This file is indexed.

/usr/include/d/std/uni.d is in libphobos2-ldc-dev 1:0.17.1-1ubuntu1.

This file is owned by root:root, with mode 0o644.

The actual contents of the file can be viewed below.

   1
   2
   3
   4
   5
   6
   7
   8
   9
  10
  11
  12
  13
  14
  15
  16
  17
  18
  19
  20
  21
  22
  23
  24
  25
  26
  27
  28
  29
  30
  31
  32
  33
  34
  35
  36
  37
  38
  39
  40
  41
  42
  43
  44
  45
  46
  47
  48
  49
  50
  51
  52
  53
  54
  55
  56
  57
  58
  59
  60
  61
  62
  63
  64
  65
  66
  67
  68
  69
  70
  71
  72
  73
  74
  75
  76
  77
  78
  79
  80
  81
  82
  83
  84
  85
  86
  87
  88
  89
  90
  91
  92
  93
  94
  95
  96
  97
  98
  99
 100
 101
 102
 103
 104
 105
 106
 107
 108
 109
 110
 111
 112
 113
 114
 115
 116
 117
 118
 119
 120
 121
 122
 123
 124
 125
 126
 127
 128
 129
 130
 131
 132
 133
 134
 135
 136
 137
 138
 139
 140
 141
 142
 143
 144
 145
 146
 147
 148
 149
 150
 151
 152
 153
 154
 155
 156
 157
 158
 159
 160
 161
 162
 163
 164
 165
 166
 167
 168
 169
 170
 171
 172
 173
 174
 175
 176
 177
 178
 179
 180
 181
 182
 183
 184
 185
 186
 187
 188
 189
 190
 191
 192
 193
 194
 195
 196
 197
 198
 199
 200
 201
 202
 203
 204
 205
 206
 207
 208
 209
 210
 211
 212
 213
 214
 215
 216
 217
 218
 219
 220
 221
 222
 223
 224
 225
 226
 227
 228
 229
 230
 231
 232
 233
 234
 235
 236
 237
 238
 239
 240
 241
 242
 243
 244
 245
 246
 247
 248
 249
 250
 251
 252
 253
 254
 255
 256
 257
 258
 259
 260
 261
 262
 263
 264
 265
 266
 267
 268
 269
 270
 271
 272
 273
 274
 275
 276
 277
 278
 279
 280
 281
 282
 283
 284
 285
 286
 287
 288
 289
 290
 291
 292
 293
 294
 295
 296
 297
 298
 299
 300
 301
 302
 303
 304
 305
 306
 307
 308
 309
 310
 311
 312
 313
 314
 315
 316
 317
 318
 319
 320
 321
 322
 323
 324
 325
 326
 327
 328
 329
 330
 331
 332
 333
 334
 335
 336
 337
 338
 339
 340
 341
 342
 343
 344
 345
 346
 347
 348
 349
 350
 351
 352
 353
 354
 355
 356
 357
 358
 359
 360
 361
 362
 363
 364
 365
 366
 367
 368
 369
 370
 371
 372
 373
 374
 375
 376
 377
 378
 379
 380
 381
 382
 383
 384
 385
 386
 387
 388
 389
 390
 391
 392
 393
 394
 395
 396
 397
 398
 399
 400
 401
 402
 403
 404
 405
 406
 407
 408
 409
 410
 411
 412
 413
 414
 415
 416
 417
 418
 419
 420
 421
 422
 423
 424
 425
 426
 427
 428
 429
 430
 431
 432
 433
 434
 435
 436
 437
 438
 439
 440
 441
 442
 443
 444
 445
 446
 447
 448
 449
 450
 451
 452
 453
 454
 455
 456
 457
 458
 459
 460
 461
 462
 463
 464
 465
 466
 467
 468
 469
 470
 471
 472
 473
 474
 475
 476
 477
 478
 479
 480
 481
 482
 483
 484
 485
 486
 487
 488
 489
 490
 491
 492
 493
 494
 495
 496
 497
 498
 499
 500
 501
 502
 503
 504
 505
 506
 507
 508
 509
 510
 511
 512
 513
 514
 515
 516
 517
 518
 519
 520
 521
 522
 523
 524
 525
 526
 527
 528
 529
 530
 531
 532
 533
 534
 535
 536
 537
 538
 539
 540
 541
 542
 543
 544
 545
 546
 547
 548
 549
 550
 551
 552
 553
 554
 555
 556
 557
 558
 559
 560
 561
 562
 563
 564
 565
 566
 567
 568
 569
 570
 571
 572
 573
 574
 575
 576
 577
 578
 579
 580
 581
 582
 583
 584
 585
 586
 587
 588
 589
 590
 591
 592
 593
 594
 595
 596
 597
 598
 599
 600
 601
 602
 603
 604
 605
 606
 607
 608
 609
 610
 611
 612
 613
 614
 615
 616
 617
 618
 619
 620
 621
 622
 623
 624
 625
 626
 627
 628
 629
 630
 631
 632
 633
 634
 635
 636
 637
 638
 639
 640
 641
 642
 643
 644
 645
 646
 647
 648
 649
 650
 651
 652
 653
 654
 655
 656
 657
 658
 659
 660
 661
 662
 663
 664
 665
 666
 667
 668
 669
 670
 671
 672
 673
 674
 675
 676
 677
 678
 679
 680
 681
 682
 683
 684
 685
 686
 687
 688
 689
 690
 691
 692
 693
 694
 695
 696
 697
 698
 699
 700
 701
 702
 703
 704
 705
 706
 707
 708
 709
 710
 711
 712
 713
 714
 715
 716
 717
 718
 719
 720
 721
 722
 723
 724
 725
 726
 727
 728
 729
 730
 731
 732
 733
 734
 735
 736
 737
 738
 739
 740
 741
 742
 743
 744
 745
 746
 747
 748
 749
 750
 751
 752
 753
 754
 755
 756
 757
 758
 759
 760
 761
 762
 763
 764
 765
 766
 767
 768
 769
 770
 771
 772
 773
 774
 775
 776
 777
 778
 779
 780
 781
 782
 783
 784
 785
 786
 787
 788
 789
 790
 791
 792
 793
 794
 795
 796
 797
 798
 799
 800
 801
 802
 803
 804
 805
 806
 807
 808
 809
 810
 811
 812
 813
 814
 815
 816
 817
 818
 819
 820
 821
 822
 823
 824
 825
 826
 827
 828
 829
 830
 831
 832
 833
 834
 835
 836
 837
 838
 839
 840
 841
 842
 843
 844
 845
 846
 847
 848
 849
 850
 851
 852
 853
 854
 855
 856
 857
 858
 859
 860
 861
 862
 863
 864
 865
 866
 867
 868
 869
 870
 871
 872
 873
 874
 875
 876
 877
 878
 879
 880
 881
 882
 883
 884
 885
 886
 887
 888
 889
 890
 891
 892
 893
 894
 895
 896
 897
 898
 899
 900
 901
 902
 903
 904
 905
 906
 907
 908
 909
 910
 911
 912
 913
 914
 915
 916
 917
 918
 919
 920
 921
 922
 923
 924
 925
 926
 927
 928
 929
 930
 931
 932
 933
 934
 935
 936
 937
 938
 939
 940
 941
 942
 943
 944
 945
 946
 947
 948
 949
 950
 951
 952
 953
 954
 955
 956
 957
 958
 959
 960
 961
 962
 963
 964
 965
 966
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174
6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414
6415
6416
6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444
6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458
6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472
6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564
6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
6602
6603
6604
6605
6606
6607
6608
6609
6610
6611
6612
6613
6614
6615
6616
6617
6618
6619
6620
6621
6622
6623
6624
6625
6626
6627
6628
6629
6630
6631
6632
6633
6634
6635
6636
6637
6638
6639
6640
6641
6642
6643
6644
6645
6646
6647
6648
6649
6650
6651
6652
6653
6654
6655
6656
6657
6658
6659
6660
6661
6662
6663
6664
6665
6666
6667
6668
6669
6670
6671
6672
6673
6674
6675
6676
6677
6678
6679
6680
6681
6682
6683
6684
6685
6686
6687
6688
6689
6690
6691
6692
6693
6694
6695
6696
6697
6698
6699
6700
6701
6702
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
6728
6729
6730
6731
6732
6733
6734
6735
6736
6737
6738
6739
6740
6741
6742
6743
6744
6745
6746
6747
6748
6749
6750
6751
6752
6753
6754
6755
6756
6757
6758
6759
6760
6761
6762
6763
6764
6765
6766
6767
6768
6769
6770
6771
6772
6773
6774
6775
6776
6777
6778
6779
6780
6781
6782
6783
6784
6785
6786
6787
6788
6789
6790
6791
6792
6793
6794
6795
6796
6797
6798
6799
6800
6801
6802
6803
6804
6805
6806
6807
6808
6809
6810
6811
6812
6813
6814
6815
6816
6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
6833
6834
6835
6836
6837
6838
6839
6840
6841
6842
6843
6844
6845
6846
6847
6848
6849
6850
6851
6852
6853
6854
6855
6856
6857
6858
6859
6860
6861
6862
6863
6864
6865
6866
6867
6868
6869
6870
6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882
6883
6884
6885
6886
6887
6888
6889
6890
6891
6892
6893
6894
6895
6896
6897
6898
6899
6900
6901
6902
6903
6904
6905
6906
6907
6908
6909
6910
6911
6912
6913
6914
6915
6916
6917
6918
6919
6920
6921
6922
6923
6924
6925
6926
6927
6928
6929
6930
6931
6932
6933
6934
6935
6936
6937
6938
6939
6940
6941
6942
6943
6944
6945
6946
6947
6948
6949
6950
6951
6952
6953
6954
6955
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
6971
6972
6973
6974
6975
6976
6977
6978
6979
6980
6981
6982
6983
6984
6985
6986
6987
6988
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
7004
7005
7006
7007
7008
7009
7010
7011
7012
7013
7014
7015
7016
7017
7018
7019
7020
7021
7022
7023
7024
7025
7026
7027
7028
7029
7030
7031
7032
7033
7034
7035
7036
7037
7038
7039
7040
7041
7042
7043
7044
7045
7046
7047
7048
7049
7050
7051
7052
7053
7054
7055
7056
7057
7058
7059
7060
7061
7062
7063
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
7077
7078
7079
7080
7081
7082
7083
7084
7085
7086
7087
7088
7089
7090
7091
7092
7093
7094
7095
7096
7097
7098
7099
7100
7101
7102
7103
7104
7105
7106
7107
7108
7109
7110
7111
7112
7113
7114
7115
7116
7117
7118
7119
7120
7121
7122
7123
7124
7125
7126
7127
7128
7129
7130
7131
7132
7133
7134
7135
7136
7137
7138
7139
7140
7141
7142
7143
7144
7145
7146
7147
7148
7149
7150
7151
7152
7153
7154
7155
7156
7157
7158
7159
7160
7161
7162
7163
7164
7165
7166
7167
7168
7169
7170
7171
7172
7173
7174
7175
7176
7177
7178
7179
7180
7181
7182
7183
7184
7185
7186
7187
7188
7189
7190
7191
7192
7193
7194
7195
7196
7197
7198
7199
7200
7201
7202
7203
7204
7205
7206
7207
7208
7209
7210
7211
7212
7213
7214
7215
7216
7217
7218
7219
7220
7221
7222
7223
7224
7225
7226
7227
7228
7229
7230
7231
7232
7233
7234
7235
7236
7237
7238
7239
7240
7241
7242
7243
7244
7245
7246
7247
7248
7249
7250
7251
7252
7253
7254
7255
7256
7257
7258
7259
7260
7261
7262
7263
7264
7265
7266
7267
7268
7269
7270
7271
7272
7273
7274
7275
7276
7277
7278
7279
7280
7281
7282
7283
7284
7285
7286
7287
7288
7289
7290
7291
7292
7293
7294
7295
7296
7297
7298
7299
7300
7301
7302
7303
7304
7305
7306
7307
7308
7309
7310
7311
7312
7313
7314
7315
7316
7317
7318
7319
7320
7321
7322
7323
7324
7325
7326
7327
7328
7329
7330
7331
7332
7333
7334
7335
7336
7337
7338
7339
7340
7341
7342
7343
7344
7345
7346
7347
7348
7349
7350
7351
7352
7353
7354
7355
7356
7357
7358
7359
7360
7361
7362
7363
7364
7365
7366
7367
7368
7369
7370
7371
7372
7373
7374
7375
7376
7377
7378
7379
7380
7381
7382
7383
7384
7385
7386
7387
7388
7389
7390
7391
7392
7393
7394
7395
7396
7397
7398
7399
7400
7401
7402
7403
7404
7405
7406
7407
7408
7409
7410
7411
7412
7413
7414
7415
7416
7417
7418
7419
7420
7421
7422
7423
7424
7425
7426
7427
7428
7429
7430
7431
7432
7433
7434
7435
7436
7437
7438
7439
7440
7441
7442
7443
7444
7445
7446
7447
7448
7449
7450
7451
7452
7453
7454
7455
7456
7457
7458
7459
7460
7461
7462
7463
7464
7465
7466
7467
7468
7469
7470
7471
7472
7473
7474
7475
7476
7477
7478
7479
7480
7481
7482
7483
7484
7485
7486
7487
7488
7489
7490
7491
7492
7493
7494
7495
7496
7497
7498
7499
7500
7501
7502
7503
7504
7505
7506
7507
7508
7509
7510
7511
7512
7513
7514
7515
7516
7517
7518
7519
7520
7521
7522
7523
7524
7525
7526
7527
7528
7529
7530
7531
7532
7533
7534
7535
7536
7537
7538
7539
7540
7541
7542
7543
7544
7545
7546
7547
