTR_PCUR_AFTER
|| relative_position == BTR_PCUR_AFTER_LAST_IN_TREE);
return(TRUE);
}
/*********************************************************************//**
Resets a plan cursor to a closed state. */
UNIV_INLINE
void
plan_reset_cursor(
/*==============*/
plan_t* plan) /*!< in: plan */
{
plan->pcur_is_open = FALSE;
plan->cursor_at_end = FALSE;
plan->n_rows_fetched = 0;
plan->n_rows_prefetched = 0;
}
#ifdef BTR_CUR_HASH_ADAPT
/*********************************************************************//**
Tries to do a shortcut to fetch a clustered index record with a unique key,
using the hash index if possible (not always).
@return SEL_FOUND, SEL_EXHAUSTED, SEL_RETRY */
static
ulint
row_sel_try_search_shortcut(
/*========================*/
sel_node_t* node, /*!< in: select node for a consistent read */
plan_t* plan, /*!< in: plan for a unique search in clustered
index */
mtr_t* mtr) /*!< in: mtr */
{
dict_index_t* index = plan->index;
ut_ad(node->read_view);
ut_ad(plan->unique_search);
ut_ad(!plan->must_get_clust);
rw_lock_t* ahi_latch = btr_get_search_latch(index);
rw_lock_s_lock(ahi_latch);
row_sel_open_pcur(plan, ahi_latch, mtr);
const rec_t* rec = btr_pcur_get_rec(&(plan->pcur));
if (!page_rec_is_user_rec(rec) || rec_is_metadata(rec, *index)) {
retry:
rw_lock_s_unlock(ahi_latch);
return(SEL_RETRY);
}
ut_ad(plan->mode == PAGE_CUR_GE);
/* As the cursor is now placed on a user record after a search with
the mode PAGE_CUR_GE, the up_match field in the cursor tells how many
fields in the user record matched to the search tuple */
if (btr_pcur_get_up_match(&(plan->pcur)) < plan->n_exact_match) {
exhausted:
rw_lock_s_unlock(ahi_latch);
return(SEL_EXHAUSTED);
}
/* This is a non-locking consistent read: if necessary, fetch
a previous version of the record */
mem_heap_t* heap = NULL;
ulint offsets_[REC_OFFS_NORMAL_SIZE];
ulint* offsets = offsets_;
rec_offs_init(offsets_);
offsets = rec_get_offsets(rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
if (dict_index_is_clust(index)) {
if (!lock_clust_rec_cons_read_sees(rec, index, offsets,
node->read_view)) {
goto retry;
}
} else if (!srv_read_only_mode
&& !lock_sec_rec_cons_read_sees(
rec, index, node->read_view)) {
goto retry;
}
if (rec_get_deleted_flag(rec, dict_table_is_comp(plan->table))) {
goto exhausted;
}
/* Fetch the columns needed in test conditions. The index
record is protected by a page latch that was acquired when
plan->pcur was positioned. The latch will not be released
until mtr->commit(). */
row_sel_fetch_columns(index, rec, offsets,
UT_LIST_GET_FIRST(plan->columns));
/* Test the rest of search conditions */
if (!row_sel_test_other_conds(plan)) {
goto exhausted;
}
ut_ad(plan->pcur.latch_mode == BTR_SEARCH_LEAF);
plan->n_rows_fetched++;
rw_lock_s_unlock(ahi_latch);
if (UNIV_LIKELY_NULL(heap)) {
mem_heap_free(heap);
}
return(SEL_FOUND);
}
#endif /* BTR_CUR_HASH_ADAPT */
/*********************************************************************//**
Performs a select step.
@return DB_SUCCESS or error code */
static MY_ATTRIBUTE((warn_unused_result))
dberr_t
row_sel(
/*====*/
sel_node_t* node, /*!< in: select node */
que_thr_t* thr) /*!< in: query thread */
{
dict_index_t* index;
plan_t* plan;
mtr_t mtr;
ibool moved;
rec_t* rec;
rec_t* old_vers;
rec_t* clust_rec;
ibool consistent_read;
/* The following flag becomes TRUE when we are doing a
consistent read from a non-clustered index and we must look
at the clustered index to find out the previous delete mark
state of the non-clustered record: */
ibool cons_read_requires_clust_rec = FALSE;
ulint cost_counter = 0;
ibool cursor_just_opened;
ibool must_go_to_next;
ibool mtr_has_extra_clust_latch = FALSE;
/* TRUE if the search was made using
a non-clustered index, and we had to
access the clustered record: now &mtr
contains a clustered index latch, and
&mtr must be committed before we move
to the next non-clustered record */
dberr_t err;
mem_heap_t* heap = NULL;
ulint offsets_[REC_OFFS_NORMAL_SIZE];
ulint* offsets = offsets_;
rec_offs_init(offsets_);
ut_ad(thr->run_node == node);
if (node->read_view) {
/* In consistent reads, we try to do with the hash index and
not to use the buffer page get. This is to reduce memory bus
load resulting from semaphore operations. The search latch
will be s-locked when we access an index with a unique search
condition, but not locked when we access an index with a
less selective search condition. */
consistent_read = TRUE;
} else {
consistent_read = FALSE;
}
table_loop:
/* TABLE LOOP
----------
This is the outer major loop in calculating a join. We come here when
node->fetch_table changes, and after adding a row to aggregate totals
and, of course, when this function is called. */
ut_ad(mtr_has_extra_clust_latch == FALSE);
plan = sel_node_get_nth_plan(node, node->fetch_table);
index = plan->index;
if (plan->n_rows_prefetched > 0) {
sel_dequeue_prefetched_row(plan);
goto next_table_no_mtr;
}
if (plan->cursor_at_end) {
/* The cursor has already reached the result set end: no more
rows to process for this table cursor, as also the prefetch
stack was empty */
ut_ad(plan->pcur_is_open);
goto table_exhausted_no_mtr;
}
/* Open a cursor to index, or restore an open cursor position */
mtr.start();
#ifdef BTR_CUR_HASH_ADAPT
if (consistent_read && plan->unique_search && !plan->pcur_is_open
&& !plan->must_get_clust) {
switch (row_sel_try_search_shortcut(node, plan, &mtr)) {
case SEL_FOUND:
goto next_table;
case SEL_EXHAUSTED:
goto table_exhausted;
default:
ut_ad(0);
/* fall through */
case SEL_RETRY:
break;
}
plan_reset_cursor(plan);
mtr.commit();
mtr.start();
}
#endif /* BTR_CUR_HASH_ADAPT */
if (!plan->pcur_is_open) {
/* Evaluate the expressions to build the search tuple and
open the cursor */
row_sel_open_pcur(plan,
#ifdef BTR_CUR_HASH_ADAPT
NULL,
#endif /* BTR_CUR_HASH_ADAPT */
&mtr);
cursor_just_opened = TRUE;
/* A new search was made: increment the cost counter */
cost_counter++;
} else {
/* Restore pcur position to the index */
must_go_to_next = row_sel_restore_pcur_pos(plan, &mtr);
cursor_just_opened = FALSE;
if (must_go_to_next) {
/* We have already processed the cursor record: move
to the next */
goto next_rec;
}
}
rec_loop:
/* RECORD LOOP
-----------
In this loop we use pcur and try to fetch a qualifying row, and
also fill the prefetch buffer for this table if n_rows_fetched has
exceeded a threshold. While we are inside this loop, the following
holds:
(1) &mtr is started,
(2) pcur is positioned and open.
NOTE that if cursor_just_opened is TRUE here, it means that we came
to this point right after row_sel_open_pcur. */
ut_ad(mtr_has_extra_clust_latch == FALSE);
rec = btr_pcur_get_rec(&(plan->pcur));
/* PHASE 1: Set a lock if specified */
if (!node->asc && cursor_just_opened
&& !page_rec_is_supremum(rec)) {
/* Do not support "descending search" for Spatial index */
ut_ad(!dict_index_is_spatial(index));
/* When we open a cursor for a descending search, we must set
a next-key lock on the successor record: otherwise it would
be possible to insert new records next to the cursor position,
and it might be that these new records should appear in the
search result set, resulting in the phantom problem. */
if (!consistent_read) {
rec_t* next_rec = page_rec_get_next(rec);
ulint lock_type;
trx_t* trx;
trx = thr_get_trx(thr);
offsets = rec_get_offsets(next_rec, index, offsets,
true,
ULINT_UNDEFINED, &heap);
/* At READ UNCOMMITTED or READ COMMITTED
isolation level, we lock only the record,
i.e., next-key locking is not used. */
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED) {
if (page_rec_is_supremum(next_rec)) {
goto skip_lock;
}
lock_type = LOCK_REC_NOT_GAP;
} else {
lock_type = LOCK_ORDINARY;
}
err = sel_set_rec_lock(&plan->pcur,
next_rec, index, offsets,
node->row_lock_mode,
lock_type, thr, &mtr);
switch (err) {
case DB_SUCCESS_LOCKED_REC:
err = DB_SUCCESS;
/* fall through */
case DB_SUCCESS:
break;
default:
/* Note that in this case we will store in pcur
the PREDECESSOR of the record we are waiting
the lock for */
goto lock_wait_or_error;
}
}
}
skip_lock:
if (page_rec_is_infimum(rec)) {
/* The infimum record on a page cannot be in the result set,
and neither can a record lock be placed on it: we skip such
a record. We also increment the cost counter as we may have
processed yet another page of index. */
cost_counter++;
goto next_rec;
}
if (rec_is_metadata(rec, *index)) {
/* Skip the metadata pseudo-record. */
cost_counter++;
goto next_rec;
}
if (!consistent_read) {
/* Try to place a lock on the index record */
ulint lock_type;
trx_t* trx;
offsets = rec_get_offsets(rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
trx = thr_get_trx(thr);
/* At READ UNCOMMITTED or READ COMMITTED isolation level,
we lock only the record, i.e., next-key locking is
not used. */
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED
|| dict_index_is_spatial(index)) {
if (page_rec_is_supremum(rec)) {
goto next_rec;
}
lock_type = LOCK_REC_NOT_GAP;
} else {
lock_type = LOCK_ORDINARY;
}
err = sel_set_rec_lock(&plan->pcur,
rec, index, offsets,
node->row_lock_mode, lock_type,
thr, &mtr);
switch (err) {
case DB_SUCCESS_LOCKED_REC:
err = DB_SUCCESS;
/* fall through */
case DB_SUCCESS:
break;
default:
goto lock_wait_or_error;
}
}
if (page_rec_is_supremum(rec)) {
/* A page supremum record cannot be in the result set: skip
it now when we have placed a possible lock on it */
goto next_rec;
}
ut_ad(page_rec_is_user_rec(rec));
if (cost_counter > SEL_COST_LIMIT) {
/* Now that we have placed the necessary locks, we can stop
for a while and store the cursor position; NOTE that if we
would store the cursor position BEFORE placing a record lock,
it might happen that the cursor would jump over some records
that another transaction could meanwhile insert adjacent to
the cursor: this would result in the phantom problem. */
goto stop_for_a_while;
}
/* PHASE 2: Check a mixed index mix id if needed */
if (plan->unique_search && cursor_just_opened) {
ut_ad(plan->mode == PAGE_CUR_GE);
/* As the cursor is now placed on a user record after a search
with the mode PAGE_CUR_GE, the up_match field in the cursor
tells how many fields in the user record matched to the search
tuple */
if (btr_pcur_get_up_match(&(plan->pcur))
< plan->n_exact_match) {
goto table_exhausted;
}
/* Ok, no need to test end_conds or mix id */
}
/* We are ready to look at a possible new index entry in the result
set: the cursor is now placed on a user record */
/* PHASE 3: Get previous version in a consistent read */
cons_read_requires_clust_rec = FALSE;
offsets = rec_get_offsets(rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
if (consistent_read) {
/* This is a non-locking consistent read: if necessary, fetch
a previous version of the record */
if (dict_index_is_clust(index)) {
if (!lock_clust_rec_cons_read_sees(
rec, index, offsets, node->read_view)) {
err = row_sel_build_prev_vers(
node->read_view, index, rec,
&offsets, &heap, &plan->old_vers_heap,
&old_vers, &mtr);
if (err != DB_SUCCESS) {
goto lock_wait_or_error;
}
if (old_vers == NULL) {
/* The record does not exist
in our read view. Skip it, but
first attempt to determine
whether the index segment we
are searching through has been
exhausted. */
offsets = rec_get_offsets(
rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
/* Fetch the columns needed in
test conditions. The clustered
index record is protected by a
page latch that was acquired
by row_sel_open_pcur() or
row_sel_restore_pcur_pos().
The latch will not be released
until mtr.commit(). */
row_sel_fetch_columns(
index, rec, offsets,
UT_LIST_GET_FIRST(
plan->columns));
if (!row_sel_test_end_conds(plan)) {
goto table_exhausted;
}
goto next_rec;
}
rec = old_vers;
}
} else if (!srv_read_only_mode
&& !lock_sec_rec_cons_read_sees(
rec, index, node->read_view)) {
cons_read_requires_clust_rec = TRUE;
}
}
/* PHASE 4: Test search end conditions and deleted flag */
/* Fetch the columns needed in test conditions. The record is
protected by a page latch that was acquired by
row_sel_open_pcur() or row_sel_restore_pcur_pos(). The latch
will not be released until mtr.commit(). */
row_sel_fetch_columns(index, rec, offsets,
UT_LIST_GET_FIRST(plan->columns));
/* Test the selection end conditions: these can only contain columns
which already are found in the index, even though the index might be
non-clustered */
if (plan->unique_search && cursor_just_opened) {
/* No test necessary: the test was already made above */
} else if (!row_sel_test_end_conds(plan)) {
goto table_exhausted;
}
if (rec_get_deleted_flag(rec, dict_table_is_comp(plan->table))
&& !cons_read_requires_clust_rec) {
/* The record is delete marked: we can skip it if this is
not a consistent read which might see an earlier version
of a non-clustered index record */
if (plan->unique_search) {
goto table_exhausted;
}
goto next_rec;
}
/* PHASE 5: Get the clustered index record, if needed and if we did
not do the search using the clustered index */
if (plan->must_get_clust || cons_read_requires_clust_rec) {
/* It was a non-clustered index and we must fetch also the
clustered index record */
err = row_sel_get_clust_rec(node, plan, rec, thr, &clust_rec,
&mtr);
mtr_has_extra_clust_latch = TRUE;
if (err != DB_SUCCESS) {
goto lock_wait_or_error;
}
/* Retrieving the clustered record required a search:
increment the cost counter */
cost_counter++;
if (clust_rec == NULL) {
/* The record did not exist in the read view */
ut_ad(consistent_read);
goto next_rec;
}
if (rec_get_deleted_flag(clust_rec,
dict_table_is_comp(plan->table))) {
/* In delete-marked records, DB_TRX_ID must
always refer to an existing update_undo log record. */
ut_ad(rec_get_trx_id(clust_rec,
dict_table_get_first_index(
plan->table)));
/* The record is delete marked: we can skip it */
goto next_rec;
}
if (node->can_get_updated) {
btr_pcur_store_position(&(plan->clust_pcur), &mtr);
}
}
/* PHASE 6: Test the rest of search conditions */
if (!row_sel_test_other_conds(plan)) {
if (plan->unique_search) {
goto table_exhausted;
}
goto next_rec;
}
/* PHASE 7: We found a new qualifying row for the current table; push
the row if prefetch is on, or move to the next table in the join */
plan->n_rows_fetched++;
ut_ad(plan->pcur.latch_mode == BTR_SEARCH_LEAF);
if ((plan->n_rows_fetched <= SEL_PREFETCH_LIMIT)
|| plan->unique_search || plan->no_prefetch) {
/* No prefetch in operation: go to the next table */
goto next_table;
}
sel_enqueue_prefetched_row(plan);
if (plan->n_rows_prefetched == SEL_MAX_N_PREFETCH) {
/* The prefetch buffer is now full */
sel_dequeue_prefetched_row(plan);
goto next_table;
}
next_rec:
if (mtr_has_extra_clust_latch) {
/* We must commit &mtr if we are moving to the next
non-clustered index record, because we could break the
latching order if we would access a different clustered
index page right away without releasing the previous. */
goto commit_mtr_for_a_while;
}
if (node->asc) {
moved = btr_pcur_move_to_next(&(plan->pcur), &mtr);
} else {
moved = btr_pcur_move_to_prev(&(plan->pcur), &mtr);
}
if (!moved) {
goto table_exhausted;
}
cursor_just_opened = FALSE;
/* END OF RECORD LOOP
------------------ */
goto rec_loop;
next_table:
/* We found a record which satisfies the conditions: we can move to
the next table or return a row in the result set */
ut_ad(btr_pcur_is_on_user_rec(&plan->pcur));
if (plan->unique_search && !node->can_get_updated) {
plan->cursor_at_end = TRUE;
} else {
plan->stored_cursor_rec_processed = TRUE;
btr_pcur_store_position(&(plan->pcur), &mtr);
}
mtr.commit();
mtr_has_extra_clust_latch = FALSE;
next_table_no_mtr:
/* If we use 'goto' to this label, it means that the row was popped
from the prefetched rows stack, and &mtr is already committed */
if (node->fetch_table + 1 == node->n_tables) {
sel_eval_select_list(node);
if (node->is_aggregate) {
goto table_loop;
}
sel_assign_into_var_values(node->into_list, node);
thr->run_node = que_node_get_parent(node);
err = DB_SUCCESS;
goto func_exit;
}
node->fetch_table++;
/* When we move to the next table, we first reset the plan cursor:
we do not care about resetting it when we backtrack from a table */
plan_reset_cursor(sel_node_get_nth_plan(node, node->fetch_table));
goto table_loop;
table_exhausted:
/* The table cursor pcur reached the result set end: backtrack to the
previous table in the join if we do not have cached prefetched rows */
plan->cursor_at_end = TRUE;
mtr.commit();
mtr_has_extra_clust_latch = FALSE;
if (plan->n_rows_prefetched > 0) {
/* The table became exhausted during a prefetch */
sel_dequeue_prefetched_row(plan);
goto next_table_no_mtr;
}
table_exhausted_no_mtr:
if (node->fetch_table == 0) {
err = DB_SUCCESS;
if (node->is_aggregate && !node->aggregate_already_fetched) {
node->aggregate_already_fetched = TRUE;
sel_assign_into_var_values(node->into_list, node);
thr->run_node = que_node_get_parent(node);
} else {
node->state = SEL_NODE_NO_MORE_ROWS;
thr->run_node = que_node_get_parent(node);
}
goto func_exit;
}
node->fetch_table--;
goto table_loop;
stop_for_a_while:
/* Return control for a while to que_run_threads, so that runaway
queries can be canceled. NOTE that when we come here, we must, in a
locking read, have placed the necessary (possibly waiting request)
record lock on the cursor record or its successor: when we reposition
the cursor, this record lock guarantees that nobody can meanwhile have
inserted new records which should have appeared in the result set,
which would result in the phantom problem. */
plan->stored_cursor_rec_processed = FALSE;
btr_pcur_store_position(&(plan->pcur), &mtr);
mtr.commit();
ut_ad(!sync_check_iterate(sync_check()));
err = DB_SUCCESS;
goto func_exit;
commit_mtr_for_a_while:
/* Stores the cursor position and commits &mtr; this is used if
&mtr may contain latches which would break the latching order if
&mtr would not be committed and the latches released. */
plan->stored_cursor_rec_processed = TRUE;
btr_pcur_store_position(&(plan->pcur), &mtr);
mtr.commit();
mtr_has_extra_clust_latch = FALSE;
ut_ad(!sync_check_iterate(dict_sync_check()));
goto table_loop;
lock_wait_or_error:
/* See the note at stop_for_a_while: the same holds for this case */
ut_ad(!btr_pcur_is_before_first_on_page(&plan->pcur) || !node->asc);
plan->stored_cursor_rec_processed = FALSE;
btr_pcur_store_position(&(plan->pcur), &mtr);
mtr.commit();
func_exit:
ut_ad(!sync_check_iterate(dict_sync_check()));
if (heap != NULL) {
mem_heap_free(heap);
}
return(err);
}
/**********************************************************************//**
Performs a select step. This is a high-level function used in SQL execution
graphs.
