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
|
<!doctype HTML public "-//W3C//DTD HTML 4.0 Frameset//EN">
<html>
<head>
<title>Chapter 5: HDF5 Datasets</title>
<!--(Meta)==========================================================-->
<!--(Links)=========================================================-->
<!--( Begin styles definition )=====================================-->
<link href="ed_styles/NewUGelect.css" rel="stylesheet" type="text/css">
<!--( End styles definition )=======================================-->
</head>
<body>
<!-- #BeginLibraryItem "/ed_libs/styles_UG.lbi" -->
<!--
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* Copyright by The HDF Group. *
* Copyright by the Board of Trustees of the University of Illinois. *
* All rights reserved. *
* *
* This file is part of HDF5. The full HDF5 copyright notice, including *
* terms governing use, modification, and redistribution, is contained in *
* the files COPYING and Copyright.html. COPYING can be found at the root *
* of the source code distribution tree; Copyright.html can be found at the *
* root level of an installed copy of the electronic HDF5 document set and *
* is linked from the top-level documents page. It can also be found at *
* http://hdfgroup.org/HDF5/doc/Copyright.html. If you do not have *
* access to either file, you may request a copy from help@hdfgroup.org. *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
-->
<!-- #EndLibraryItem --><!-- HEADER LEFT "HDF5 User's Guide" -->
<!-- HEADER RIGHT "HDF5 Datasets" -->
<!--( TOC )=========================================================-->
<!--<SCRIPT language="JavaScript">-->
<!--
document.writeln ('\
<table x-use-null-cells\
align=right\
width=240\
cellspacing="0"\
class="tocTable">\
-->
<!-- Table Version 3 --><!-- \ -->
<!--
<tr valign="top"> \
<td class="tocTableHeaderCell" colspan="2"> \
<span class="TableHead">Chapter Contents</span></td>\
</tr>\
<tr valign="top"> \
<td class="tocTableContentCell2"> \
<a href="#Intro">1.</a></td>\
<td class="tocTableContentCell3">\
<a href="#Intro">Introduction</a></td> \
</tr>\
<tr valign="top"> \
<td class="tocTableContentCell2"> \
<a href="#FileFunctSums">2.</a></td>\
<td class="tocTableContentCell3">\
<a href="#FileFunctSums">Dataset (H5D) Function Summaries</a></td>\
</tr>\
<tr valign="top"> \
<td class="tocTableContentCell2"> \
<a href="#PModel">3.</a></td>\
<td class="tocTableContentCell3">\
<a href="#PModel">Programming Model</a></td> \
</tr>\
<tr valign="top"> \
<td class="tocTableContentCell2"> \
<a href="#DTransfer">4.</a></td>\
<td class="tocTableContentCell3">\
<a href="#DTransfer">Data Transfer</a></td>\
</tr>\
<tr valign="top"> \
<td class="tocTableContentCell2"> \
<a href="#Allocation">5.</a></td>\
<td class="tocTableContentCell3">\
<a href="#Allocation">Allocation of Space</a></td>\
</tr>\
<tr valign="top"> \
<td class="tocTableContentCell"> \
<a href="#UseFilters">6.</a></td>\
<td class="tocTableContentCell4">\
<a href="#UseFilters">Specialized Filters</a>\
<br> \
<a href="#N-Bit">N-bit</a>\
<br> \
<a href="#ScaleOffset">Scale-offset</a></td>\
</tr>\
</td></tr>\
</table>\
')
-->
<!--</SCRIPT>-->
<!--(End TOC)=======================================================-->
<!-- editingComment
<span class="editingComment">[ [ [
] ] ]</span>
-->
<!-- editingComment
-->
<div align="center">
<a name="TOP">
<h2>Chapter 5<br><font size="7">HDF5 Datasets</font></h2>
</a>
</div>
<!-- editingComment
<span class="editingComment">[ [ [
Original title. Which is proper?
<h2>10. Datasets I/O</h2>
] ] ]</span>
-->
<br />
<a name="Intro">
<h3>5.1. Introduction</h3>
</a>
<p>An HDF5 dataset is an object composed of a collection of data elements,
or raw data, and metadata that stores a description of the data elements,
data layout, and all other information necessary to write, read, and interpret
the stored data. From the viewpoint of the application the raw data is stored
as a one-dimensional or multi-dimensional array of elements (the <em>raw
data</em>), those elements can be any of several numerical or character
types, small arrays, or even compound types similar to C structs. The
dataset object may have attribute objects. See the figure below.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig1.JPG">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 1. Application view of a dataset</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<!-- editingComment
<span class="editingComment">
<p>Datatypes are described in [ [ [ "Datatypes" ] ] ]. and in the [ [ [ "HDF5
Datatypes" chapter in this guide ] ] ], Dataspace objects are described in
[ [ [ Dataspace ] ] ], and Attributes are described in [ [ [ Attributes ] ] ].
</span>
-->
<p>A dataset object is stored in a file in two parts: a header and a data
array. The header contains information that is needed to interpret the
array portion of the dataset, as well as metadata (or pointers to metadata)
that describes or annotates the dataset. Header information includes the
name of the object, its dimensionality, its number-type, information about
how the data itself is stored on disk (the <em>storage layout</em>), and
other information used by the library to speed up access to the dataset
or maintain the file’s integrity. </p>
<p>The HDF5 dataset interface, comprising the H5D functions, provides a
mechanism for managing HDF5 datasets including the transfer of data
between memory and disk and the description of dataset properties. </p>
<p>A dataset is used by other HDF5 APIs, either by name or by an
identifier (e.g., returned by <code>H5Dopen</code>).</p>
<h4><em>Link/Unlink</em></h4>
<p>A dataset can be added to a group with one of the <code>H5Lcreate</code>
calls, and deleted from a group with <code>H5Ldelete</code>. The link and
unlink operations use the name of an object, which may be a dataset. The
dataset does not have to open to be linked or unlinked.</p>
<h4><em>Object reference</em></h4>
<p>A dataset may be the target of an object reference. The object
reference is created by <code>H5Rcreate</code> with the name of an object
which may be a dataset and the reference type <code>H5R_OBJECT</code>.
The dataset does not have to be open to create a reference to it.</p>
<p>An object reference may also refer to a region (selection) of a dataset.
The reference is created with <code>H5Rcreate</code> and a reference type of
<code>H5R_DATASET_REGION</code>.</p>
<p>An object reference can be accessed by a call to <code>H5Rdereference</code>. When the
reference is to a dataset or dataset region, the <code>H5Rdeference</code>
call returns an identifier to the dataset just as if <code>H5Dopen</code>
has been called.</p>
<h4><em>Adding attributes</em></h4>
<p>A dataset may have user-defined attributes which are created with
<code>H5Acreate</code> and accessed through the H5A API. To create an
attribute for a dataset, the dataset must be open, and the identifier is
passed to <code>H5Acreate</code>. The attributes of a dataset are
discovered and opened using <code>H5Aopen_name</code>,
<code>H5Aopen_idx</code>, or <code>H5Aiterate</code>; these functions
use the identifier of the dataset. An attribute can be deleted with
<code>H5Adelete</code> which also uses the identifier of the dataset.</p>
<!-- editingComment
<span class="editingComment">
<p>The remaining sections of this chapter discuss... [To be written last.]</p>
</span>
-->
<br>
<SCRIPT language="JavaScript">
<!--
document.writeln ("
<a name="FileFunctSums">
<div align=right>
<a href="#TOP"><font size="-1">(Top)</font></a>
</div>
</a>
");
-->
</SCRIPT>
<!-- NEW PAGE -->
<a name="FileFunctSums">
<h3 class=pagebefore>5.2. Dataset Function Summaries</h3>
</a>
<p>Functions that can be used with datasets (H5D functions) and property
list functions that can used with datasets (H5P functions) are listed below.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="3" align="left" valign="bottom">
<b>Function Listing 1. Dataset functions (H5D)
</b></td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td>
<b>C Function<br>Fortran Function</b>
</td><td> </td>
<td>
<b>Purpose</b>
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dcreate<br>h5dcreate_f</code>
</td><td> </td>
<td>
Creates a dataset at the specified location. The C function is a
macro: see <a href="../RM/APICompatMacros.html">“API
Compatibility Macros in HDF5.”</a>
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dcreate_anon<br>h5dcreate_anon_f</code>
</td><td> </td>
<td>
Creates a dataset in a file without linking it into the file structure.
</td>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dopen<br>h5dopen_f</code>
</td><td> </td>
<td>
Opens an existing dataset. The C function is a
macro: see <a href="../RM/APICompatMacros.html">“API
Compatibility Macros in HDF5.”</a>
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dclose<br>h5dclose_f</code>
</td><td> </td>
<td>
Closes the specified dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dget_space<br>h5dget_space_f</code>
</td><td> </td>
<td>
Returns an identifier for a copy of the dataspace for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dget_space_status<br>h5dget_space_status_f</code>
</td><td> </td>
<td>
Determines whether space has been allocated for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dget_type<br>h5dget_type_f</code>
</td><td> </td>
<td>
Returns an identifier for a copy of the datatype for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dget_create_plist<br>h5dget_create_plist_f</code>
</td><td> </td>
<td>
Returns an identifier for a copy of the dataset creation property
list for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dget_access_plist<br>(none)</code>
</td><td> </td>
<td>
Returns the dataset access property list associated with a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dget_offset<br>h5dget_offset_f</code>
</td><td> </td>
<td>
Returns the dataset address in a file.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dget_storage_size<br>h5dget_storage_size_f</code>
</td><td> </td>
<td>
Returns the amount of storage required for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dvlen_get_buf_size<br>h5dvlen_get_max_len_f</code>
</td><td> </td>
<td>
Determines the number of bytes required to store variable-length (VL)
data.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dvlen_reclaim<br>h5dvlen_reclaim_f</code>
</td><td> </td>
<td>
Reclaims VL datatype memory buffers.
</td>
</tr>
<!-- NEW PAGE -->
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dread<br>h5dread_f</code>
</td><td> </td>
<td>
Reads raw data from a dataset into a buffer.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dwrite<br>h5dwrite_f</code>
</td><td> </td>
<td>
Writes raw data from a buffer to a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Diterate<br>(none)</code>
</td><td> </td>
<td>
Iterates over all selected elements in a dataspace.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dgather<br>(none)</code>
</td><td> </td>
<td>
Gathers data from a selection within a memory buffer.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dscatter<br>(none)</code>
</td><td> </td>
<td>
Scatters data into a selection within a memory buffer.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dfill<br>h5dfill_f</code>
</td><td> </td>
<td>
Fills dataspace elements with a fill value in a memory buffer.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Dset_extent<br>h5dset_extent_f</code>
</td><td> </td>
<td>
Changes the sizes of a dataset’s dimensions.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
</table>
<br />
<br />
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="3" align="left" valign="bottom">
<b>Function Listing 2. Dataset creation property list functions (H5P)
</b></td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td>
<b>C Function<br>Fortran Function</b>
</td><td> </td>
<td>
<b>Purpose</b>
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_layout<br>h5pset_layout_f</code>
</td><td> </td>
<td>
Sets the type of storage used to store the raw data for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_layout<br>h5pget_layout_f</code>
</td><td> </td>
<td>
Returns the layout of the raw data for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_chunk<br>h5pset_chunk_f</code>
</td><td> </td>
<td>
Sets the size of the chunks used to store a chunked layout dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_chunk<br>h5pget_chunk_f</code>
</td><td> </td>
<td>
Retrieves the size of chunks for the raw data of a chunked layout
dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_deflate<br>h5pset_deflate_f</code>
</td><td> </td>
<td>
Sets compression method and compression level.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_fill_value<br>h5pset_fill_value_f</code>
</td><td> </td>
<td>
Sets the fill value for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_fill_value<br>h5pget_fill_value_f</code>
</td><td> </td>
<td>
Retrieves a dataset fill value.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pfill_value_defined<br>(none)</code>
</td><td> </td>
<td>
Determines whether the fill value is defined.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_fill_time<br>h5pset_fill_time_f</code>
</td><td> </td>
<td>
Sets the time when fill values are written to a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_fill_time<br>h5pget_fill_time_f</code>
</td><td> </td>
<td>
Retrieves the time when fill value are written to a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_alloc_time<br>h5pset_alloc_time_f</code>
</td><td> </td>
<td>
Sets the timing for storage space allocation.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_alloc_time<br>h5pget_alloc_time_f</code>
</td><td> </td>
<td>
Retrieves the timing for storage space allocation.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_filter<br>h5pset_filter_f</code>
</td><td> </td>
<td>
Adds a filter to the filter pipeline.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pall_filters_avail<br>(none)</code>
</td><td> </td>
<td>
Verifies that all required filters are available.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_nfilters<br>h5pget_nfilters_f</code>
</td><td> </td>
<td>
Returns the number of filters in the pipeline.
</td>
</tr>
<!-- NEW PAGE -->
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_filter<br>h5pget_filter_f</code>
</td><td> </td>
<td>
Returns information about a filter in a pipeline. The C function is a
macro: see <a href="../RM/APICompatMacros.html">“API
Compatibility Macros in HDF5.”</a>
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_filter_by_id<br>h5pget_filter_by_id_f</code>
</td><td> </td>
<td>
Returns information about the specified filter. The C function is a
macro: see <a href="../RM/APICompatMacros.html">“API
Compatibility Macros in HDF5.”</a>
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pmodify_filter<br>h5pmodify_filter_f</code>
</td><td> </td>
<td>
Modifies a filter in the filter pipeline.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Premove_filter<br>h5premove_filter_f</code>
</td><td> </td>
<td>
Deletes one or more filters in the filter pipeline.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_fletcher32<br>h5pset_fletcher32_f</code>
</td><td> </td>
<td>
Sets up use of the Fletcher32 checksum filter.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_nbit<br>h5pset_nbit_f</code>
</td><td> </td>
<td>
Sets up use of the n-bit filter.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_scaleoffset<br>h5pset_scaleoffset_f</code>
</td><td> </td>
<td>
Sets up use of the scale-offset filter.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_shuffle<br>h5pset_shuffle_f</code>
</td><td> </td>
<td>
Sets up use of the shuffle filter.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_szip<br>h5pset_szip_f</code>
</td><td> </td>
<td>
Sets up use of the Szip compression filter.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_external<br>h5pset_external_f</code>
</td><td> </td>
<td>
Adds an external file to the list of external files.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_external_count<br>h5pget_external_count_f</code>
</td><td> </td>
<td>
Returns the number of external files for a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_external<br>h5pget_external_f</code>
</td><td> </td>
<td>
Returns information about an external file.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_char_encoding<br>h5pset_char_encoding_f</code>
</td><td> </td>
<td>
Sets the character encoding used to encode a string.
Use to set ASCII or UTF-8 character encoding for object names.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_char_encoding<br>h5pget_char_encoding_f</code>
</td><td> </td>
<td>
Retrieves the character encoding used to create a string.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
</table>
<br />
<br/>
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="3" align="left" valign="bottom">
<b>Function Listing 3. Dataset access property list functions (H5P)
</b></td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td>
<b>C Function<br>Fortran Function</b>
</td><td> </td>
<td>
<b>Purpose</b>
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_buffer<br>h5pset_buffer_f</code>
</td><td> </td>
<td>
Sets type conversion and background buffers.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_buffer<br>h5pget_buffer_f</code>
</td><td> </td>
<td>
Reads buffer settings.