7548
7549
7550
7551
7552
7553
7554
7555
7556
7557
7558
7559
7560
7561
7562
7563
7564
7565
7566
7567
7568
7569
7570
7571
7572
7573
7574
7575
7576
7577
7578
7579
7580
7581
7582
7583
7584
7585
7586
7587
7588
7589
7590
7591
7592
7593
7594
7595
7596
7597
7598
7599
7600
7601
7602
7603
7604
7605
7606
7607
7608
7609
7610
7611
7612
7613
7614
7615
7616
7617
7618
7619
7620
7621
7622
7623
7624
7625
7626
7627
7628
7629
7630
7631
7632
7633
7634
7635
7636
7637
7638
7639
7640
7641
7642
7643
7644
7645
7646
7647
7648
7649
7650
7651
7652
7653
7654
7655
7656
7657
7658
7659
7660
7661
7662
7663
7664
7665
7666
7667
7668
7669
7670
7671
7672
7673
7674
7675
7676
7677
7678
7679
7680
7681
7682
7683
7684
7685
7686
7687
7688
7689
7690
7691
7692
7693
7694
7695
7696
7697
7698
7699
7700
7701
7702
7703
7704
7705
7706
7707
7708
7709
7710
7711
7712
7713
7714
7715
7716
7717
7718
7719
7720
7721
7722
7723
7724
7725
7726
7727
7728
7729
7730
7731
7732
7733
7734
7735
7736
7737
7738
7739
7740
7741
7742
7743
7744
7745
7746
7747
7748
7749
7750
7751
7752
7753
7754
7755
7756
7757
7758
7759
7760
7761
7762
7763
7764
7765
7766
7767
7768
7769
7770
7771
7772
7773
7774
7775
7776
7777
7778
7779
7780
7781
7782
7783
7784
7785
7786
7787
7788
7789
7790
7791
7792
7793
7794
7795
7796
7797
7798
7799
7800
7801
7802
7803
7804
7805
7806
7807
7808
7809
7810
7811
7812
7813
7814
7815
7816
7817
7818
7819
7820
7821
7822
7823
7824
7825
7826
7827
7828
7829
7830
7831
7832
7833
7834
7835
7836
7837
7838
7839
7840
7841
7842
7843
7844
7845
7846
7847
7848
7849
7850
7851
7852
7853
7854
7855
7856
7857
7858
7859
7860
7861
7862
7863
7864
7865
7866
7867
7868
7869
7870
7871
7872
7873
7874
7875
7876
7877
7878
7879
7880
7881
7882
7883
7884
7885
7886
7887
7888
7889
7890
7891
7892
7893
7894
7895
7896
7897
7898
7899
7900
7901
7902
7903
7904
7905
7906
7907
7908
7909
7910
7911
7912
7913
7914
7915
7916
7917
7918
7919
7920
7921
7922
7923
7924
7925
7926
7927
7928
7929
7930
7931
7932
7933
7934
7935
7936
7937
7938
7939
7940
7941
7942
7943
7944
7945
7946
7947
7948
7949
7950
7951
7952
7953
7954
7955
7956
7957
7958
7959
7960
7961
7962
7963
7964
7965
7966
7967
7968
7969
7970
7971
7972
7973
7974
7975
7976
7977
7978
7979
7980
7981
7982
7983
7984
7985
7986
7987
7988
7989
7990
7991
7992
7993
7994
7995
7996
7997
7998
7999
8000
8001
8002
8003
8004
8005
8006
8007
8008
8009
8010
8011
8012
8013
8014
8015
8016
8017
8018
8019
8020
8021
8022
8023
8024
8025
8026
8027
8028
8029
8030
8031
8032
8033
8034
8035
8036
8037
8038
8039
8040
8041
8042
8043
8044
8045
8046
8047
8048
8049
8050
8051
8052
8053
8054
8055
8056
8057
8058
8059
8060
8061
8062
8063
8064
8065
8066
8067
8068
8069
8070
8071
8072
8073
8074
8075
8076
8077
8078
8079
8080
8081
8082
8083
8084
8085
8086
8087
8088
8089
8090
8091
8092
8093
8094
8095
8096
8097
8098
8099
8100
8101
8102
8103
8104
8105
8106
8107
8108
8109
8110
8111
8112
8113
8114
8115
8116
8117
8118
8119
8120
8121
8122
8123
8124
8125
8126
8127
8128
8129
8130
8131
8132
8133
8134
8135
8136
8137
8138
8139
8140
8141
8142
8143
8144
8145
8146
8147
8148
8149
8150
8151
8152
8153
8154
8155
8156
8157
8158
8159
8160
8161
8162
8163
8164
8165
8166
8167
8168
8169
8170
8171
8172
8173
8174
8175
8176
8177
8178
8179
8180
8181
8182
8183
8184
8185
8186
8187
8188
8189
8190
8191
8192
8193
8194
8195
8196
8197
8198
8199
8200
8201
8202
8203
8204
8205
8206
8207
8208
8209
8210
8211
8212
8213
8214
8215
8216
8217
8218
8219
8220
8221
8222
8223
8224
8225
8226
8227
8228
8229
8230
8231
8232
8233
8234
8235
8236
8237
8238
8239
8240
8241
8242
8243
8244
8245
8246
8247
8248
8249
8250
8251
8252
8253
8254
8255
8256
8257
8258
8259
8260
8261
8262
8263
8264
8265
8266
8267
8268
8269
8270
8271
8272
8273
8274
8275
8276
8277
8278
8279
8280
8281
8282
8283
8284
8285
8286
8287
8288
8289
8290
8291
8292
8293
8294
8295
8296
8297
8298
8299
8300
8301
8302
8303
8304
8305
8306
8307
8308
8309
8310
8311
8312
8313
8314
8315
8316
8317
8318
8319
8320
8321
8322
8323
8324
8325
8326
8327
8328
8329
8330
8331
8332
8333
8334
8335
8336
8337
8338
8339
8340
8341
8342
8343
8344
8345
8346
8347
8348
8349
8350
8351
8352
8353
8354
8355
8356
8357
8358
8359
8360
8361
8362
8363
8364
8365
8366
8367
8368
8369
8370
8371
8372
8373
8374
8375
8376
8377
8378
8379
8380
8381
8382
8383
8384
8385
8386
8387
8388
8389
8390
8391
8392
8393
8394
8395
8396
8397
8398
8399
8400
8401
8402
8403
8404
8405
8406
8407
8408
8409
8410
8411
8412
8413
8414
8415
8416
8417
8418
8419
8420
8421
8422
8423
8424
8425
8426
8427
8428
8429
8430
8431
8432
8433
8434
8435
8436
8437
8438
8439
8440
8441
8442
8443
8444
8445
8446
8447
8448
8449
8450
8451
8452
8453
8454
8455
8456
8457
8458
8459
8460
8461
8462
8463
8464
8465
8466
8467
8468
8469
8470
8471
8472
8473
8474
8475
8476
8477
8478
8479
8480
8481
8482
8483
8484
8485
8486
8487
8488
8489
8490
8491
8492
8493
8494
8495
8496
8497
8498
8499
8500
8501
8502
8503
8504
8505
8506
8507
8508
8509
8510
8511
8512
8513
8514
8515
8516
8517
8518
8519
8520
8521
8522
8523
8524
8525
8526
8527
8528
8529
8530
8531
8532
8533
8534
8535
8536
8537
8538
8539
8540
8541
8542
8543
8544
8545
8546
8547
8548
8549
8550
8551
8552
8553
8554
8555
8556
8557
8558
8559
8560
8561
8562
8563
8564
8565
8566
8567
8568
8569
8570
8571
8572
8573
8574
8575
8576
8577
8578
8579
8580
8581
8582
8583
8584
8585
8586
8587
8588
8589
8590
8591
8592
8593
8594
8595
8596
8597
8598
8599
8600
8601
8602
8603
8604
8605
8606
8607
8608
8609
8610
8611
8612
8613
8614
8615
8616
8617
8618
8619
8620
8621
8622
8623
8624
8625
8626
8627
8628
8629
8630
8631
8632
8633
8634
8635
8636
8637
8638
8639
8640
8641
8642
8643
8644
8645
8646
8647
8648
8649
8650
8651
8652
8653
8654
8655
8656
8657
8658
8659
8660
8661
8662
8663
8664
8665
8666
8667
8668
8669
8670
8671
8672
8673
8674
8675
8676
8677
8678
8679
8680
8681
8682
8683
8684
8685
8686
8687
8688
8689
8690
8691
8692
8693
8694
8695
8696
8697
8698
8699
8700
8701
8702
8703
8704
8705
8706
8707
8708
8709
8710
8711
8712
8713
8714
8715
8716
8717
8718
8719
8720
8721
8722
8723
8724
8725
8726
8727
8728
8729
8730
8731
8732
8733
8734
8735
8736
8737
8738
8739
8740
8741
8742
8743
8744
8745
8746
8747
8748
8749
8750
8751
8752
8753
8754
8755
8756
8757
8758
8759
8760
8761
8762
8763
8764
8765
8766
8767
8768
8769
8770
8771
8772
8773
8774
8775
8776
8777
8778
8779
8780
8781
8782
8783
8784
8785
8786
8787
8788
8789
8790
8791
8792
8793
8794
8795
8796
8797
8798
8799
8800
8801
8802
8803
8804
8805
8806
8807
8808
8809
8810
8811
8812
8813
8814
8815
8816
8817
8818
8819
8820
8821
8822
8823
8824
8825
8826
8827
8828
8829
8830
8831
8832
8833
8834
8835
8836
8837
8838
8839
8840
8841
8842
8843
8844
8845
8846
8847
8848
8849
8850
8851
8852
8853
8854
8855
8856
8857
8858
8859
8860
8861
8862
8863
8864
8865
8866
8867
8868
8869
8870
8871
8872
8873
8874
8875
8876
8877
8878
8879
8880
8881
8882
8883
8884
8885
8886
8887
8888
8889
8890
8891
8892
8893
8894
8895
8896
8897
8898
8899
8900
8901
8902
8903
8904
8905
8906
8907
8908
8909
8910
8911
8912
8913
8914
8915
8916
8917
8918
8919
8920
8921
8922
8923
8924
8925
8926
8927
8928
8929
8930
8931
8932
8933
8934
8935
8936
8937
8938
8939
8940
8941
8942
8943
8944
8945
8946
8947
8948
8949
8950
8951
8952
8953
8954
8955
8956
8957
8958
8959
8960
8961
8962
8963
8964
8965
8966
8967
8968
8969
8970
8971
8972
8973
8974
8975
8976
8977
8978
8979
8980
8981
8982
8983
8984
8985
8986
8987
8988
8989
8990
8991
8992
8993
8994
8995
8996
8997
8998
8999
9000
9001
9002
9003
9004
9005
9006
9007
9008
9009
9010
9011
9012
9013
9014
9015
9016
9017
9018
9019
9020
9021
9022
9023
9024
9025
9026
9027
9028
9029
9030
9031
9032
9033
9034
9035
9036
9037
9038
9039
9040
9041
9042
9043
9044
9045
9046
9047
9048
9049
9050
9051
9052
9053
9054
9055
9056
9057
9058
9059
9060
9061
9062
9063
9064
9065
9066
9067
9068
9069
9070
9071
9072
9073
9074
9075
9076
9077
9078
9079
9080
9081
9082
9083
9084
9085
9086
9087
9088
9089
9090
9091
9092
9093
9094
9095
9096
9097
9098
9099
9100
9101
9102
9103
9104
9105
9106
9107
9108
9109
9110
9111
9112
9113
9114
9115
9116
9117
9118
9119
9120
9121
9122
9123
9124
9125
9126
9127
9128
9129
9130
9131
9132
9133
9134
9135
9136
9137
9138
9139
9140
9141
9142
9143
9144
9145
9146
9147
9148
9149
9150
9151
9152
9153
9154
9155
9156
9157
9158
9159
9160
9161
9162
9163
9164
9165
9166
9167
9168
9169
9170
9171
9172
9173
9174
9175
9176
9177
9178
9179
9180
9181
9182
9183
9184
9185
9186
9187
9188
9189
9190
9191
9192
9193
9194
9195
9196
9197
9198
9199
9200
9201
9202
9203
9204
9205
9206
9207
9208
9209
9210
9211
9212
9213
9214
9215
9216
9217
9218
9219
9220
9221
9222
9223
9224
9225
9226
9227
9228
9229
9230
9231
9232
9233
9234
9235
9236
9237
9238
9239
9240
9241
9242
9243
9244
9245
9246
9247
9248
9249
9250
9251
9252
9253
9254
9255
9256
9257
9258
9259
9260
9261
9262
9263
9264
9265
9266
9267
9268
9269
9270
9271
9272
9273
9274
9275
9276
9277
9278
9279
9280
9281
9282
9283
9284
9285
9286
9287
9288
9289
9290
9291
9292
9293
9294
9295
9296
9297
9298
9299
9300
9301
9302
9303
9304
9305
9306
9307
9308
9309
9310
9311
9312
9313
9314
9315
9316
9317
9318
9319
9320
9321
9322
9323
9324
9325
9326
9327
9328
9329
9330
9331
9332
9333
9334
9335
9336
9337
9338
9339
9340
9341
9342
9343
9344
9345
9346
9347
9348
9349
9350
9351
9352
9353
9354
9355
9356
9357
9358
9359
9360
9361
9362
9363
9364
9365
9366
9367
9368
9369
9370
9371
9372
9373
9374
9375
9376
9377
9378
9379
9380
9381
9382
9383
9384
9385
9386
9387
9388
9389
9390
9391
9392
9393
9394
9395
9396
9397
9398
9399
9400
9401
9402
9403
9404
9405
9406
9407
9408
9409
9410
9411
9412
9413
9414
9415
9416
9417
9418
9419
9420
9421
9422
9423
9424
9425
9426
9427
9428
9429
9430
9431
9432
9433
9434
9435
9436
9437
9438
9439
9440
9441
9442
9443
9444
9445
9446
9447
9448
9449
9450
9451
9452
9453
9454
9455
9456
9457
9458
9459
9460
9461
9462
9463
9464
9465
9466
9467
9468
9469
9470
9471
9472
9473
9474
9475
9476
9477
9478
9479
9480
9481
9482
9483
9484
9485
9486
9487
9488
9489
9490
9491
9492
9493
9494
9495
9496
9497
9498
9499
9500
9501
9502
9503
9504
9505
9506
9507
9508
9509
9510
9511
9512
9513
9514
9515
9516
9517
9518
9519
9520
9521
9522
9523
9524
9525
9526
9527
9528
9529
9530
9531
9532
9533
9534
9535
9536
9537
9538
9539
9540
9541
9542
// Written in the D programming language.