@return query thread to run next or NULL */
que_thr_t*
row_sel_step(
/*=========*/
que_thr_t* thr) /*!< in: query thread */
{
sel_node_t* node;
ut_ad(thr);
node = static_cast<sel_node_t*>(thr->run_node);
ut_ad(que_node_get_type(node) == QUE_NODE_SELECT);
/* If this is a new time this node is executed (or when execution
resumes after wait for a table intention lock), set intention locks
on the tables, or assign a read view */
if (node->into_list && (thr->prev_node == que_node_get_parent(node))) {
node->state = SEL_NODE_OPEN;
}
if (node->state == SEL_NODE_OPEN) {
/* It may be that the current session has not yet started
its transaction, or it has been committed: */
trx_start_if_not_started_xa(thr_get_trx(thr), false);
plan_reset_cursor(sel_node_get_nth_plan(node, 0));
if (node->consistent_read) {
trx_t *trx = thr_get_trx(thr);
/* Assign a read view for the query */
trx->read_view.open(trx);
node->read_view = trx->read_view.is_open() ?
&trx->read_view : NULL;
} else {
sym_node_t* table_node;
lock_mode i_lock_mode;
if (node->set_x_locks) {
i_lock_mode = LOCK_IX;
} else {
i_lock_mode = LOCK_IS;
}
for (table_node = node->table_list;
table_node != 0;
table_node = static_cast<sym_node_t*>(
que_node_get_next(table_node))) {
dberr_t err = lock_table(
0, table_node->table, i_lock_mode,
thr);
if (err != DB_SUCCESS) {
trx_t* trx;
trx = thr_get_trx(thr);
trx->error_state = err;
return(NULL);
}
}
}
/* If this is an explicit cursor, copy stored procedure
variable values, so that the values cannot change between
fetches (currently, we copy them also for non-explicit
cursors) */
if (node->explicit_cursor
&& UT_LIST_GET_FIRST(node->copy_variables)) {
row_sel_copy_input_variable_vals(node);
}
node->state = SEL_NODE_FETCH;
node->fetch_table = 0;
if (node->is_aggregate) {
/* Reset the aggregate total values */
sel_reset_aggregate_vals(node);
}
}
dberr_t err = row_sel(node, thr);
/* NOTE! if queries are parallelized, the following assignment may
have problems; the assignment should be made only if thr is the
only top-level thr in the graph: */
thr->graph->last_sel_node = node;
if (err != DB_SUCCESS) {
thr_get_trx(thr)->error_state = err;
return(NULL);
}
return(thr);
}
/**********************************************************************//**
Performs a fetch for a cursor.
@return query thread to run next or NULL */
que_thr_t*
fetch_step(
/*=======*/
que_thr_t* thr) /*!< in: query thread */
{
sel_node_t* sel_node;
fetch_node_t* node;
ut_ad(thr);
node = static_cast<fetch_node_t*>(thr->run_node);
sel_node = node->cursor_def;
ut_ad(que_node_get_type(node) == QUE_NODE_FETCH);
if (thr->prev_node != que_node_get_parent(node)) {
if (sel_node->state != SEL_NODE_NO_MORE_ROWS) {
if (node->into_list) {
sel_assign_into_var_values(node->into_list,
sel_node);
} else {
ibool ret = (*node->func->func)(
sel_node, node->func->arg);
if (!ret) {
sel_node->state
= SEL_NODE_NO_MORE_ROWS;
}
}
}
thr->run_node = que_node_get_parent(node);
return(thr);
}
/* Make the fetch node the parent of the cursor definition for
the time of the fetch, so that execution knows to return to this
fetch node after a row has been selected or we know that there is
no row left */
sel_node->common.parent = node;
if (sel_node->state == SEL_NODE_CLOSED) {
ib::error() << "fetch called on a closed cursor";
thr_get_trx(thr)->error_state = DB_ERROR;
return(NULL);
}
thr->run_node = sel_node;
return(thr);
}
/***********************************************************//**
Prints a row in a select result.
@return query thread to run next or NULL */
que_thr_t*
row_printf_step(
/*============*/
que_thr_t* thr) /*!< in: query thread */
{
row_printf_node_t* node;
sel_node_t* sel_node;
que_node_t* arg;
ut_ad(thr);
node = static_cast<row_printf_node_t*>(thr->run_node);
sel_node = node->sel_node;
ut_ad(que_node_get_type(node) == QUE_NODE_ROW_PRINTF);
if (thr->prev_node == que_node_get_parent(node)) {
/* Reset the cursor */
sel_node->state = SEL_NODE_OPEN;
/* Fetch next row to print */
thr->run_node = sel_node;
return(thr);
}
if (sel_node->state != SEL_NODE_FETCH) {
ut_ad(sel_node->state == SEL_NODE_NO_MORE_ROWS);
/* No more rows to print */
thr->run_node = que_node_get_parent(node);
return(thr);
}
arg = sel_node->select_list;
while (arg) {
dfield_print_also_hex(que_node_get_val(arg));
fputs(" ::: ", stderr);
arg = que_node_get_next(arg);
}
putc('\n', stderr);
/* Fetch next row to print */
thr->run_node = sel_node;
return(thr);
}
/****************************************************************//**
Converts a key value stored in MySQL format to an Innobase dtuple. The last
field of the key value may be just a prefix of a fixed length field: hence
the parameter key_len. But currently we do not allow search keys where the
last field is only a prefix of the full key field len and print a warning if
such appears. A counterpart of this function is
ha_innobase::store_key_val_for_row() in ha_innodb.cc. */
void
row_sel_convert_mysql_key_to_innobase(
/*==================================*/
dtuple_t* tuple, /*!< in/out: tuple where to build;
NOTE: we assume that the type info
in the tuple is already according
to index! */
byte* buf, /*!< in: buffer to use in field
conversions; NOTE that dtuple->data
may end up pointing inside buf so
do not discard that buffer while
the tuple is being used. See
row_mysql_store_col_in_innobase_format()
in the case of DATA_INT */
ulint buf_len, /*!< in: buffer length */
dict_index_t* index, /*!< in: index of the key value */
const byte* key_ptr, /*!< in: MySQL key value */
ulint key_len) /*!< in: MySQL key value length */
{
byte* original_buf = buf;
const byte* original_key_ptr = key_ptr;
dict_field_t* field;
dfield_t* dfield;
ulint data_offset;
ulint data_len;
ulint data_field_len;
ibool is_null;
const byte* key_end;
ulint n_fields = 0;
/* For documentation of the key value storage format in MySQL, see
ha_innobase::store_key_val_for_row() in ha_innodb.cc. */
key_end = key_ptr + key_len;
/* Permit us to access any field in the tuple (ULINT_MAX): */
dtuple_set_n_fields(tuple, ULINT_MAX);
dfield = dtuple_get_nth_field(tuple, 0);
field = dict_index_get_nth_field(index, 0);
if (UNIV_UNLIKELY(dfield_get_type(dfield)->mtype == DATA_SYS)) {
/* A special case: we are looking for a position in the
generated clustered index which InnoDB automatically added
to a table with no primary key: the first and the only
ordering column is ROW_ID which InnoDB stored to the key_ptr
buffer. */
ut_a(key_len == DATA_ROW_ID_LEN);
dfield_set_data(dfield, key_ptr, DATA_ROW_ID_LEN);
dtuple_set_n_fields(tuple, 1);
return;
}
while (key_ptr < key_end) {
ulint type = dfield_get_type(dfield)->mtype;
ut_a(field->col->mtype == type);
data_offset = 0;
is_null = FALSE;
if (!(dfield_get_type(dfield)->prtype & DATA_NOT_NULL)) {
/* The first byte in the field tells if this is
an SQL NULL value */
data_offset = 1;
if (*key_ptr != 0) {
dfield_set_null(dfield);
is_null = TRUE;
}
}
/* Calculate data length and data field total length */
if (DATA_LARGE_MTYPE(type) || DATA_GEOMETRY_MTYPE(type)) {
/* For R-tree index, data length should be the
total size of the wkb data.*/
if (dict_index_is_spatial(index)) {
ut_ad(DATA_GEOMETRY_MTYPE(type));
data_len = key_len;
data_field_len = data_offset + data_len;
} else {
/* The key field is a column prefix of a BLOB
or TEXT. */
ut_a(field->prefix_len > 0);
/* MySQL stores the actual data length to the
first 2 bytes after the optional SQL NULL
marker byte. The storage format is
little-endian, that is, the most significant
byte at a higher address. In UTF-8, MySQL
seems to reserve field->prefix_len bytes for
storing this field in the key value buffer,
even though the actual value only takes data
len bytes from the start. */
data_len = ulint(key_ptr[data_offset])
| ulint(key_ptr[data_offset + 1]) << 8;
data_field_len = data_offset + 2
+ field->prefix_len;
data_offset += 2;
/* Now that we know the length, we store the
column value like it would be a fixed char
field */
}
} else if (field->prefix_len > 0) {
/* Looks like MySQL pads unused end bytes in the
prefix with space. Therefore, also in UTF-8, it is ok
to compare with a prefix containing full prefix_len
bytes, and no need to take at most prefix_len / 3
UTF-8 characters from the start.
If the prefix is used as the upper end of a LIKE
'abc%' query, then MySQL pads the end with chars
0xff. TODO: in that case does it any harm to compare
with the full prefix_len bytes. How do characters
0xff in UTF-8 behave? */
data_len = field->prefix_len;
data_field_len = data_offset + data_len;
} else {
data_len = dfield_get_type(dfield)->len;
data_field_len = data_offset + data_len;
}
if ((dtype_get_mysql_type(dfield_get_type(dfield))
== DATA_MYSQL_TRUE_VARCHAR)
&& (type != DATA_INT)) {
/* In a MySQL key value format, a true VARCHAR is
always preceded by 2 bytes of a length field.
dfield_get_type(dfield)->len returns the maximum
'payload' len in bytes. That does not include the
2 bytes that tell the actual data length.
We added the check != DATA_INT to make sure we do
not treat MySQL ENUM or SET as a true VARCHAR! */
data_len += 2;
data_field_len += 2;
}
/* Storing may use at most data_len bytes of buf */
if (UNIV_LIKELY(!is_null)) {
buf = row_mysql_store_col_in_innobase_format(
dfield, buf,
FALSE, /* MySQL key value format col */
key_ptr + data_offset, data_len,
dict_table_is_comp(index->table));
ut_a(buf <= original_buf + buf_len);
}
key_ptr += data_field_len;
if (UNIV_UNLIKELY(key_ptr > key_end)) {
/* The last field in key was not a complete key field
but a prefix of it.
Print a warning about this! HA_READ_PREFIX_LAST does
not currently work in InnoDB with partial-field key
value prefixes. Since MySQL currently uses a padding
trick to calculate LIKE 'abc%' type queries there
should never be partial-field prefixes in searches. */
ib::warn() << "Using a partial-field key prefix in"
" search, index " << index->name
<< " of table " << index->table->name
<< ". Last data field length "
<< data_field_len << " bytes, key ptr now"
" exceeds key end by " << (key_ptr - key_end)
<< " bytes. Key value in the MySQL format:";
ut_print_buf(stderr, original_key_ptr, key_len);
putc('\n', stderr);
if (!is_null) {
ulint len = dfield_get_len(dfield);
dfield_set_len(dfield, len
- (ulint) (key_ptr - key_end));
}
ut_ad(0);
}
n_fields++;
field++;
dfield++;
}
ut_a(buf <= original_buf + buf_len);
/* We set the length of tuple to n_fields: we assume that the memory
area allocated for it is big enough (usually bigger than n_fields). */
dtuple_set_n_fields(tuple, n_fields);
}
/**************************************************************//**
Stores a non-SQL-NULL field in the MySQL format. The counterpart of this
function is row_mysql_store_col_in_innobase_format() in row0mysql.cc. */
void
row_sel_field_store_in_mysql_format_func(
byte* dest,
const mysql_row_templ_t* templ,
#ifdef UNIV_DEBUG
const dict_index_t* index,
ulint field_no,
#endif /* UNIV_DEBUG */
const byte* data,
ulint len)
{
#ifdef UNIV_DEBUG
const dict_field_t* field
= templ->is_virtual
? NULL : dict_index_get_nth_field(index, field_no);
#endif /* UNIV_DEBUG */
ut_ad(len != UNIV_SQL_NULL);
UNIV_MEM_ASSERT_RW(data, len);
UNIV_MEM_ASSERT_W(dest, templ->mysql_col_len);
UNIV_MEM_INVALID(dest, templ->mysql_col_len);
byte* pad = dest + len;
switch (templ->type) {
const byte* field_end;
case DATA_VARCHAR:
case DATA_VARMYSQL:
case DATA_BINARY:
field_end = dest + templ->mysql_col_len;
if (templ->mysql_type == DATA_MYSQL_TRUE_VARCHAR) {
/* This is a >= 5.0.3 type true VARCHAR. Store the
length of the data to the first byte or the first
two bytes of dest. */
dest = row_mysql_store_true_var_len(
dest, len, templ->mysql_length_bytes);
/* Copy the actual data. Leave the rest of the
buffer uninitialized. */
memcpy(dest, data, len);
break;
}
/* Copy the actual data */
ut_memcpy(dest, data, len);
/* Pad with trailing spaces. */
if (pad == field_end) {
break;
}
if (UNIV_UNLIKELY(templ->type == DATA_FIXBINARY)) {
memset(pad, 0, field_end - pad);
break;
}
ut_ad(templ->mbminlen <= templ->mbmaxlen);
/* We treat some Unicode charset strings specially. */
switch (templ->mbminlen) {
case 4:
/* InnoDB should never have stripped partial
UTF-32 characters. */
ut_a(!(len & 3));
break;
case 2:
/* A space char is two bytes,
0x0020 in UCS2 and UTF-16 */
if (UNIV_UNLIKELY(len & 1)) {
/* A 0x20 has been stripped from the column.