</td>
</tr>
<!-- 8.10.10, MEE: I removed two dataset access property list functions:
H5Pset_preserve and H5Pget_preserve. -->
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_chunk_cache<br>h5pset_chunk_cache_f</code>
</td><td> </td>
<td>
Sets the raw data chunk cache parameters.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_chunk_cache<br>h5pget_chunk_cache_f</code>
</td><td> </td>
<td>
Retrieves the raw data chunk cache parameters.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_edc_check<br>h5pset_edc_check_f</code>
</td><td> </td>
<td>
Sets whether to enable error-detection when reading a dataset.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_edc_check<br>h5pget_edc_check_f</code>
</td><td> </td>
<td>
Determines whether error-detection is enabled for dataset reads.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_filter_callback<br>(none)</code>
</td><td> </td>
<td>
Sets user-defined filter callback function.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_data_transform<br>h5pset_data_transform_f</code>
</td><td> </td>
<td>
Sets a data transform expression.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_data_transform<br>h5pget_data_transform_f</code>
</td><td> </td>
<td>
Retrieves a data transform expression.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_type_conv_cb<br>(none)</code>
</td><td> </td>
<td>
Sets user-defined datatype conversion callback function.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_type_conv_cb<br>(none)</code>
</td><td> </td>
<td>
Gets user-defined datatype conversion callback function.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_hyper_vector_size<br>h5pset_hyper_vector_size_f</code>
</td><td> </td>
<td>
Sets number of I/O vectors to be read/written in hyperslab I/O.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_hyper_vector_size<br>h5pget_hyper_vector_size_f</code>
</td><td> </td>
<td>
Retrieves number of I/O vectors to be read/written in hyperslab I/O.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_btree_ratios<br>h5pset_btree_ratios_f</code>
</td><td> </td>
<td>
Sets B-tree split ratios for a dataset transfer property list.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_btree_ratios<br>h5pget_btree_ratios_f</code>
</td><td> </td>
<td>
Gets B-tree split ratios for a dataset transfer property list.
</td>
</tr>
<!-- NEW PAGE -->
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_vlen_mem_manager<br>(none)</code>
</td><td> </td>
<td>
Sets the memory manager for variable-length datatype allocation in
<code>H5Dread</code> and <code>H5Dvlen_reclaim</code>.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_vlen_mem_manager<br>(none)</code>
</td><td> </td>
<td>
Gets the memory manager for variable-length datatype allocation in
<code>H5Dread</code> and <code>H5Dvlen_reclaim</code>.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_dxpl_mpio<br>h5pset_dxpl_mpio_f</code>
</td><td> </td>
<td>
Sets data transfer mode.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_dxpl_mpio<br>h5pget_dxpl_mpio_f</code>
</td><td> </td>
<td>
Returns the data transfer mode.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_dxpl_mpio_chunk_opt<br>(none)</code>
</td><td> </td>
<td>
Sets a flag specifying linked-chunk I/O or multi-chunk I/O.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_dxpl_mpio_chunk_opt_num<br>(none)</code>
</td><td> </td>
<td>
Sets a numeric threshold for linked-chunk I/O.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_dxpl_mpio_chunk_opt_ratio<br>(none)</code>
</td><td> </td>
<td>
Sets a ratio threshold for collective I/O.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_dxpl_mpio_collective_opt<br>(none)</code>
</td><td> </td>
<td>
Sets a flag governing the use of independent versus collective I/O.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_multi_type<br>(none)</code>
</td><td> </td>
<td>
Sets the type of data property for the MULTI driver.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_multi_type<br>(none)</code>
</td><td> </td>
<td>
Retrieves the type of data property for the MULTI driver.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pset_small_data_block_size<br>h5pset_small_data_block_size_f</code>
</td><td> </td>
<td>
Sets the size of a contiguous block reserved for small data.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>
<code>H5Pget_small_data_block_size<br>h5pget_small_data_block_size_f</code>
</td><td> </td>
<td>
Retrieves the current small data block size setting.
</td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
</table>
<br />
<br />
<SCRIPT language="JavaScript">
<!--
document.writeln ("
<a name="PModel">
<div align=right>
<a href="#TOP"><font size="-1">(Top)</font></a>
</div>
</a>
");
-->
</SCRIPT>
<!-- NEW PAGE -->
<a name="PModel">
<h3 class=pagebefore>5.3. Programming Model</h3>
</a>
<p>This section explains the programming model for datasets.</p>
<br />
<h4>5.3.1. General Model</h4>
<p>The programming model for using a dataset has three main phases:</p>
<ul>
<li>Obtain access to the dataset </li>
<li>Operate on the dataset using the dataset identifier returned
at access </li>
<li>Release the dataset</li>
</ul>
<p>These three phases or steps are described in more detail below the
figure.</p>
<p>A dataset may be opened several times and operations performed
with several different identifiers to the same dataset. All the
operations affect the dataset although the calling program must
synchronize if necessary to serialize accesses.</p>
<p>Note that the dataset remains open until every identifier is closed.
The figure below shows the basic sequence of operations.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig2.JPG">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 2. Dataset programming sequence</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>Creation and data access operations may have optional parameters
which are set with property lists. The general programming model is:</p>
<ul>
<li>Create property list of appropriate class (dataset create,
dataset transfer)</li>
<li>Set properties as needed; each type of property has its own
format and datatype</li>
<li>Pass the property list as a parameter of the API call</li>
</ul>
<p>The steps below describe the programming phases or steps for using a
dataset.</p>
<h4><em>Step 1. Obtain Access</em></h4>
<p>A new dataset is created by a call to <code>H5Dcreate</code>. If
successful, the call returns an identifier for the newly created dataset.</p>
<p>Access to an existing dataset is obtained by a call to <code>H5Dopen</code>. This call
returns an identifier for the existing dataset.</p>
<p>An object reference
<!-- editingComment
<span class="editingComment">[ [ [
(Chapter ???)
] ] ]</span>
-->
may be dereferenced to obtain an identifier to
the dataset it points to.</p>
<p>In each of these cases, the successful call returns an identifier
to the dataset. The identifier is used in subsequent operations until
the dataset is closed.</p>
<h4><em>Step 2. Operate on the Dataset</em></h4>
<p>The dataset identifier can be used to write and read data to the dataset,
to query and set properties, and to perform other operations such as
adding attributes, linking in groups, and creating references.</p>
<p>The dataset identifier can be used for any number of
operations until the dataset is closed.</p>
<h4><em>Step 3. Close the Dataset</em></h4>
<p>When all operations are completed, the dataset identifier should
be closed. This releases the dataset. </p>
<!-- editingComment
<span class="editingComment">[ [ [
{ and writes all metadata to the file? }
] ] ]</span>
-->
<p>After the identifier is closed, it cannot be used for further operations.</p>
<h4>5.3.2. Create Dataset</h4>
<p>A dataset is created and initialized with a call to <code>H5Dcreate</code>. The dataset
create operation sets permanent properties of the dataset:</p>
<ul>
<li>Name</li>
<li>Dataspace</li>
<li>Datatype</li>
<li>Storage properties</li>
</ul>
<p>These properties cannot be changed for the life of the dataset,
although the dataspace may be expanded up to its maximum dimensions.</p>
<h4><em>Name</em></h4>
<p>A dataset name is a sequence of alphanumeric ASCII characters. The
full name would include a tracing of the group hierarchy from the root
group of the file, e.g., /rootGroup/groupA/subgroup23/dataset1. The
local name or relative name within the lowest-level group containing
the dataset would include none of the group hierarchy. e.g., Dataset1.</p>
<h4><em>Dataspace</em></h4>
<p>The dataspace of a dataset defines the number of dimensions and the
size of each dimension.
<!-- editingComment
<span class="editingComment">[ [ [
[[Dataspace]].
] ] ]</span>
-->
The dataspace defines the number of dimensions,
and the maximum dimension sizes and current size of each dimension.
The maximum dimension size can be a fixed value or the constant
<code>H5D_UNLIMITED</code>, in which case the actual dimension size
can be changed with calls to <code>H5Dset_extent</code>, up to the
maximum set with the <code>maxdims</code> parameter in the
<a href="../RM/RM_H5S.html#Dataspace-CreateSimple" target=RMwindow>
<code>H5Screate_simple</code></a> call that established the
dataset’s original dimensions. The maximum dimension size is set
when the dataset is created and cannot be changed.</p>
<h4><em>Datatype</em></h4>
<p>Raw data has a datatype which describes the layout of the raw data
stored in the file.
<!-- editingComment
<span class="editingComment">[ [ [
(See [[Datatype]].
] ] ]</span>
-->
The datatype is set when the dataset is created
and can never be changed. When data is transferred to and from the dataset,
the HDF5 Library will assure that the data is transformed to and
from the stored format.</p>
<h4><em>Storage Properties</em></h4>
<p>Storage properties of the dataset are set when it is created. The
required inputs table
below shows the categories of storage properties. The storage properties
cannot be changed after the dataset is created. </p>
<!-- editingComment
<span class="editingComment">[ [ [
in [[storage properties]]
] ] ]</span>
-->
<h4><em>Filters</em></h4>
<p>When a dataset is created, optional filters are specified. The
filters are added to the data transfer pipeline when data is read or
written. The standard library includes filters to implement compression,
data shuffling, and error detection code. Additional user-defined
filters may also be used. </p>
<!-- editingComment
<span class="editingComment">[ [ [
See [[[filter]]].
] ] ]</span>
-->
<p>The required filters are stored as part of the dataset, and the list may
not be changed after the dataset is created. The HDF5 Library automatically
applies the filters whenever data is transferred.</p>
<h4><em>Summary</em></h4>
<p>A newly created dataset has no attributes and no data values. The
dimensions, datatype, storage properties, and selected filters are set.
The table below lists the required inputs, and the second table below lists
the optional inputs.</p>
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 1. Required inputs</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="25%"><b>Required Inputs</b></td>
<td width="75%"><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Dataspace</td>
<td>The shape of the array.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Datatype</td>
<td>The layout of the stored elements.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Name</td>
<td>The name of the dataset in the group.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<br />
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 2. Optional inputs</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="25%"><b>Optional Inputs</b></td>
<td width="75%"><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Storage Layout</td>
<td>How the data is organized in the file including chunking.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> Fill Value</td>
<td>The behavior and value for uninitialized data.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> External Storage</td>
<td>Option to store the raw data in an external file.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> Filters</td>
<td>Select optional filters to be applied. One of the filters
that might be applied is compression.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<!-- NEW PAGE -->
<h4><em>Example</em></h4>
<p>To create a new dataset, go through the following general steps:</p>
<ul>
<li>Set dataset characteristics (optional where default settings are
acceptable)</li>
<ul>
<li>Datatype</li>
<li>Dataspace</li>
<li>Dataset creation property list</li>
</ul>
<li>Create the dataset</li>
<li>Close the datatype, dataspace, and property list (as necessary)</li>
<li>Close the dataset</li>
</ul>
<p>Example 1 below shows example code to create an empty dataset. The
dataspace is 7 x 8, and the datatype is a big-endian integer. The dataset
is created with the name “dset1” and is a member of the root
group, “/”.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
hid_t dataset, datatype, dataspace;
/*
* Create dataspace: Describe the size of the array and
* create the dataspace for fixed-size dataset.
*/
dimsf[0] = 7;
dimsf[1] = 8;
dataspace = H5Screate_simple(2, dimsf, NULL);
/*
* Define datatype for the data in the file.
* For this example, store little-endian integer numbers.
*/
datatype = H5Tcopy(H5T_NATIVE_INT);
status = H5Tset_order(datatype, H5T_ORDER_LE);
/*
* Create a new dataset within the file using defined
* dataspace and datatype. No properties are set.
*/
dataset = H5Dcreate(file, "/dset", datatype, dataspace, H5P_DEFAULT,
H5P_DEFAULT, H5P_DEFAULT);
H5Dclose(dataset);
H5Sclose(dataspace);
H5Tclose(datatype);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 1. Create an empty dataset</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<!-- NEW PAGE -->
<p>Example 2 below shows example code to create a similar dataset with a
fill value of ‘-1’.
This code has the same steps as in the example above, but uses a non-default
property list. A file creation property list is created, and then the
fill value is set to the desired value. Then the property list is passed
to the <code>H5Dcreate</code> call.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
hid_t dataset, datatype, dataspace;
hid_t plist; /* property list */
int fillval = -1;
dimsf[0] = 7;
dimsf[1] = 8;
dataspace = H5Screate_simple(2, dimsf, NULL);
datatype = H5Tcopy(H5T_NATIVE_INT);
status = H5Tset_order(datatype, H5T_ORDER_LE);
/*
* Example of Dataset Creation property list: set fill value to '-1'
*/
plist = H5Pcreate(H5P_DATASET_CREATE);
status = H5Pset_fill_value(plist, datatype, &fillval);
/* Same as above, but use the property list */
dataset = H5Dcreate(file, "/dset", datatype, dataspace, H5P_DEFAULT,
plist, H5P_DEFAULT);
H5Dclose(dataset);
H5Sclose(dataspace);
H5Tclose(datatype);
H5Pclose(plist);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 2. Create a dataset with fill value set</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>After this code is executed, the dataset has been created and written to
the file. The data array is uninitialized. Depending on the storage
strategy and fill value options that have been selected, some or all of the
space may be allocated in the file, and fill values may be written in the
file.</p>
<!-- editingComment
<span class="editingComment">[ [ [
See <<below>>.
] ] ]</span>
-->
<h4>5.3.3. Data Transfer Operations on a Dataset</h4>
<p>Data is transferred between memory and the raw data array of the dataset
through <code>H5Dwrite</code> and <code>H5Dread</code> operations. A data
transfer has the following basic steps:</p>
<ol>
<li>Allocate and initialize memory space as needed</li>
<li>Define the datatype of the memory elements</li>
<li>Define the elements to be transferred (a selection, or all the
elements)</li>
<li>Set data transfer properties (including parameters for filters or
file drivers) as needed</li>
<li>Call the H5D API</li>
</ol>
<p>Note that the location of the data in the file, the datatype of the data in
the file, the storage properties, and the filters do not need to be specified
because these are stored as a permanent part of the dataset. A selection of
elements from the dataspace is specified; the selected elements may be the
whole dataspace.</p>
<!-- NEW PAGE -->
<p>The figure below shows a diagram of a write operation which transfers a
data array from memory to a dataset in the file (usually on disk). A read
operation has similar parameters with the data flowing the other direction.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig5.JPG" width="670">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 3. A write operation</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<h4><em>Memory Space</em></h4>
<p>The calling program must allocate sufficient memory to store the data
elements to be transferred. For a write (from memory to the file), the
memory must be initialized with the data to be written to the file. For
a read, the memory must be large enough to store the elements that
will be read. The amount of storage needed can be computed from the
memory datatype (which defines the size of each data element) and the
number of elements in the selection.</p>
<!-- NEW PAGE -->
<h4><em>Memory Datatype</em></h4>
<p>The memory layout of a single data element is specified by the memory
datatype. This specifies the size, alignment, and byte order of the
element as well as the datatype class. Note that the memory datatype
must be the same datatype class as the file, but may have different byte
order and other properties. The HDF5 Library automatically transforms
data elements between the source and destination layouts. See the chapter
“<a href="11_Datatypes.html">HDF5 Datatypes</a>”
for more details.</p>
<p>For a write, the memory datatype defines the layout of the data to be
written; an example is IEEE floating-point numbers in native byte order.
If the file datatype (defined when the dataset is created) is different
but compatible, the HDF5 Library will transform each data element when
it is written. For example, if the file byte order is different than
the native byte order, the HDF5 Library will swap the bytes.</p>
<p>For a read, the memory datatype defines the desired layout of the
data to be read. This must be compatible with the file datatype, but
should generally use native formats, e.g., byte orders. The HDF5 Library
will transform each data element as it is read.</p>
<h4><em>Selection</em></h4>
<p>The data transfer will transfer some or all of the elements of the
dataset depending on the dataspace selection. The selection has two
dataspace objects: one for the source, and one for the destination.
These objects describe which elements of the dataspace to be transferred.
Some (partial I/O) or all of the data may be transferred. Partial I/O
is defined by defining hyperslabs or lists of elements in a dataspace
object.</p>
<p>The dataspace selection for the source defines the indices of the elements
to be read or written. The two selections must define the same number of
points, but the order and layout may be different. The HDF5 Library
automatically selects and distributes the elements according to the
selections. It might, for example, perform a scatter-gather or
sub-set of the data. </p>
<!-- editingComment
<span class="editingComment">[ [ [
See [[Selections]].