/++
    $(P The $(D std.uni) module provides an implementation
    of fundamental Unicode algorithms and data structures.
    This doesn't include UTF encoding and decoding primitives,
    see $(XREF _utf, decode) and $(XREF _utf, encode) in std.utf
    for this functionality. )

    $(P All primitives listed operate on Unicode characters and
    sets of characters. For functions which operate on ASCII characters
    and ignore Unicode $(CHARACTERS), see $(LINK2 std_ascii.html, std.ascii).
    For definitions of Unicode $(CHARACTER), $(CODEPOINT) and other terms
    used throughout this module see the $(S_LINK Terminology, terminology) section
    below.
    )

    $(P The focus of this module is the core needs of developing Unicode-aware
    applications. To that effect it provides the following optimized primitives:
    )
    $(UL
    $(LI Character classification by category and common properties:
        $(LREF isAlpha), $(LREF isWhite) and others.
    )
    $(LI
        Case-insensitive string comparison ($(LREF sicmp), $(LREF icmp)).
    )
    $(LI
        Converting text to any of the four normalization forms via $(LREF normalize).
    )
    $(LI
        Decoding ($(LREF decodeGrapheme))  and iteration ($(LREF byGrapheme), $(LREF graphemeStride))
        by user-perceived characters, that is by $(LREF Grapheme) clusters.
    )
    $(LI
        Decomposing and composing of individual character(s) according to canonical
        or compatibility rules, see $(LREF compose) and $(LREF decompose),
        including the specific version for Hangul syllables $(LREF composeJamo)
        and $(LREF decomposeHangul).
    )
    )
    $(P It's recognized that an application may need further enhancements
    and extensions, such as less commonly known algorithms,
    or tailoring existing ones for region specific needs. To help users
    with building any extra functionality beyond the core primitives,
    the module provides:
    )
    $(UL
    $(LI
        $(LREF CodepointSet), a type for easy manipulation of sets of characters.
        Besides the typical set algebra it provides an unusual feature:
        a D source code generator for detection of $(CODEPOINTS) in this set.
        This is a boon for meta-programming parser frameworks,
        and is used internally to power classification in small
        sets like $(LREF isWhite).
    )
    $(LI
        A way to construct optimal packed multi-stage tables also known as a
        special case of $(LUCKY Trie).
        The functions $(LREF codepointTrie), $(LREF codepointSetTrie)
        construct custom tries that map dchar to value.
        The end result is a fast and predictable $(BIGOH 1) lookup that powers
        functions like $(LREF isAlpha) and $(LREF combiningClass),
        but for user-defined data sets.
    )
    $(LI
        A useful technique for Unicode-aware parsers that perform
        character classification of encoded $(CODEPOINTS)
        is to avoid unnecassary decoding at all costs.
        $(LREF utfMatcher) provides an improvement over the usual workflow
        of decode-classify-process, combining the decoding and classification
        steps. By extracting necessary bits directly from encoded
        $(S_LINK Code unit, code units) matchers achieve
        significant performance improvements. See $(LREF MatcherConcept) for
        the common interface of UTF matchers.
    )
    $(LI
        Generally useful building blocks for customized normalization:
        $(LREF combiningClass) for querying combining class
        and $(LREF allowedIn) for testing the Quick_Check
        property of a given normalization form.
    )
    $(LI
        Access to a large selection of commonly used sets of $(CODEPOINTS).
        $(S_LINK Unicode properties, Supported sets) include Script,
        Block and General Category. The exact contents of a set can be
        observed in the CLDR utility, on the
        $(WEB www.unicode.org/cldr/utility/properties.jsp, property index) page
        of the Unicode website.
        See $(LREF unicode) for easy and (optionally) compile-time checked set
        queries.
    )
    )
    $(SECTION Synopsis)
    ---
    import std.uni;
    void main()
    {
        // initialize code point sets using script/block or property name
        // now 'set' contains code points from both scripts.
        auto set = unicode("Cyrillic") | unicode("Armenian");
        // same thing but simpler and checked at compile-time
        auto ascii = unicode.ASCII;
        auto currency = unicode.Currency_Symbol;