Pad it back. */
if (pad < field_end) {
*pad++ = 0x20;
}
}
}
row_mysql_pad_col(templ->mbminlen, pad,
ulint(field_end - pad));
break;
case DATA_BLOB:
/* Store a pointer to the BLOB buffer to dest: the BLOB was
already copied to the buffer in row_sel_store_mysql_rec */
row_mysql_store_blob_ref(dest, templ->mysql_col_len, data,
len);
break;
case DATA_GEOMETRY:
/* We store all geometry data as BLOB data at server layer. */
row_mysql_store_geometry(dest, templ->mysql_col_len, data, len);
break;
case DATA_MYSQL:
memcpy(dest, data, len);
ut_ad(templ->mysql_col_len >= len);
ut_ad(templ->mbmaxlen >= templ->mbminlen);
/* If field_no equals to templ->icp_rec_field_no,
we are examining a row pointed by "icp_rec_field_no".
There is possibility that icp_rec_field_no refers to
a field in a secondary index while templ->rec_field_no
points to field in a primary index. The length
should still be equal, unless the field pointed
by icp_rec_field_no has a prefix */
ut_ad(templ->mbmaxlen > templ->mbminlen
|| templ->mysql_col_len == len
|| (field_no == templ->icp_rec_field_no
&& field->prefix_len > 0));
/* The following assertion would fail for old tables
containing UTF-8 ENUM columns due to Bug #9526. */
ut_ad(!templ->mbmaxlen
|| !(templ->mysql_col_len % templ->mbmaxlen));
ut_ad(len * templ->mbmaxlen >= templ->mysql_col_len
|| (field_no == templ->icp_rec_field_no
&& field->prefix_len > 0)
|| templ->rec_field_is_prefix);
ut_ad(templ->is_virtual
|| !(field->prefix_len % templ->mbmaxlen));
if (templ->mbminlen == 1 && templ->mbmaxlen != 1) {
/* Pad with spaces. This undoes the stripping
done in row0mysql.cc, function
row_mysql_store_col_in_innobase_format(). */
memset(pad, 0x20, templ->mysql_col_len - len);
}
break;
default:
#ifdef UNIV_DEBUG
case DATA_SYS_CHILD:
case DATA_SYS:
/* These column types should never be shipped to MySQL. */
ut_ad(0);
/* fall through */
case DATA_CHAR:
case DATA_FIXBINARY:
case DATA_FLOAT:
case DATA_DOUBLE:
case DATA_DECIMAL:
#endif /* UNIV_DEBUG */
ut_ad((templ->is_virtual && !field)
|| (field && field->prefix_len
? field->prefix_len == len
: templ->mysql_col_len == len));
memcpy(dest, data, len);
break;
case DATA_INT:
/* Convert InnoDB big-endian integer to little-endian
format, sign bit restored to 2's complement form */
DBUG_ASSERT(templ->mysql_col_len == len);
byte* ptr = pad;
do *--ptr = *data++; while (ptr != dest);
if (!templ->is_unsigned) {
pad[-1] ^= 0x80;
}
}
}
/** Convert a field in the Innobase format to a field in the MySQL format.
@param[out] mysql_rec record in the MySQL format
@param[in,out] prebuilt prebuilt struct
@param[in] rec InnoDB record; must be protected
by a page latch
@param[in] index index of rec
@param[in] offsets array returned by rec_get_offsets()
@param[in] field_no templ->rec_field_no or
templ->clust_rec_field_no
or templ->icp_rec_field_no
@param[in] templ row template
*/
static MY_ATTRIBUTE((warn_unused_result))
ibool
row_sel_store_mysql_field(
byte* mysql_rec,
row_prebuilt_t* prebuilt,
const rec_t* rec,
const dict_index_t* index,
const ulint* offsets,
ulint field_no,
const mysql_row_templ_t*templ)
{
DBUG_ENTER("row_sel_store_mysql_field_func");
const byte* data;
ulint len;
ut_ad(prebuilt->default_rec);
ut_ad(templ);
ut_ad(templ >= prebuilt->mysql_template);
ut_ad(templ < &prebuilt->mysql_template[prebuilt->n_template]);
ut_ad(field_no == templ->clust_rec_field_no
|| field_no == templ->rec_field_no
|| field_no == templ->icp_rec_field_no);
ut_ad(rec_offs_validate(rec, index, offsets));
if (UNIV_UNLIKELY(rec_offs_nth_extern(offsets, field_no) != 0)) {
mem_heap_t* heap;
/* Copy an externally stored field to a temporary heap */
ut_ad(field_no == templ->clust_rec_field_no);
if (DATA_LARGE_MTYPE(templ->type)) {
if (prebuilt->blob_heap == NULL) {
prebuilt->blob_heap = mem_heap_create(
srv_page_size);
}
heap = prebuilt->blob_heap;
} else {
heap = mem_heap_create(srv_page_size);
}
/* NOTE: if we are retrieving a big BLOB, we may
already run out of memory in the next call, which
causes an assert */
data = btr_rec_copy_externally_stored_field(
rec, offsets, prebuilt->table->space->zip_size(),
field_no, &len, heap);
if (UNIV_UNLIKELY(!data)) {
/* The externally stored field was not written
yet. This record should only be seen by
recv_recovery_rollback_active() or any
TRX_ISO_READ_UNCOMMITTED transactions. */
if (heap != prebuilt->blob_heap) {
mem_heap_free(heap);
}
ut_a(prebuilt->trx->isolation_level
== TRX_ISO_READ_UNCOMMITTED);
DBUG_RETURN(FALSE);
}
ut_a(len != UNIV_SQL_NULL);
row_sel_field_store_in_mysql_format(
mysql_rec + templ->mysql_col_offset,
templ, index, field_no, data, len);
if (heap != prebuilt->blob_heap) {
mem_heap_free(heap);
}
} else {
/* The field is stored in the index record, or
in the metadata for instant ADD COLUMN. */
data = rec_get_nth_cfield(rec, index, offsets, field_no, &len);
if (len == UNIV_SQL_NULL) {
/* MySQL assumes that the field for an SQL
NULL value is set to the default value. */
ut_ad(templ->mysql_null_bit_mask);
UNIV_MEM_ASSERT_RW(prebuilt->default_rec
+ templ->mysql_col_offset,
templ->mysql_col_len);
mysql_rec[templ->mysql_null_byte_offset]
|= (byte) templ->mysql_null_bit_mask;
memcpy(mysql_rec + templ->mysql_col_offset,
(const byte*) prebuilt->default_rec
+ templ->mysql_col_offset,
templ->mysql_col_len);
DBUG_RETURN(TRUE);
}
if (DATA_LARGE_MTYPE(templ->type)
|| DATA_GEOMETRY_MTYPE(templ->type)) {
/* It is a BLOB field locally stored in the
InnoDB record: we MUST copy its contents to
prebuilt->blob_heap here because
row_sel_field_store_in_mysql_format() stores a
pointer to the data, and the data passed to us
will be invalid as soon as the
mini-transaction is committed and the page
latch on the clustered index page is
released. */
if (prebuilt->blob_heap == NULL) {
prebuilt->blob_heap = mem_heap_create(
srv_page_size);
DBUG_PRINT("anna", ("blob_heap allocated: %p",
prebuilt->blob_heap));
}
data = static_cast<byte*>(
mem_heap_dup(prebuilt->blob_heap, data, len));
}
row_sel_field_store_in_mysql_format(
mysql_rec + templ->mysql_col_offset,
templ, index, field_no, data, len);
}
ut_ad(len != UNIV_SQL_NULL);
if (templ->mysql_null_bit_mask) {
/* It is a nullable column with a non-NULL
value */
mysql_rec[templ->mysql_null_byte_offset]
&= ~(byte) templ->mysql_null_bit_mask;
}
DBUG_RETURN(TRUE);
}
/** Convert a row in the Innobase format to a row in the MySQL format.
Note that the template in prebuilt may advise us to copy only a few
columns to mysql_rec, other columns are left blank. All columns may not
be needed in the query.
@param[out] mysql_rec row in the MySQL format
@param[in] prebuilt cursor
@param[in] rec Innobase record in the index
which was described in prebuilt's
template, or in the clustered index;
must be protected by a page latch
@param[in] vrow virtual columns
@param[in] rec_clust whether index must be the clustered index
@param[in] index index of rec
@param[in] offsets array returned by rec_get_offsets(rec)
@retval true on success
@retval false if not all columns could be retrieved */
MY_ATTRIBUTE((warn_unused_result))
static bool row_sel_store_mysql_rec(
byte* mysql_rec,
row_prebuilt_t* prebuilt,
const rec_t* rec,
const dtuple_t* vrow,
bool rec_clust,
const dict_index_t* index,
const ulint* offsets)
{
DBUG_ENTER("row_sel_store_mysql_rec");
ut_ad(rec_clust || index == prebuilt->index);
ut_ad(!rec_clust || dict_index_is_clust(index));
if (UNIV_LIKELY_NULL(prebuilt->blob_heap)) {
row_mysql_prebuilt_free_blob_heap(prebuilt);
}
for (ulint i = 0; i < prebuilt->n_template; i++) {
const mysql_row_templ_t*templ = &prebuilt->mysql_template[i];
if (templ->is_virtual && dict_index_is_clust(index)) {
/* Skip virtual columns if it is not a covered
search or virtual key read is not requested. */
if (!rec_clust
|| !prebuilt->index->has_virtual()
|| (!prebuilt->read_just_key
&& !prebuilt->m_read_virtual_key)) {
/* Initialize the NULL bit. */
if (templ->mysql_null_bit_mask) {
mysql_rec[templ->mysql_null_byte_offset]
|= (byte) templ->mysql_null_bit_mask;
}
continue;
}
dict_v_col_t* col;
col = dict_table_get_nth_v_col(
index->table, templ->clust_rec_field_no);
ut_ad(vrow);
const dfield_t* dfield = dtuple_get_nth_v_field(
vrow, col->v_pos);
/* If this is a partitioned table, it might request
InnoDB to fill out virtual column data for serach
index key values while other non key columns are also
getting selected. The non-key virtual columns may
not be materialized and we should skip them. */
if (dfield_get_type(dfield)->mtype == DATA_MISSING) {
#ifdef UNIV_DEBUG
ulint prefix;
#endif /* UNIV_DEBUG */
ut_ad(prebuilt->m_read_virtual_key);
/* If it is part of index key the data should
have been materialized. */
ut_ad(dict_index_get_nth_col_or_prefix_pos(
prebuilt->index, col->v_pos, false,
true, &prefix) == ULINT_UNDEFINED);
continue;
}
if (dfield->len == UNIV_SQL_NULL) {
mysql_rec[templ->mysql_null_byte_offset]
|= (byte) templ->mysql_null_bit_mask;
memcpy(mysql_rec
+ templ->mysql_col_offset,
(const byte*) prebuilt->default_rec
+ templ->mysql_col_offset,
templ->mysql_col_len);
} else {
row_sel_field_store_in_mysql_format(
mysql_rec + templ->mysql_col_offset,
templ, index, templ->clust_rec_field_no,
(const byte*)dfield->data, dfield->len);
if (templ->mysql_null_bit_mask) {
mysql_rec[
templ->mysql_null_byte_offset]
&= ~(byte) templ->mysql_null_bit_mask;
}
}
continue;
}
const ulint field_no
= rec_clust
? templ->clust_rec_field_no
: templ->rec_field_no;
/* We should never deliver column prefixes to the SQL layer,
except for evaluating handler_index_cond_check()
or handler_rowid_filter_check(). */
/* ...actually, we do want to do this in order to
support the prefix query optimization.
ut_ad(dict_index_get_nth_field(index, field_no)->prefix_len
== 0);
...so we disable this assert. */
if (!row_sel_store_mysql_field(mysql_rec, prebuilt,
rec, index, offsets,
field_no, templ)) {
DBUG_RETURN(false);
}
}
/* FIXME: We only need to read the doc_id if an FTS indexed
column is being updated.
NOTE, the record can be cluster or secondary index record.
if secondary index is used then FTS_DOC_ID column should be part
of this index. */
if (dict_table_has_fts_index(prebuilt->table)) {
if (dict_index_is_clust(index)
|| prebuilt->fts_doc_id_in_read_set) {
prebuilt->fts_doc_id = fts_get_doc_id_from_rec(
rec, index, offsets);
}
}
DBUG_RETURN(true);
}
/*********************************************************************//**
Builds a previous version of a clustered index record for a consistent read
@return DB_SUCCESS or error code */
static MY_ATTRIBUTE((warn_unused_result))
dberr_t
row_sel_build_prev_vers_for_mysql(
/*==============================*/
ReadView* read_view, /*!< in: read view */
dict_index_t* clust_index, /*!< in: clustered index */
row_prebuilt_t* prebuilt, /*!< in: prebuilt struct */
const rec_t* rec, /*!< in: record in a clustered index */
ulint** offsets, /*!< in/out: offsets returned by
rec_get_offsets(rec, clust_index) */
mem_heap_t** offset_heap, /*!< in/out: memory heap from which
the offsets are allocated */
rec_t** old_vers, /*!< out: old version, or NULL if the
record does not exist in the view:
i.e., it was freshly inserted
afterwards */
dtuple_t** vrow, /*!< out: dtuple to hold old virtual
column data */
mtr_t* mtr) /*!< in: mtr */
{
dberr_t err;
if (prebuilt->old_vers_heap) {
mem_heap_empty(prebuilt->old_vers_heap);
} else {
prebuilt->old_vers_heap = mem_heap_create(200);
}
err = row_vers_build_for_consistent_read(
rec, mtr, clust_index, offsets, read_view, offset_heap,
prebuilt->old_vers_heap, old_vers, vrow);
return(err);
}
/** Helper class to cache clust_rec and old_ver */
class Row_sel_get_clust_rec_for_mysql
{
const rec_t *cached_clust_rec;
rec_t *cached_old_vers;
public:
Row_sel_get_clust_rec_for_mysql() :
cached_clust_rec(NULL), cached_old_vers(NULL) {}
dberr_t operator()(row_prebuilt_t *prebuilt, dict_index_t *sec_index,
const rec_t *rec, que_thr_t *thr, const rec_t **out_rec,
ulint **offsets, mem_heap_t **offset_heap,
dtuple_t **vrow, mtr_t *mtr);
};
/*********************************************************************//**
Retrieves the clustered index record corresponding to a record in a
non-clustered index. Does the necessary locking. Used in the MySQL
interface.