] ] ]</span>
-->
<h4><em>Data Transfer Properties</em></h4>
<p>For some data transfers, additional parameters should be set using the
transfer property list. The table below lists the categories of transfer
properties. These properties set parameters for the HDF5 Library and may
be used to pass parameters for optional filters and file drivers. For
example, transfer properties are used to select independent or collective
operation when using MPI-I/O.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 3. Categories of transfer properties</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td><b>Properties</b></td>
<td><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Library parameters</td>
<td>Internal caches, buffers, B-Trees, etc.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Memory management</td>
<td>Variable-length memory management, data overwrite</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>File driver management</td>
<td>Parameters for file drivers</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Filter management</td>
<td>Parameters for filters</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<h4><em>Data Transfer Operation (Read or Write)</em></h4>
<p>The data transfer is done by calling <code>H5Dread</code> or
<code>H5Dwrite</code> with the parameters described above. The HDF5
Library constructs the required pipeline, which will scatter-gather,
transform datatypes, apply the requested filters, and use the correct
file driver.</p>
<p>During the data transfer, the transformations and filters are applied to
each element of the data in the required order until all the
data is transferred.</p>
<!-- editingComment
<span class="editingComment">[ [ [
<p>[[See Data Transfer Below]]
] ] ]</span>
-->
<h4><em>Summary</em></h4>
<p>To perform a data transfer, it is necessary to allocate and initialize
memory, describe the source and destination, set required and optional
transfer properties, and call the H5D API. </p>
<h4><em>Examples</em></h4>
<p>The basic procedure to <b>write</b> to a dataset is the following:</p>
<dir>
Open the dataset.<br>
Set the dataset dataspace for the write (optional if dataspace is
<code>H5S_SELECT_ALL</code>).<br>
Write data.<br>
Close the datatype, dataspace, and property list (as necessary).<br>
Close the dataset.<br>
</dir>
<p>Example 3 below shows example code to write a 4 x 6 array of integers.
In the example, the data is initialized in the memory array dset_data.
The dataset has already been created in the file, so it is opened
with <code>H5Dopen</code>.</p>
<p>The data is written with <code>H5Dwrite</code>. The arguments are the
dataset identifier, the memory datatype (<code>H5T_NATIVE_INT</code>), the
memory and file selections (<code>H5S_ALL</code> in this case:
the whole array), and the default (empty) property list. The last argument
is the data to be transferred.</p>
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
hid_t file_id, dataset_id; /* identifiers */
herr_t status;
int i, j, dset_data[4][6];
/* Initialize the dataset. */
for (i = 0; i < 4; i++)
for (j = 0; j < 6; j++)
dset_data[i][j] = i * 6 + j + 1;
/* Open an existing file. */
file_id = H5Fopen("dset.h5", H5F_ACC_RDWR, H5P_DEFAULT);
/* Open an existing dataset. */
dataset_id = H5Dopen(file_id, "/dset", H5P_DEFAULT);
/* Write the entire dataset, using 'dset_data':
memory type is 'native int'
write the entire dataspace to the entire dataspace,
no transfer properties,
*/
status = H5Dwrite(dataset_id, H5T_NATIVE_INT, H5S_ALL,
H5S_ALL, H5P_DEFAULT, dset_data);
status = H5Dclose(dataset_id);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 3. Write an array of integers</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>Example 4 below shows a similar write except for setting a non-default
value for the transfer buffer.
<!-- editingComment
<span class="editingComment">[ [ [
<<explain what this does>>.
] ] ]</span>
-->
The code is the same as Example 3, but a transfer
property list is created, and the desired buffer size is set. The
<code>H5Dwrite</code> function has the same arguments, but uses the
property list to set the buffer.</p>
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
hid_t file_id, dataset_id;
hid_t xferplist;
herr_t status;
int i, j, dset_data[4][6];
file_id = H5Fopen("dset.h5", H5F_ACC_RDWR, H5P_DEFAULT);
dataset_id = H5Dopen(file_id, "/dset", H5P_DEFAULT);
/*
* Example: set type conversion buffer to 64MB
*/
xferplist = H5Pcreate(H5P_DATASET_XFER);
status = H5Pset_buffer( xferplist, 64 * 1024 *1024, NULL, NULL);
/* Write the entire dataset, using 'dset_data':
memory type is 'native int'
write the entire dataspace to the entire dataspace,
set the buffer size with the property list,
*/
status = H5Dwrite(dataset_id, H5T_NATIVE_INT, H5S_ALL,
H5S_ALL, xferplist, dset_data);
status = H5Dclose(dataset_id);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 4. Write an array using a property list</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<!-- editingComment
<span class="editingComment">[ [ [
<p>Partial writes (i.e., of selected data elements, such as a hyperslab)
are explained below [[partial I/O]]
] ] ]</span>
-->
<p>The basic procedure to <b>read</b> from a dataset is the
following:</p>
<dir>
Define the memory dataspace of the read (optional if dataspace is
<code>H5S_SELECT_ALL</code>).<br>
Open the dataset.<br>
Get the dataset dataspace (if using <code>H5S_SELECT_ALL</code> above).<br>
<dir>Else define dataset dataspace of read.</dir>
Define the memory datatype (optional).<br>
Define the memory buffer.<br>
Open the dataset.<br>
Read data.<br>
Close the datatype, dataspace, and property list (as necessary).<br>
Close the dataset.
</dir>
<p>The example below shows code that reads a 4 x 6 array of integers from
a dataset called “dset1”. First, the dataset is opened.
The <code>H5Dread</code> call has parameters:</p>
<!-- NEW PAGE -->
<ul>
<li>The dataset identifier (from <code>H5Dopen</code>)</li>
<li>The memory datatype (<code>H5T_NATVE_INT</code>)</li>
<li>The memory and file dataspace (<code>H5S_ALL</code>, the whole array)</li>
<li>A default (empty) property list</li>
<li>The memory to be filled</li>
</ul>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
hid_t file_id, dataset_id;
herr_t status;
int i, j, dset_data[4][6];
/* Open an existing file. */
file_id = H5Fopen("dset.h5", H5F_ACC_RDWR, H5P_DEFAULT);
/* Open an existing dataset. */
dataset_id = H5Dopen(file_id, "/dset", H5P_DEFAULT);
/* read the entire dataset, into 'dset_data':
memory type is 'native int'
read the entire dataspace to the entire dataspace,
no transfer properties,
*/
status = H5Dread(dataset_id, H5T_NATIVE_INT, H5S_ALL,
H5S_ALL, H5P_DEFAULT, dset_data);
status = H5Dclose(dataset_id);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 5. Read an array from a dataset</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<h4>5.3.4. Retrieve the Properties of a Dataset</h4>
<p>
The functions listed below allow the user to retrieve
information regarding a dataset including the datatype,
the dataspace, the dataset creation property list,
and the total stored size of the data.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="3" align="left" valign="bottom">
<b>Function Listing 4. Retrieve dataset information
</b></td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td>
<b>Query Function</b></td><td> </td>
<td>
<b>Description</b></td></tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Dget_space</code></td><td> </td>
<td> Retrieve the dataspace of the dataset
as stored in the file.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Dget_type</code></td><td> </td>
<td> Retrieve the datatype of the dataset
as stored in the file.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Dget_create_plist</code></td><td> </td>
<td> Retrieve the
dataset creation properties.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Dget_storage_size</code></td><td> </td>
<td>
Retrieve the total bytes for all the data of the dataset.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Dvlen_get_buf_size</code></td><td> </td>
<td>
Retrieve the total bytes for all the variable-length
data of the dataset.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
</table>
<br />
<!-- NEW PAGE -->
<p>The example below illustrates how to retrieve dataset information.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
hid_t file_id, dataset_id;
hid_t dspace_id, dtype_id, plist_id;
herr_t status;
/* Open an existing file. */
file_id = H5Fopen("dset.h5", H5F_ACC_RDWR, H5P_DEFAULT);
/* Open an existing dataset. */
dataset_id = H5Dopen(file_id, "/dset", H5P_DEFAULT);
dspace_id = H5Dget_space(dataset_id);
dtype_id = H5Dget_type(dataset_id);
plist_id = H5Dget_create_plist(dataset_id);
/* use the objects to discover the properties of the dataset */
status = H5Dclose(dataset_id);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 6. Retrieve dataset</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<SCRIPT language="JavaScript">
<!--
document.writeln ("
<a name="DTransfer">
<div align=right>
<a href="#TOP"><font size="-1">(Top)</font></a>
</div>
</a>
");
-->
</SCRIPT>
<a name="DTransfer">
<h3 class=pagebefore>5.4. Data Transfer</h3>
</a>
<p>The HDF5 Library implements data transfers through a pipeline which
implements data transformations (according to the datatype and selections),
chunking (as requested), and I/O operations using different mechanisms
(file drivers). The pipeline is automatically configured by the HDF5
Library. Metadata is stored in the file so that the correct pipeline
can be constructed to retrieve the data. In addition, optional filters
such as compression may be added to the standard pipeline. </p>
<p>The figure below illustrates data layouts for different layers of an
application using HDF5. The application data is organized as a
multidimensional array of elements. The HDF5 format specification
<!-- editingComment
<span class="editingComment">[ [ [
[[cite it]]
] ] ]</span>
-->
defines the stored layout of the data and metadata. The storage layout
properties define the organization of the abstract data. This data is
written and read to and from some storage medium.</p>
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig9.JPG" width="95%">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 4. Data layouts in an application </b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>The last stage of a write (and first stage of a read) is managed by
an HDF5 file driver module. The virtual file layer of the HDF5 Library
implements a standard interface to alternative I/O methods, including
memory (AKA “core”) files, single serial file I/O, multiple
file I/O, and parallel I/O. The file driver maps a simple abstract HDF5
file to the specific access methods.</p>
<p>The raw data of an HDF5 dataset is conceived to be a multidimensional
array of data elements. This array may be stored in the file according to
several storage strategies:</p>
<ul>
<li>Contiguous</li>
<li>Chunked</li>
<li>Compact</li>
</ul>
<p>The storage strategy does not affect data access methods except that
certain operations may be more or less efficient depending on the storage
strategy and the access patterns.</p>
<p>Overall, the data transfer operations (<code>H5Dread</code> and
<code>H5Dwrite</code>) work identically for any storage method, for any
file driver, and for any filters and transformations. The HDF5 Library
automatically manages the data transfer process. In some cases, transfer
properties should or must be used to pass additional parameters such as
MPI/IO directives when used the parallel file driver.</p>
<h4>5.4.1. The Data Pipeline</h4>
<p>When data is written or read to or from an HDF5 file, the HDF5
Library passes the data through a sequence of processing steps which
are known as the HDF5 data pipeline. This data pipeline performs
operations on the data in memory such as byte swapping, alignment,
scatter-gather, and hyperslab selections. The HDF5 Library automatically
determines which operations are needed and manages the organization
of memory operations such as extracting selected elements from a
data block. The data pipeline modules operate on data buffers: each
module processes a buffer and passes the transformed buffer to the
next stage.</p>
<p>The table below lists the stages of the data pipeline. The figure below
the table shows the order of processing
during a read or write.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 4. Stages of the data pipeline</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="35%"><b>Layers</b></td>
<td width="65%"><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>I/O initiation</td>
<td>Initiation of HDF5 I/O activities (<code>H5Dwrite</code> and
<code>H5Dread</code>) in a user’s application program. </td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Memory hyperslab operation</td>
<td>Data is scattered to (for read), or gathered from (for write)
the application’s memory buffer (bypassed if no datatype
conversion is needed).</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Datatype conversion</td>
<td>Datatype is converted if it is different between memory and
storage (bypassed if no datatype conversion is needed).</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>File hyperslab operation</td>
<td>Data is gathered from (for read), or scattered to (for write) to
file space in memory (bypassed if no datatype conversion is needed).</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Filter pipeline</td>
<td>Data is processed by filters when it passes. Data can be
modified and restored here (bypassed if no datatype conversion
is needed, no filter is enabled, or dataset is not chunked).</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Virtual File Layer</td>
<td>Facilitate easy plug-in file drivers such as MPIO or
POSIX I/O.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Actual I/O</td>
<td>Actual file driver used by the library such as MPIO or STDIO.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<br />
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig10.JPG"><br>
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 5. The processing order in the data pipeline</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>The HDF5 Library automatically applies the stages as needed. </p>
<p>When the memory dataspace selection is other than the whole dataspace,
the memory hyperslab stage scatters/gathers the data elements between
the application memory (described by the selection) and a contiguous
memory buffer for the pipeline. On a write, this is a gather operation;
on a read, this is a scatter operation.</p>
<p>When the memory datatype is different from the file datatype, the
datatype conversion stage transforms each data element. For example, if
data is written from 32-bit big-endian memory, and the file datatype is
32-bit little-endian, the datatype conversion stage will swap the bytes
of every elements. Similarly, when data is read from the file to
native memory, byte swapping will be applied automatically when needed.</p>
<p>The file hyperslab stage is similar to the memory hyperslab stage,
but is managing the arrangement of the elements according to the
dataspace selection. When data is read, data elements are gathered from
the data blocks from the file to fill the contiguous buffers which are then
processed by the pipeline. When data is read, the elements from a buffer
are scattered to the data blocks of the file.</p>
<h4>5.4.2. Data Pipeline Filters</h4>
<p>In addition to the standard pipeline, optional stages, called filters,
can be inserted in the pipeline.
<!-- editingComment
<span class="editingComment">[ [ [
see [[chunked]])
] ] ]</span>
-->
The standard distribution includes optional filters to
implement compression and error checking. User applications may
add custom filters as well.</p>
<p>The HDF5 Library distribution includes or employs
several optional filters. These are listed in the table below.
The filters are applied in the pipeline between the virtual file layer and
the file hyperslab operation. See the figure above. The application can
use any number of filters in any order.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 5. Data pipeline filters</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="35%"><b>Filter</b></td>
<td width="65%"><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>gzip compression</td>
<td>Data compression using <code>zlib</code>.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Szip compression</td>
<td>Data compression using the Szip library. See The HDF Group
website for more information regarding the
<a href="http://www.hdfgroup.org/doc_resource/SZIP/"
target="Ext1">Szip</a> filter.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>N-bit compression</td>
<td>Data compression using an algorithm specialized for
n-bit datatypes.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Scale-offset compression</td>
<td>Data compression using using a “scale and
offset” algorithm.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Shuffling</td>
<td>To improve compression performance, data is regrouped by
its byte position in the data unit. In other words, the
1<sup><font size="-1">st</font></sup>,
2<sup><font size="-1">nd</font></sup>,
3<sup><font size="-1">rd</font></sup>, and
4<sup><font size="-1">th</font></sup> bytes of integers are
stored together respectively.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Fletcher32</td>
<td>Fletcher32 checksum for error-detection.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>Filters may be used only for chunked data and are applied to chunks of
data between the file hyperslab stage and the virtual file layer. At this
stage in the pipeline, the data is organized as fixed-size blocks of
elements, and the filter stage processes each chunk separately.</p>
<p>Filters are selected by dataset creation properties, and some behavior may
be controlled by data transfer properties. The library determines what
filters must be applied and applies them in the order in which they were
set by the application. That is, if an application calls
<code>H5Pset_shuffle</code> and then <code>H5Pset_deflate</code> when
creating
a dataset’s creation property list, the library will apply the
shuffle filter first and then the deflate filter.</p>
<p>Information regarding the n-bit and scale-offset filters
can be found in the “<a href="#N-Bit">Using the N-bit Filter</a>”
and “<a href="#ScaleOffset">Using the Scale-offset Filter</a>”
sections, respectively.</p>
<h4>5.4.3. File Drivers</h4>
<p>I/O is performed by the HDF5 virtual file layer. The file driver
interface writes and reads blocks of data; each driver module implements
the interface using different I/O mechanisms. The table below lists the
file drivers currently supported. Note that the I/O mechanisms are
separated from the pipeline processing: the pipeline and filter
operations are identical no matter what data access mechanism is used.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 6. I/O file drivers</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td><b>File Driver</b></td>
<td><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>H5FD_CORE</code></td>
<td>Store in memory (optional backing store to disk file).</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>H5FD_FAMILY</code></td>
<td>Store in a set of files.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>H5FD_LOG</code></td>
<td>Store in logging file.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>H5FD_MPIO</code></td>
<td>Store using MPI/IO.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>H5FD_MULTI</code></td>
<td>Store in multiple files. There are several options to control
layout.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>H5FD_SEC2</code></td>
<td>Serial I/O to file using Unix “section 2” functions.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>H5FD_STDIO</code></td>
<td>Serial I/O to file using Unix “stdio” functions.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>Each file driver writes/reads contiguous blocks of bytes from a logically
contiguous address space. The file driver is responsible for managing the
details of the different physical storage methods.</p>
<p>In serial environments, everything above the virtual file layer tends
to work identically no matter what storage method is used. </p>
<p>Some options may have substantially different performance depending
on the file driver that is used. In particular, multi-file and parallel
I/O may perform considerably differently from serial drivers depending
on chunking and other settings.</p>
<h4>5.4.4. Data Transfer Properties to Manage the Pipeline</h4>
<p>Data transfer properties set optional parameters that control parts of the
data pipeline. The function listing below shows transfer properties
that control the behavior of the library.</p>
<!-- editingComment
<span class="editingComment">[ [ [
<<Developers: explain what these do!>></p>
] ] ]</span>
-->
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="3" align="left" valign="bottom">
<b>Function Listing 5. Data transfer property list functions
</b></td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td>
<b>Property</b></td><td> </td>
<td>
<b>Description</b></td></tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Pset_buffer</code></td><td> </td>
<td>
Maximum size for the type conversion buffer and the background
buffer. May also
supply pointers to application-allocated buffers.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Pset_hyper_cache</code></td><td> </td>
<td>
Whether to cache hyperslab blocks during I/O.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Pset_btree_ratios</code></td><td> </td>
<td>
Set the B-tree split ratios for a dataset transfer property list.