        // easy set ops
        auto a = set & ascii;
        assert(a.empty); // as it has no intersection with ascii
        a = set | ascii;
        auto b = currency - a; // subtract all ASCII, Cyrillic and Armenian

        // some properties of code point sets
        assert(b.length > 45); // 46 items in Unicode 6.1, even more in 6.2
        // testing presence of a code point in a set
        // is just fine, it is O(logN)
        assert(!b['$']);
        assert(!b['\u058F']); // Armenian dram sign
        assert(b['¥']);

        // building fast lookup tables, these guarantee O(1) complexity
        // 1-level Trie lookup table essentially a huge bit-set ~262Kb
        auto oneTrie = toTrie!1(b);
        // 2-level far more compact but typically slightly slower
        auto twoTrie = toTrie!2(b);
        // 3-level even smaller, and a bit slower yet
        auto threeTrie = toTrie!3(b);
        assert(oneTrie['£']);
        assert(twoTrie['£']);
        assert(threeTrie['£']);

        // build the trie with the most sensible trie level
        // and bind it as a functor
        auto cyrillicOrArmenian = toDelegate(set);
        auto balance = find!(cyrillicOrArmenian)("Hello ընկեր!");
        assert(balance == "ընկեր!");
        // compatible with bool delegate(dchar)
        bool delegate(dchar) bindIt = cyrillicOrArmenian;

        // Normalization
        string s = "Plain ascii (and not only), is always normalized!";
        assert(s is normalize(s));// is the same string

        string nonS = "A\u0308ffin"; // A ligature
        auto nS = normalize(nonS); // to NFC, the W3C endorsed standard
        assert(nS == "Äffin");
        assert(nS != nonS);
        string composed = "Äffin";

        assert(normalize!NFD(composed) == "A\u0308ffin");
        // to NFKD, compatibility decomposition useful for fuzzy matching/searching
        assert(normalize!NFKD("2¹⁰") == "210");
    }
    ---
    $(SECTION Terminology)
    $(P The following is a list of important Unicode notions
    and definitions. Any conventions used specifically in this
    module alone are marked as such. The descriptions are based on the formal
    definition as found in $(WEB www.unicode.org/versions/Unicode6.2.0/ch03.pdf,
    chapter three of The Unicode Standard Core Specification.)
    )

    $(P $(DEF Abstract character) A unit of information used for the organization,
    control, or representation of textual data.
    Note that:
        $(UL
        $(LI When representing data, the nature of that data
        is generally symbolic as opposed to some other
        kind of data (for example, visual).)

        $(LI An abstract character has no concrete form
        and should not be confused with a $(S_LINK Glyph, glyph).)

        $(LI An abstract character does not necessarily
        correspond to what a user thinks of as a “character”
         and should not be confused with a $(LREF Grapheme).)

        $(LI The abstract characters encoded (see Encoded character)
        are known as Unicode abstract characters.)

        $(LI Abstract characters not directly
        encoded by the Unicode Standard can often be
        represented by the use of combining character sequences.)
        )
    )

    $(P $(DEF Canonical decomposition)
    The decomposition of a character or character sequence
    that results from recursively applying the canonical
    mappings found in the Unicode Character Database
    and these described in Conjoining Jamo Behavior
    (section 12 of
    $(WEB www.unicode.org/uni2book/ch03.pdf, Unicode Conformance)).
    )

    $(P $(DEF Canonical composition)
    The precise definition of the Canonical composition
    is the algorithm as specified in $(WEB www.unicode.org/uni2book/ch03.pdf,
    Unicode Conformance) section 11.
    Informally it's the process that does the reverse of the canonical
    decomposition with the addition of certain rules
    that e.g. prevent legacy characters from appearing in the composed result.
    )

    $(P $(DEF Canonical equivalent)
    Two character sequences are said to be canonical equivalents if
    their full canonical decompositions are identical.
    )

    $(P $(DEF Character) Typically differs by context.
    For the purpose of this documentation the term $(I character)
    implies $(I encoded character), that is, a code point having
    an assigned abstract character (a symbolic meaning).
    )

    $(P $(DEF Code point) Any value in the Unicode codespace;
    that is, the range of integers from 0 to 10FFFF (hex).
    Not all code points are assigned to encoded characters.
    )

    $(P $(DEF Code unit) The minimal bit combination that can represent
    a unit of encoded text for processing or interchange.
    Depending on the encoding this could be:
    8-bit code units in the UTF-8 ($(D char)),
    16-bit code units in the UTF-16 ($(D wchar)),
    and 32-bit code units in the UTF-32 ($(D dchar)).
    $(I Note that in UTF-32, a code unit is a code point
    and is represented by the D $(D dchar) type.)
    )

    $(P $(DEF Combining character) A character with the General Category
     of Combining Mark(M).
        $(UL
            $(LI All characters with non-zero canonical combining class
            are combining characters, but the reverse is not the case:
            there are combining characters with a zero combining class.
            )
            $(LI These characters are not normally used in isolation
            unless they are being described. They include such characters
            as accents, diacritics, Hebrew points, Arabic vowel signs,
            and Indic matras.
            )
        )
    )

    $(P $(DEF Combining class)
        A numerical value used by the Unicode Canonical Ordering Algorithm
        to determine which sequences of combining marks are to be
        considered canonically equivalent and  which are not.
    )

    $(P $(DEF Compatibility decomposition)
    The decomposition of a character or character sequence that results
    from recursively applying both the compatibility mappings and
    the canonical mappings found in the Unicode Character Database, and those
    described in Conjoining Jamo Behavior no characters
    can be further decomposed.
    )

    $(P $(DEF Compatibility equivalent)
    Two character sequences are said to be compatibility
    equivalents if their full compatibility decompositions are identical.
    )

    $(P $(DEF Encoded character) An association (or mapping)
    between an abstract character and a code point.
    )

    $(P $(DEF Glyph) The actual, concrete image of a glyph representation
    having been rasterized or otherwise imaged onto some display surface.
    )

    $(P $(DEF Grapheme base) A character with the property
     Grapheme_Base, or any standard Korean syllable block.
    )

    $(P $(DEF Grapheme cluster) Defined as the text between
        grapheme boundaries  as specified by Unicode Standard Annex #29,
        $(WEB www.unicode.org/reports/tr29/, Unicode text segmentation).
        Important general properties of a grapheme:
        $(UL
            $(LI The grapheme cluster represents a horizontally segmentable
            unit of text, consisting of some grapheme base (which may
            consist of a Korean syllable) together with any number of
            nonspacing marks applied to it.
            )
            $(LI  A grapheme cluster typically starts with a grapheme base
            and then extends across any subsequent sequence of nonspacing marks.
            A grapheme cluster is most directly relevant to text rendering and
            processes such as cursor placement and text selection in editing,
            but may also be relevant to comparison and searching.
            )
            $(LI For many processes, a grapheme cluster behaves as if it was a
            single character with the same properties as its grapheme base.
            Effectively, nonspacing marks apply $(I graphically) to the base,
            but do not change its properties.
            )
        )
        $(P This module defines a number of primitives that work with graphemes:
        $(LREF Grapheme), $(LREF decodeGrapheme) and $(LREF graphemeStride).
        All of them are using $(I extended grapheme) boundaries
        as defined in the aforementioned standard annex.
        )
    )


    $(P $(DEF Nonspacing mark) A combining character with the
        General Category of Nonspacing Mark (Mn) or Enclosing Mark (Me).
    )

    $(P $(DEF Spacing mark) A combining character that is not a nonspacing mark.)


    $(SECTION Normalization)

    $(P The concepts of $(S_LINK Canonical equivalent, canonical equivalent)
     or $(S_LINK Compatibility equivalent, compatibility equivalent)
    characters in the Unicode Standard make it necessary to have a full, formal
    definition of equivalence for Unicode strings.
    String equivalence is determined by a process called normalization,
    whereby strings are converted into forms which are compared
    directly for identity. This is the primary goal of the normalization process,
    see the function $(LREF normalize) to convert into any of
    the four defined forms.
    )

    $(P A very important attribute of the Unicode Normalization Forms
    is that they must remain stable between versions of the Unicode Standard.
    A Unicode string normalized to a particular Unicode Normalization Form
    in one version of the standard is guaranteed to remain in that Normalization
    Form for implementations of future versions of the standard.
    )

    $(P The Unicode Standard specifies four normalization forms.
    Informally, two of these forms are defined by maximal decomposition
    of equivalent sequences, and two of these forms are defined
    by maximal $(I composition) of equivalent sequences.
        $(UL
        $(LI Normalization Form D (NFD): The $(S_LINK Canonical decomposition,
            canonical decomposition) of a character sequence.)
        $(LI Normalization Form KD (NFKD): The $(S_LINK Compatibility decomposition,
            compatibility decomposition) of a character sequence.)
        $(LI Normalization Form C (NFC): The canonical composition of the
            $(S_LINK Canonical decomposition, canonical decomposition)
            of a coded character sequence.)
        $(LI Normalization Form KC (NFKC): The canonical composition
        of the $(S_LINK Compatibility decomposition,
            compatibility decomposition) of a character sequence)
        )
    )

    $(P The choice of the normalization form depends on the particular use case.
    NFC is the best form for general text, since it's more compatible with
    strings converted from legacy encodings. NFKC is the preferred form for
    identifiers, especially where there are security concerns. NFD and NFKD
    are the most useful for internal processing.
    )

    $(SECTION Construction of lookup tables)

    $(P The Unicode standard describes a set of algorithms that
    depend on having the ability to quickly look up various properties
    of a code point. Given the the codespace of about 1 million $(CODEPOINTS),
    it is not a trivial task to provide a space-efficient solution for
    the multitude of properties.)