@return DB_SUCCESS, DB_SUCCESS_LOCKED_REC, or error code */
dberr_t
Row_sel_get_clust_rec_for_mysql::operator()(
/*============================*/
row_prebuilt_t* prebuilt,/*!< in: prebuilt struct in the handle */
dict_index_t* sec_index,/*!< in: secondary index where rec resides */
const rec_t* rec, /*!< in: record in a non-clustered index; if
this is a locking read, then rec is not
allowed to be delete-marked, and that would
not make sense either */
que_thr_t* thr, /*!< in: query thread */
const rec_t** out_rec,/*!< out: clustered record or an old version of
it, NULL if the old version did not exist
in the read view, i.e., it was a fresh
inserted version */
ulint** offsets,/*!< in: offsets returned by
rec_get_offsets(rec, sec_index);
out: offsets returned by
rec_get_offsets(out_rec, clust_index) */
mem_heap_t** offset_heap,/*!< in/out: memory heap from which
the offsets are allocated */
dtuple_t** vrow, /*!< out: virtual column to fill */
mtr_t* mtr) /*!< in: mtr used to get access to the
non-clustered record; the same mtr is used to
access the clustered index */
{
dict_index_t* clust_index;
const rec_t* clust_rec;
rec_t* old_vers;
dberr_t err;
trx_t* trx;
*out_rec = NULL;
trx = thr_get_trx(thr);
srv_stats.n_sec_rec_cluster_reads.inc(
thd_get_thread_id(trx->mysql_thd));
row_build_row_ref_in_tuple(prebuilt->clust_ref, rec,
sec_index, *offsets);
clust_index = dict_table_get_first_index(sec_index->table);
btr_pcur_open_with_no_init(clust_index, prebuilt->clust_ref,
PAGE_CUR_LE, BTR_SEARCH_LEAF,
prebuilt->clust_pcur, 0, mtr);
clust_rec = btr_pcur_get_rec(prebuilt->clust_pcur);
prebuilt->clust_pcur->trx_if_known = trx;
/* Note: only if the search ends up on a non-infimum record is the
low_match value the real match to the search tuple */
if (!page_rec_is_user_rec(clust_rec)
|| btr_pcur_get_low_match(prebuilt->clust_pcur)
< dict_index_get_n_unique(clust_index)) {
btr_cur_t* btr_cur = btr_pcur_get_btr_cur(prebuilt->pcur);
/* If this is a spatial index scan, and we are reading
from a shadow buffer, the record could be already
deleted (due to rollback etc.). So get the original
page and verify that */
if (dict_index_is_spatial(sec_index)
&& btr_cur->rtr_info->matches
&& (page_align(rec)
== btr_cur->rtr_info->matches->block.frame
|| rec != btr_pcur_get_rec(prebuilt->pcur))) {
#ifdef UNIV_DEBUG
rtr_info_t* rtr_info = btr_cur->rtr_info;
mutex_enter(&rtr_info->matches->rtr_match_mutex);
/* The page could be deallocated (by rollback etc.) */
if (!rtr_info->matches->valid) {
mutex_exit(&rtr_info->matches->rtr_match_mutex);
clust_rec = NULL;
err = DB_SUCCESS;
goto func_exit;
}
mutex_exit(&rtr_info->matches->rtr_match_mutex);
if (rec_get_deleted_flag(rec,
dict_table_is_comp(sec_index->table))
&& prebuilt->select_lock_type == LOCK_NONE) {
clust_rec = NULL;
err = DB_SUCCESS;
goto func_exit;
}
if (rec != btr_pcur_get_rec(prebuilt->pcur)) {
clust_rec = NULL;
err = DB_SUCCESS;
goto func_exit;
}
/* FIXME: Why is this block not the
same as btr_pcur_get_block(prebuilt->pcur),
and is it not unsafe to use RW_NO_LATCH here? */
buf_block_t* block = buf_page_get_gen(
btr_pcur_get_block(prebuilt->pcur)->page.id,
btr_pcur_get_block(prebuilt->pcur)->zip_size(),
RW_NO_LATCH, NULL, BUF_GET,
__FILE__, __LINE__, mtr, &err);
mem_heap_t* heap = mem_heap_create(256);
dtuple_t* tuple = dict_index_build_data_tuple(
rec, sec_index, true,
sec_index->n_fields, heap);
page_cur_t page_cursor;
ulint low_match = page_cur_search(
block, sec_index, tuple,
PAGE_CUR_LE, &page_cursor);
ut_ad(low_match < dtuple_get_n_fields_cmp(tuple));
mem_heap_free(heap);
clust_rec = NULL;
err = DB_SUCCESS;
goto func_exit;
#endif /* UNIV_DEBUG */
} else if (!rec_get_deleted_flag(rec,
dict_table_is_comp(sec_index->table))
|| prebuilt->select_lock_type != LOCK_NONE) {
/* In a rare case it is possible that no clust
rec is found for a delete-marked secondary index
record: if in row0umod.cc in
row_undo_mod_remove_clust_low() we have already removed
the clust rec, while purge is still cleaning and
removing secondary index records associated with
earlier versions of the clustered index record.
In that case we know that the clustered index
record did not exist in the read view of trx. */
ib::error() << "Clustered record for sec rec not found"
" index " << sec_index->name
<< " of table " << sec_index->table->name;
fputs("InnoDB: sec index record ", stderr);
rec_print(stderr, rec, sec_index);
fputs("\n"
"InnoDB: clust index record ", stderr);
rec_print(stderr, clust_rec, clust_index);
putc('\n', stderr);
trx_print(stderr, trx, 600);
fputs("\n"
"InnoDB: Submit a detailed bug report"
" to https://jira.mariadb.org/\n", stderr);
ut_ad(0);
}
clust_rec = NULL;
err = DB_SUCCESS;
goto func_exit;
}
*offsets = rec_get_offsets(clust_rec, clust_index, *offsets, true,
ULINT_UNDEFINED, offset_heap);
if (prebuilt->select_lock_type != LOCK_NONE) {
/* Try to place a lock on the index record; we are searching
the clust rec with a unique condition, hence
we set a LOCK_REC_NOT_GAP type lock */
err = lock_clust_rec_read_check_and_lock(
0, btr_pcur_get_block(prebuilt->clust_pcur),
clust_rec, clust_index, *offsets,
static_cast<lock_mode>(prebuilt->select_lock_type),
LOCK_REC_NOT_GAP,
thr);
switch (err) {
case DB_SUCCESS:
case DB_SUCCESS_LOCKED_REC:
break;
default:
goto err_exit;
}
} else {
/* This is a non-locking consistent read: if necessary, fetch
a previous version of the record */
old_vers = NULL;
/* If the isolation level allows reading of uncommitted data,
then we never look for an earlier version */
if (trx->isolation_level > TRX_ISO_READ_UNCOMMITTED
&& !lock_clust_rec_cons_read_sees(
clust_rec, clust_index, *offsets,
&trx->read_view)) {
if (clust_rec != cached_clust_rec) {
/* The following call returns 'offsets' associated with
'old_vers' */
err = row_sel_build_prev_vers_for_mysql(
&trx->read_view, clust_index, prebuilt,
clust_rec, offsets, offset_heap, &old_vers,
vrow, mtr);
if (err != DB_SUCCESS) {
goto err_exit;
}
cached_clust_rec = clust_rec;
cached_old_vers = old_vers;
} else {
err = DB_SUCCESS;
old_vers = cached_old_vers;
/* The offsets need not be same for the latest
version of clust_rec and its old version
old_vers. Re-calculate the offsets for old_vers. */
if (old_vers != NULL) {
*offsets = rec_get_offsets(
old_vers, clust_index, *offsets,
true, ULINT_UNDEFINED, offset_heap);
}
}
if (old_vers == NULL) {
goto err_exit;
}
clust_rec = old_vers;
}
/* If we had to go to an earlier version of row or the
secondary index record is delete marked, then it may be that
the secondary index record corresponding to clust_rec
(or old_vers) is not rec; in that case we must ignore
such row because in our snapshot rec would not have existed.
Remember that from rec we cannot see directly which transaction
id corresponds to it: we have to go to the clustered index
record. A query where we want to fetch all rows where
the secondary index value is in some interval would return
a wrong result if we would not drop rows which we come to
visit through secondary index records that would not really
exist in our snapshot. */
/* And for spatial index, since the rec is from shadow buffer,
so we need to check if it's exactly match the clust_rec. */
if (clust_rec
&& (old_vers
|| trx->isolation_level <= TRX_ISO_READ_UNCOMMITTED
|| dict_index_is_spatial(sec_index)
|| rec_get_deleted_flag(rec, dict_table_is_comp(
sec_index->table)))
&& !row_sel_sec_rec_is_for_clust_rec(
rec, sec_index, clust_rec, clust_index, thr)) {
clust_rec = NULL;
}
err = DB_SUCCESS;
}
func_exit:
*out_rec = clust_rec;
if (prebuilt->select_lock_type != LOCK_NONE) {
/* We may use the cursor in update or in unlock_row():
store its position */
btr_pcur_store_position(prebuilt->clust_pcur, mtr);
}
err_exit:
return(err);
}
/********************************************************************//**
Restores cursor position after it has been stored. We have to take into
account that the record cursor was positioned on may have been deleted.
Then we may have to move the cursor one step up or down.
@return true if we may need to process the record the cursor is now
positioned on (i.e. we should not go to the next record yet) */
static
bool
sel_restore_position_for_mysql(
/*===========================*/
ibool* same_user_rec, /*!< out: TRUE if we were able to restore
the cursor on a user record with the
same ordering prefix in in the
B-tree index */
ulint latch_mode, /*!< in: latch mode wished in
restoration */
btr_pcur_t* pcur, /*!< in: cursor whose position
has been stored */
ibool moves_up, /*!< in: TRUE if the cursor moves up
in the index */
mtr_t* mtr) /*!< in: mtr; CAUTION: may commit
mtr temporarily! */
{
ibool success;
success = btr_pcur_restore_position(latch_mode, pcur, mtr);
*same_user_rec = success;
ut_ad(!success || pcur->rel_pos == BTR_PCUR_ON);
#ifdef UNIV_DEBUG
if (pcur->pos_state == BTR_PCUR_IS_POSITIONED_OPTIMISTIC) {
ut_ad(pcur->rel_pos == BTR_PCUR_BEFORE
|| pcur->rel_pos == BTR_PCUR_AFTER);
} else {
ut_ad(pcur->pos_state == BTR_PCUR_IS_POSITIONED);
ut_ad((pcur->rel_pos == BTR_PCUR_ON)
== btr_pcur_is_on_user_rec(pcur));
}
#endif /* UNIV_DEBUG */
/* The position may need be adjusted for rel_pos and moves_up. */
switch (pcur->rel_pos) {
case BTR_PCUR_ON:
if (!success && moves_up) {
next:
if (btr_pcur_move_to_next(pcur, mtr)
&& rec_is_metadata(btr_pcur_get_rec(pcur),
*pcur->btr_cur.index)) {
btr_pcur_move_to_next(pcur, mtr);
}
return true;
}
return(!success);
case BTR_PCUR_AFTER_LAST_IN_TREE:
case BTR_PCUR_BEFORE_FIRST_IN_TREE:
return true;
case BTR_PCUR_AFTER:
/* positioned to record after pcur->old_rec. */
pcur->pos_state = BTR_PCUR_IS_POSITIONED;
prev:
if (btr_pcur_is_on_user_rec(pcur) && !moves_up
&& !rec_is_metadata(btr_pcur_get_rec(pcur),
*pcur->btr_cur.index)) {
btr_pcur_move_to_prev(pcur, mtr);
}
return true;
case BTR_PCUR_BEFORE:
/* For non optimistic restoration:
The position is now set to the record before pcur->old_rec.
For optimistic restoration:
The position also needs to take the previous search_mode into
consideration. */
switch (pcur->pos_state) {
case BTR_PCUR_IS_POSITIONED_OPTIMISTIC:
pcur->pos_state = BTR_PCUR_IS_POSITIONED;
if (pcur->search_mode == PAGE_CUR_GE) {
/* Positioned during Greater or Equal search
with BTR_PCUR_BEFORE. Optimistic restore to
the same record. If scanning for lower then
we must move to previous record.
This can happen with:
HANDLER READ idx a = (const);
HANDLER READ idx PREV; */
goto prev;
}
return true;
case BTR_PCUR_IS_POSITIONED:
if (moves_up && btr_pcur_is_on_user_rec(pcur)) {
goto next;
}
return true;
case BTR_PCUR_WAS_POSITIONED:
case BTR_PCUR_NOT_POSITIONED:
break;
}
}
ut_ad(0);
return true;
}
/********************************************************************//**
Copies a cached field for MySQL from the fetch cache. */
static
void
row_sel_copy_cached_field_for_mysql(
/*================================*/
byte* buf, /*!< in/out: row buffer */
const byte* cache, /*!< in: cached row */
const mysql_row_templ_t*templ) /*!< in: column template */
{
ulint len;
buf += templ->mysql_col_offset;
cache += templ->mysql_col_offset;
UNIV_MEM_ASSERT_W(buf, templ->mysql_col_len);
if (templ->mysql_type == DATA_MYSQL_TRUE_VARCHAR
&& (templ->type != DATA_INT)) {
/* Check for != DATA_INT to make sure we do
not treat MySQL ENUM or SET as a true VARCHAR!
Find the actual length of the true VARCHAR field. */
row_mysql_read_true_varchar(
&len, cache, templ->mysql_length_bytes);
len += templ->mysql_length_bytes;
UNIV_MEM_INVALID(buf, templ->mysql_col_len);
} else {
len = templ->mysql_col_len;
}
ut_memcpy(buf, cache, len);
}
/** Copy used fields from cached row.
Copy cache record field by field, don't touch fields that
are not covered by current key.
@param[out] buf Where to copy the MySQL row.
@param[in] cached_rec What to copy (in MySQL row format).
@param[in] prebuilt prebuilt struct. */
void
row_sel_copy_cached_fields_for_mysql(
byte* buf,
const byte* cached_rec,
row_prebuilt_t* prebuilt)
{
const mysql_row_templ_t*templ;
ulint i;
for (i = 0; i < prebuilt->n_template; i++) {
templ = prebuilt->mysql_template + i;
/* Skip virtual columns */
if (templ->is_virtual) {
continue;
}
row_sel_copy_cached_field_for_mysql(
buf, cached_rec, templ);
/* Copy NULL bit of the current field from cached_rec
to buf */
if (templ->mysql_null_bit_mask) {
buf[templ->mysql_null_byte_offset]
^= (buf[templ->mysql_null_byte_offset]
^ cached_rec[templ->mysql_null_byte_offset])
& (byte) templ->mysql_null_bit_mask;
}
}
}
/********************************************************************//**
Pops a cached row for MySQL from the fetch cache. */
UNIV_INLINE
void
row_sel_dequeue_cached_row_for_mysql(
/*=================================*/
byte* buf, /*!< in/out: buffer where to copy the
row */
row_prebuilt_t* prebuilt) /*!< in: prebuilt struct */
{
ulint i;
const mysql_row_templ_t*templ;
const byte* cached_rec;
ut_ad(prebuilt->n_fetch_cached > 0);
ut_ad(prebuilt->mysql_prefix_len <= prebuilt->mysql_row_len);
UNIV_MEM_ASSERT_W(buf, prebuilt->mysql_row_len);
cached_rec = prebuilt->fetch_cache[prebuilt->fetch_cache_first];
if (UNIV_UNLIKELY(prebuilt->keep_other_fields_on_keyread)) {
row_sel_copy_cached_fields_for_mysql(buf, cached_rec, prebuilt);
} else if (prebuilt->mysql_prefix_len > 63) {
/* The record is long. Copy it field by field, in case
there are some long VARCHAR column of which only a
small length is being used. */
UNIV_MEM_INVALID(buf, prebuilt->mysql_prefix_len);
/* First copy the NULL bits. */
ut_memcpy(buf, cached_rec, prebuilt->null_bitmap_len);
/* Then copy the requested fields. */
for (i = 0; i < prebuilt->n_template; i++) {
templ = prebuilt->mysql_template + i;
/* Skip virtual columns */
if (templ->is_virtual
&& !(dict_index_has_virtual(prebuilt->index)
&& prebuilt->read_just_key)) {
continue;
}
row_sel_copy_cached_field_for_mysql(
buf, cached_rec, templ);
}
} else {
ut_memcpy(buf, cached_rec, prebuilt->mysql_prefix_len);
}
prebuilt->n_fetch_cached--;
prebuilt->fetch_cache_first++;
if (prebuilt->n_fetch_cached == 0) {
prebuilt->fetch_cache_first = 0;
}
}
/********************************************************************//**
Initialise the prefetch cache. */
UNIV_INLINE
void
row_sel_prefetch_cache_init(
/*========================*/
row_prebuilt_t* prebuilt) /*!< in/out: prebuilt struct */
{
ulint i;
ulint sz;
byte* ptr;
/* Reserve space for the magic number. */
sz = UT_ARR_SIZE(prebuilt->fetch_cache) * (prebuilt->mysql_row_len + 8);
ptr = static_cast<byte*>(ut_malloc_nokey(sz));
for (i = 0; i < UT_ARR_SIZE(prebuilt->fetch_cache); i++) {
/* A user has reported memory corruption in these
buffers in Linux. Put magic numbers there to help
to track a possible bug. */
mach_write_to_4(ptr, ROW_PREBUILT_FETCH_MAGIC_N);
ptr += 4;
prebuilt->fetch_cache[i] = ptr;
ptr += prebuilt->mysql_row_len;
mach_write_to_4(ptr, ROW_PREBUILT_FETCH_MAGIC_N);
ptr += 4;
}
}
/********************************************************************//**
Get the last fetch cache buffer from the queue.