The split ratios determine what percent of children go in the
first node when a node splits.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>Some filters and file drivers require or use additional parameters
from the application program. These can be passed in the data transfer
property list. The table below shows file driver property list functions.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="3" align="left" valign="bottom">
<b>Function Listing 6. File driver property list functions
</b></td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td>
<b>Property</b></td><td> </td>
<td>
<b>Description</b></td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Pset_dxpl_mpio</code></td><td> </td>
<td>
Control the MPI I/O transfer mode (independent or collective)
during data I/O operations.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Pset_small_data_block_size</code></td><td> </td>
<td>
Reserves blocks of size bytes for the contiguous storage of the raw
data portion of small datasets. The HDF5 Library then writes the raw data
from small datasets to this reserved space which reduces unnecessary
discontinuities within blocks of metadata and
improves I/O performance.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td> <code>H5Pset_edc_check</code></td><td> </td>
<td>
Disable/enable EDC checking for read. When selected, EDC
is always written.</td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>The transfer properties are set in a property list which is passed as a
parameter of the <code>H5Dread</code> or <code>H5Dwrite</code> call. The
transfer properties are passed to each pipeline stage. Each stage may use
or ignore any property in the list. In short, there is one property list
that contains all the properties.</p>
<h4>5.4.5. Storage Strategies</h4>
<p>The raw data is conceptually a multi-dimensional array of elements that
is stored as a contiguous array of bytes. The data may be physically stored
in the file in several ways. The table below lists the storage strategies
for a dataset.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 7. Dataset storage strategies</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td><b>Storage Strategy</b></td>
<td><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Contiguous</td>
<td>The dataset is stored as one continuous array of bytes.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Chunked</td>
<td>The dataset is stored as fixed-size chunks.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Compact</td>
<td>A small dataset is stored in the metadata header.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>The different storage strategies do not affect the data transfer
operations of the dataset: reads and writes work the same for any
storage strategy.</p>
<!-- editingComment
<span class="editingComment">[ [ [
<p><<Relationship between storage strategies, and pipeline, filters,
and file drivers.??>>
] ] ]</span>
-->
<p>These strategies are described in the following sections.</p>
<h4><em>Contiguous</em></h4>
<p>A contiguous dataset is stored in the file as a header and a single
continuous array of bytes. See the figure below. In the case of a
multi-dimensional array, the data is serialized in row major order.
By default, data is stored contiguously.</p>
<table width="400" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig12.JPG">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 6. Contiguous data storage</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>Contiguous storage is the simplest model. It has several limitations.
First, the dataset must be a fixed-size: it is not possible to extend
the limit of the dataset or to have unlimited dimensions. In other
words, if the number of dimensions of the array might change over
time, then chunking storage must be used instead of contiguous.
Second, because data is passed through the pipeline as fixed-size
blocks, compression and other filters cannot be used with contiguous
data.</p>
<!-- NEW PAGE -->
<h4><em>Chunked</em></h4>
<p>The data of a dataset may be stored as fixed-size chunks. See the
figure below.
A chunk is a hyper-rectangle of any shape.
When a dataset is chunked, each chunk is read or written as a single I/O
operation, and individually passed from stage to stage of the data pipeline.
</p>
<table width="400" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig13.JPG">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 7. Chunked data storage</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>Chunks may be any size and shape that fits in the dataspace of the dataset.
For example, a three dimensional dataspace can be chunked as 3-D cubes,
2-D planes, or 1-D lines. The chunks may extend beyond the size of the
dataspace. For example, a 3 x 3 dataset might by chunked in 2 x 2 chunks.
Sufficient chunks will be allocated to store the array, and any extra space
will not be accessible. So, to store the 3 x 3 array, four 2 x 2 chunks would
be allocated with 5 unused elements stored.</p>
<p>Chunked datasets can be unlimited in any direction
and can be compressed or filtered.</p>
<p>Since the data is read or written by chunks, chunking can have a dramatic
effect on performance by optimizing what is read and written. Note, too,
that for specific access patterns such as parallel I/O, decomposition into
chunks can have a large impact on performance.</p>
<p>Two restrictions have been placed on chunk shape and size:</p>
<ul>
<li>The rank of a chunk must be less than or equal to
the rank of the dataset</li>
<li>Chunk size cannot exceed the size of a fixed-size dataset;
for example, a dataset consisting of a 5 x 4 fixed-size array
cannot be defined with 10 x 10 chunks</li>
</ul>
<!-- NEW PAGE -->
<h4><em>Compact</em></h4>
<p>For contiguous and chunked storage, the dataset header information and data
are stored in two (or more) blocks. Therefore, at least two I/O operations
are required to access the data: one to access the header, and one (or more)
to access data. For a small dataset, this is considerable overhead.</p>
<p>A small dataset may be stored in a continuous array of bytes in the
header block using the compact storage option. This dataset can be read
entirely in one operation which retrieves the header and data.
The dataset must fit in the header. This may vary depending on the
metadata that is stored. In general, a compact dataset should be
approximately 30 KB or less total size.
See the figure below.</p>
<table width="400" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig14.JPG">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 8. Compact data storage</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<h4>5.4.6. Partial I/O Sub-setting and Hyperslabs</h4>
<p>Data transfers can write or read some of the data elements of the dataset.
This is controlled by specifying two selections: one for the source and
one for the destination. Selections are specified by creating a dataspace
with selections. </p>
<!-- editingComment
<span class="editingComment">[ [ [
(see [[dataspace chapter]])
] ] ]</span>
-->
<p>Selections may be a union of hyperslabs or a list of points.
A hyperslab is a contiguous hyper-rectangle from the dataspace.
Selected fields of a compound datatype may be read or written.
In this case, the selection is controlled by the memory and file
datatypes.</p>
<p>Summary of procedure:</p>
<ol>
<li>Open the dataset</li>
<li>Define the memory datatype</li>
<li>Define the memory dataspace selection and file dataspace
selection</li>
<li>Transfer data (<code>H5Dread</code> or <code>H5Dwrite</code>)</li>
</ol>
<p>For a detailed explanation of selections, see the chapter
“<a href="12_Dataspaces.html">HDF5 Dataspaces and Partial I/O</a>.
”</p>
<SCRIPT language="JavaScript">
<!--
document.writeln ("
<a name="Allocation">
<div align=right>
<a href="#TOP"><font size="-1">(Top)</font></a>
</div>
</a>
");
-->
</SCRIPT>
<br />
<!-- NEW PAGE -->
<a name="Allocation">
<h3 class=pagebefore>5.5. Allocation of Space in the File</h3>
</a>
<p>When a dataset is created, space is allocated in the file for its
header and initial data. The amount of space allocated when the dataset
is created depends on the storage properties. When the dataset is
modified (data is written, attributes added, or other changes),
additional storage may be allocated if necessary.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 8. Initial dataset size</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="15%"><b>Object</b></td>
<td width="85%"><b>Size</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Header</td>
<td>Variable, but typically around 256 bytes at the creation of
a simple dataset with a simple datatype.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Data</td>
<td>Size of the data array (number of elements x size of element).
Space allocated in the file depends on the storage strategy
and the allocation strategy.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<h4><em>Header</em></h4>
<p>A dataset header consists of one or more header messages containing
persistent metadata describing various aspects of the dataset.
These records are defined in the <a href="../H5.format.html"><i>HDF5 File
Format Specification</i></a>. The amount of storage required for the
metadata depends on the metadata to be stored. The table below
summarizes the metadata.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 9. Metadata storage sizes</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="25%"><b>Header Information</b></td>
<td width="70%"><b>Approximate Storage Size</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Datatype (required)</td>
<td>Bytes or more. Depends on type.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Dataspace (required)</td>
<td>Bytes or more. Depends on number of dimensions and hsize_t.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Layout (required)</td>
<td>Points to the stored data. Bytes or more. Depends on hsize_t and
number of dimensions.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Filters</td>
<td>Depends on the number of filters. The size of the filter message
depends on the name and data that will be passed.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>The header blocks also store the name and values of attributes, so
the total storage depends on the number and size of the attributes.</p>
<p>In addition, the dataset must have at least one link, including a name,
which is stored in the file and in the group it is linked from.</p>
<p>The different storage strategies determine when and how much space is
allocated for the data array. See the discussion of fill values below
<!-- editingComment
<span class="editingComment">[ [ [
Link
] ] ]</span>
-->
for a detailed explanation of the storage allocation.</p>
<!-- NEW PAGE -->
<h4><em>Contiguous Storage</em></h4>
<p>For a continuous storage option, the data is stored in a single,
contiguous block in the file. The data is nominally a fixed-size,
(number of elements x size of element). The figure below shows an example
of a two dimensional array stored as a contiguous dataset.</p>
<p>Depending on the fill value properties, the space may be allocated
when the dataset is created or when first written (default), and filled
with fill values if specified. For parallel I/O, by default the space
is allocated when the dataset is created.</p>
<table width="400" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig15.JPG">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 9. A two dimensional array stored as a contiguous dataset</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<h4><em>Chunked</em></h4>
<p>For chunked storage, the data is stored in one or more chunks. Each chunk
is a continuous block in the file, but chunks are not necessarily stored
contiguously. Each chunk has the same size. The data array has the same
nominal size as a contiguous array (number of elements x size of element),
but the storage is allocated in chunks, so the total size in the file can
be larger that the nominal size of the array. See the figure below.</p>
<p>If a fill value is defined, each chunk will be filled with the fill value.
Chunks must be allocated when data is written, but they may be allocated when
the file is created, as the file expands, or when data is written. </p>
<p>For serial I/O, by default chunks are allocated incrementally, as data is
written to the chunk. For a sparse dataset, chunks are allocated only for the
parts of the dataset that are written. In this case, if the dataset is
extended, no storage is allocated.</p>
<p>For parallel I/O, by default chunks are allocated when the dataset is
created or extended with fill values written to the chunk.</p>
<p>In either case, the default can be changed using fill value properties.
For example, using serial I/O, the properties can select to allocate
chunks when the dataset is created.</p>
<table width="400" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig16.JPG">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 10. A two dimensional array stored in chunks</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<h4><em>Changing Dataset Dimensions</em></h4>
<p><code>H5Dset_extent</code> is used to change the current dimensions
of the dataset within the limits of the dataspace. Each dimension can
be extended up to its maximum or unlimited. Extending the dataspace may
or may not allocate space in the file and may or may not write fill
values, if they are defined. See the example code below.</p>
<p>The dimensions of the dataset can also reduced. If the sizes specified
are smaller than the datasets current dimension sizes,
<code>H5Dset_extent</code> will reduce the datasets dimension sizes to
the specified values. It is the users responsibility to ensure that
valuable data is not lost; <code>H5Dset_extent</code> does not check.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
hid_t file_id, dataset_id;
Herr_t status;
size_t newdims[2];
/* Open an existing file. */
file_id = H5Fopen("dset.h5", H5F_ACC_RDWR, H5P_DEFAULT);
/* Open an existing dataset. */
dataset_id = H5Dopen(file_id, "/dset", H5P_DEFAULT);
/* Example: dataset is 2 x 3, each dimension is UNLIMITED */
/* extend to 2 x 7 */
newdims[0] = 2;
newdims[1] = 7;
status = H5Dset_extent(dataset_id, newdims);
/* dataset is now 2 x 7 */
status = H5Dclose(dataset_id);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 7. Using <code>H5Dset_extent</code>
to increase the size of a dataset</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<h4>5.5.1. Storage Allocation in the File: Early, Incremental, Late</h4>
<p>The HDF5 Library implements several strategies for when storage is
allocated if and when it is filled with fill values for elements not
yet written by the user. Different strategies are recommended for
different storage layouts and file drivers. In particular, a parallel
program needs storage allocated during a collective call (for example,
create or extend) while serial programs may benefit from delaying the
allocation until the data is written.</p>
<p>Two file creation properties control when to allocate space, when to
write the fill value, and the actual fill value to write. </p>
<h4><em>When to Allocate Space</em></h4>
<p>The table below shows the options for when data is allocated in the
file. “Early” allocation is done during the dataset create
call. Certain file drivers (especially MPI-I/O and MPI-POSIX) require
space to be allocated when a dataset is created, so all processors will
have the correct view of the data.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 10. File storage allocation options</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="20%"><b>Strategy</b></td>
<td width="80%"><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Early</td>
<td>Allocate storage for the dataset immediately when the dataset
is created. </td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Late</td>
<td>Defer allocating space for storing the dataset until the
dataset is written.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Incremental</td>
<td>Defer allocating space for storing each chunk until
the chunk is written.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Default</td>
<td>Use the strategy (Early, Late, or Incremental) for the storage method
and access method. This is the recommended strategy.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>“Late” allocation is done at the time of the first write to
dataset. Space for the whole dataset is allocated at the first write.</p>
<p>“Incremental” allocation (chunks only) is done at the time
of the first write to the chunk. Chunks that have never been written are
not allocated in the file. In a sparsely populated dataset, this option
allocates chunks only where data is actually written.</p>
<p>The “Default” property selects the option recommended as
appropriate for the storage method and access method. The defaults are
shown in the table below. Note that “Early” allocation is
recommended for all Parallel I/O, while other options are recommended
as the default for serial I/O cases.</p>
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="3" align="left" valign="bottom">
<b>Table 11. Default storage options</b></td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="33%">
<b> </b></td>
<td width="34%">
<b>Serial I/O</b></td>
<td width="33%">
<b>Parallel I/O</b></td>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Contiguous Storage</td>
<td>Late</td>
<td>Early</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Chunked Storage</td>
<td>Incremental</td>
<td>Early</td>
</tr>
<tr><td colspan="3"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Compact Storage</td>
<td>Early</td>
<td>Early</td>
</tr>
<tr><td colspan="3"><hr color="green" size="3" /></td></tr>
</table>
<br />
<h4><em>When to Write the Fill Value</em></h4>
<p>The second property is when to write the fill value. The possible values
are “Never” and “Allocation”.
The table below shows these options.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 12. When to write fill values</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="15%"><b>When</b></td>
<td width="85%"><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Never</td>
<td>Fill value will never be written.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Allocation</td>
<td>Fill value is written when space is allocated. (Default for
chunked and contiguous data storage.)</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<h4><em>Fill Values</em></h4>
<p>The third property is the fill value to write. The table below shows the
values. By default, the data is filled with zeroes. The application may
choose no fill value (Undefined). In this case, uninitialized data may have
random values. The application may define a fill value of an
appropriate type. See the chapter “<a href="11_Datatypes.html">HDF5
Datatypes</a>” for more information regarding fill values.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 13. Fill values</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td><b>What to Write</b></td>
<td><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Default</td>
<td>By default, the library fills allocated space with zeroes.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Undefined</td>
<td>Allocated space is filled with random values.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>User-defined</td>
<td>The application specifies the fill value.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>Together these three properties control the library’s behavior.