    $(P Common approaches such as hash-tables or binary search over
     sorted code point intervals (as in $(LREF InversionList)) are insufficient.
     Hash-tables have enormous memory footprint and binary search
     over intervals is not fast enough for some heavy-duty algorithms.
     )

    $(P The recommended solution (see Unicode Implementation Guidelines)
    is using multi-stage tables that are an implementation of the
    $(WEB en.wikipedia.org/wiki/Trie, Trie) data structure with integer
    keys and a fixed number of stages. For the remainder of the section
    this will be called a fixed trie. The following describes a particular
    implementation that is aimed for the speed of access at the expense
    of ideal size savings.
    )

    $(P Taking a 2-level Trie as an example the principle of operation is as follows.
        Split the number of bits in a key (code point, 21 bits) into 2 components
        (e.g. 15 and 8).  The first is the number of bits in the index of the trie
         and the other is number of bits in each page of the trie.
        The layout of the trie is then an array of size 2^^bits-of-index followed
        an array of memory chunks of size 2^^bits-of-page/bits-per-element.
    )

    $(P The number of pages is variable (but not less then 1)
        unlike the number of entries in the index. The slots of the index
        all have to contain a number of a page that is present. The lookup is then
        just a couple of operations - slice the upper bits,
        lookup an index for these, take a page at this index and use
        the lower bits as an offset within this page.

        Assuming that pages are laid out consequently
        in one array at $(D pages), the pseudo-code is:
    )
    ---
    auto elemsPerPage = (2 ^^ bits_per_page) / Value.sizeOfInBits;
    pages[index[n >> bits_per_page]][n & (elemsPerPage - 1)];
    ---
    $(P Where if $(D elemsPerPage) is a power of 2 the whole process is
    a handful of simple instructions and 2 array reads. Subsequent levels
    of the trie are introduced by recursing on this notion - the index array
    is treated as values. The number of bits in index is then again
    split into 2 parts, with pages over 'current-index' and the new 'upper-index'.
    )

    $(P For completeness a level 1 trie is simply an array.
    The current implementation takes advantage of bit-packing values
    when the range is known to be limited in advance (such as $(D bool)).
    See also $(LREF BitPacked) for enforcing it manually.
    The major size advantage however comes from the fact
    that multiple $(B identical pages on every level are merged) by construction.
    )

    $(P The process of constructing a trie is more involved and is hidden from
    the user in a form of the convenience functions $(LREF codepointTrie),
    $(LREF codepointSetTrie) and the even more convenient $(LREF toTrie).
    In general a set or built-in AA with $(D dchar) type
    can be turned into a trie. The trie object in this module
    is read-only (immutable); it's effectively frozen after construction.
    )

    $(SECTION Unicode properties)

    $(P This is a full list of Unicode properties accessible through $(LREF unicode)
    with specific helpers per category nested within. Consult the
    $(WEB www.unicode.org/cldr/utility/properties.jsp, CLDR utility)
    when in doubt about the contents of a particular set.)

    $(P General category sets listed below are only accessible with the
    $(LREF unicode) shorthand accessor.)
    $(BOOKTABLE $(B General category ),
         $(TR $(TH Abb.) $(TH Long form)
            $(TH Abb.) $(TH Long form)$(TH Abb.) $(TH Long form))
        $(TR $(TD L) $(TD Letter)
            $(TD Cn) $(TD Unassigned)  $(TD Po) $(TD Other_Punctuation))
        $(TR $(TD Ll) $(TD Lowercase_Letter)
            $(TD Co) $(TD Private_Use) $(TD Ps) $(TD Open_Punctuation))
        $(TR $(TD Lm) $(TD Modifier_Letter)
            $(TD Cs) $(TD Surrogate)   $(TD S) $(TD Symbol))
        $(TR $(TD Lo) $(TD Other_Letter)
            $(TD N) $(TD Number)  $(TD Sc) $(TD Currency_Symbol))
        $(TR $(TD Lt) $(TD Titlecase_Letter)
          $(TD Nd) $(TD Decimal_Number)  $(TD Sk) $(TD Modifier_Symbol))
        $(TR $(TD Lu) $(TD Uppercase_Letter)
          $(TD Nl) $(TD Letter_Number)   $(TD Sm) $(TD Math_Symbol))
        $(TR $(TD M) $(TD Mark)
          $(TD No) $(TD Other_Number)    $(TD So) $(TD Other_Symbol))
        $(TR $(TD Mc) $(TD Spacing_Mark)
          $(TD P) $(TD Punctuation) $(TD Z) $(TD Separator))
        $(TR $(TD Me) $(TD Enclosing_Mark)
          $(TD Pc) $(TD Connector_Punctuation)   $(TD Zl) $(TD Line_Separator))
        $(TR $(TD Mn) $(TD Nonspacing_Mark)
          $(TD Pd) $(TD Dash_Punctuation)    $(TD Zp) $(TD Paragraph_Separator))
        $(TR $(TD C) $(TD Other)
          $(TD Pe) $(TD Close_Punctuation) $(TD Zs) $(TD Space_Separator))
        $(TR $(TD Cc) $(TD Control) $(TD Pf)
          $(TD Final_Punctuation)   $(TD -) $(TD Any))
        $(TR $(TD Cf) $(TD Format)
          $(TD Pi) $(TD Initial_Punctuation) $(TD -) $(TD ASCII))
    )
    $(P Sets for other commonly useful properties that are
    accessible with $(LREF unicode):)
    $(BOOKTABLE $(B Common binary properties),
        $(TR $(TH Name) $(TH Name) $(TH Name))
        $(TR $(TD Alphabetic)  $(TD Ideographic) $(TD Other_Uppercase))
        $(TR $(TD ASCII_Hex_Digit) $(TD IDS_Binary_Operator) $(TD Pattern_Syntax))
        $(TR $(TD Bidi_Control)    $(TD ID_Start)    $(TD Pattern_White_Space))
        $(TR $(TD Cased)   $(TD IDS_Trinary_Operator)    $(TD Quotation_Mark))
        $(TR $(TD Case_Ignorable)  $(TD Join_Control)    $(TD Radical))
        $(TR $(TD Dash)    $(TD Logical_Order_Exception) $(TD Soft_Dotted))
        $(TR $(TD Default_Ignorable_Code_Point)    $(TD Lowercase)   $(TD STerm))
        $(TR $(TD Deprecated)  $(TD Math)    $(TD Terminal_Punctuation))
        $(TR $(TD Diacritic)   $(TD Noncharacter_Code_Point) $(TD Unified_Ideograph))
        $(TR $(TD Extender)    $(TD Other_Alphabetic)    $(TD Uppercase))
        $(TR $(TD Grapheme_Base)   $(TD Other_Default_Ignorable_Code_Point)  $(TD Variation_Selector))
        $(TR $(TD Grapheme_Extend) $(TD Other_Grapheme_Extend)   $(TD White_Space))
        $(TR $(TD Grapheme_Link)   $(TD Other_ID_Continue)   $(TD XID_Continue))
        $(TR $(TD Hex_Digit)   $(TD Other_ID_Start)  $(TD XID_Start))
        $(TR $(TD Hyphen)  $(TD Other_Lowercase) )
        $(TR $(TD ID_Continue) $(TD Other_Math)  )
    )
    $(P Bellow is the table with block names accepted by $(LREF unicode.block).
    Note that the shorthand version $(LREF unicode) requires "In"
    to be prepended to the names of blocks so as to disambiguate
    scripts and blocks.)