@return pointer to buffer. */
UNIV_INLINE
byte*
row_sel_fetch_last_buf(
/*===================*/
row_prebuilt_t* prebuilt) /*!< in/out: prebuilt struct */
{
ut_ad(!prebuilt->templ_contains_blob);
ut_ad(prebuilt->n_fetch_cached < MYSQL_FETCH_CACHE_SIZE);
if (prebuilt->fetch_cache[0] == NULL) {
/* Allocate memory for the fetch cache */
ut_ad(prebuilt->n_fetch_cached == 0);
row_sel_prefetch_cache_init(prebuilt);
}
ut_ad(prebuilt->fetch_cache_first == 0);
UNIV_MEM_INVALID(prebuilt->fetch_cache[prebuilt->n_fetch_cached],
prebuilt->mysql_row_len);
return(prebuilt->fetch_cache[prebuilt->n_fetch_cached]);
}
/********************************************************************//**
Pushes a row for MySQL to the fetch cache. */
UNIV_INLINE
void
row_sel_enqueue_cache_row_for_mysql(
/*================================*/
byte* mysql_rec, /*!< in/out: MySQL record */
row_prebuilt_t* prebuilt) /*!< in/out: prebuilt struct */
{
/* For non ICP code path the row should already exist in the
next fetch cache slot. */
if (prebuilt->pk_filter || prebuilt->idx_cond) {
byte* dest = row_sel_fetch_last_buf(prebuilt);
ut_memcpy(dest, mysql_rec, prebuilt->mysql_row_len);
}
++prebuilt->n_fetch_cached;
}
#ifdef BTR_CUR_HASH_ADAPT
/*********************************************************************//**
Tries to do a shortcut to fetch a clustered index record with a unique key,
using the hash index if possible (not always). We assume that the search
mode is PAGE_CUR_GE, it is a consistent read, there is a read view in trx,
btr search latch has been locked in S-mode if AHI is enabled.
@return SEL_FOUND, SEL_EXHAUSTED, SEL_RETRY */
static
ulint
row_sel_try_search_shortcut_for_mysql(
/*==================================*/
const rec_t** out_rec,/*!< out: record if found */
row_prebuilt_t* prebuilt,/*!< in: prebuilt struct */
ulint** offsets,/*!< in/out: for rec_get_offsets(*out_rec) */
mem_heap_t** heap, /*!< in/out: heap for rec_get_offsets() */
mtr_t* mtr) /*!< in: started mtr */
{
dict_index_t* index = prebuilt->index;
const dtuple_t* search_tuple = prebuilt->search_tuple;
btr_pcur_t* pcur = prebuilt->pcur;
trx_t* trx = prebuilt->trx;
const rec_t* rec;
ut_ad(dict_index_is_clust(index));
ut_ad(!prebuilt->templ_contains_blob);
rw_lock_t* ahi_latch = btr_get_search_latch(index);
rw_lock_s_lock(ahi_latch);
btr_pcur_open_with_no_init(index, search_tuple, PAGE_CUR_GE,
BTR_SEARCH_LEAF, pcur, ahi_latch, mtr);
rec = btr_pcur_get_rec(pcur);
if (!page_rec_is_user_rec(rec) || rec_is_metadata(rec, *index)) {
retry:
rw_lock_s_unlock(ahi_latch);
return(SEL_RETRY);
}
/* As the cursor is now placed on a user record after a search with
the mode PAGE_CUR_GE, the up_match field in the cursor tells how many
fields in the user record matched to the search tuple */
if (btr_pcur_get_up_match(pcur) < dtuple_get_n_fields(search_tuple)) {
exhausted:
rw_lock_s_unlock(ahi_latch);
return(SEL_EXHAUSTED);
}
/* This is a non-locking consistent read: if necessary, fetch
a previous version of the record */
*offsets = rec_get_offsets(rec, index, *offsets, true,
ULINT_UNDEFINED, heap);
if (!lock_clust_rec_cons_read_sees(rec, index, *offsets,
&trx->read_view)) {
goto retry;
}
if (rec_get_deleted_flag(rec, dict_table_is_comp(index->table))) {
/* In delete-marked records, DB_TRX_ID must
always refer to an existing undo log record. */
ut_ad(row_get_rec_trx_id(rec, index, *offsets));
goto exhausted;
}
*out_rec = rec;
rw_lock_s_unlock(ahi_latch);
return(SEL_FOUND);
}
#endif /* BTR_CUR_HASH_ADAPT */
/*********************************************************************//**
Check a pushed-down index condition.
@return ICP_NO_MATCH, ICP_MATCH, or ICP_OUT_OF_RANGE */
static
ICP_RESULT
row_search_idx_cond_check(
/*======================*/
byte* mysql_rec, /*!< out: record
in MySQL format (invalid unless
prebuilt->idx_cond!=NULL and
we return ICP_MATCH) */
row_prebuilt_t* prebuilt, /*!< in/out: prebuilt struct
for the table handle */
const rec_t* rec, /*!< in: InnoDB record */
const ulint* offsets) /*!< in: rec_get_offsets() */
{
ulint i;
ut_ad(rec_offs_validate(rec, prebuilt->index, offsets));
if (!prebuilt->idx_cond) {
if (!handler_rowid_filter_is_active(prebuilt->pk_filter)) {
return(ICP_MATCH);
}
} else {
MONITOR_INC(MONITOR_ICP_ATTEMPTS);
}
/* Convert to MySQL format those fields that are needed for
evaluating the index condition. */
if (UNIV_LIKELY_NULL(prebuilt->blob_heap)) {
mem_heap_empty(prebuilt->blob_heap);
}
for (i = 0; i < prebuilt->idx_cond_n_cols; i++) {
const mysql_row_templ_t*templ = &prebuilt->mysql_template[i];
/* Skip virtual columns */
if (templ->is_virtual) {
continue;
}
if (!row_sel_store_mysql_field(mysql_rec, prebuilt,
rec, prebuilt->index, offsets,
templ->icp_rec_field_no,
templ)) {
return(ICP_NO_MATCH);
}
}
/* We assume that the index conditions on
case-insensitive columns are case-insensitive. The
case of such columns may be wrong in a secondary
index, if the case of the column has been updated in
the past, or a record has been deleted and a record
inserted in a different case. */
ICP_RESULT result = prebuilt->idx_cond
? handler_index_cond_check(prebuilt->idx_cond)
: ICP_MATCH;
switch (result) {
case ICP_MATCH:
if (handler_rowid_filter_is_active(prebuilt->pk_filter)) {
ut_ad(!prebuilt->index->is_primary());
if (prebuilt->clust_index_was_generated) {
ulint len;
dict_index_t* index = prebuilt->index;
const byte* data = rec_get_nth_field(
rec, offsets, index->n_fields - 1,
&len);
ut_ad(dict_index_get_nth_col(index,
index->n_fields - 1)
->prtype == (DATA_ROW_ID | DATA_NOT_NULL));
ut_ad(len == DATA_ROW_ID_LEN);
memcpy(prebuilt->row_id, data, DATA_ROW_ID_LEN);
}
if (!handler_rowid_filter_check(prebuilt->pk_filter)) {
MONITOR_INC(MONITOR_ICP_MATCH);
return(ICP_NO_MATCH);
}
}
/* Convert the remaining fields to MySQL format.
If this is a secondary index record, we must defer
this until we have fetched the clustered index record. */
if (!prebuilt->need_to_access_clustered
|| dict_index_is_clust(prebuilt->index)) {
if (!row_sel_store_mysql_rec(
mysql_rec, prebuilt, rec, NULL, false,
prebuilt->index, offsets)) {
ut_ad(dict_index_is_clust(prebuilt->index));
return(ICP_NO_MATCH);
}
}
MONITOR_INC(MONITOR_ICP_MATCH);
return(result);
case ICP_NO_MATCH:
MONITOR_INC(MONITOR_ICP_NO_MATCH);
return(result);
case ICP_OUT_OF_RANGE:
MONITOR_INC(MONITOR_ICP_OUT_OF_RANGE);
return(result);
case ICP_ERROR:
case ICP_ABORTED_BY_USER:
return(result);
}
ut_error;
return(result);
}
/** Extract virtual column data from a virtual index record and fill a dtuple
@param[in] rec the virtual (secondary) index record
@param[in] index the virtual index
@param[in,out] vrow the dtuple where data extract to
@param[in] heap memory heap to allocate memory
*/
static
void
row_sel_fill_vrow(
const rec_t* rec,
dict_index_t* index,
dtuple_t** vrow,
mem_heap_t* heap)
{
ulint offsets_[REC_OFFS_NORMAL_SIZE];
ulint* offsets = offsets_;
rec_offs_init(offsets_);
ut_ad(!(*vrow));
ut_ad(heap);
ut_ad(!dict_index_is_clust(index));
ut_ad(!index->is_instant());
ut_ad(page_rec_is_leaf(rec));
offsets = rec_get_offsets(rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
*vrow = dtuple_create_with_vcol(
heap, 0, dict_table_get_n_v_cols(index->table));
/* Initialize all virtual row's mtype to DATA_MISSING */
dtuple_init_v_fld(*vrow);
for (ulint i = 0; i < dict_index_get_n_fields(index); i++) {
const dict_field_t* field;
const dict_col_t* col;
field = dict_index_get_nth_field(index, i);
col = dict_field_get_col(field);
if (col->is_virtual()) {
const byte* data;
ulint len;
data = rec_get_nth_field(rec, offsets, i, &len);
const dict_v_col_t* vcol = reinterpret_cast<
const dict_v_col_t*>(col);
dfield_t* dfield = dtuple_get_nth_v_field(
*vrow, vcol->v_pos);
dfield_set_data(dfield, data, len);
dict_col_copy_type(col, dfield_get_type(dfield));
}
}
}
/** Return the record field length in characters.
@param[in] col table column of the field
@param[in] field_no field number
@param[in] rec physical record
@param[in] offsets field offsets in the physical record
@return field length in characters. */
static
size_t
rec_field_len_in_chars(
const dict_col_t* col,
const ulint field_no,
const rec_t* rec,
const ulint* offsets)
{
const ulint cset = dtype_get_charset_coll(col->prtype);
const CHARSET_INFO* cs = all_charsets[cset];
ulint rec_field_len;
const char* rec_field = reinterpret_cast<const char *>(
rec_get_nth_field(
rec, offsets, field_no, &rec_field_len));
if (UNIV_UNLIKELY(!cs)) {
ib::warn() << "Missing collation " << cset;
return SIZE_T_MAX;
}
return(cs->cset->numchars(cs, rec_field, rec_field + rec_field_len));
}
/** Avoid the clustered index lookup if all the following conditions
are true:
1) all columns are in secondary index
2) all values for columns that are prefix-only indexes are shorter
than the prefix size. This optimization can avoid many IOs for certain schemas.
@return true, to avoid clustered index lookup. */
static
bool row_search_with_covering_prefix(
row_prebuilt_t* prebuilt,
const rec_t* rec,
const ulint* offsets)
{
const dict_index_t* index = prebuilt->index;
ut_ad(!dict_index_is_clust(index));
if (!srv_prefix_index_cluster_optimization) {
return false;
}
/** Optimization only applicable if there the number of secondary index
fields are greater than or equal to number of clustered index fields. */
if (prebuilt->n_template > index->n_fields) {
return false;
}
for (ulint i = 0; i < prebuilt->n_template; i++) {
mysql_row_templ_t* templ = prebuilt->mysql_template + i;
ulint j = templ->rec_prefix_field_no;
/** Condition (1) : is the field in the index. */
if (j == ULINT_UNDEFINED) {
return false;
}
/** Condition (2): If this is a prefix index then
row's value size shorter than prefix length. */
if (!templ->rec_field_is_prefix) {
continue;
}
ulint rec_size = rec_offs_nth_size(offsets, j);
const dict_field_t* field = dict_index_get_nth_field(index, j);
ulint max_chars = field->prefix_len / templ->mbmaxlen;
ut_a(field->prefix_len > 0);
if (rec_size < max_chars) {
/* Record in bytes shorter than the index
prefix length in char. */
continue;
}
if (rec_size * templ->mbminlen >= field->prefix_len) {
/* Shortest representation string by the
byte length of the record is longer than the
maximum possible index prefix. */
return false;
}
size_t num_chars = rec_field_len_in_chars(
field->col, j, rec, offsets);
if (num_chars >= max_chars) {
/* No of chars to store the record exceeds
the index prefix character length. */
return false;
}
}
/* If prefix index optimization condition satisfied then
for all columns above, use rec_prefix_field_no instead of
rec_field_no, and skip the clustered lookup below. */
for (ulint i = 0; i < prebuilt->n_template; i++) {
mysql_row_templ_t* templ = prebuilt->mysql_template + i;
templ->rec_field_no = templ->rec_prefix_field_no;
ut_a(templ->rec_field_no != ULINT_UNDEFINED);
}
srv_stats.n_sec_rec_cluster_reads_avoided.inc();
return true;
}
/** Searches for rows in the database using cursor.
Function is mainly used for tables that are shared across connections and
so it employs technique that can help re-construct the rows that
transaction is suppose to see.
It also has optimization such as pre-caching the rows, using AHI, etc.
@param[out] buf buffer for the fetched row in MySQL format
@param[in] mode search mode PAGE_CUR_L
@param[in,out] prebuilt prebuilt struct for the table handler;
this contains the info to search_tuple,
index; if search tuple contains 0 field then
we position the cursor at start or the end of
index, depending on 'mode'
@param[in] match_mode 0 or ROW_SEL_EXACT or ROW_SEL_EXACT_PREFIX
@param[in] direction 0 or ROW_SEL_NEXT or ROW_SEL_PREV;
Note: if this is != 0, then prebuilt must has a
pcur with stored position! In opening of a
cursor 'direction' should be 0.
@return DB_SUCCESS or error code */
dberr_t
row_search_mvcc(
byte* buf,
page_cur_mode_t mode,
row_prebuilt_t* prebuilt,
ulint match_mode,
ulint direction)
{
DBUG_ENTER("row_search_mvcc");
DBUG_ASSERT(prebuilt->index->table == prebuilt->table);
dict_index_t* index = prebuilt->index;
ibool comp = dict_table_is_comp(prebuilt->table);
const dtuple_t* search_tuple = prebuilt->search_tuple;
btr_pcur_t* pcur = prebuilt->pcur;
trx_t* trx = prebuilt->trx;
dict_index_t* clust_index;
que_thr_t* thr;
const rec_t* UNINIT_VAR(rec);
dtuple_t* vrow = NULL;
const rec_t* result_rec = NULL;
const rec_t* clust_rec;
Row_sel_get_clust_rec_for_mysql row_sel_get_clust_rec_for_mysql;
ibool unique_search = FALSE;
ibool mtr_has_extra_clust_latch = FALSE;
ibool moves_up = FALSE;
/* if the returned record was locked and we did a semi-consistent
read (fetch the newest committed version), then this is set to
TRUE */
ulint next_offs;
ibool same_user_rec;
ibool table_lock_waited = FALSE;
byte* next_buf = 0;
bool spatial_search = false;
ut_ad(index && pcur && search_tuple);
ut_a(prebuilt->magic_n == ROW_PREBUILT_ALLOCATED);
ut_a(prebuilt->magic_n2 == ROW_PREBUILT_ALLOCATED);
/* We don't support FTS queries from the HANDLER interfaces, because
we implemented FTS as reversed inverted index with auxiliary tables.
So anything related to traditional index query would not apply to
it. */
if (prebuilt->index->type & DICT_FTS) {
DBUG_RETURN(DB_END_OF_INDEX);
}
ut_ad(!sync_check_iterate(sync_check()));
if (!prebuilt->table->space) {
DBUG_RETURN(DB_TABLESPACE_DELETED);
} else if (!prebuilt->table->is_readable()) {
DBUG_RETURN(prebuilt->table->space
? DB_DECRYPTION_FAILED
: DB_TABLESPACE_NOT_FOUND);
} else if (!prebuilt->index_usable) {
DBUG_RETURN(DB_MISSING_HISTORY);
} else if (prebuilt->index->is_corrupted()) {
DBUG_RETURN(DB_CORRUPTION);
}
/* We need to get the virtual column values stored in secondary
index key, if this is covered index scan or virtual key read is
requested. */
bool need_vrow = dict_index_has_virtual(prebuilt->index)
&& (prebuilt->read_just_key
|| prebuilt->m_read_virtual_key);
/* Reset the new record lock info if READ UNCOMMITTED or
READ COMMITED isolation level is used. Then
we are able to remove the record locks set here on an individual
row. */
prebuilt->new_rec_locks = 0;
/*-------------------------------------------------------------*/
/* PHASE 1: Try to pop the row from the prefetch cache */
if (UNIV_UNLIKELY(direction == 0)) {
trx->op_info = "starting index read";
prebuilt->n_rows_fetched = 0;
prebuilt->n_fetch_cached = 0;
prebuilt->fetch_cache_first = 0;
if (prebuilt->sel_graph == NULL) {
/* Build a dummy select query graph */
row_prebuild_sel_graph(prebuilt);
}
} else {
trx->op_info = "fetching rows";
if (prebuilt->n_rows_fetched == 0) {
prebuilt->fetch_direction = direction;
}
if (UNIV_UNLIKELY(direction != prebuilt->fetch_direction)) {
if (UNIV_UNLIKELY(prebuilt->n_fetch_cached > 0)) {
ut_error;
/* TODO: scrollable cursor: restore cursor to
the place of the latest returned row,
or better: prevent caching for a scroll
cursor! */
}
prebuilt->n_rows_fetched = 0;
prebuilt->n_fetch_cached = 0;
prebuilt->fetch_cache_first = 0;
} else if (UNIV_LIKELY(prebuilt->n_fetch_cached > 0)) {
row_sel_dequeue_cached_row_for_mysql(buf, prebuilt);
prebuilt->n_rows_fetched++;
trx->op_info = "";
DBUG_RETURN(DB_SUCCESS);
}
if (prebuilt->fetch_cache_first > 0
&& prebuilt->fetch_cache_first < MYSQL_FETCH_CACHE_SIZE) {
early_not_found:
/* The previous returned row was popped from the fetch
cache, but the cache was not full at the time of the
popping: no more rows can exist in the result set */
trx->op_info = "";
DBUG_RETURN(DB_RECORD_NOT_FOUND);
}
prebuilt->n_rows_fetched++;
if (prebuilt->n_rows_fetched > 1000000000) {
/* Prevent wrap-over */
prebuilt->n_rows_fetched = 500000000;
}
mode = pcur->search_mode;
}
/* In a search where at most one record in the index may match, we
can use a LOCK_REC_NOT_GAP type record lock when locking a
non-delete-marked matching record.