The table below summarizes the possibilities during the dataset
create-write-close cycle.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="4" align="left" valign="bottom">
<b>Table 14. Storage allocation and fill summary</b></td>
</tr>
<tr><td colspan="4"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="17%">
<b>When to<br />allocate<br />space</b></td>
<td width="17%">
<b>When to<br />write fill<br />value</b></td>
<td width="17%">
<b>What fill<br />value to<br />write</b></td>
<td width="49%">
<b>Library create-write-close behavior</b></td>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Early</td>
<td>Never</td>
<td>-</td>
<td>Library allocates space when dataset is created, but never
writes a fill value to dataset. A read of unwritten data returns
undefined values.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Late</td>
<td>Never</td>
<td>-</td>
<td>Library allocates space when dataset is written to, but never
writes a fill value to the dataset. A read of unwritten data
returns undefined values.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Incremental</td>
<td>Never</td>
<td>-</td>
<td>Library allocates space when a dataset or chunk (whichever
is the smallest unit of space) is written to, but it never writes
a fill value to a dataset or a chunk. A read of unwritten data
returns undefined values.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>-</td>
<td>Allocation</td>
<td>Undefined</td>
<td><b>Error</b> on creating the dataset. The dataset is not
created.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Early</td>
<td>Allocation</td>
<td>Default or User-defined</td>
<td>Allocate space for the dataset when the dataset is created.
Write the fill value (default or user-defined) to the entire
dataset when the dataset is created.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Late</td>
<td>Allocation</td>
<td>Default or User-defined</td>
<td>Allocate space for the dataset when the application first
writes data values to the dataset. Write the fill value to the
entire dataset before writing application data values.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Incremental</td>
<td>Allocation</td>
<td>Default or User-defined</td>
<td>Allocate space for the dataset when the application first
writes data values to the dataset or chunk (whichever is the
smallest unit of space). Write the fill value to the entire dataset
or chunk before writing application data values. </td>
</tr>
<tr><td colspan="4"><hr color="green" size="3" /></td></tr>
</table>
<br />
<!-- NEW PAGE -->
<p>During the <code>H5Dread</code> function call, the library behavior
depends on whether space has been allocated, whether the fill value has
been written to storage, how the fill value is defined, and when to
write the fill value. The table below summarizes the different behaviors.</p>
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="4" align="left" valign="bottom">
<b>Table 15. <code>H5Dread</code> summary</b></td>
</tr>
<tr><td colspan="4"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td width="17%">
<b>Is space<br />allocated<br />in the file?</b></td>
<td width="17%">
<b>What is the<br />fill value?</b></td>
<td width="17%">
<b>When to<br />write the<br />fill value?</b></td>
<td width="49%">
<b>Library read behavior</b></td>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>No</td>
<td>Undefined</td>
<td><<any>></td>
<td><b>Error</b>. Cannot create this dataset.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>No</td>
<td>Default or User-defined</td>
<td><<any>></td>
<td>Fill the memory buffer with the fill value.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Yes</td>
<td>Undefined</td>
<td><<any>></td>
<td>Return data from storage (dataset). Trash is possible if
the application has not written data to the portion of the
dataset being read.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Yes</td>
<td>Default or User-defined</td>
<td>Never</td>
<td>Return data from storage (dataset). Trash is possible if the
application has not written data to the portion of the dataset being
read.</td>
</tr>
<tr><td colspan="4"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td>Yes</td>
<td>Default or User-defined</td>
<td>Allocation</td>
<td>Return data from storage (dataset).</td>
</tr>
<tr><td colspan="4"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>There are two cases to consider depending on whether the space in the
file has been allocated before the read or not. When space has not yet
been allocated and if a fill value is defined, the memory buffer will
be filled with the fill values and returned. In other words, no data
has been read from the disk. If space has been allocated, the values
are returned from the stored data. The unwritten elements will be
filled according to the fill value.</p>
<h4>5.5.2. Deleting a Dataset from a File and Reclaiming Space</h4>
<p>HDF5 does not at this time provide an easy mechanism to remove a dataset
from a file or to reclaim the storage space occupied by a deleted object. </p>
<p>Removing a dataset and reclaiming the space it used can be done with
the <code>H5Ldelete</code> function and the
<a href="../RM/Tools.html#Tools-Repack">h5repack</a> utility program.
With the <code>H5Ldelete</code> function, links to a dataset can be
removed from the file structure. After all the links have been removed,
the dataset becomes inaccessible to any application and is effectively
removed from the file. The way to recover the space occupied by an
unlinked dataset is to write all of the objects of the file into a
new file. Any unlinked object is inaccessible to the application and
will not be included in the new file. Writing objects to a new file
can be done with a custom program or with the h5repack utility program.</p>
<!-- 8.11.10, MEE: in the paragraph below, the link should be changed.
Links are now done separately from groups, but there is no HDF5 Links
chapter yet. -->
<p>See the chapter “<a href="09_Groups.html">HDF5 Groups</a>” for
further discussion of HDF5 file structures and the use of links. </p>
<h4>5.5.3. Releasing Memory Resources</h4>
<p>The system resources required for HDF5 objects such as datasets,
datatypes, and dataspaces should be released once access to the object is
no longer needed. This is accomplished via the appropriate close function.
This is not unique to datasets but a general requirement when working
with the HDF5 Library; failure to close objects will result in resource leaks. </p>
<p>In the case where a dataset is created or data has been transferred,
there are several objects that must be closed. These objects
<!-- editingComment
<span class="editingComment">
[ [ [
(T? above)
originally appeared here. On the full editorial pass,
see if there is any apparent reason for the question.
] ] ]
</span>
-->
include datasets,
datatypes, dataspaces, and property lists. </p>
<p>The application program must free any memory variables and buffers it
allocates. When accessing data from the file, the amount of memory required
can be determined by calculating the size of the memory datatype and the
number of elements in the memory selection.</p>
<p>Variable-length data are organized in two or more areas of memory.
See “<a href="11_Datatypes.html">HDF5 Datatypes</a>” for
more information. When writing data, the application creates an array
of <code>vl_info_t</code> which contains pointers to the elements.
The elements might be, for example, strings. In the file, the
variable-length data is stored in two parts: a heap with the
variable-length values of the data elements and an array of
<code>vlinfo_t</code> elements. When the data is read, the amount of
memory required for the heap can be determined with the
<code>H5Dget_vlen_buf_size</code> call.</p>
<p>The data transfer property may be used to set a custom memory manager
for allocating variable-length data for a <code>H5Dread</code>. This is
set with the <code>H5Pset_vlen_mem_manager</code> call.</p>
<p>To free the memory for variable-length data, it is necessary to visit
each element, free the variable-length data, and reset the element. The
application must free the memory it has allocated. For memory allocated
by the HDF5 Library during a read, the <code>H5Dvlen_reclaim</code>
function can be used to perform this operation.</p>
<h4>5.5.4. External Storage Properties</h4>
<p>The external storage format allows data to be stored across a set of
non-HDF5 files. A set of segments (offsets and sizes) in one or more files
is defined as an external file list, or EFL, and the contiguous logical
addresses of the data storage are mapped onto these segments. Currently,
only the <code>H5D_CONTIGUOUS</code> storage format allows external storage.
External storage is enabled by a dataset creation property. The table
below shows the API.</p>
<!-- NEW PAGE -->
<table width="600" cellspacing="0" align="center" cellpadding="0">
<tr valign="bottom">
<td colspan="2" align="left" valign="bottom">
<b>Table 16. External storage API</b></td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
<tr valign="top">
<td><b>Function</b></td>
<td><b>Description</b></td>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td width="50%">
<code>herr_t H5Pset_external (hid_t plist, const char *name,
off_t offset, hsize_t size)</code></td>
<td width="50%">This function adds a new segment to the end of
the external file list of the specified dataset creation property
list. The segment begins a byte offset of file name and continues
for size bytes. The space represented by this segment is adjacent
to the space already represented by the external file list. The
last segment in a file list may have the size
<code>H5F_UNLIMITED</code>, in which case the external file may
be of unlimited size and no more files can be added to the
external files list.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>int H5Pget_external_count (hid_t plist)</code></td>
<td>Calling this function returns the number of segments in an
external file list. If the dataset creation property list has no
external data, then zero is returned.</td>
</tr>
<tr><td colspan="2"><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td><code>herr_t H5Pget_external (hid_t plist, int idx, size_t
name_size,<br />char *name, off_t *offset,<br />hsize_t *size)</code></td>
<td>This is the counterpart for the <code>H5Pset_external()</code>
function. Given a dataset creation property list and a zero-based
index into that list, the file name, byte offset, and segment
size are returned through non-null arguments. At most name_size
characters are copied into the name argument which is not null
terminated if the file name is longer than the supplied name
buffer (this is similar to <code>strncpy()</code>). </td>
</tr>
<tr><td colspan="2"><hr color="green" size="3" /></td></tr>
</table>
<br />
<p>The figure below shows an example of how a contiguous, one-dimensional
dataset is partitioned into three parts and each of those parts is stored
in a segment of an external file. The top rectangle represents the logical
address space of the dataset while the bottom rectangle represents an
external file.</p>
<table width="500" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig19.JPG">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 11. External file storage</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>The example below shows code that defines the external storage for the
example. Note that the segments are defined in order of the logical
addresses they represent, not their order within the external file. It
would also have been possible to put the segments in separate files.
Care should be taken when setting up segments in a single file since
the library does not automatically check for segments that overlap.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
Plist = H5Pcreate (H5P_DATASET_CREATE);
H5Pset_external (plist, "velocity.data", 3000, 1000);
H5Pset_external (plist, "velocity.data", 0, 2500);
H5Pset_external (plist, "velocity.data", 4500, 1500);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 8. External storage</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>The figure below shows an example of how a contiguous, two-dimensional
dataset is partitioned into three parts and each of those parts is
stored in a separate external file. The top rectangle represents the
logical address space of the dataset while the bottom rectangles
represent external files.</p>
<table width="500" cellspacing="0" align="center">
<tr valign="top">
<td align="center">
<hr color="green" size="3"/>
<img src="Images/Dsets_fig20.jpg">
</td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left" >
<b>Figure 12. Partitioning a 2-D dataset for external storage</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>The example below shows code for the partitioning described above.
In this example, the library maps the multi-dimensional array onto a linear
address space as defined by the HDF5 format specification, and then maps that
address space into the segments defined in the external file list. </p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
Plist = H5Pcreate (H5P_DATASET_CREATE);
H5Pset_external (plist, "scan1.data", 0, 24);
H5Pset_external (plist, "scan2.data", 0, 24);
H5Pset_external (plist, "scan3.data", 0, 16);</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 9. Partitioning a 2-D dataset for external storage</b>
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>The segments of an external file can exist beyond the end of the
(external) file. The library reads that part of a segment as zeros. When
writing to a segment that exists beyond the end of a file, the external
file is automatically extended. Using this feature, one can create a
segment (or set of segments) which is larger than the current size of
the dataset. This allows the dataset to be extended at a future time
(provided the dataspace also allows the extension).</p>
<p>All referenced external data files must exist before performing raw
data I/O on the dataset. This is normally not a problem since those files
are being managed directly by the application or indirectly through some
other library. However, if the file is transferred from its original
context, care must be taken to assure that all the external files are
accessible in the new location.</p>
<br />
<!-- NEW PAGE -->
<a name="UseFilters">
<h3 class=pagebefore>5.6. Using HDF5 Filters</h3>
</a>
<p>This section describes in detail how to use the n-bit and
scale-offset filters. </p>
<a name="N-Bit">
<h3>5.6.1. Using the N-bit Filter</h3>
</a>
<p>N-bit data has <i>n</i> significant bits,
where <i>n</i> may not correspond to a precise number of bytes.
On the other hand, computing systems and applications universally,
or nearly so, run most efficiently when manipulating data as
whole bytes or multiple bytes.</p>
<p>Consider the case of 12-bit integer data.
In memory, that data will be handled in at least 2 bytes, or 16 bits,
and on some platforms in 4 or even 8 bytes.
The size of such a dataset can be significantly reduced when written
to disk if the unused bits are stripped out.</p>
<p>The <i>n-bit filter</i> is provided for this purpose,
<i>packing</i> n-bit data on output by stripping off all unused bits
and <i>unpacking</i> on input, restoring the extra bits required
by the computational processor.</p>
<h4><em>N-bit Datatype</em></h4>
<p>An <i>n-bit datatype</i> is a datatype of <i>n</i> significant bits.
Unless it is packed, an <i>n</i>-bit datatype is presented as an
<i>n</i>-bit bitfield within a larger-sized value.
For example, a 12-bit datatype might be presented as a 12-bit field
in a 16-bit, or 2-byte, value.</p>
<p>Currently, the datatype classes of n-bit datatype or n-bit field of a
compound datatype or an array datatype are limited to integer or
floating-point.</p>
<p>The HDF5 user can create an n-bit datatype through a series of
of function calls.
For example, the following calls create a 16-bit datatype
that is stored in a 32-bit value with a 4-bit offset:</p>
<dir><pre>
hid_t nbit_datatype = H5Tcopy(H5T_STD_I32LE);
H5Tset_precision(nbit_datatype, 16);
H5Tset_offset(nbit_datatype, 4);
</pre></dir>
<p>In memory, one value of the above example n-bit datatype would be stored on
a little-endian machine as follows:</p>
<dl>
<dt>
<table border="1" width="80%" align="center">
<tr>
<td width="25%" align="center">byte 3</td>
<td width="25%" align="center">byte 2</td>
<td width="25%" align="center">byte 1</td>
<td width="25%" align="center">byte 0</td>
</tr>
<tr>
<td width="25%" align="center"><code>????????</code></td>
<td width="25%" align="center"><code>????SPPP</code></td>
<td width="25%" align="center"><code>PPPPPPPP</code></td>
<td width="25%" align="center"><code>PPPP????</code></td>
</tr>
</table>
<table width="80%" border="0" align="center">
<tr>
<td colspan="4"><font size="-1">
Key:
<code>S</code> - sign bit,
<code>P</code> - significant bit,
<code>?</code> - padding bit
<br>
Sign bit is included in signed integer datatype precision.
</font></td>
</tr>
</table>
<br>
</dt>
</dl>
<br />
<!-- NEW PAGE -->
<h4><em>N-bit Filter</em></h4>
<p>When data of an n-bit datatype is stored on disk using the
n-bit filter, the filter <i>packs</i> the data by stripping off the
padding bits; only the significant bits are retained and stored.
The values on disk will appear as follows:</p>
<dl>
<dt>
<table border="1" width="80%" align="center">
<tr>
<td width="45%" align="center">1st value</td>
<td width="45%" align="center">2nd value</td>
<td width="10%" align="center"> </td>
</tr>
<tr>
<td width="45%" align="center"><code>SPPPPPPP PPPPPPPP</code></td>
<td width="45%" align="center"><code>SPPPPPPP PPPPPPPP</code></td>
<td width="10%" align="center">...</td>
</tr>
</table>
<table width="80%" border="0" align="center">
<tr>
<td colspan="4"><font size="-1">
Key:
<code>S</code> - sign bit,
<code>P</code> - significant bit,
<code>?</code> - padding bit
<br>
Sign bit is included in signed integer datatype precision.
</font></td>
</tr>
</table>
</dt>
</dl>
<br />
<p>The n-bit filter can be used effectively for compressing data of an n-bit
datatype, including arrays and the n-bit fields of compound datatypes.
The filter supports complex situations where a compound datatype
contains member(s) of a compound datatype or an array datatype has
a compound datatype as the base type.</p>
<p>At present, the n-bit filter supports all datatypes.
For datatypes of class time, string,
opaque, reference, <small>ENUM</small>, and
variable-length, the n-bit filter acts as a no-op which is short for no
operation.