    $(BOOKTABLE $(B Blocks),
        $(TR $(TD Aegean Numbers)    $(TD Ethiopic Extended) $(TD Mongolian))
        $(TR $(TD Alchemical Symbols)    $(TD Ethiopic Extended-A)   $(TD Musical Symbols))
        $(TR $(TD Alphabetic Presentation Forms) $(TD Ethiopic Supplement)   $(TD Myanmar))
        $(TR $(TD Ancient Greek Musical Notation)    $(TD General Punctuation)   $(TD Myanmar Extended-A))
        $(TR $(TD Ancient Greek Numbers) $(TD Geometric Shapes)  $(TD New Tai Lue))
        $(TR $(TD Ancient Symbols)   $(TD Georgian)  $(TD NKo))
        $(TR $(TD Arabic)    $(TD Georgian Supplement)   $(TD Number Forms))
        $(TR $(TD Arabic Extended-A) $(TD Glagolitic)    $(TD Ogham))
        $(TR $(TD Arabic Mathematical Alphabetic Symbols)    $(TD Gothic)    $(TD Ol Chiki))
        $(TR $(TD Arabic Presentation Forms-A)   $(TD Greek and Coptic)  $(TD Old Italic))
        $(TR $(TD Arabic Presentation Forms-B)   $(TD Greek Extended)    $(TD Old Persian))
        $(TR $(TD Arabic Supplement) $(TD Gujarati)  $(TD Old South Arabian))
        $(TR $(TD Armenian)  $(TD Gurmukhi)  $(TD Old Turkic))
        $(TR $(TD Arrows)    $(TD Halfwidth and Fullwidth Forms) $(TD Optical Character Recognition))
        $(TR $(TD Avestan)   $(TD Hangul Compatibility Jamo) $(TD Oriya))
        $(TR $(TD Balinese)  $(TD Hangul Jamo)   $(TD Osmanya))
        $(TR $(TD Bamum) $(TD Hangul Jamo Extended-A)    $(TD Phags-pa))
        $(TR $(TD Bamum Supplement)  $(TD Hangul Jamo Extended-B)    $(TD Phaistos Disc))
        $(TR $(TD Basic Latin)   $(TD Hangul Syllables)  $(TD Phoenician))
        $(TR $(TD Batak) $(TD Hanunoo)   $(TD Phonetic Extensions))
        $(TR $(TD Bengali)   $(TD Hebrew)    $(TD Phonetic Extensions Supplement))
        $(TR $(TD Block Elements)    $(TD High Private Use Surrogates)   $(TD Playing Cards))
        $(TR $(TD Bopomofo)  $(TD High Surrogates)   $(TD Private Use Area))
        $(TR $(TD Bopomofo Extended) $(TD Hiragana)  $(TD Rejang))
        $(TR $(TD Box Drawing)   $(TD Ideographic Description Characters)    $(TD Rumi Numeral Symbols))
        $(TR $(TD Brahmi)    $(TD Imperial Aramaic)  $(TD Runic))
        $(TR $(TD Braille Patterns)  $(TD Inscriptional Pahlavi) $(TD Samaritan))
        $(TR $(TD Buginese)  $(TD Inscriptional Parthian)    $(TD Saurashtra))
        $(TR $(TD Buhid) $(TD IPA Extensions)    $(TD Sharada))
        $(TR $(TD Byzantine Musical Symbols) $(TD Javanese)  $(TD Shavian))
        $(TR $(TD Carian)    $(TD Kaithi)    $(TD Sinhala))
        $(TR $(TD Chakma)    $(TD Kana Supplement)   $(TD Small Form Variants))
        $(TR $(TD Cham)  $(TD Kanbun)    $(TD Sora Sompeng))
        $(TR $(TD Cherokee)  $(TD Kangxi Radicals)   $(TD Spacing Modifier Letters))
        $(TR $(TD CJK Compatibility) $(TD Kannada)   $(TD Specials))
        $(TR $(TD CJK Compatibility Forms)   $(TD Katakana)  $(TD Sundanese))
        $(TR $(TD CJK Compatibility Ideographs)  $(TD Katakana Phonetic Extensions)  $(TD Sundanese Supplement))
        $(TR $(TD CJK Compatibility Ideographs Supplement)   $(TD Kayah Li)  $(TD Superscripts and Subscripts))
        $(TR $(TD CJK Radicals Supplement)   $(TD Kharoshthi)    $(TD Supplemental Arrows-A))
        $(TR $(TD CJK Strokes)   $(TD Khmer) $(TD Supplemental Arrows-B))
        $(TR $(TD CJK Symbols and Punctuation)   $(TD Khmer Symbols) $(TD Supplemental Mathematical Operators))
        $(TR $(TD CJK Unified Ideographs)    $(TD Lao)   $(TD Supplemental Punctuation))
        $(TR $(TD CJK Unified Ideographs Extension A)    $(TD Latin-1 Supplement)    $(TD Supplementary Private Use Area-A))
        $(TR $(TD CJK Unified Ideographs Extension B)    $(TD Latin Extended-A)  $(TD Supplementary Private Use Area-B))
        $(TR $(TD CJK Unified Ideographs Extension C)    $(TD Latin Extended Additional) $(TD Syloti Nagri))
        $(TR $(TD CJK Unified Ideographs Extension D)    $(TD Latin Extended-B)  $(TD Syriac))
        $(TR $(TD Combining Diacritical Marks)   $(TD Latin Extended-C)  $(TD Tagalog))
        $(TR $(TD Combining Diacritical Marks for Symbols)   $(TD Latin Extended-D)  $(TD Tagbanwa))
        $(TR $(TD Combining Diacritical Marks Supplement)    $(TD Lepcha)    $(TD Tags))
        $(TR $(TD Combining Half Marks)  $(TD Letterlike Symbols)    $(TD Tai Le))
        $(TR $(TD Common Indic Number Forms) $(TD Limbu) $(TD Tai Tham))
        $(TR $(TD Control Pictures)  $(TD Linear B Ideograms)    $(TD Tai Viet))
        $(TR $(TD Coptic)    $(TD Linear B Syllabary)    $(TD Tai Xuan Jing Symbols))
        $(TR $(TD Counting Rod Numerals) $(TD Lisu)  $(TD Takri))
        $(TR $(TD Cuneiform) $(TD Low Surrogates)    $(TD Tamil))
        $(TR $(TD Cuneiform Numbers and Punctuation) $(TD Lycian)    $(TD Telugu))
        $(TR $(TD Currency Symbols)  $(TD Lydian)    $(TD Thaana))
        $(TR $(TD Cypriot Syllabary) $(TD Mahjong Tiles) $(TD Thai))
        $(TR $(TD Cyrillic)  $(TD Malayalam) $(TD Tibetan))
        $(TR $(TD Cyrillic Extended-A)   $(TD Mandaic)   $(TD Tifinagh))
        $(TR $(TD Cyrillic Extended-B)   $(TD Mathematical Alphanumeric Symbols) $(TD Transport And Map Symbols))
        $(TR $(TD Cyrillic Supplement)   $(TD Mathematical Operators)    $(TD Ugaritic))
        $(TR $(TD Deseret)   $(TD Meetei Mayek)  $(TD Unified Canadian Aboriginal Syllabics))
        $(TR $(TD Devanagari)    $(TD Meetei Mayek Extensions)   $(TD Unified Canadian Aboriginal Syllabics Extended))
        $(TR $(TD Devanagari Extended)   $(TD Meroitic Cursive)  $(TD Vai))
        $(TR $(TD Dingbats)  $(TD Meroitic Hieroglyphs)  $(TD Variation Selectors))
        $(TR $(TD Domino Tiles)  $(TD Miao)  $(TD Variation Selectors Supplement))
        $(TR $(TD Egyptian Hieroglyphs)  $(TD Miscellaneous Mathematical Symbols-A)  $(TD Vedic Extensions))
        $(TR $(TD Emoticons) $(TD Miscellaneous Mathematical Symbols-B)  $(TD Vertical Forms))
        $(TR $(TD Enclosed Alphanumerics)    $(TD Miscellaneous Symbols) $(TD Yijing Hexagram Symbols))
        $(TR $(TD Enclosed Alphanumeric Supplement)  $(TD Miscellaneous Symbols and Arrows)  $(TD Yi Radicals))
        $(TR $(TD Enclosed CJK Letters and Months)   $(TD Miscellaneous Symbols And Pictographs) $(TD Yi Syllables))
        $(TR $(TD Enclosed Ideographic Supplement)   $(TD Miscellaneous Technical)   )
        $(TR $(TD Ethiopic)  $(TD Modifier Tone Letters) )
    )

    $(P Bellow is the table with script names accepted by $(LREF unicode.script)
    and by the shorthand version $(LREF unicode):)
    $(BOOKTABLE $(B Scripts),
        $(TR $(TD Arabic)  $(TD Hanunoo) $(TD Old_Italic))
        $(TR $(TD Armenian)    $(TD Hebrew)  $(TD Old_Persian))
        $(TR $(TD Avestan) $(TD Hiragana)    $(TD Old_South_Arabian))
        $(TR $(TD Balinese)    $(TD Imperial_Aramaic)    $(TD Old_Turkic))
        $(TR $(TD Bamum)   $(TD Inherited)   $(TD Oriya))
        $(TR $(TD Batak)   $(TD Inscriptional_Pahlavi)   $(TD Osmanya))
        $(TR $(TD Bengali) $(TD Inscriptional_Parthian)  $(TD Phags_Pa))
        $(TR $(TD Bopomofo)    $(TD Javanese)    $(TD Phoenician))
        $(TR $(TD Brahmi)  $(TD Kaithi)  $(TD Rejang))
        $(TR $(TD Braille) $(TD Kannada) $(TD Runic))
        $(TR $(TD Buginese)    $(TD Katakana)    $(TD Samaritan))
        $(TR $(TD Buhid)   $(TD Kayah_Li)    $(TD Saurashtra))
        $(TR $(TD Canadian_Aboriginal) $(TD Kharoshthi)  $(TD Sharada))
        $(TR $(TD Carian)  $(TD Khmer)   $(TD Shavian))
        $(TR $(TD Chakma)  $(TD Lao) $(TD Sinhala))
        $(TR $(TD Cham)    $(TD Latin)   $(TD Sora_Sompeng))
        $(TR $(TD Cherokee)    $(TD Lepcha)  $(TD Sundanese))
        $(TR $(TD Common)  $(TD Limbu)   $(TD Syloti_Nagri))
        $(TR $(TD Coptic)  $(TD Linear_B)    $(TD Syriac))
        $(TR $(TD Cuneiform)   $(TD Lisu)    $(TD Tagalog))
        $(TR $(TD Cypriot) $(TD Lycian)  $(TD Tagbanwa))
        $(TR $(TD Cyrillic)    $(TD Lydian)  $(TD Tai_Le))
        $(TR $(TD Deseret) $(TD Malayalam)   $(TD Tai_Tham))
        $(TR $(TD Devanagari)  $(TD Mandaic) $(TD Tai_Viet))
        $(TR $(TD Egyptian_Hieroglyphs)    $(TD Meetei_Mayek)    $(TD Takri))
        $(TR $(TD Ethiopic)    $(TD Meroitic_Cursive)    $(TD Tamil))
        $(TR $(TD Georgian)    $(TD Meroitic_Hieroglyphs)    $(TD Telugu))
        $(TR $(TD Glagolitic)  $(TD Miao)    $(TD Thaana))
        $(TR $(TD Gothic)  $(TD Mongolian)   $(TD Thai))
        $(TR $(TD Greek)   $(TD Myanmar) $(TD Tibetan))
        $(TR $(TD Gujarati)    $(TD New_Tai_Lue) $(TD Tifinagh))
        $(TR $(TD Gurmukhi)    $(TD Nko) $(TD Ugaritic))
        $(TR $(TD Han) $(TD Ogham)   $(TD Vai))
        $(TR $(TD Hangul)  $(TD Ol_Chiki)    $(TD Yi))
    )

    $(P Bellow is the table of names accepted by $(LREF unicode.hangulSyllableType).)
    $(BOOKTABLE $(B Hangul syllable type),
        $(TR $(TH Abb.) $(TH Long form))
        $(TR $(TD L)   $(TD Leading_Jamo))
        $(TR $(TD LV)  $(TD LV_Syllable))
        $(TR $(TD LVT) $(TD LVT_Syllable) )
        $(TR $(TD T)   $(TD Trailing_Jamo))
        $(TR $(TD V)   $(TD Vowel_Jamo))
    )
    References:
        $(WEB www.digitalmars.com/d/ascii-table.html, ASCII Table),
        $(WEB en.wikipedia.org/wiki/Unicode, Wikipedia),
        $(WEB www.unicode.org, The Unicode Consortium),
        $(WEB www.unicode.org/reports/tr15/, Unicode normalization forms),
        $(WEB www.unicode.org/reports/tr29/, Unicode text segmentation)
        $(WEB www.unicode.org/uni2book/ch05.pdf,
            Unicode Implementation Guidelines)
        $(WEB www.unicode.org/uni2book/ch03.pdf,
            Unicode Conformance)
    Trademarks:
        Unicode(tm) is a trademark of Unicode, Inc.