Note that in a unique secondary index there may be different
delete-marked versions of a record where only the primary key
values differ: thus in a secondary index we must use next-key
locks when locking delete-marked records. */
if (match_mode == ROW_SEL_EXACT
&& dict_index_is_unique(index)
&& dtuple_get_n_fields(search_tuple)
== dict_index_get_n_unique(index)
&& (dict_index_is_clust(index)
|| !dtuple_contains_null(search_tuple))) {
/* Note above that a UNIQUE secondary index can contain many
rows with the same key value if one of the columns is the SQL
null. A clustered index under MySQL can never contain null
columns because we demand that all the columns in primary key
are non-null. */
unique_search = TRUE;
/* Even if the condition is unique, MySQL seems to try to
retrieve also a second row if a primary key contains more than
1 column. Return immediately if this is not a HANDLER
command. */
if (UNIV_UNLIKELY(direction != 0
&& !prebuilt->used_in_HANDLER)) {
goto early_not_found;
}
}
/* We don't support sequencial scan for Rtree index, because it
is no meaning to do so. */
if (dict_index_is_spatial(index) && !RTREE_SEARCH_MODE(mode)) {
trx->op_info = "";
DBUG_RETURN(DB_END_OF_INDEX);
}
/* if the query is a plain locking SELECT, and the isolation level
is <= TRX_ISO_READ_COMMITTED, then this is set to FALSE */
bool did_semi_consistent_read = false;
mtr_t mtr;
mtr.start();
mem_heap_t* heap = NULL;
ulint offsets_[REC_OFFS_NORMAL_SIZE];
ulint* offsets = offsets_;
rec_offs_init(offsets_);
#ifdef BTR_CUR_HASH_ADAPT
/*-------------------------------------------------------------*/
/* PHASE 2: Try fast adaptive hash index search if possible */
/* Next test if this is the special case where we can use the fast
adaptive hash index to try the search. Since we must release the
search system latch when we retrieve an externally stored field, we
cannot use the adaptive hash index in a search in the case the row
may be long and there may be externally stored fields */
if (UNIV_UNLIKELY(direction == 0)
&& unique_search
&& btr_search_enabled
&& dict_index_is_clust(index)
&& !prebuilt->templ_contains_blob
&& !prebuilt->used_in_HANDLER
&& (prebuilt->mysql_row_len < srv_page_size / 8)) {
mode = PAGE_CUR_GE;
if (prebuilt->select_lock_type == LOCK_NONE
&& trx->isolation_level > TRX_ISO_READ_UNCOMMITTED
&& trx->read_view.is_open()) {
/* This is a SELECT query done as a consistent read,
and the read view has already been allocated:
let us try a search shortcut through the hash
index. */
dberr_t err = DB_SUCCESS;
switch (row_sel_try_search_shortcut_for_mysql(
&rec, prebuilt, &offsets, &heap,
&mtr)) {
case SEL_FOUND:
/* At this point, rec is protected by
a page latch that was acquired by
row_sel_try_search_shortcut_for_mysql().
The latch will not be released until
mtr.commit(). */
ut_ad(!rec_get_deleted_flag(rec, comp));
if (prebuilt->pk_filter || prebuilt->idx_cond) {
switch (row_search_idx_cond_check(
buf, prebuilt,
rec, offsets)) {
case ICP_NO_MATCH:
case ICP_OUT_OF_RANGE:
case ICP_ABORTED_BY_USER:
case ICP_ERROR:
err = DB_RECORD_NOT_FOUND;
goto shortcut_done;
case ICP_MATCH:
goto shortcut_done;
}
}
if (!row_sel_store_mysql_rec(
buf, prebuilt,
rec, NULL, false, index,
offsets)) {
/* Only fresh inserts may contain
incomplete externally stored
columns. Pretend that such
records do not exist. Such
records may only be accessed
at the READ UNCOMMITTED
isolation level or when
rolling back a recovered
transaction. Rollback happens
at a lower level, not here. */
/* Proceed as in case SEL_RETRY. */
break;
}
goto shortcut_done;
case SEL_EXHAUSTED:
err = DB_RECORD_NOT_FOUND;
shortcut_done:
mtr.commit();
/* NOTE that we do NOT store the cursor
position */
trx->op_info = "";
ut_ad(!sync_check_iterate(sync_check()));
ut_ad(!did_semi_consistent_read);
DBUG_RETURN(err);
case SEL_RETRY:
break;
default:
ut_ad(0);
}
mtr.commit();
mtr.start();
}
}
#endif /* BTR_CUR_HASH_ADAPT */
/*-------------------------------------------------------------*/
/* PHASE 3: Open or restore index cursor position */
spatial_search = dict_index_is_spatial(index)
&& mode >= PAGE_CUR_CONTAIN;
/* The state of a running trx can only be changed by the
thread that is currently serving the transaction. Because we
are that thread, we can read trx->state without holding any
mutex. */
ut_ad(prebuilt->sql_stat_start
|| trx->state == TRX_STATE_ACTIVE
|| (prebuilt->table->no_rollback()
&& trx->state == TRX_STATE_NOT_STARTED));
ut_ad(!trx_is_started(trx) || trx->state == TRX_STATE_ACTIVE);
ut_ad(prebuilt->sql_stat_start
|| prebuilt->select_lock_type != LOCK_NONE
|| trx->read_view.is_open()
|| prebuilt->table->no_rollback()
|| srv_read_only_mode);
/* Do not lock gaps for plain SELECT
at READ UNCOMMITTED or READ COMMITTED isolation level */
const bool set_also_gap_locks =
prebuilt->select_lock_type != LOCK_NONE
&& (trx->isolation_level > TRX_ISO_READ_COMMITTED
|| !thd_is_select(trx->mysql_thd))
#ifdef WITH_WSREP
&& !wsrep_thd_skip_locking(trx->mysql_thd)
#endif /* WITH_WSREP */
;
/* Note that if the search mode was GE or G, then the cursor
naturally moves upward (in fetch next) in alphabetical order,
otherwise downward */
if (UNIV_UNLIKELY(direction == 0)) {
if (mode == PAGE_CUR_GE || mode == PAGE_CUR_G
|| mode >= PAGE_CUR_CONTAIN) {
moves_up = TRUE;
}
} else if (direction == ROW_SEL_NEXT) {
moves_up = TRUE;
}
thr = que_fork_get_first_thr(prebuilt->sel_graph);
que_thr_move_to_run_state_for_mysql(thr, trx);
clust_index = dict_table_get_first_index(prebuilt->table);
dberr_t err = DB_SUCCESS;
/* Do some start-of-statement preparations */
if (prebuilt->table->no_rollback()) {
/* NO_ROLLBACK tables do not support MVCC or locking. */
prebuilt->select_lock_type = LOCK_NONE;
prebuilt->sql_stat_start = FALSE;
} else if (!prebuilt->sql_stat_start) {
/* No need to set an intention lock or assign a read view */
ut_a(prebuilt->select_lock_type != LOCK_NONE
|| srv_read_only_mode || trx->read_view.is_open());
} else {
prebuilt->sql_stat_start = FALSE;
trx_start_if_not_started(trx, false);
if (prebuilt->select_lock_type == LOCK_NONE) {
trx->read_view.open(trx);
} else {
wait_table_again:
err = lock_table(0, prebuilt->table,
prebuilt->select_lock_type == LOCK_S
? LOCK_IS : LOCK_IX, thr);
if (err != DB_SUCCESS) {
table_lock_waited = TRUE;
goto lock_table_wait;
}
}
}
/* Open or restore index cursor position */
if (UNIV_LIKELY(direction != 0)) {
if (spatial_search) {
/* R-Tree access does not need to do
cursor position and resposition */
goto next_rec;
}
bool need_to_process = sel_restore_position_for_mysql(
&same_user_rec, BTR_SEARCH_LEAF,
pcur, moves_up, &mtr);
if (UNIV_UNLIKELY(need_to_process)) {
if (UNIV_UNLIKELY(prebuilt->row_read_type
== ROW_READ_DID_SEMI_CONSISTENT)) {
/* We did a semi-consistent read,
but the record was removed in
the meantime. */
prebuilt->row_read_type
= ROW_READ_TRY_SEMI_CONSISTENT;
}
} else if (UNIV_LIKELY(prebuilt->row_read_type
!= ROW_READ_DID_SEMI_CONSISTENT)) {
/* The cursor was positioned on the record
that we returned previously. If we need
to repeat a semi-consistent read as a
pessimistic locking read, the record
cannot be skipped. */
goto next_rec;
}
} else if (dtuple_get_n_fields(search_tuple) > 0) {
pcur->btr_cur.thr = thr;
if (dict_index_is_spatial(index)) {
if (!prebuilt->rtr_info) {
prebuilt->rtr_info = rtr_create_rtr_info(
set_also_gap_locks, true,
btr_pcur_get_btr_cur(pcur), index);
prebuilt->rtr_info->search_tuple = search_tuple;
prebuilt->rtr_info->search_mode = mode;
rtr_info_update_btr(btr_pcur_get_btr_cur(pcur),
prebuilt->rtr_info);
} else {
rtr_info_reinit_in_cursor(
btr_pcur_get_btr_cur(pcur),
index, set_also_gap_locks);
prebuilt->rtr_info->search_tuple = search_tuple;
prebuilt->rtr_info->search_mode = mode;
}
}
err = btr_pcur_open_with_no_init(index, search_tuple, mode,
BTR_SEARCH_LEAF,
pcur, 0, &mtr);
if (err != DB_SUCCESS) {
rec = NULL;
goto page_read_error;
}
pcur->trx_if_known = trx;
rec = btr_pcur_get_rec(pcur);
ut_ad(page_rec_is_leaf(rec));
if (!moves_up
&& set_also_gap_locks
&& !page_rec_is_supremum(rec)
&& !dict_index_is_spatial(index)) {
/* Try to place a gap lock on the next index record
to prevent phantoms in ORDER BY ... DESC queries */
const rec_t* next_rec = page_rec_get_next_const(rec);
offsets = rec_get_offsets(next_rec, index, offsets,
true,
ULINT_UNDEFINED, &heap);
err = sel_set_rec_lock(pcur,
next_rec, index, offsets,
prebuilt->select_lock_type,
LOCK_GAP, thr, &mtr);
switch (err) {
case DB_SUCCESS_LOCKED_REC:
err = DB_SUCCESS;
/* fall through */
case DB_SUCCESS:
break;
default:
goto lock_wait_or_error;
}
}
} else if (mode == PAGE_CUR_G || mode == PAGE_CUR_L) {
err = btr_pcur_open_at_index_side(
mode == PAGE_CUR_G, index, BTR_SEARCH_LEAF,
pcur, false, 0, &mtr);
if (err != DB_SUCCESS) {
if (err == DB_DECRYPTION_FAILED) {
ib_push_warning(trx->mysql_thd,
DB_DECRYPTION_FAILED,
"Table %s is encrypted but encryption service or"
" used key_id is not available. "
" Can't continue reading table.",
prebuilt->table->name.m_name);
index->table->file_unreadable = true;
}
rec = NULL;
goto page_read_error;
}
}
rec_loop:
DEBUG_SYNC_C("row_search_rec_loop");
if (trx_is_interrupted(trx)) {
if (!spatial_search) {
btr_pcur_store_position(pcur, &mtr);
}
err = DB_INTERRUPTED;
goto normal_return;
}
/*-------------------------------------------------------------*/
/* PHASE 4: Look for matching records in a loop */
rec = btr_pcur_get_rec(pcur);
if (!index->table->is_readable()) {
err = DB_DECRYPTION_FAILED;
goto page_read_error;
}
ut_ad(!!page_rec_is_comp(rec) == comp);
ut_ad(page_rec_is_leaf(rec));
if (page_rec_is_infimum(rec)) {
/* The infimum record on a page cannot be in the result set,
and neither can a record lock be placed on it: we skip such
a record. */
goto next_rec;
}
if (page_rec_is_supremum(rec)) {
if (set_also_gap_locks
&& trx->isolation_level > TRX_ISO_READ_COMMITTED
&& !dict_index_is_spatial(index)) {
/* Try to place a lock on the index record */
/* If the transaction isolation level is
READ UNCOMMITTED or READ COMMITTED,
we do not lock gaps. Supremum record is really
a gap and therefore we do not set locks there. */
offsets = rec_get_offsets(rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
err = sel_set_rec_lock(pcur,
rec, index, offsets,
prebuilt->select_lock_type,
LOCK_ORDINARY, thr, &mtr);
switch (err) {
case DB_SUCCESS_LOCKED_REC:
err = DB_SUCCESS;
/* fall through */
case DB_SUCCESS:
break;
default:
goto lock_wait_or_error;
}
}
/* A page supremum record cannot be in the result set: skip
it now that we have placed a possible lock on it */
goto next_rec;
}
/*-------------------------------------------------------------*/
/* Do sanity checks in case our cursor has bumped into page
corruption */
if (comp) {
if (rec_get_info_bits(rec, true) & REC_INFO_MIN_REC_FLAG) {
/* Skip the metadata pseudo-record. */
ut_ad(index->is_instant());
goto next_rec;
}
next_offs = rec_get_next_offs(rec, TRUE);
if (UNIV_UNLIKELY(next_offs < PAGE_NEW_SUPREMUM)) {
goto wrong_offs;
}
} else {
if (rec_get_info_bits(rec, false) & REC_INFO_MIN_REC_FLAG) {
/* Skip the metadata pseudo-record. */
ut_ad(index->is_instant());
goto next_rec;
}
next_offs = rec_get_next_offs(rec, FALSE);
if (UNIV_UNLIKELY(next_offs < PAGE_OLD_SUPREMUM)) {
goto wrong_offs;
}
}
if (UNIV_UNLIKELY(next_offs >= srv_page_size - PAGE_DIR)) {
wrong_offs:
if (srv_force_recovery == 0 || moves_up == FALSE) {
ib::error() << "Rec address "
<< static_cast<const void*>(rec)
<< ", buf block fix count "
<< btr_cur_get_block(
btr_pcur_get_btr_cur(pcur))->page
.buf_fix_count;
ib::error() << "Index corruption: rec offs "
<< page_offset(rec) << " next offs "
<< next_offs << ", page no "
<< page_get_page_no(page_align(rec))
<< ", index " << index->name
<< " of table " << index->table->name
<< ". Run CHECK TABLE. You may need to"
" restore from a backup, or dump + drop +"
" reimport the table.";
ut_ad(0);
err = DB_CORRUPTION;
goto page_read_error;
} else {
/* The user may be dumping a corrupt table. Jump
over the corruption to recover as much as possible. */
ib::info() << "Index corruption: rec offs "
<< page_offset(rec) << " next offs "
<< next_offs << ", page no "
<< page_get_page_no(page_align(rec))
<< ", index " << index->name
<< " of table " << index->table->name
<< ". We try to skip the rest of the page.";
page_cur_set_after_last(btr_pcur_get_block(pcur),
btr_pcur_get_page_cur(pcur));
pcur->old_stored = false;
goto next_rec;
}
}
/*-------------------------------------------------------------*/
/* Calculate the 'offsets' associated with 'rec' */
ut_ad(fil_page_index_page_check(btr_pcur_get_page(pcur)));
ut_ad(btr_page_get_index_id(btr_pcur_get_page(pcur)) == index->id);
offsets = rec_get_offsets(rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
if (UNIV_UNLIKELY(srv_force_recovery > 0)) {
if (!rec_validate(rec, offsets)
|| !btr_index_rec_validate(rec, index, FALSE)) {
ib::error() << "Index corruption: rec offs "
<< page_offset(rec) << " next offs "
<< next_offs << ", page no "
<< page_get_page_no(page_align(rec))
<< ", index " << index->name
<< " of table " << index->table->name
<< ". We try to skip the record.";
goto next_rec;
}
}
/* Note that we cannot trust the up_match value in the cursor at this
place because we can arrive here after moving the cursor! Thus
we have to recompare rec and search_tuple to determine if they
match enough. */
if (match_mode == ROW_SEL_EXACT) {
/* Test if the index record matches completely to search_tuple
in prebuilt: if not, then we return with DB_RECORD_NOT_FOUND */
/* fputs("Comparing rec and search tuple\n", stderr); */
if (0 != cmp_dtuple_rec(search_tuple, rec, offsets)) {
if (set_also_gap_locks
&& !dict_index_is_spatial(index)) {
err = sel_set_rec_lock(
pcur,
rec, index, offsets,
prebuilt->select_lock_type, LOCK_GAP,
thr, &mtr);
switch (err) {
case DB_SUCCESS_LOCKED_REC:
case DB_SUCCESS:
break;
default:
goto lock_wait_or_error;
}
}
btr_pcur_store_position(pcur, &mtr);
/* The found record was not a match, but may be used
as NEXT record (index_next). Set the relative position
to BTR_PCUR_BEFORE, to reflect that the position of
the persistent cursor is before the found/stored row
(pcur->old_rec). */
ut_ad(pcur->rel_pos == BTR_PCUR_ON);
pcur->rel_pos = BTR_PCUR_BEFORE;
err = DB_RECORD_NOT_FOUND;
goto normal_return;
}
} else if (match_mode == ROW_SEL_EXACT_PREFIX) {
if (!cmp_dtuple_is_prefix_of_rec(search_tuple, rec, offsets)) {
if (set_also_gap_locks
&& !dict_index_is_spatial(index)) {
err = sel_set_rec_lock(
pcur,
rec, index, offsets,
prebuilt->select_lock_type, LOCK_GAP,
thr, &mtr);
switch (err) {
case DB_SUCCESS_LOCKED_REC:
case DB_SUCCESS:
break;
default:
goto lock_wait_or_error;
}
}
btr_pcur_store_position(pcur, &mtr);
/* The found record was not a match, but may be used
as NEXT record (index_next). Set the relative position
to BTR_PCUR_BEFORE, to reflect that the position of
the persistent cursor is before the found/stored row
(pcur->old_rec). */
ut_ad(pcur->rel_pos == BTR_PCUR_ON);
pcur->rel_pos = BTR_PCUR_BEFORE;
err = DB_RECORD_NOT_FOUND;
goto normal_return;
}
}
/* We are ready to look at a possible new index entry in the result
set: the cursor is now placed on a user record */
if (prebuilt->select_lock_type != LOCK_NONE) {
/* Try to place a lock on the index record; note that delete
marked records are a special case in a unique search. If there
is a non-delete marked record, then it is enough to lock its
existence with LOCK_REC_NOT_GAP. */
ulint lock_type;
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED) {
/* At READ COMMITTED or READ UNCOMMITTED
isolation levels, do not lock committed
delete-marked records. */
if (!rec_get_deleted_flag(rec, comp)) {
goto no_gap_lock;
}
/* At most one transaction can be active
for temporary table. */
if (clust_index->table->is_temporary()) {
goto no_gap_lock;
}
if (index == clust_index) {
trx_id_t trx_id = row_get_rec_trx_id(
rec, index, offsets);
/* In delete-marked records, DB_TRX_ID must
always refer to an existing undo log record. */
ut_ad(trx_id);
if (!trx_sys.is_registered(trx, trx_id)) {
/* The clustered index record
was delete-marked in a committed
transaction. Ignore the record. */
goto locks_ok_del_marked;
}
} else if (trx_t* t = row_vers_impl_x_locked(
trx, rec, index, offsets)) {
/* The record belongs to an active
transaction. We must acquire a lock. */
t->release_reference();
} else {
/* The secondary index record does not
point to a delete-marked clustered index
record that belongs to an active transaction.