For convenience, the rest of this section refers to such datatypes
as <i>no-op datatypes</i>.</p>
<p>As is the case with all HDF5 filters, an application using
the n-bit filter must store data with chunked storage.</p>
<h4><em>How Does the N-bit Filter Work?</em></h4>
<p>The n-bit filter always compresses and decompresses according to
dataset properties supplied by the HDF5 Library in the
datatype, dataspace, or dataset creation property list.</p>
<p>The dataset datatype refers to how data is stored in an HDF5 file while
the memory datatype refers to how data is stored in memory.
The HDF5 Library will do datatype conversion when writing data
in memory to the dataset or reading data from the dataset to memory if
the memory datatype differs from the dataset datatype.
Datatype conversion is performed by HDF5 Library before n-bit compression
and after n-bit decompression.</p>
<p>The following sub-sections examine the common cases:</p>
<ul>
<li>N-bit integer conversions</li>
<li>N-bit floating-point conversions</li>
</ul>
<h5><em>N-bit Integer Conversions</em></h5>
<p>Integer data with a dataset of integer datatype of less than
full precision and a memory datatype of <code>H5T_NATIVE_INT</code>,
provides the simplest application of the n-bit filter.</p>
<p>The precision of <code>H5T_NATIVE_INT</code> is 8 muliplied by
<code>sizeof(int)</code>.
This value, the size of an <code>int</code> in bytes, differs from
platform to platform; we assume a value of <code>4</code>
for the following illustration.
We further assume the memory byte order to be little-endian.</p>
<!-- NEW PAGE -->
<p>In memory, therefore, the precision of <code>H5T_NATIVE_INT</code>
is 32 and the offset is 0.
One value of <code>H5T_NATIVE_INT</code> is laid out in memory
as follows:</p>
<table border="0" width="80%" align="center">
<tr>
<td>
<pre>
| byte 3 | byte 2 | byte 1 | byte 0 |
|SPPPPPPP|PPPPPPPP|PPPPPPPP|PPPPPPPP|
</pre>
</td>
</tr>
<tr>
<td colspan="4"><font size="-1">
Key:
<code>S</code> - sign bit,
<code>P</code> - significant bit,
<code>?</code> - padding bit
<br>
Sign bit is included in signed integer datatype precision.
</font></td>
</tr>
</table>
<p>Suppose the dataset datatype has a precision of 16 and an offset of 4.
After HDF5 converts values from the memory datatype to the dataset datatype,
it passes something like the following to the n-bit filter for
compression:</p>
<table border="0" width="80%" align="center">
<tr>
<td>
<pre>
| byte 3 | byte 2 | byte 1 | byte 0 |
| |
|????????|????S|PPP|PPPPPPPP|PPPP|????|
|_________________|
truncated bits
</pre>
</td>
</tr>
<tr>
<td colspan="4"><font size="-1">
Key:
<code>S</code> - sign bit,
<code>P</code> - significant bit,
<code>?</code> - padding bit
<br>
Sign bit is included in signed integer datatype precision.
</font></td>
</tr>
</table>
<p>Notice that only the specified 16 bits (15 significant bits and the
sign bit) are retained in the conversion. All other significant bits
of the memory datatype are discarded because the dataset datatype
calls for only 16 bits of precision.
After n-bit compression, none of these discarded bits, known as
<i>padding bits</i> will be stored on disk.</p>
<h5><em>N-bit Floating-point Conversions</em></h5>
<p>Things get more complicated in the case of a floating-point dataset
datatype class. This sub-section provides an example that
illustrates the conversion from a memory datatype of
<code>H5T_NATIVE_FLOAT</code> to a dataset datatype of class
floating-point.</p>
<p>As before, let the <code>H5T_NATIVE_FLOAT</code> be 4 bytes long,
and let the memory byte order be little-endian.
Per the IEEE standard, one value of <code>H5T_NATIVE_FLOAT</code>
is laid out in memory as follows:</p>
<table border="0" width="80%" align="center">
<tr>
<td>
<pre>
| byte 3 | byte 2 | byte 1 | byte 0 |
|SEEEEEEE|EMMMMMMM|MMMMMMMM|MMMMMMMM|
</pre>
</td></tr>
<tr>
<td colspan="4"><font size="-1">
Key:
<code>S</code> - sign bit,
<code>E</code> - exponent bit,
<code>M</code> - mantissa bit,
<code>?</code> - padding bit
<br>
Sign bit is included in floating-point datatype precision.
</font></td>
</tr>
</table>
<br>
<p>Suppose the dataset datatype has a precision of 20, offset of 7,
mantissa size of 13, mantissa position of 7,
exponent size of 6, exponent position of 20,
and sign position of 26.
(See “Definition of Datatypes,” section 4.3 of the
“<a href="UG_frame11Datatypes.html">Datatypes</a>” chapter in
the <a href="index.html"><cite>HDF5 User’s Guide</cite></a>
for a discussion of creating and modifying datatypes.)</p>
<p>After HDF5 converts values from the memory datatype to the dataset datatype,
it passes something like the following to the n-bit filter for
compression:</p>
<table border="0" width="80%" align="center">
<tr>
<td>
<pre>
| byte 3 | byte 2 | byte 1 | byte 0 |
| |
|?????SEE|EEEE|MMMM|MMMMMMMM|M|???????|
|_______________|
truncated mantissa
</pre>
</td></tr>
<tr>
<td colspan="4"><font size="-1">
Key:
<code>S</code> - sign bit,
<code>E</code> - exponent bit,
<code>M</code> - mantissa bit,
<code>?</code> - padding bit
<br>
Sign bit is included in floating-point datatype precision.
</font></td>
</tr>
</table>
<p>The sign bit and truncated mantissa bits are not changed during
datatype conversion by the HDF5 Library. On the other hand,
the conversion of the 8-bit exponent to a 6-bit exponent
is a little tricky:</p>
<dir>
<p>The bias for the new exponent in the n-bit datatype is: </p>
<code>2<sup>(n-1)</sup>-1</code>
<br>
<p>The following formula is used for this exponent conversion:</p>
<code>exp8 - (2<sup>(8-1)</sup>-1)</code> =
<code>exp6 - (2<sup>(6-1)</sup>-1)</code> =
<i>actual exponent value</i>
<br /><br />
where <code>exp8</code> is the stored decimal value
as represented by the 8-bit exponent,
<br>
and <code>exp6</code> is the stored decimal value
as represented by the 6-bit exponent
</dir>
<p>In this example, caution must be taken to ensure that,
after conversion, the actual exponent value is
within the range that can be represented by a 6-bit exponent.
For example,
an 8-bit exponent can represent values from -127 to 128 while
a 6-bit exponent can represent values only from -31 to 32.</p>
<a name="Design">
<h4><em>N-bit Filter Behavior</em></h4>
</a>
<p>The n-bit filter was designed to treat the incoming data byte by byte at
the lowest level. The purpose was to make the n-bit filter as generic as
possible so that no pointer cast related to the datatype is needed.</p>
<p>Bitwise operations are employed for packing and unpacking at the byte
level.</p>
<p>Recursive function calls are used to treat compound and array datatypes.</p>
<h5><em>N-bit Compression</em></h5>
<p>The main idea of n-bit compression is to use a loop to compress each
data element in a chunk. Depending on the datatype of each element,
the n-bit filter will call one of four functions. Each of these functions
performs one of the following tasks: </p>
<ul>
<li>Compress a data element of a no-op datatype</li>
<li>Compress a data element of an atomic datatype</li>
<li>Compress a data element of a compound datatype</li>
<li>Compress a data element of an array datatype</li>
</ul>
<p><b>No-op datatypes:</b>
The n-bit filter does not actually compress no-op datatypes.
Rather, it copies the data buffer of the no-op datatype from the
noncompressed buffer to the proper location in the compressed buffer;
the compressed buffer has no holes. The term “compress”
is used here simply to distinguish this function from the function
that performs the reverse operation during decompression.</p>
<p><b>Atomic datatypes:</b>
The n-bit filter will find the bytes where significant bits are
located and try to compress these bytes, one byte at a time, using a loop.
At this level, the filter needs the following information:</p>
<ul>
<li>The byte offset of the beginning of the current data element with
respect to the beginning of the input data buffer</li>
<li>Datatype size, precision, offset, and byte order </li>
</ul>
<p>The n-bit filter compresses from the most significant byte containing
significant bits to the least significant byte.
For big-endian data, therefore, the loop index progresses from smaller
to larger while for little-endian, the loop index progresses from larger
to smaller.</p>
<p>In the extreme case of when the n-bit datatype has full precision,
this function copies the content of the entire noncompressed datatype
to the compressed output buffer.</p>
<p><b>Compound datatypes:</b>
The n-bit filter will compress each data member of the compound datatype.
If the member datatype is of an integer or floating-point datatype,
the n-bit filter will call the function described above<!-- in section 2.1.2-->.
If the member datatype is of a no-op datatype,
the filter will call the function described above<!-- in section 2.1.1-->.
If the member datatype is of a compound datatype, the filter will make a
recursive call to itself.
<!--
(i.e., to the function described in this section, 2.1.3).
-->
If the member datatype is of an array datatype, the filter will call the
function described below<!-- in section 2.1.4.--></p>
<p><b>Array datatypes:</b>
The n-bit filter will use a loop to compress each array element in
the array. If the base datatype of array element is of an integer or
floating-point datatype, the n-bit filter will call the function described
above<!-- in section 2.1.2.-->
If the base datatype is of a no-op datatype, the filter will call the
function described above<!-- in section 2.1.1.-->
If the base datatype is of a compound datatype, the filter will call the
function described above<!-- in section 2.1.3-->.
If the member datatype is of an array datatype, the filter will make a
recursive call of itself.</p>
<!--
(i.e., to the function described in this section, 2.1.4).
-->
<h5><em>N-bit Decompression</em></h5>
<p>The n-bit decompression algorithm is very similar to n-bit compression.
The only difference is that at the byte level, compression packs out all
padding bits and stores only significant bits into a continous buffer
(unsigned char) while decompression unpacks significant bits and inserts
padding bits (zeros) at the proper positions to recover the data bytes
as they existed before compression.</p>
<h5><em>Storing N-bit Parameters to Array</em> <code>cd_value[]</code></h5>
<p>All of the information, or parameters, required by the n-bit filter
are gathered and stored in the array <code>cd_values[]</code> by the
private function <code>H5Z_set_local_nbit</code> and are passed
to another private function,
<code>H5Z_filter_nbit</code>, by the HDF5 Library. </p>
<p>These parameters are as follows:</p>
<ol>
<li>Parameters related to the datatype</li>
<li>The number of elements within the chunk</li>
<li>A flag indicating whether compression is needed</li>
</ol>
<p>The first and second parameters can be obtained using the HDF5 dataspace
and datatype interface calls. </p>
<!--
The third parameter is set during the storing process as described
in section 3.2.
-->
<p>A compound datatype can have members of array or compound datatype.
An array datatype’s base datatype can be a complex compound datatype.
Recursive calls are required to set parameters for these complex situations.</p>
<p>Before setting the parameters, the number of parameters should be
calculated to dynamically allocate the array <code>cd_values[]</code>,
which will be passed to the HDF5 Library.
This also requires recursive calls.</p>
<p>For an atomic datatype (integer or floating-point), parameters that will
be stored include the datatype’s size, endianness, precision, and
offset. </p>
<p>For a no-op datatype, only the size is required.</p>
<p>For a compound datatype, parameters that will be stored include the
datatype’s total size and number of members. For each member,
its member offset needs to be stored. Other parameters for members
will depends on the respective datatype class.</p>
<p>For an array datatype, the total size parameter should be stored.
Other parameters for the array’s base type depend on the base
type’s datatype class. </p>
<p>Further, to correctly retrieve the parameter for use of n-bit
compression or decompression later, parameters for distinguishing
between datatype classes should be stored.</p>
<a name="implementation">
<h4><em>Implementation</em></h4>
</a>
<p>Three filter callback functions were written for the n-bit filter:</p>
<ul>
<li><code>H5Z_can_apply_nbit</code></li>
<li><code>H5Z_set_local_nbit</code></li>
<li><code>H5Z_filter_nbit</code></li>
</ul>
<p>These functions are called internally by the HDF5 Library.
A number of utility functions were written for the function
<code>H5Z_set_local_nbit</code>. Compression and decompression functions
were written and are called by function <code>H5Z_filter_nbit</code>.
All these functions are included in the file <code>H5Znbit.c</code>.</p>
<p>The public function <code>H5Pset_nbit</code> is called by
the application to set up the use of the n-bit filter.
This function is included in the file <code>H5Pdcpl.c</code>.
The application does not need to supply any parameters.</p>
<h5><em>How N-bit Parameters are Stored</em></h5>
<p>A scheme of storing parameters required by the n-bit filter in the
array <code>cd_values[]</code> was developed utilizing recursive
function calls.</p>
<p>Four private utility functions were written for storing the parameters
associated with atomic (integer or floating-point), no-op, array, and
compound datatypes:</p>
<ul>
<li><code>H5Z_set_parms_atomic</code></li>
<li><code>H5Z_set_parms_array</code></li>
<li><code>H5Z_set_parms_nooptype</code></li>
<li><code>H5Z_set_parms_compound</code> </li>
</ul>
<!-- NEW PAGE -->
<p>The scheme is briefly described below.</p>
<dir>
First, assign a numeric code for datatype class atomic (integer or float),
no-op, array, and compound datatype. The code is stored before other
datatype related parameters are stored.
<dl>
<dd>
<dt>The first three parameters of <code>cd_values[]</code> are reserved for:
<ol>
<li>The number of valid entries in the array <code>cd_values[]</code></li>
<li>A flag indicating whether compression is needed</li>
<li>The number of elements in the chunk</li>
</ol>
<dt>Throughout the balance of this explanation,
<code>i</code> represents the index of <code>cd_values[]</code>.
<br>
<dt>In the function <code>H5Z_set_local_nbit</code>:
<dd>
<ol>
<li><code>i</code> = 2</li>
<li>Get the number of elements in the chunk and store in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the class of the datatype:
<br> For an integer or floating-point datatype, call
<code>H5Z_set_parms_atomic</code>
<br> For an array datatype, call
<code>H5Z_set_parms_array</code>
<br> For a compound datatype, call
<code>H5Z_set_parms_compound</code>
<br> For none of the above, call
<code>H5Z_set_parms_noopdatatype</code></li>
<li>Store <code>i</code> in <code>cd_value[0]</code> and
flag in <code>cd_values[1]</code></li>
</ol>
</dd>
</dl>
<dl>
<dt>In the function <code>H5Z_set_parms_atomic</code>:</dt>
<dd>
<ol>
<li>Store the assigned numeric code for the atomic datatype in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the size of the atomic datatype and store in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the order of the atomic datatype and store in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the precision of the atomic datatype and store in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the offset of the atomic datatype and store in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Determine the need to do compression at this point</li>
</ol>
</dd>
</dl>
<dl>
<dt>In the function <code>H5Z_set_parms_nooptype</code>:
<dd>
<ol>
<li>Store the assigned numeric code for the no-op datatype in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the size of the no-op datatype and store in
<code>cd_value[i]</code>; increment <code>i</code></li>
</ol>
</dd>
</dl>
<dl>
<dt>In the function <code>H5Z_set_parms_array</code>:
<dd>
<ol>
<li>Store the assigned numeric code for the array datatype in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the size of the array datatype and store in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the class of the array'’s base datatype.
<br> For an integer or floating-point datatype,
call <code>H5Z_set_parms_atomic</code>
<br> For an array datatype, call
<code>H5Z_set_parms_array</code>
<br> For a compound datatype, call
<code>H5Z_set_parms_compound</code>
<br> If none of the above,
call <code>H5Z_set_parms_noopdatatype</code></li>
</ol>
</dd>
</dl>
<dl>
<dt>In the function <code>H5Z_set_parms_compound</code>:
<dd>
<ol>
<li>Store the assigned numeric code for the compound datatype in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the size of the compound datatype and store in
<code>cd_value[i]</code>; increment <code>i</code></li>
<li>Get the number of members and store in
<code>cd_values[i]</code>; increment <code>i</code></li>
<li>For each member
<br> Get the member offset and store in
<code>cd_values[i]</code>; increment <code>i</code>
<br> Get the class of the member datatype
<br> For an integer or floating-point datatype,
call <code>H5Z_set_parms_atomic</code>
<br> For an array datatype,
call <code>H5Z_set_parms_array</code>
<br> For a compound datatype,
call <code>H5Z_set_parms_compound</code>
<br> If none of the above,
call <code>H5Z_set_parms_noopdatatype</code></li>
</ol>
</dd>
</dl>
</dir>
<h5><em>N-bit Compression and Decompression Functions</em></h5>
<p>The n-bit compression and decompression functions above are called
by the private HDF5 function <code>H5Z_filter_nbit</code>.