    Macros:
        WIKI=Phobos/StdUni

    Copyright: Copyright 2013 -
    License:   $(WEB www.boost.org/LICENSE_1_0.txt, Boost License 1.0).
    Authors:   Dmitry Olshansky
    Source:    $(PHOBOSSRC std/_uni.d)
    Standards: $(WEB www.unicode.org/versions/Unicode6.2.0/, Unicode v6.2)

Macros:

SECTION = <h3><a id="$1">$0</a></h3>
DEF = <div><a id="$1"><i>$0</i></a></div>
S_LINK = <a href="#$1">$+</a>
CODEPOINT = $(S_LINK Code point, code point)
CODEPOINTS = $(S_LINK Code point, code points)
CHARACTER = $(S_LINK Character, character)
CHARACTERS = $(S_LINK Character, characters)
CLUSTER = $(S_LINK Grapheme cluster, grapheme cluster)
+/
module std.uni;

import core.stdc.stdlib;
import std.traits, std.typetuple;
import std.range.primitives;


// debug = std_uni;

debug(std_uni) import std.stdio;

private:

version(std_uni_bootstrap){}
else
{
    import std.internal.unicode_tables; // generated file
}

void copyBackwards(T,U)(T[] src, U[] dest)
{
    assert(src.length == dest.length);
    for(size_t i=src.length; i-- > 0; )
        dest[i] = src[i];
}

void copyForward(T,U)(T[] src, U[] dest)
{
    assert(src.length == dest.length);
    for(size_t i=0; i<src.length; i++)
        dest[i] = src[i];
}

// TODO: update to reflect all major CPUs supporting unaligned reads
version(X86)
    enum hasUnalignedReads = true;
else version(X86_64)
    enum hasUnalignedReads = true;
else
    enum hasUnalignedReads = false; // better be safe then sorry

public enum dchar lineSep = '\u2028'; /// Constant $(CODEPOINT) (0x2028) - line separator.
public enum dchar paraSep = '\u2029'; /// Constant $(CODEPOINT) (0x2029) - paragraph separator.
public enum dchar nelSep  = '\u0085'; /// Constant $(CODEPOINT) (0x0085) - next line.

// test the intro example
@safe unittest
{
    import std.algorithm : find;
    // initialize code point sets using script/block or property name
    // set contains code points from both scripts.
    auto set = unicode("Cyrillic") | unicode("Armenian");
    // or simpler and statically-checked look
    auto ascii = unicode.ASCII;
    auto currency = unicode.Currency_Symbol;

    // easy set ops
    auto a = set & ascii;
    assert(a.empty); // as it has no intersection with ascii
    a = set | ascii;
    auto b = currency - a; // subtract all ASCII, Cyrillic and Armenian

    // some properties of code point sets
    assert(b.length > 45); // 46 items in Unicode 6.1, even more in 6.2
    // testing presence of a code point in a set
    // is just fine, it is O(logN)
    assert(!b['$']);
    assert(!b['\u058F']); // Armenian dram sign
    assert(b['¥']);

    // building fast lookup tables, these guarantee O(1) complexity
    // 1-level Trie lookup table essentially a huge bit-set ~262Kb
    auto oneTrie = toTrie!1(b);
    // 2-level far more compact but typically slightly slower
    auto twoTrie = toTrie!2(b);
    // 3-level even smaller, and a bit slower yet
    auto threeTrie = toTrie!3(b);
    assert(oneTrie['£']);
    assert(twoTrie['£']);
    assert(threeTrie['£']);

    // build the trie with the most sensible trie level
    // and bind it as a functor
    auto cyrillicOrArmenian = toDelegate(set);
    auto balance = find!(cyrillicOrArmenian)("Hello ընկեր!");
    assert(balance == "ընկեր!");
    // compatible with bool delegate(dchar)
    bool delegate(dchar) bindIt = cyrillicOrArmenian;

    // Normalization
    string s = "Plain ascii (and not only), is always normalized!";
    assert(s is normalize(s));// is the same string

    string nonS = "A\u0308ffin"; // A ligature
    auto nS = normalize(nonS); // to NFC, the W3C endorsed standard
    assert(nS == "Äffin");
    assert(nS != nonS);
    string composed = "Äffin";

    assert(normalize!NFD(composed) == "A\u0308ffin");
    // to NFKD, compatibility decomposition useful for fuzzy matching/searching
    assert(normalize!NFKD("2¹⁰") == "210");
}

enum lastDchar = 0x10FFFF;

auto force(T, F)(F from)
    if(isIntegral!T && !is(T == F))
{
    assert(from <= T.max && from >= T.min);
    return cast(T)from;
}

auto force(T, F)(F from)
    if(isBitPacked!T && !is(T == F))
{
    assert(from <= 2^^bitSizeOf!T-1);
    return T(cast(TypeOfBitPacked!T)from);
}

auto force(T, F)(F from)
    if(is(T == F))
{
    return from;
}

// repeat X times the bit-pattern in val assuming it's length is 'bits'
size_t replicateBits(size_t times, size_t bits)(size_t val) @safe pure nothrow @nogc
{
    static if(times == 1)
        return val;
    else static if(bits == 1)
    {
        static if(times == size_t.sizeof*8)
            return val ? size_t.max : 0;
        else
            return val ? (1<<times)-1 : 0;
    }
    else static if(times % 2)
        return (replicateBits!(times-1, bits)(val)<<bits) | val;
    else
        return replicateBits!(times/2, bits*2)((val<<bits) | val);
}

@safe pure nothrow @nogc unittest // for replicate
{
    import std.algorithm : sum, map;
    import std.range : iota;
    size_t m = 0b111;
    size_t m2 = 0b01;
    foreach(i; TypeTuple!(1, 2, 3, 4, 5, 6, 7, 8, 9, 10))
    {
        assert(replicateBits!(i, 3)(m)+1 == (1<<(3*i)));
        assert(replicateBits!(i, 2)(m2) == iota(0, i).map!"2^^(2*a)"().sum());
    }
}

// multiple arrays squashed into one memory block
struct MultiArray(Types...)
{
    this(size_t[] sizes...) @safe pure nothrow
    {
        assert(dim == sizes.length);
        size_t full_size;
        foreach(i, v; Types)
        {
            full_size += spaceFor!(bitSizeOf!v)(sizes[i]);
            sz[i] = sizes[i];
            static if(i >= 1)
                offsets[i] = offsets[i-1] +
                    spaceFor!(bitSizeOf!(Types[i-1]))(sizes[i-1]);
        }

        storage = new size_t[full_size];
    }

    this(const(size_t)[] raw_offsets,
        const(size_t)[] raw_sizes, const(size_t)[] data)const @safe pure nothrow @nogc
    {
        offsets[] = raw_offsets[];
        sz[] = raw_sizes[];
        storage = data;
    }

    @property auto slice(size_t n)()inout pure nothrow @nogc
    {
        auto ptr = raw_ptr!n;
        return packedArrayView!(Types[n])(ptr, sz[n]);
    }

    @property auto ptr(size_t n)()inout pure nothrow @nogc
    {
        auto ptr = raw_ptr!n;
        return inout(PackedPtr!(Types[n]))(ptr);
    }

    template length(size_t n)
    {
        @property size_t length()const @safe pure nothrow @nogc{ return sz[n]; }

        @property void length(size_t new_size)
        {
            if(new_size > sz[n])
            {// extend
                size_t delta = (new_size - sz[n]);
                sz[n] += delta;
                delta = spaceFor!(bitSizeOf!(Types[n]))(delta);
                storage.length +=  delta;// extend space at end
                // raw_slice!x must follow resize as it could be moved!
                // next stmts move all data past this array, last-one-goes-first
                static if(n != dim-1)
                {
                    auto start = raw_ptr!(n+1);
                    // len includes delta
                    size_t len = (storage.ptr+storage.length-start);

                    copyBackwards(start[0..len-delta], start[delta..len]);

                    start[0..delta] = 0;
                    // offsets are used for raw_slice, ptr etc.
                    foreach(i; n+1..dim)
                        offsets[i] += delta;
                }
            }
            else if(new_size < sz[n])
            {// shrink
                size_t delta = (sz[n] - new_size);
                sz[n] -= delta;
                delta = spaceFor!(bitSizeOf!(Types[n]))(delta);
                // move all data past this array, forward direction
                static if(n != dim-1)
                {
                    auto start = raw_ptr!(n+1);
                    size_t len = (storage.ptr+storage.length-start);
                    copyForward(start[0..len-delta], start[delta..len]);

                    // adjust offsets last, they affect raw_slice
                    foreach(i; n+1..dim)
                        offsets[i] -= delta;
                }
                storage.length -= delta;
            }
            // else - NOP
        }
    }

    @property size_t bytes(size_t n=size_t.max)() const
    {
        static if(n == size_t.max)
            return storage.length*size_t.sizeof;
        else static if(n != Types.length-1)
            return (raw_ptr!(n+1)-raw_ptr!n)*size_t.sizeof;
        else
            return (storage.ptr+storage.length - raw_ptr!n)*size_t.sizeof;
    }

    void store(OutRange)(scope OutRange sink) const
        if(isOutputRange!(OutRange, char))
    {
        import std.format;
        formattedWrite(sink, "[%( 0x%x, %)]", offsets[]);
        formattedWrite(sink, ", [%( 0x%x, %)]", sz[]);
        formattedWrite(sink, ", [%( 0x%x, %)]", storage);
    }

private:
    @property auto raw_ptr(size_t n)()inout
    {
        static if(n == 0)
            return storage.ptr;
        else
        {
            return storage.ptr+offsets[n];
        }
    }
    enum dim = Types.length;
    size_t[dim] offsets;// offset for level x
    size_t[dim] sz;// size of level x
    alias bitWidth = staticMap!(bitSizeOf, Types);
    size_t[] storage;
}

unittest
{
    import std.conv;
    enum dg = (){
        // sizes are:
        // lvl0: 3, lvl1 : 2, lvl2: 1
        auto m = MultiArray!(int, ubyte, int)(3,2,1);

        static void check(size_t k, T)(ref T m, int n)
        {
            foreach(i; 0..n)
                assert(m.slice!(k)[i] == i+1, text("level:",i," : ",m.slice!(k)[0..n]));
        }

        static void checkB(size_t k, T)(ref T m, int n)
        {
            foreach(i; 0..n)
                assert(m.slice!(k)[i] == n-i, text("level:",i," : ",m.slice!(k)[0..n]));
        }

        static void fill(size_t k, T)(ref T m, int n)
        {
            foreach(i; 0..n)
                m.slice!(k)[i] = force!ubyte(i+1);
        }

        static void fillB(size_t k, T)(ref T m, int n)
        {
            foreach(i; 0..n)
                m.slice!(k)[i] = force!ubyte(n-i);
        }

        m.length!1 = 100;
        fill!1(m, 100);
        check!1(m, 100);

        m.length!0 = 220;
        fill!0(m, 220);
        check!1(m, 100);
        check!0(m, 220);

        m.length!2 = 17;
        fillB!2(m, 17);
        checkB!2(m, 17);
        check!0(m, 220);
        check!1(m, 100);

        m.length!2 = 33;
        checkB!2(m, 17);
        fillB!2(m, 33);
        checkB!2(m, 33);
        check!0(m, 220);
        check!1(m, 100);

        m.length!1 = 195;
        fillB!1(m, 195);
        checkB!1(m, 195);
        checkB!2(m, 33);
        check!0(m, 220);