Ignore the secondary index record, because
it is not locked. */
goto next_rec;
}
goto no_gap_lock;
}
if (!set_also_gap_locks
|| (unique_search && !rec_get_deleted_flag(rec, comp))
|| dict_index_is_spatial(index)) {
goto no_gap_lock;
} else {
lock_type = LOCK_ORDINARY;
}
/* If we are doing a 'greater or equal than a primary key
value' search from a clustered index, and we find a record
that has that exact primary key value, then there is no need
to lock the gap before the record, because no insert in the
gap can be in our search range. That is, no phantom row can
appear that way.
An example: if col1 is the primary key, the search is WHERE
col1 >= 100, and we find a record where col1 = 100, then no
need to lock the gap before that record. */
if (index == clust_index
&& mode == PAGE_CUR_GE
&& direction == 0
&& dtuple_get_n_fields_cmp(search_tuple)
== dict_index_get_n_unique(index)
&& 0 == cmp_dtuple_rec(search_tuple, rec, offsets)) {
no_gap_lock:
lock_type = LOCK_REC_NOT_GAP;
}
err = sel_set_rec_lock(pcur,
rec, index, offsets,
prebuilt->select_lock_type,
lock_type, thr, &mtr);
switch (err) {
const rec_t* old_vers;
case DB_SUCCESS_LOCKED_REC:
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED) {
/* Note that a record of
prebuilt->index was locked. */
prebuilt->new_rec_locks = 1;
}
err = DB_SUCCESS;
/* fall through */
case DB_SUCCESS:
break;
case DB_LOCK_WAIT:
/* Lock wait for R-tree should already
be handled in sel_set_rtr_rec_lock() */
ut_ad(!dict_index_is_spatial(index));
/* Never unlock rows that were part of a conflict. */
prebuilt->new_rec_locks = 0;
if (UNIV_LIKELY(prebuilt->row_read_type
!= ROW_READ_TRY_SEMI_CONSISTENT)
|| unique_search
|| index != clust_index) {
goto lock_wait_or_error;
}
/* The following call returns 'offsets'
associated with 'old_vers' */
row_sel_build_committed_vers_for_mysql(
clust_index, prebuilt, rec,
&offsets, &heap, &old_vers, need_vrow ? &vrow : NULL,
&mtr);
/* Check whether it was a deadlock or not, if not
a deadlock and the transaction had to wait then
release the lock it is waiting on. */
err = lock_trx_handle_wait(trx);
switch (err) {
case DB_SUCCESS:
/* The lock was granted while we were
searching for the last committed version.
Do a normal locking read. */
offsets = rec_get_offsets(
rec, index, offsets, true,
ULINT_UNDEFINED, &heap);
goto locks_ok;
case DB_DEADLOCK:
goto lock_wait_or_error;
case DB_LOCK_WAIT:
ut_ad(!dict_index_is_spatial(index));
err = DB_SUCCESS;
break;
default:
ut_error;
}
if (old_vers == NULL) {
/* The row was not yet committed */
goto next_rec;
}
did_semi_consistent_read = true;
rec = old_vers;
break;
case DB_RECORD_NOT_FOUND:
if (dict_index_is_spatial(index)) {
goto next_rec;
} else {
goto lock_wait_or_error;
}
default:
goto lock_wait_or_error;
}
} else {
/* This is a non-locking consistent read: if necessary, fetch
a previous version of the record */
if (trx->isolation_level == TRX_ISO_READ_UNCOMMITTED
|| prebuilt->table->no_rollback()) {
/* Do nothing: we let a non-locking SELECT read the
latest version of the record */
} else if (index == clust_index) {
/* Fetch a previous version of the row if the current
one is not visible in the snapshot; if we have a very
high force recovery level set, we try to avoid crashes
by skipping this lookup */
if (!lock_clust_rec_cons_read_sees(
rec, index, offsets, &trx->read_view)) {
ut_ad(srv_force_recovery
< SRV_FORCE_NO_UNDO_LOG_SCAN);
rec_t* old_vers;
/* The following call returns 'offsets'
associated with 'old_vers' */
err = row_sel_build_prev_vers_for_mysql(
&trx->read_view, clust_index,
prebuilt, rec, &offsets, &heap,
&old_vers, need_vrow ? &vrow : NULL,
&mtr);
if (err != DB_SUCCESS) {
goto lock_wait_or_error;
}
if (old_vers == NULL) {
/* The row did not exist yet in
the read view */
goto next_rec;
}
rec = old_vers;
}
} else {
/* We are looking into a non-clustered index,
and to get the right version of the record we
have to look also into the clustered index: this
is necessary, because we can only get the undo
information via the clustered index record. */
ut_ad(!dict_index_is_clust(index));
if (!srv_read_only_mode
&& !lock_sec_rec_cons_read_sees(
rec, index, &trx->read_view)) {
/* We should look at the clustered index.
However, as this is a non-locking read,
we can skip the clustered index lookup if
the condition does not match the secondary
index entry. */
switch (row_search_idx_cond_check(
buf, prebuilt, rec, offsets)) {
case ICP_NO_MATCH:
goto next_rec;
case ICP_OUT_OF_RANGE:
case ICP_ABORTED_BY_USER:
case ICP_ERROR:
err = DB_RECORD_NOT_FOUND;
goto idx_cond_failed;
case ICP_MATCH:
goto requires_clust_rec;
}
ut_error;
}
}
}
locks_ok:
/* NOTE that at this point rec can be an old version of a clustered
index record built for a consistent read. We cannot assume after this
point that rec is on a buffer pool page. Functions like
page_rec_is_comp() cannot be used! */
if (rec_get_deleted_flag(rec, comp)) {
locks_ok_del_marked:
/* In delete-marked records, DB_TRX_ID must
always refer to an existing undo log record. */
ut_ad(index != clust_index
|| row_get_rec_trx_id(rec, index, offsets));
/* The record is delete-marked: we can skip it */
/* This is an optimization to skip setting the next key lock
on the record that follows this delete-marked record. This
optimization works because of the unique search criteria
which precludes the presence of a range lock between this
delete marked record and the record following it.
For now this is applicable only to clustered indexes while
doing a unique search except for HANDLER queries because
HANDLER allows NEXT and PREV even in unique search on
clustered index. There is scope for further optimization
applicable to unique secondary indexes. Current behaviour is
to widen the scope of a lock on an already delete marked record
if the same record is deleted twice by the same transaction */
if (index == clust_index && unique_search
&& !prebuilt->used_in_HANDLER) {
err = DB_RECORD_NOT_FOUND;
goto normal_return;
}
goto next_rec;
}
/* Check if the record matches the index condition. */
switch (row_search_idx_cond_check(buf, prebuilt, rec, offsets)) {
case ICP_NO_MATCH:
if (did_semi_consistent_read) {
row_unlock_for_mysql(prebuilt, TRUE);
}
goto next_rec;
case ICP_OUT_OF_RANGE:
case ICP_ABORTED_BY_USER:
case ICP_ERROR:
err = DB_RECORD_NOT_FOUND;
goto idx_cond_failed;
case ICP_MATCH:
break;
}
if (index != clust_index && prebuilt->need_to_access_clustered) {
if (row_search_with_covering_prefix(prebuilt, rec, offsets)) {
goto use_covering_index;
}
requires_clust_rec:
ut_ad(index != clust_index);
/* We use a 'goto' to the preceding label if a consistent
read of a secondary index record requires us to look up old
versions of the associated clustered index record. */
ut_ad(rec_offs_validate(rec, index, offsets));
/* It was a non-clustered index and we must fetch also the
clustered index record */
mtr_has_extra_clust_latch = TRUE;
ut_ad(!vrow);
/* The following call returns 'offsets' associated with
'clust_rec'. Note that 'clust_rec' can be an old version
built for a consistent read. */
err = row_sel_get_clust_rec_for_mysql(prebuilt, index, rec,
thr, &clust_rec,
&offsets, &heap,
need_vrow ? &vrow : NULL,
&mtr);
switch (err) {
case DB_SUCCESS:
if (clust_rec == NULL) {
/* The record did not exist in the read view */
ut_ad(prebuilt->select_lock_type == LOCK_NONE
|| dict_index_is_spatial(index));
goto next_rec;
}
break;
case DB_SUCCESS_LOCKED_REC:
ut_a(clust_rec != NULL);
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED) {
/* Note that the clustered index record
was locked. */
prebuilt->new_rec_locks = 2;
}
err = DB_SUCCESS;
break;
default:
vrow = NULL;
goto lock_wait_or_error;
}
if (rec_get_deleted_flag(clust_rec, comp)) {
/* The record is delete marked: we can skip it */
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED
&& prebuilt->select_lock_type != LOCK_NONE) {
/* No need to keep a lock on a delete-marked
record if we do not want to use next-key
locking. */
row_unlock_for_mysql(prebuilt, TRUE);
}
goto next_rec;
}
if (need_vrow && !vrow) {
if (!heap) {
heap = mem_heap_create(100);
}
row_sel_fill_vrow(rec, index, &vrow, heap);
}
result_rec = clust_rec;
ut_ad(rec_offs_validate(result_rec, clust_index, offsets));
if (prebuilt->pk_filter || prebuilt->idx_cond) {
/* Convert the record to MySQL format. We were
unable to do this in row_search_idx_cond_check(),
because the condition is on the secondary index
and the requested column is in the clustered index.
We convert all fields, including those that
may have been used in ICP, because the
secondary index may contain a column prefix
rather than the full column. Also, as noted
in Bug #56680, the column in the secondary
index may be in the wrong case, and the
authoritative case is in result_rec, the
appropriate version of the clustered index record. */
if (!row_sel_store_mysql_rec(
buf, prebuilt, result_rec, vrow,
true, clust_index, offsets)) {
goto next_rec;
}
}
} else {
use_covering_index:
result_rec = rec;
}
/* We found a qualifying record 'result_rec'. At this point,
'offsets' are associated with 'result_rec'. */
ut_ad(rec_offs_validate(result_rec,
result_rec != rec ? clust_index : index,
offsets));
ut_ad(!rec_get_deleted_flag(result_rec, comp));
/* Decide whether to prefetch extra rows.
At this point, the clustered index record is protected
by a page latch that was acquired when pcur was positioned.
The latch will not be released until mtr.commit(). */
if ((match_mode == ROW_SEL_EXACT
|| prebuilt->n_rows_fetched >= MYSQL_FETCH_CACHE_THRESHOLD)
&& prebuilt->select_lock_type == LOCK_NONE
&& !prebuilt->m_no_prefetch
&& !prebuilt->templ_contains_blob
&& !prebuilt->clust_index_was_generated
&& !prebuilt->used_in_HANDLER
&& prebuilt->template_type != ROW_MYSQL_DUMMY_TEMPLATE
&& !prebuilt->in_fts_query) {
/* Inside an update, for example, we do not cache rows,
since we may use the cursor position to do the actual
update, that is why we require ...lock_type == LOCK_NONE.
Since we keep space in prebuilt only for the BLOBs of
a single row, we cannot cache rows in the case there
are BLOBs in the fields to be fetched. In HANDLER we do
not cache rows because there the cursor is a scrollable
cursor. */
ut_a(prebuilt->n_fetch_cached < MYSQL_FETCH_CACHE_SIZE);
/* We only convert from InnoDB row format to MySQL row
format when ICP is disabled. */
if (!prebuilt->pk_filter && !prebuilt->idx_cond) {
/* We use next_buf to track the allocation of buffers
where we store and enqueue the buffers for our
pre-fetch optimisation.
If next_buf == 0 then we store the converted record
directly into the MySQL record buffer (buf). If it is
!= 0 then we allocate a pre-fetch buffer and store the
converted record there.