The compress and decompress functions retrieve the n-bit parameters
from <code>cd_values[]</code> as it was passed by
<code>H5Z_filter_nbit</code>. Parameters are retrieved in exactly the
same order in which they are stored and lower-level compression and
decompression functions for different datatype classes are called. </p>
<!--
These functions are implementated according to the descriptions
in sections 2.1 and 2.2.
-->
<p>N-bit compression is not implemented in place. Due to the
difficulty of calculating actual output buffer size after compression,
the same space as that of the input buffer is allocated for the output
buffer as passed to the compression function. However, the size of the
output buffer passed by reference to the compression function will
be changed (smaller) after the compression is complete.</p>
<a name="examples">
<h4><em>Usage Examples</em></h4>
</a>
<p>The following code example illustrates the use of the n-bit filter
for writing and reading n-bit integer data.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
#include "hdf5.h"
#include "stdlib.h"
#include "math.h"
#define H5FILE_NAME "nbit_test_int.h5"
#define DATASET_NAME "nbit_int"
#define NX 200
#define NY 300
#define CH_NX 10
#define CH_NY 15
int main(void)
{
hid_t file, dataspace, dataset, datatype, mem_datatype, dset_create_props;
hsize_t dims[2], chunk_size[2];
int orig_data[NX][NY];
int new_data[NX][NY];
int i, j;
size_t precision, offset;
/* Define dataset datatype (integer), and set precision, offset */
datatype = H5Tcopy(H5T_NATIVE_INT);
precision = 17; /* precision includes sign bit */
if(H5Tset_precision(datatype,precision)<0) {
printf("Error: fail to set precision\n");
return -1;
}
offset = 4;
if(H5Tset_offset(datatype,offset)<0) {
printf("Error: fail to set offset\n");
return -1;
}
/* Copy to memory datatype */
mem_datatype = H5Tcopy(datatype);
/* Set order of dataset datatype */
if(H5Tset_order(datatype, H5T_ORDER_BE)<0) {
printf("Error: fail to set endianness\n");
return -1;
}
/* Initiliaze data buffer with random data within correct range
* corresponding to the memory datatype's precision and offset.
*/
for (i=0; i < NX; i++)
for (j=0; j < NY; j++)
orig_data[i][j] = rand() % (int)pow(2, precision-1) <<offset;
/* Describe the size of the array. */
dims[0] = NX;
dims[1] = NY;
if((dataspace = H5Screate_simple (2, dims, NULL))<0) {
printf("Error: fail to create dataspace\n");
return -1;
}
/*
* Create a new file using read/write access, default file
* creation properties, and default file access properties.
*/
if((file = H5Fcreate (H5FILE_NAME, H5F_ACC_TRUNC,
H5P_DEFAULT, H5P_DEFAULT))<0) {
printf("Error: fail to create file\n");
return -1;
}
/*
* Set the dataset creation property list to specify that
* the raw data is to be partitioned into 10 x 15 element
* chunks and that each chunk is to be compressed.
*/
chunk_size[0] = CH_NX;
chunk_size[1] = CH_NY;
if((dset_create_props = H5Pcreate (H5P_DATASET_CREATE))<0) {
printf("Error: fail to create dataset property\n");
return -1;
}
if(H5Pset_chunk (dset_create_props, 2, chunk_size)<0) {
printf("Error: fail to set chunk\n");
return -1;
}
</pre><!-- NEW PAGE -->
<pre>
/*
* Set parameters for n-bit compression; check the description of
* the H5Pset_nbit function in the HDF5 Reference Manual for more
* information.
*/
if(H5Pset_nbit (dset_create_props)<0) {
printf("Error: fail to set nbit filter\n");
return -1;
}
/*
* Create a new dataset within the file. The datatype
* and dataspace describe the data on disk, which may
* be different from the format used in the application's
* memory.
*/
if((dataset = H5Dcreate(file, DATASET_NAME, datatype,
dataspace, H5P_DEFAULT,
dset_create_props, H5P_DEFAULT))<0) {
printf("Error: fail to create dataset\n");
return -1;
}
/*
* Write the array to the file. The datatype and dataspace
* describe the format of the data in the 'orig_data' buffer.
* The raw data is translated to the format required on disk,
* as defined above. We use default raw data transfer properties.
*/
if(H5Dwrite (dataset, mem_datatype, H5S_ALL, H5S_ALL,
H5P_DEFAULT, orig_data)<0) {
printf("Error: fail to write to dataset\n");
return -1;
}
H5Dclose (dataset);
if((dataset = H5Dopen(file, DATASET_NAME, H5P_DEFAULT))<0) {
printf("Error: fail to open dataset\n");
return -1;
}
/*
* Read the array. This is similar to writing data,
* except the data flows in the opposite direction.
* Note: Decompression is automatic.
*/
if(H5Dread (dataset, mem_datatype, H5S_ALL, H5S_ALL,
H5P_DEFAULT, new_data)<0) {
printf("Error: fail to read from dataset\n");
return -1;
}
</pre><!-- NEW PAGE -->
<pre>
H5Tclose (datatype);
H5Tclose (mem_datatype);
H5Dclose (dataset);
H5Sclose (dataspace);
H5Pclose (dset_create_props);
H5Fclose (file);
return 0;
}</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 10. N-bit compression for integer data</b><br />
Illustrates the use of the n-bit filter for writing and reading
n-bit integer data.
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<p>The following code example illustrates the use of the n-bit filter
for writing and reading n-bit floating-point data.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
#include "hdf5.h"
#define H5FILE_NAME "nbit_test_float.h5"
#define DATASET_NAME "nbit_float"
#define NX 2
#define NY 5
#define CH_NX 2
#define CH_NY 5
int main(void)
{
hid_t file, dataspace, dataset, datatype, dset_create_props;
hsize_t dims[2], chunk_size[2];
/* orig_data[] are initialized to be within the range that can be
* represented by dataset datatype (no precision loss during
* datatype conversion)
*/
float orig_data[NX][NY] = {{188384.00, 19.103516, -1.0831790e9,
-84.242188, 5.2045898}, {-49140.000, 2350.2500, -3.2110596e-1,
6.4998865e-5, -0.0000000}};
float new_data[NX][NY];
size_t precision, offset;
/* Define single-precision floating-point type for dataset
*-------------------------------------------------------------------
* size=4 byte, precision=20 bits, offset=7 bits,
* mantissa size=13 bits, mantissa position=7,
* exponent size=6 bits, exponent position=20,
* exponent bias=31.
* It can be illustrated in little-endian order as:
* (S - sign bit, E - exponent bit, M - mantissa bit,
* ? - padding bit)
*
* 3 2 1 0
* ?????SEE EEEEMMMM MMMMMMMM M???????
*
* To create a new floating-point type, the following
* properties must be set in the order of
* set fields -> set offset -> set precision -> set size.
* All these properties must be set before the type can function.
* Other properties can be set anytime. Derived type size cannot
* be expanded bigger than original size but can be decreased.
* There should be no holes among the significant bits. Exponent
* bias usually is set 2^(n-1)-1, where n is the exponent size.
*-------------------------------------------------------------------*/
datatype = H5Tcopy(H5T_IEEE_F32BE);
if(H5Tset_fields(datatype, 26, 20, 6, 7, 13)<0) {
printf("Error: fail to set fields\n");
return -1;
}
offset = 7;
if(H5Tset_offset(datatype,offset)<0) {
printf("Error: fail to set offset\n");
return -1;
}
precision = 20;
if(H5Tset_precision(datatype,precision)<0) {
printf("Error: fail to set precision\n");
return -1;
}
if(H5Tset_size(datatype, 4)<0) {
printf("Error: fail to set size\n");
return -1;
}
if(H5Tset_ebias(datatype, 31)<0) {
printf("Error: fail to set exponent bias\n");
return -1;
}
/* Describe the size of the array. */
dims[0] = NX;
dims[1] = NY;
if((dataspace = H5Screate_simple (2, dims, NULL))<0) {
printf("Error: fail to create dataspace\n");
return -1;
}
/*
* Create a new file using read/write access, default file
* creation properties, and default file access properties.
*/
if((file = H5Fcreate (H5FILE_NAME, H5F_ACC_TRUNC,
H5P_DEFAULT, H5P_DEFAULT))<0) {
printf("Error: fail to create file\n");
return -1;
}
/*
* Set the dataset creation property list to specify that
* the raw data is to be partitioned into 2 x 5 element
* chunks and that each chunk is to be compressed.
*/
chunk_size[0] = CH_NX;
chunk_size[1] = CH_NY;
if((dset_create_props = H5Pcreate (H5P_DATASET_CREATE))<0) {
printf("Error: fail to create dataset property\n");
return -1;
}
if(H5Pset_chunk (dset_create_props, 2, chunk_size)<0) {
printf("Error: fail to set chunk\n");
return -1;
}
/*
* Set parameters for n-bit compression; check the description
* of the H5Pset_nbit function in the HDF5 Reference Manual
* for more information.
*/
if(H5Pset_nbit (dset_create_props)<0) {
printf("Error: fail to set nbit filter\n");
return -1;
}
/*
* Create a new dataset within the file. The datatype
* and dataspace describe the data on disk, which may
* be different from the format used in the application's
* memory.
*/
if((dataset = H5Dcreate(file, DATASET_NAME, datatype,
dataspace, H5P_DEFAULT,
dset_creat_plists, H5P_DEFAULT))<0) {
printf("Error: fail to create dataset\n");
return -1;
}
/*
* Write the array to the file. The datatype and dataspace
* describe the format of the data in the 'orig_data' buffer.
* The raw data is translated to the format required on disk,
* as defined above. We use default raw data transfer properties.
*/
if(H5Dwrite (dataset, H5T_NATIVE_FLOAT, H5S_ALL, H5S_ALL,
H5P_DEFAULT, orig_data)<0) {
printf("Error: fail to write to dataset\n");
return -1;
}
H5Dclose (dataset);
if((dataset = H5Dopen(file, DATASET_NAME, H5P_DEFAULT))<0) {
printf("Error: fail to open dataset\n");
return -1;
}
/*
* Read the array. This is similar to writing data,
* except the data flows in the opposite direction.
* Note: Decompression is automatic.
*/
if(H5Dread (dataset, H5T_NATIVE_FLOAT, H5S_ALL, H5S_ALL,
H5P_DEFAULT, new_data)<0) {
printf("Error: fail to read from dataset\n");
return -1;
}
H5Tclose (datatype);
H5Dclose (dataset);
H5Sclose (dataspace);
H5Pclose (dset_create_props);
H5Fclose (file);
return 0;
}</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 11. N-bit compression for floating-point data</b><br />
Illustrates the use of the n-bit filter for writing and reading
n-bit floating-point data.
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<a name="limitations">
<h4><em>Limitations</em></h4>
</a>
<p>Because the array <code>cd_values[]</code> has to fit into an object
header message of 64K, the n-bit filter has an upper limit on the number
of n-bit parameters that can be stored in it. To be conservative, a maximum
of 4K is allowed for the number of parameters.</p>
<p>The n-bit filter currently only compresses n-bit datatypes or fields derived
from integer or floating-point datatypes. The n-bit filter assumes padding
bits of zero. This may not be true since the HDF5 user can set padding bit
to be zero, one, or leave the background alone. However, it is expected
the n-bit filter will be modified to adjust to such situations.</p>
<p>The n-bit filter does not have a way to handle the situation where the
fill value of a dataset is defined and the fill value is not of an n-bit
datatype although the dataset datatype is.</p>
<!-- NEW PAGE -->
<a name="ScaleOffset">
<h3>5.6.2. Using the Scale-offset Filter</h3>
</a>
<p>Generally speaking, scale-offset compression performs a scale and/or
offset operation on each data value and truncates the resulting value
to a minimum number of bits (minimum-bits) before storing it. </p>
<p>The current scale-offset filter supports integer and floating-point
datatypes only. For the floating-point datatype, float and double are
supported, but long double is not supported.</p>
<p>Integer data compression uses a straight-forward algorithm. Floating-point
data compression adopts the GRiB data packing mechanism which offers
two alternate methods: a fixed minimum-bits method, and a variable
minimum-bits method. Currently, only the variable minimum-bits method
is implemented. <!-- 9.3.10, MEE: according to Kent, the fixed minimum-bits
method has not yet been implemented, and they do not have any plans to
implement it. --></p>
<p>Like other I/O filters supported by the HDF5 Library, applications
using the scale-offset filter must store data with chunked storage.</p>
<p><b><i>Integer type:</i></b>
The minimum-bits of integer data can be determined by the filter.
For example, if the maximum value of data to be compressed is 7065
and the minimum value is 2970. Then the “span” of dataset
values is equal to (max-min+1), which is 4676. If no fill value is
defined for the dataset, the minimum-bits is:
<code>ceiling(log2(span)) = 12</code>. With fill value set, the
minimum-bits is: <code>ceiling(log2(span+1)) = 13</code>.</p>
<p>HDF5 users can also set the minimum-bits. However, if the user gives
a minimum-bits that is less than that calculated by the filter,
the compression will be lossy.</p>
<p><b><i>Floating-point type:</i></b>
The basic idea of the scale-offset filter for the floating-point type is
to transform the data by some kind of scaling to integer data, and
then to follow the procedure of the scale-offset filter for the integer
type to do the data compression. Due to the data transformation from
floating-point to integer, the scale-offset filter is
lossy in nature. </p>
<p>Two methods of scaling the floating-point data are used: the so-called
D-scaling and E-scaling. D-scaling is more straightforward and easy to
understand. For HDF5 1.8 release, only the D-scaling method has been
implemented. <!-- 9.3.10, MEE: According to Kent, E-scaling has not yet
been implemented, and they have no plans to implement it in the future. --></p>
<h4><em>Design</em></h4>
<p>Before the filter does any real work, it needs to gather some information
from the HDF5 Library through API calls. The parameters the filter needs
are: </p>
<ul>
<li>The minimum-bits of the data value</li>
<li>The number of data elements in the chunk</li>
<li>The datatype class, size, sign (only for integer type), byte order,
and fill value
if defined</li>
</ul>
<p>Size and sign are needed to determine what kind of pointer
cast to use when retrieving values from the data buffer.</p>
<p>The pipeline of the filter can be divided into four parts:
(1)pre-compression; (2)compression; (3)decompression;
(4)post-decompression.</p>
<p>Depending on whether a fill value is defined or not, the filter will
handle pre-compression and post-decompression differently. </p>
<p>The scale-offset filter only needs the memory byte order, size of
datatype, and minimum-bits for compression and decompression.</p>
<p>Since decompression has no access to the original data, the minimum-bits
and the minimum value need to be stored with the compressed data for
decompression and post-decompression.</p>
<h5><em>Integer Type</em></h5>
<p><i>Pre-compression: </i>
During pre-compression minimum-bits is calculated if it is not
set by the user. For more information on how minimum-bits are calculated,
see section 6.1. “The N-bit Filter.” </p>
<p>If the fill value is defined, finding the maximum and minimum values
should ignore the data element whose value is equal to the fill value. </p>
<p>If no fill value is defined, the value of each data element is subtracted
by the minimum value during this stage.</p>
<p>If the fill value is defined, the fill value is assigned to the maximum
value. In this way minimum-bits can represent a data element whose value
is equal to the fill value and subtracts the minimum value from a data
element whose value is not equal to the fill value.</p>
<!-- 8.19.10, MEE: the paragraph belowis is Frank's revision of my editing -->
<!-- 9.3.10, MEE: Kent reviewed the paragraph below and said it was clear. -->
<p>The fill value (if defined), the number of elements in a chunk, the
class of the datatype, the size of the datatype, the memory order of the
datatype, and other similar elements will be stored in the HDF5 object
header for the post-decompression usage.</p>
<p>After pre-compression, all values are non-negative and are within the
range that can be stored by minimum-bits.</p>
<p><i>Compression: </i>
All modified data values after pre-compression are packed together
into the compressed data buffer. The number of bits for each data value
decreases from the number of bits of integer (32 for most platforms) to
minimum-bits. The value of minimum-bits and the minimum value are added to
the data buffer and the whole buffer is sent back to the library. In this
way, the number of bits for each modified value is no more than
the size of minimum-bits.</p>
<p><i>Decompression: </i>
In this stage, the number of bits for each data value is resumed from
minimum-bits to the number of bits of integer.</p>
<p><i>Post-decompression: </i>
For the post-decompression stage, the filter does the opposite
of what it does during pre-compression except that it does not calculate
the minimum-bits or the minimum value. These values were saved during
compression and can be retrieved through the resumed data buffer. If
no fill value is defined, the filter adds the minimum value back to
each data element.</p>
<p>If the fill value is defined, the filter assigns the fill value to the
data element whose value is equal to the maximum value that minimum-bits can
represent and adds the minimum value back to each data element whose value
is not equal to the maximum value that minimum-bits can represent.</p>
<a name="SO_FloatingPoint"><p> </p></a>
<h5><em>Floating-point Type</em></h5>
<p>The filter will do data transformation from floating-point type to
integer type and then handle the data by using the procedure for handling
the integer data inside the filter.
Insignificant bits of floating-point data will be cut off
during data transformation, so this filter is a lossy compression method.</p>
<p>There are two scaling methods: D-scaling and E-scaling.
The HDF5 1.8 release only supports D-scaling. D-scaling is short for
decimal scaling. E-scaling should be similar conceptually. In order
to transform data from floating-point to
integer, a scale factor is introduced. The minimum value will be calculated.
Each data element value will subtract the minimum value. The modified data
will be multiplied by 10 (Decimal) to the power of <code>scale_factor</code>,
and only the integer part will be kept and manipulated through the routines
for the integer type of the filter during pre-compression and compression.
Integer data will be divided by 10 to the power of
<code>scale_factor</code> to transform back to floating-point data
during decompression and post-decompression.
Each data element value will then add the minimum value, and the
floating-point data are resumed. However, the resumed data will lose some
insignificant bits compared with the original value.</p>
<p>For example, the following floating-point data are manipulated by the
filter, and the D-scaling factor is 2.</p>
<code>{104.561, 99.459, 100.545, 105.644}</code>
<p>The minimum value is 99.459, each data element subtracts 99.459, the
modified data is </p>
<code>{5.102, 0, 1.086, 6.185}</code>
<p>Since the D-scaling factor is 2, all floating-point data will be
multiplied by 10^2 with this result: </p>
<code>{510.2, 0, 108.6, 618.5}</code>
<p>The digit after decimal point will be rounded off, and then the set looks
like: </p>
<code>{510 , 0, 109, 619}</code>
<p>After decompression, each value will be divided by 10^2 and will be added
to the offset 99.459.</p>
<p>The floating-point data becomes </p>
<code>{104.559, 99.459, 100.549, 105.649}</code>.
<p>The relative error for each value should be no more than
5* (10^(D-scaling factor +1)). D-scaling sometimes is also referred
as a variable minimum-bits method since for different datasets the
minimum-bits to represent the same decimal precision will vary. The
data value is modified to 2 to power of <code>scale_factor</code> for
E-scaling. E-scaling is also called fixed-bits method since for different
datasets the minimum-bits will always be fixed to the scale factor of
E-scaling.
Currently HDF5 ONLY supports D-scaling (variable minimum-bits) method.</p>
<h4><em>Implementation</em></h4>
<p>The scale-offset filter implementation was written and included in the file
<code>H5Zscaleoffset.c</code>. Function <code>H5Pset_scaleoffset</code> was
written and included in the file “<code>H5Pdcpl.c</code>”. The
HDF5 user can supply minimum-bits by calling function
<code>H5Pset_scaleoffset</code>.</p>
<!-- NEW PAGE -->
<p>The scale-offset filter was implemented based on the design outlined in
this section. However, the following factors need to be considered:</p>
<dl>
<dd>
<ol>
<li>The filter needs the appropriate cast pointer whenever it needs
to retrieve data values.</li>
<li>The HDF5 Library passes to the filter the to-be-compressed data
in the format of the dataset datatype, and the filter passes back the
decompressed data in the same format. If a fill value is defined,
it is also in dataset datatype format.
For example, if the byte order of the dataset datatype is different
from that of the memory datatype of the platform, compression or
decompression performs an endianness conversion of data buffer.
Moreover, it should be aware that memory byte order can be different
during compression and decompression.</li>
<li>The difference of endianness and datatype between file and memory
should be considered when saving and retrieval of minimum-bits,
minimum value, and fill value.</li>
<li>If the user sets the minimum-bits to full precision of the datatype,
no operation is needed at the filter side. If the full precision is
a result of calculation by the filter, then the minimum-bits needs
to be saved for decompression but no compression or decompression
is needed (only a copy of the input buffer is needed).</li>
<li>If by calculation of the filter, the minimum-bits is equal to zero,
special handling is needed. Since it means all values are the same,
no compression or decompression is needed. But the minimum-bits
and minimum value still need to be saved during compression.</li>
<li>For floating-point data, the minimum value of the dataset should
be calculated at first. Each data element value will then subtract
the minimum value to obtain the “offset” data.
The offset data will then follow the steps outlined above in the
discussion of <a href="#SO_FloatingPoint">floating-point types</a>
to do data transformation to integer and rounding.</li>
</ol>
</dd>
</dl>
<h4><em>Usage Examples</em></h4>
<p>The following code example illustrates the use of the scale-offset filter
for writing and reading integer data.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
#include "hdf5.h"
#include "stdlib.h"
#define H5FILE_NAME "scaleoffset_test_int.h5"
#define DATASET_NAME "scaleoffset_int"
#define NX 200
#define NY 300
#define CH_NX 10
#define CH_NY 15
</pre>
<pre>
int main(void)
{
hid_t file, dataspace, dataset, datatype, dset_create_props;
hsize_t dims[2], chunk_size[2];
int orig_data[NX][NY];
int new_data[NX][NY];
int i, j, fill_val;
/* Define dataset datatype */
datatype = H5Tcopy(H5T_NATIVE_INT);
/* Initiliaze data buffer */
for (i=0; i < NX; i++)
for (j=0; j < NY; j++)
orig_data[i][j] = rand() % 10000;
/* Describe the size of the array. */
dims[0] = NX;
dims[1] = NY;
if((dataspace = H5Screate_simple (2, dims, NULL))<0) {
printf("Error: fail to create dataspace\n");
return -1;
}
/*
* Create a new file using read/write access, default file
* creation properties, and default file access properties.
*/
if((file = H5Fcreate (H5FILE_NAME, H5F_ACC_TRUNC,
H5P_DEFAULT, H5P_DEFAULT))<0) {
printf("Error: fail to create file\n");
return -1;
}
/*
* Set the dataset creation property list to specify that
* the raw data is to be partitioned into 10 x 15 element
* chunks and that each chunk is to be compressed.
*/
chunk_size[0] = CH_NX;
chunk_size[1] = CH_NY;
if((dset_create_props = H5Pcreate (H5P_DATASET_CREATE))<0) {
printf("Error: fail to create dataset property\n");
return -1;
}
if(H5Pset_chunk (dset_create_props, 2, chunk_size)<0) {
printf("Error: fail to set chunk\n");
return -1;
}
/* Set the fill value of dataset */
fill_val = 10000;
if (H5Pset_fill_value(dset_create_props, H5T_NATIVE_INT,
&fill_val)<0) {
printf("Error: can not set fill value for dataset\n");
return -1;
}
/*
* Set parameters for scale-offset compression. Check the
* description of the H5Pset_scaleoffset function in the
* HDF5 Reference Manual for more information [3].
*/
if(H5Pset_scaleoffset (dset_create_props, H5Z_SO_INT,
H5Z_SO_INT_MINIMUMBITS_DEFAULT)<0) {
printf("Error: fail to set scaleoffset filter\n");
return -1;
}
/*
* Create a new dataset within the file. The datatype
* and dataspace describe the data on disk, which may
* or may not be different from the format used in the
* application's memory. The link creation and
* dataset access property list parameters are passed
* with default values.
*/
if((dataset = H5Dcreate (file, DATASET_NAME, datatype,
dataspace, H5P_DEFAULT,
dset_create_props, H5P_DEFAULT))<0) {
printf("Error: fail to create dataset\n");
return -1;
}
/*
* Write the array to the file. The datatype and dataspace
* describe the format of the data in the 'orig_data' buffer.
* We use default raw data transfer properties.
*/
if(H5Dwrite (dataset, H5T_NATIVE_INT, H5S_ALL, H5S_ALL,
H5P_DEFAULT, orig_data)<0) {
printf("Error: fail to write to dataset\n");
return -1;
}
H5Dclose (dataset);
if((dataset = H5Dopen(file, DATASET_NAME, H5P_DEFAULT))<0) {
printf("Error: fail to open dataset\n");
return -1;
}
/*
* Read the array. This is similar to writing data,
* except the data flows in the opposite direction.
* Note: Decompression is automatic.
*/
if(H5Dread (dataset, H5T_NATIVE_INT, H5S_ALL, H5S_ALL,
H5P_DEFAULT, new_data)<0) {
printf("Error: fail to read from dataset\n");
return -1;
}
H5Tclose (datatype);
H5Dclose (dataset);
H5Sclose (dataspace);
H5Pclose (dset_create_props);
H5Fclose (file);
return 0;
}</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 12. Scale-offset compression integer data</b><br />
Illustrates the use of the scale-offset filter for writing
and reading integer data.
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<!-- NEW PAGE -->
<p>The following code example illustrates the use of the scale-offset filter
(set for variable minimum-bits method) for writing and reading
floating-point data.</p>
<table width="600" cellspacing="0" align="center">
<tr valign="top">
<td align="left">
<hr color="green" size="3"/>
<pre>
#include "hdf5.h"
#include "stdlib.h"
#define H5FILE_NAME "scaleoffset_test_float_Dscale.h5"
#define DATASET_NAME "scaleoffset_float_Dscale"
#define NX 200
#define NY 300
#define CH_NX 10
#define CH_NY 15
</pre>
<pre>
int main(void)
{
hid_t file, dataspace, dataset, datatype, dset_create_props;
hsize_t dims[2], chunk_size[2];
float orig_data[NX][NY];
float new_data[NX][NY];
float fill_val;
int i, j;
/* Define dataset datatype */
datatype = H5Tcopy(H5T_NATIVE_FLOAT);
/* Initiliaze data buffer */
for (i=0; i < NX; i++)
for (j=0; j < NY; j++)
orig_data[i][j] = (rand() % 10000) / 1000.0;
/* Describe the size of the array. */
dims[0] = NX;
dims[1] = NY;
if((dataspace = H5Screate_simple (2, dims, NULL))<0) {
printf("Error: fail to create dataspace\n");
return -1;
}
/*
* Create a new file using read/write access, default file
* creation properties, and default file access properties.
*/
if((file = H5Fcreate (H5FILE_NAME, H5F_ACC_TRUNC,
H5P_DEFAULT, H5P_DEFAULT))<0) {
printf("Error: fail to create file\n");
return -1;
}
/*
* Set the dataset creation property list to specify that
* the raw data is to be partitioned into 10 x 15 element
* chunks and that each chunk is to be compressed.
*/
chunk_size[0] = CH_NX;
chunk_size[1] = CH_NY;
if((dset_create_props = H5Pcreate (H5P_DATASET_CREATE))<0) {
printf("Error: fail to create dataset property\n");
return -1;
}
if(H5Pset_chunk (dset_create_props, 2, chunk_size)<0) {
printf("Error: fail to set chunk\n");
return -1;
}
/* Set the fill value of dataset */
fill_val = 10000.0;
if (H5Pset_fill_value(dset_create_props, H5T_NATIVE_FLOAT,
&fill_val)<0) {
printf("Error: can not set fill value for dataset\n");
return -1;
}
/*
* Set parameters for scale-offset compression; use variable
* minimum-bits method, set decimal scale factor to 3. Check the
* description of the H5Pset_scaleoffset function in the HDF5
* Reference Manual for more information [3].
*/
if(H5Pset_scaleoffset (dset_create_props, H5Z_SO_FLOAT_DSCALE, 3)<0) {
printf("Error: fail to set scaleoffset filter\n");
return -1;
}
/*
* Create a new dataset within the file. The datatype
* and dataspace describe the data on disk, which may
* or may not be different from the format used in the
* application's memory.
*/
if((dataset = H5Dcreate (file, DATASET_NAME, datatype,
dataspace, H5P_DEFAULT,
dset_create_props, H5P_DEFAULT))<0) {
printf("Error: fail to create dataset\n");
return -1;
}
/*
* Write the array to the file. The datatype and dataspace
* describe the format of the data in the 'orig_data' buffer.
* We use default raw data transfer properties.
*/
if(H5Dwrite (dataset, H5T_NATIVE_FLOAT, H5S_ALL, H5S_ALL,
H5P_DEFAULT, orig_data)<0) {
printf("Error: fail to write to dataset\n");
return -1;
}
H5Dclose (dataset);
if((dataset = H5Dopen(file, DATASET_NAME, H5P_DEFAULT))<0) {
printf("Error: fail to open dataset\n");
return -1;
}
/*
* Read the array. This is similar to writing data,
* except the data flows in the opposite direction.
* Note: Decompression is automatic.
*/
if(H5Dread (dataset, H5T_NATIVE_FLOAT, H5S_ALL, H5S_ALL,
H5P_DEFAULT, new_data)<0) {
printf("Error: fail to read from dataset\n");
return -1;
}
H5Tclose (datatype);
H5Dclose (dataset);
H5Sclose (dataspace);
H5Pclose (dset_create_props);
H5Fclose (file);
return 0;
}</pre></td>
</tr>
<tr><td><hr color="green" size="1" /></td></tr>
<tr valign="top">
<td align="left">
<b>Example 13. Scale-offset compression floating-point data</b><br />
Illustrates the use of the scale-offset filter for writing
and reading floating-point data.
<hr color="green" size="3"/></td>
</tr>
</table>
<br />
<h4><em>Limitations</em></h4>
<p>For floating-point data handling, there are some algorithmic
limitations to the GRiB data packing mechanism:</p>
<dl>
<dd>
<ol>
<li>Both the E-scaling and D-scaling methods are lossy compression</li>
<li>For the D-scaling method, since data values have been rounded to
integer values (positive) before truncating to the minimum-bits,
their range is limited by the maximum value that can be represented
by the corresponding unsigned integer type (the same size as that of
the floating-point type)</li>
</ol>
</dd>
</dl>
<h4><em>Suggestions</em></h4>
<p>The following are some suggestions for using the filter for
floating-point data:</p>
<dl>
<dd>
<ol>
<li>It is better to convert the units of data so that the units are
within certain common range (for example, 1200m to 1.2km)</li>
<li>If data values to be compressed are very near to zero, it
is strongly recommended that the user sets the fill value away
from zero (for example, a large positive number); if the user
does
nothing, the HDF5 Library will set the fill value to zero, and
this may cause undesirable compression results</li>
<li>Users are not encouraged to use a very large decimal scale
factor (e.g. 100) for the D-scaling method; this can cause the
filter not to ignore the fill value when finding maximum and minimum
values, and they will get a much larger minimum-bits (poor
compression)</li>
</ol>
</dd>
</dl>
<a name="Szip">
<h3>5.6.3. Using the Szip Filter</h3>
</a>
<p>See The HDF Group website for
<a href="http://www.hdfgroup.org/doc_resource/SZIP/" target="Ext1">further
information</a> regarding the Szip filter.</p>
<p> </p>
<p> </p>
<!-- HEADER RIGHT " " -->
</body>
</html>
|