If the conversion fails and the MySQL record buffer
was not written to then we reset next_buf so that
we can re-use the MySQL record buffer in the next
iteration. */
next_buf = next_buf
? row_sel_fetch_last_buf(prebuilt) : buf;
if (!row_sel_store_mysql_rec(
next_buf, prebuilt, result_rec, vrow,
result_rec != rec,
result_rec != rec ? clust_index : index,
offsets)) {
if (next_buf == buf) {
ut_a(prebuilt->n_fetch_cached == 0);
next_buf = 0;
}
/* Only fresh inserts may contain incomplete
externally stored columns. Pretend that such
records do not exist. Such records may only be
accessed at the READ UNCOMMITTED isolation
level or when rolling back a recovered
transaction. Rollback happens at a lower
level, not here. */
goto next_rec;
}
if (next_buf != buf) {
row_sel_enqueue_cache_row_for_mysql(
next_buf, prebuilt);
}
} else {
row_sel_enqueue_cache_row_for_mysql(buf, prebuilt);
}
if (prebuilt->n_fetch_cached < MYSQL_FETCH_CACHE_SIZE) {
goto next_rec;
}
} else {
if (UNIV_UNLIKELY
(prebuilt->template_type == ROW_MYSQL_DUMMY_TEMPLATE)) {
/* CHECK TABLE: fetch the row */
if (result_rec != rec
&& !prebuilt->need_to_access_clustered) {
/* We used 'offsets' for the clust
rec, recalculate them for 'rec' */
offsets = rec_get_offsets(rec, index, offsets,
true,
ULINT_UNDEFINED,
&heap);
result_rec = rec;
}
memcpy(buf + 4, result_rec
- rec_offs_extra_size(offsets),
rec_offs_size(offsets));
mach_write_to_4(buf,
rec_offs_extra_size(offsets) + 4);
} else if (!prebuilt->pk_filter && !prebuilt->idx_cond) {
/* The record was not yet converted to MySQL format. */
if (!row_sel_store_mysql_rec(
buf, prebuilt, result_rec, vrow,
result_rec != rec,
result_rec != rec ? clust_index : index,
offsets)) {
/* Only fresh inserts may contain
incomplete externally stored
columns. Pretend that such records do
not exist. Such records may only be
accessed at the READ UNCOMMITTED
isolation level or when rolling back a
recovered transaction. Rollback
happens at a lower level, not here. */
goto next_rec;
}
}
if (!prebuilt->clust_index_was_generated) {
} else if (result_rec != rec || index->is_primary()) {
memcpy(prebuilt->row_id, result_rec, DATA_ROW_ID_LEN);
} else {
ulint len;
const byte* data = rec_get_nth_field(
result_rec, offsets, index->n_fields - 1,
&len);
ut_ad(dict_index_get_nth_col(index,
index->n_fields - 1)
->prtype == (DATA_ROW_ID | DATA_NOT_NULL));
ut_ad(len == DATA_ROW_ID_LEN);
memcpy(prebuilt->row_id, data, DATA_ROW_ID_LEN);
}
}
/* From this point on, 'offsets' are invalid. */
/* We have an optimization to save CPU time: if this is a consistent
read on a unique condition on the clustered index, then we do not
store the pcur position, because any fetch next or prev will anyway
return 'end of file'. Exceptions are locking reads and the MySQL
HANDLER command where the user can move the cursor with PREV or NEXT
even after a unique search. */
err = DB_SUCCESS;
idx_cond_failed:
if (!unique_search
|| !dict_index_is_clust(index)
|| direction != 0
|| prebuilt->select_lock_type != LOCK_NONE
|| prebuilt->used_in_HANDLER) {
/* Inside an update always store the cursor position */
if (!spatial_search) {
btr_pcur_store_position(pcur, &mtr);
}
}
goto normal_return;
next_rec:
/* Reset the old and new "did semi-consistent read" flags. */
if (UNIV_UNLIKELY(prebuilt->row_read_type
== ROW_READ_DID_SEMI_CONSISTENT)) {
prebuilt->row_read_type = ROW_READ_TRY_SEMI_CONSISTENT;
}
did_semi_consistent_read = false;
prebuilt->new_rec_locks = 0;
vrow = NULL;
/*-------------------------------------------------------------*/
/* PHASE 5: Move the cursor to the next index record */
/* NOTE: For moves_up==FALSE, the mini-transaction will be
committed and restarted every time when switching b-tree
pages. For moves_up==TRUE in index condition pushdown, we can
scan an entire secondary index tree within a single
mini-transaction. As long as the prebuilt->idx_cond does not
match, we do not need to consult the clustered index or
return records to MySQL, and thus we can avoid repositioning
the cursor. What prevents us from buffer-fixing all leaf pages
within the mini-transaction is the btr_leaf_page_release()
call in btr_pcur_move_to_next_page(). Only the leaf page where
the cursor is positioned will remain buffer-fixed.
For R-tree spatial search, we also commit the mini-transaction
each time */
if (spatial_search) {
/* No need to do store restore for R-tree */
mtr.commit();
mtr.start();
mtr_has_extra_clust_latch = FALSE;
} else if (mtr_has_extra_clust_latch) {
/* If we have extra cluster latch, we must commit
mtr if we are moving to the next non-clustered
index record, because we could break the latching
order if we would access a different clustered
index page right away without releasing the previous. */
btr_pcur_store_position(pcur, &mtr);
mtr.commit();
mtr_has_extra_clust_latch = FALSE;
mtr.start();
if (sel_restore_position_for_mysql(&same_user_rec,
BTR_SEARCH_LEAF,
pcur, moves_up, &mtr)) {
goto rec_loop;
}
}
if (moves_up) {
if (UNIV_UNLIKELY(spatial_search)) {
if (rtr_pcur_move_to_next(
search_tuple, mode, pcur, 0, &mtr)) {
goto rec_loop;
}
} else {
const buf_block_t* block = btr_pcur_get_block(pcur);
/* This is based on btr_pcur_move_to_next(),
but avoids infinite read loop of a corrupted page. */
ut_ad(pcur->pos_state == BTR_PCUR_IS_POSITIONED);
ut_ad(pcur->latch_mode != BTR_NO_LATCHES);
pcur->old_stored = false;
if (btr_pcur_is_after_last_on_page(pcur)) {
if (btr_pcur_is_after_last_in_tree(pcur)) {
goto not_moved;
}
btr_pcur_move_to_next_page(pcur, &mtr);
if (UNIV_UNLIKELY(btr_pcur_get_block(pcur)
== block)) {
err = DB_CORRUPTION;
goto lock_wait_or_error;
}
} else {
btr_pcur_move_to_next_on_page(pcur);
}
goto rec_loop;
}
} else {
if (btr_pcur_move_to_prev(pcur, &mtr)) {
goto rec_loop;
}
}
not_moved:
if (!spatial_search) {
btr_pcur_store_position(pcur, &mtr);
}
err = match_mode ? DB_RECORD_NOT_FOUND : DB_END_OF_INDEX;
goto normal_return;
lock_wait_or_error:
if (!dict_index_is_spatial(index)) {
btr_pcur_store_position(pcur, &mtr);
}
page_read_error:
/* Reset the old and new "did semi-consistent read" flags. */
if (UNIV_UNLIKELY(prebuilt->row_read_type
== ROW_READ_DID_SEMI_CONSISTENT)) {
prebuilt->row_read_type = ROW_READ_TRY_SEMI_CONSISTENT;
}
did_semi_consistent_read = false;
lock_table_wait:
mtr.commit();
mtr_has_extra_clust_latch = FALSE;
trx->error_state = err;
/* The following is a patch for MySQL */
if (thr->is_active) {
que_thr_stop_for_mysql(thr);
}
thr->lock_state = QUE_THR_LOCK_ROW;
if (row_mysql_handle_errors(&err, trx, thr, NULL)) {
/* It was a lock wait, and it ended */
thr->lock_state = QUE_THR_LOCK_NOLOCK;
mtr.start();
/* Table lock waited, go try to obtain table lock
again */
if (table_lock_waited) {
table_lock_waited = FALSE;
goto wait_table_again;
}
if (!dict_index_is_spatial(index)) {
sel_restore_position_for_mysql(
&same_user_rec, BTR_SEARCH_LEAF, pcur,
moves_up, &mtr);
}
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED
&& !same_user_rec) {
/* Since we were not able to restore the cursor
on the same user record, we cannot use
row_unlock_for_mysql() to unlock any records, and
we must thus reset the new rec lock info. Since
in lock0lock.cc we have blocked the inheriting of gap
X-locks, we actually do not have any new record locks
set in this case.
Note that if we were able to restore on the 'same'
user record, it is still possible that we were actually
waiting on a delete-marked record, and meanwhile
it was removed by purge and inserted again by some
other user. But that is no problem, because in
rec_loop we will again try to set a lock, and
new_rec_lock_info in trx will be right at the end. */
prebuilt->new_rec_locks = 0;
}
mode = pcur->search_mode;
goto rec_loop;
}
thr->lock_state = QUE_THR_LOCK_NOLOCK;
goto func_exit;
normal_return:
/*-------------------------------------------------------------*/
{
/* handler_index_cond_check() may pull TR_table search
which initates another row_search_mvcc(). */
ulint n_active_thrs= trx->lock.n_active_thrs;
trx->lock.n_active_thrs= 1;
que_thr_stop_for_mysql_no_error(thr, trx);
trx->lock.n_active_thrs= n_active_thrs - 1;
}
mtr.commit();
DEBUG_SYNC_C("row_search_for_mysql_before_return");
if (prebuilt->pk_filter || prebuilt->idx_cond) {
/* When ICP is active we don't write to the MySQL buffer
directly, only to buffers that are enqueued in the pre-fetch
queue. We need to dequeue the first buffer and copy the contents
to the record buffer that was passed in by MySQL. */
if (prebuilt->n_fetch_cached > 0) {
row_sel_dequeue_cached_row_for_mysql(buf, prebuilt);
err = DB_SUCCESS;
}
} else if (next_buf != 0) {
/* We may or may not have enqueued some buffers to the
pre-fetch queue, but we definitely wrote to the record
buffer passed to use by MySQL. */
DEBUG_SYNC_C("row_search_cached_row");
err = DB_SUCCESS;
}
#ifdef UNIV_DEBUG
if (dict_index_is_spatial(index) && err != DB_SUCCESS
&& err != DB_END_OF_INDEX && err != DB_INTERRUPTED) {
rtr_node_path_t* path = pcur->btr_cur.rtr_info->path;
ut_ad(path->empty());
}
#endif
func_exit:
trx->op_info = "";
if (heap != NULL) {
mem_heap_free(heap);
}
/* Set or reset the "did semi-consistent read" flag on return.
The flag did_semi_consistent_read is set if and only if
the record being returned was fetched with a semi-consistent read. */
ut_ad(prebuilt->row_read_type != ROW_READ_WITH_LOCKS
|| !did_semi_consistent_read);
if (prebuilt->row_read_type != ROW_READ_WITH_LOCKS) {
if (did_semi_consistent_read) {
prebuilt->row_read_type = ROW_READ_DID_SEMI_CONSISTENT;
} else {
prebuilt->row_read_type = ROW_READ_TRY_SEMI_CONSISTENT;
}
}
ut_ad(!sync_check_iterate(sync_check()));
DEBUG_SYNC_C("innodb_row_search_for_mysql_exit");
DBUG_RETURN(err);
}
/********************************************************************//**
Count rows in a R-Tree leaf level.
@return DB_SUCCESS if successful */
dberr_t
row_count_rtree_recs(
/*=================*/
row_prebuilt_t* prebuilt, /*!< in: prebuilt struct for the
table handle; this contains the info
of search_tuple, index; if search
tuple contains 0 fields then we
position the cursor at the start or
the end of the index, depending on
'mode' */
ulint* n_rows) /*!< out: number of entries
seen in the consistent read */
{
dict_index_t* index = prebuilt->index;
dberr_t ret = DB_SUCCESS;
mtr_t mtr;
mem_heap_t* heap;
dtuple_t* entry;
dtuple_t* search_entry = prebuilt->search_tuple;
ulint entry_len;
ulint i;
byte* buf;
ut_a(dict_index_is_spatial(index));
*n_rows = 0;
heap = mem_heap_create(256);
/* Build a search tuple. */
entry_len = dict_index_get_n_fields(index);
entry = dtuple_create(heap, entry_len);
for (i = 0; i < entry_len; i++) {
const dict_field_t* ind_field
= dict_index_get_nth_field(index, i);
const dict_col_t* col
= ind_field->col;
dfield_t* dfield
= dtuple_get_nth_field(entry, i);
if (i == 0) {
double* mbr;
double tmp_mbr[SPDIMS * 2];
dfield->type.mtype = DATA_GEOMETRY;
dfield->type.prtype |= DATA_GIS_MBR;
/* Allocate memory for mbr field */
mbr = static_cast<double*>
(mem_heap_alloc(heap, DATA_MBR_LEN));
/* Set mbr field data. */
dfield_set_data(dfield, mbr, DATA_MBR_LEN);
for (uint j = 0; j < SPDIMS; j++) {
tmp_mbr[j * 2] = DBL_MAX;
tmp_mbr[j * 2 + 1] = -DBL_MAX;
}
dfield_write_mbr(dfield, tmp_mbr);
continue;
}
dfield->type.mtype = col->mtype;
dfield->type.prtype = col->prtype;
}
prebuilt->search_tuple = entry;
ulint bufsize = std::max<ulint>(srv_page_size,
prebuilt->mysql_row_len);
buf = static_cast<byte*>(ut_malloc_nokey(bufsize));
ulint cnt = 1000;
ret = row_search_for_mysql(buf, PAGE_CUR_WITHIN, prebuilt, 0, 0);
loop:
/* Check thd->killed every 1,000 scanned rows */
if (--cnt == 0) {
if (trx_is_interrupted(prebuilt->trx)) {
ret = DB_INTERRUPTED;
goto func_exit;
}
cnt = 1000;
}
switch (ret) {
case DB_SUCCESS:
break;
case DB_DEADLOCK:
case DB_LOCK_TABLE_FULL:
case DB_LOCK_WAIT_TIMEOUT:
case DB_INTERRUPTED:
goto func_exit;
default:
/* fall through (this error is ignored by CHECK TABLE) */
case DB_END_OF_INDEX:
ret = DB_SUCCESS;
func_exit:
prebuilt->search_tuple = search_entry;
ut_free(buf);
mem_heap_free(heap);
return(ret);
}
*n_rows = *n_rows + 1;
ret = row_search_for_mysql(
buf, PAGE_CUR_WITHIN, prebuilt, 0, ROW_SEL_NEXT);
goto loop;
}
/*******************************************************************//**
Read the AUTOINC column from the current row. If the value is less than
0 and the type is not unsigned then we reset the value to 0.
@return value read from the column */
static
ib_uint64_t
row_search_autoinc_read_column(
/*===========================*/
dict_index_t* index, /*!< in: index to read from */
const rec_t* rec, /*!< in: current rec */
ulint col_no, /*!< in: column number */
ulint mtype, /*!< in: column main type */
ibool unsigned_type) /*!< in: signed or unsigned flag */
{
ulint len;
const byte* data;
ib_uint64_t value;
mem_heap_t* heap = NULL;
ulint offsets_[REC_OFFS_NORMAL_SIZE];
ulint* offsets = offsets_;
rec_offs_init(offsets_);
ut_ad(page_rec_is_leaf(rec));
offsets = rec_get_offsets(rec, index, offsets, true,
col_no + 1, &heap);
if (rec_offs_nth_sql_null(offsets, col_no)) {
/* There is no non-NULL value in the auto-increment column. */
value = 0;
goto func_exit;
}
data = rec_get_nth_field(rec, offsets, col_no, &len);
value = row_parse_int(data, len, mtype, unsigned_type);
func_exit:
if (UNIV_LIKELY_NULL(heap)) {
mem_heap_free(heap);
}
return(value);
}
/** Get the maximum and non-delete-marked record in an index.
@param[in] index index tree
@param[in,out] mtr mini-transaction (may be committed and restarted)
@return maximum record, page s-latched in mtr
@retval NULL if there are no records, or if all of them are delete-marked */
static
const rec_t*
row_search_get_max_rec(
dict_index_t* index,
mtr_t* mtr)
{
btr_pcur_t pcur;
const rec_t* rec;
/* Open at the high/right end (false), and init cursor */
btr_pcur_open_at_index_side(
false, index, BTR_SEARCH_LEAF, &pcur, true, 0, mtr);
do {
const page_t* page;
page = btr_pcur_get_page(&pcur);
rec = page_find_rec_max_not_deleted(page);
if (page_rec_is_user_rec(rec)) {
break;
} else {
rec = NULL;
}
btr_pcur_move_before_first_on_page(&pcur);
} while (btr_pcur_move_to_prev(&pcur, mtr));
btr_pcur_close(&pcur);
ut_ad(!rec
|| !(rec_get_info_bits(rec, dict_table_is_comp(index->table))
& (REC_INFO_MIN_REC_FLAG | REC_INFO_DELETED_FLAG)));
return(rec);
}
/** Read the max AUTOINC value from an index.
@param[in] index index starting with an AUTO_INCREMENT column
@return the largest AUTO_INCREMENT value
@retval 0 if no records were found */
ib_uint64_t
row_search_max_autoinc(dict_index_t* index)
{
const dict_field_t* dfield = dict_index_get_nth_field(index, 0);
ib_uint64_t value = 0;
mtr_t mtr;
mtr.start();
if (const rec_t* rec = row_search_get_max_rec(index, &mtr)) {
value = row_search_autoinc_read_column(
index, rec, 0,
dfield->col->mtype,
dfield->col->prtype & DATA_UNSIGNED);
}
mtr.commit();
return(value);
}
|