        auto marr = MultiArray!(BitPacked!(uint, 4), BitPacked!(uint, 6))(20, 10);
        marr.length!0 = 15;
        marr.length!1 = 30;
        fill!1(marr, 30);
        fill!0(marr, 15);
        check!1(marr, 30);
        check!0(marr, 15);
        return 0;
    };
    enum ct = dg();
    auto rt = dg();
}

unittest
{// more bitpacking tests
    import std.conv;

    alias Bitty =
      MultiArray!(BitPacked!(size_t, 3)
                , BitPacked!(size_t, 4)
                , BitPacked!(size_t, 3)
                , BitPacked!(size_t, 6)
                , bool);
    alias fn1 = sliceBits!(13, 16);
    alias fn2 = sliceBits!( 9, 13);
    alias fn3 = sliceBits!( 6,  9);
    alias fn4 = sliceBits!( 0,  6);
    static void check(size_t lvl, MA)(ref MA arr){
        for(size_t i = 0; i< arr.length!lvl; i++)
            assert(arr.slice!(lvl)[i] == i, text("Mismatch on lvl ", lvl, " idx ", i, " value: ", arr.slice!(lvl)[i]));
    }

    static void fillIdx(size_t lvl, MA)(ref MA arr){
        for(size_t i = 0; i< arr.length!lvl; i++)
            arr.slice!(lvl)[i] = i;
    }
    Bitty m1;

    m1.length!4 = 10;
    m1.length!3 = 2^^6;
    m1.length!2 = 2^^3;
    m1.length!1 = 2^^4;
    m1.length!0 = 2^^3;

    m1.length!4 = 2^^16;

    for(size_t i = 0; i< m1.length!4; i++)
        m1.slice!(4)[i] = i % 2;

    fillIdx!1(m1);
    check!1(m1);
    fillIdx!2(m1);
    check!2(m1);
    fillIdx!3(m1);
    check!3(m1);
    fillIdx!0(m1);
    check!0(m1);
    check!3(m1);
    check!2(m1);
    check!1(m1);
    for(size_t i=0; i < 2^^16; i++)
    {
        m1.slice!(4)[i] = i % 2;
        m1.slice!(0)[fn1(i)] = fn1(i);
        m1.slice!(1)[fn2(i)] = fn2(i);
        m1.slice!(2)[fn3(i)] = fn3(i);
        m1.slice!(3)[fn4(i)] = fn4(i);
    }
    for(size_t i=0; i < 2^^16; i++)
    {
        assert(m1.slice!(4)[i] == i % 2);
        assert(m1.slice!(0)[fn1(i)] == fn1(i));
        assert(m1.slice!(1)[fn2(i)] == fn2(i));
        assert(m1.slice!(2)[fn3(i)] == fn3(i));
        assert(m1.slice!(3)[fn4(i)] == fn4(i));
    }
}

size_t spaceFor(size_t _bits)(size_t new_len) @safe pure nothrow @nogc
{
    enum bits = _bits == 1 ? 1 : ceilPowerOf2(_bits);// see PackedArrayView
    static if(bits > 8*size_t.sizeof)
    {
        static assert(bits % (size_t.sizeof*8) == 0);
        return new_len * bits/(8*size_t.sizeof);
    }
    else
    {
        enum factor = size_t.sizeof*8/bits;
        return (new_len+factor-1)/factor; // rounded up
    }
}

template isBitPackableType(T)
{
    enum isBitPackableType = isBitPacked!T
        || isIntegral!T || is(T == bool) || isSomeChar!T;
}

//============================================================================
template PackedArrayView(T)
    if((is(T dummy == BitPacked!(U, sz), U, size_t sz)
        && isBitPackableType!U) || isBitPackableType!T)
{
    private enum bits = bitSizeOf!T;
    alias PackedArrayView = PackedArrayViewImpl!(T, bits > 1 ? ceilPowerOf2(bits) : 1);
}

//unsafe and fast access to a chunk of RAM as if it contains packed values
template PackedPtr(T)
    if((is(T dummy == BitPacked!(U, sz), U, size_t sz)
        && isBitPackableType!U) || isBitPackableType!T)
{
    private enum bits = bitSizeOf!T;
    alias PackedPtr = PackedPtrImpl!(T, bits > 1 ? ceilPowerOf2(bits) : 1);
}

@trusted struct PackedPtrImpl(T, size_t bits)
{
pure nothrow:
    static assert(isPowerOf2(bits));

    this(inout(size_t)* ptr)inout @safe @nogc
    {
        origin = ptr;
    }

    private T simpleIndex(size_t n) inout
    {
        auto q = n / factor;
        auto r = n % factor;
        return cast(T)((origin[q] >> bits*r) & mask);
    }

    private void simpleWrite(TypeOfBitPacked!T val, size_t n)
    in
    {
        static if(isIntegral!T)
            assert(val <= mask);
    }
    body
    {
        auto q = n / factor;
        auto r = n % factor;
        size_t tgt_shift = bits*r;
        size_t word = origin[q];
        origin[q] = (word & ~(mask<<tgt_shift))
            | (cast(size_t)val << tgt_shift);
    }

    static if(factor == bytesPerWord// can safely pack by byte
         || factor == 1 // a whole word at a time
         || ((factor == bytesPerWord/2 || factor == bytesPerWord/4)
                && hasUnalignedReads)) // this needs unaligned reads
    {
        static if(factor == bytesPerWord)
            alias U = ubyte;
        else static if(factor == bytesPerWord/2)
            alias U = ushort;
        else static if(factor == bytesPerWord/4)
            alias U = uint;
        else static if(size_t.sizeof == 8 && factor == bytesPerWord/8)
            alias U = ulong;

        T opIndex(size_t idx) inout
        {
            return __ctfe ? simpleIndex(idx) :
                cast(inout(T))(cast(U*)origin)[idx];
        }

        static if(isBitPacked!T) // lack of user-defined implicit conversion
        {
            void opIndexAssign(T val, size_t idx)
            {
                return opIndexAssign(cast(TypeOfBitPacked!T)val, idx);
            }
        }

        void opIndexAssign(TypeOfBitPacked!T val, size_t idx)
        {
            if(__ctfe)
                simpleWrite(val, idx);
            else
                (cast(U*)origin)[idx] = cast(U)val;
        }
    }
    else
    {
        T opIndex(size_t n) inout
        {
            return simpleIndex(n);
        }

        static if(isBitPacked!T) // lack of user-defined implicit conversion
        {
            void opIndexAssign(T val, size_t idx)
            {
                return opIndexAssign(cast(TypeOfBitPacked!T)val, idx);
            }
        }

        void opIndexAssign(TypeOfBitPacked!T val, size_t n)
        {
            return simpleWrite(val, n);
        }
    }

private:
    // factor - number of elements in one machine word
    enum factor = size_t.sizeof*8/bits, mask = 2^^bits-1;
    enum bytesPerWord =  size_t.sizeof;
    size_t* origin;
}

// data is packed only by power of two sized packs per word,
// thus avoiding mul/div overhead at the cost of ultimate packing
// this construct doesn't own memory, only provides access, see MultiArray for usage
@trusted struct PackedArrayViewImpl(T, size_t bits)
{
pure nothrow:

    this(inout(size_t)* origin, size_t offset, size_t items) inout
    {
        ptr = inout(PackedPtr!(T))(origin);
        ofs = offset;
        limit = items;
    }

    bool zeros(size_t s, size_t e)
    in
    {
        assert(s <= e);
    }
    body
    {
        s += ofs;
        e += ofs;
        size_t pad_s = roundUp(s);
        if ( s >= e)
        {
            foreach (i; s..e)
                if(ptr[i])
                    return false;
            return true;
        }
        size_t pad_e = roundDown(e);
        size_t i;
        for(i=s; i<pad_s; i++)
            if(ptr[i])
                return false;
        // all in between is x*factor elements
        for(size_t j=i/factor; i<pad_e; i+=factor, j++)
            if(ptr.origin[j])
                return false;
        for(; i<e; i++)
            if(ptr[i])
                return false;
        return true;
    }

    T opIndex(size_t idx) inout
    in
    {
        assert(idx < limit);
    }
    body
    {
        return ptr[ofs + idx];
    }

    static if(isBitPacked!T) // lack of user-defined implicit conversion
    {
        void opIndexAssign(T val, size_t idx)
        {
            return opIndexAssign(cast(TypeOfBitPacked!T)val, idx);
        }
    }

    void opIndexAssign(TypeOfBitPacked!T val, size_t idx)
    in
    {
        assert(idx < limit);
    }
    body
    {
        ptr[ofs + idx] = val;
    }

    static if(isBitPacked!T) // lack of user-defined implicit conversions
    {
        void opSliceAssign(T val, size_t start, size_t end)
        {
            opSliceAssign(cast(TypeOfBitPacked!T)val, start, end);
        }
    }

    void opSliceAssign(TypeOfBitPacked!T val, size_t start, size_t end)
    in
    {
        assert(start <= end);
        assert(end <= limit);
    }
    body
    {
        // account for ofsetted view
        start += ofs;
        end += ofs;
        // rounded to factor granularity
        size_t pad_start = roundUp(start);// rounded up
        if(pad_start >= end) //rounded up >= then end of slice
        {
            //nothing to gain, use per element assignment
            foreach(i; start..end)
                ptr[i] = val;
            return;
        }
        size_t pad_end = roundDown(end); // rounded down
        size_t i;
        for(i=start; i<pad_start; i++)
            ptr[i] = val;
        // all in between is x*factor elements
        if(pad_start != pad_end)
        {
            size_t repval = replicateBits!(factor, bits)(val);
            for(size_t j=i/factor; i<pad_end; i+=factor, j++)
                ptr.origin[j] = repval;// so speed it up by factor
        }
        for(; i<end; i++)
            ptr[i] = val;
    }

    auto opSlice(size_t from, size_t to)inout
    in
    {
        assert(from <= to);
        assert(ofs + to <= limit);
    }
    body
    {
        return typeof(this)(ptr.origin, ofs + from, to - from);
    }

    auto opSlice(){ return opSlice(0, length); }

    bool opEquals(T)(auto ref T arr) const
    {
        if(limit != arr.limit)
           return false;
        size_t s1 = ofs, s2 = arr.ofs;
        size_t e1 = s1 + limit, e2 = s2 + limit;
        if(s1 % factor == 0 && s2 % factor == 0 && length % factor == 0)
        {
            return ptr.origin[s1/factor .. e1/factor]
                == arr.ptr.origin[s2/factor .. e2/factor];
        }
        for(size_t i=0;i<limit; i++)
            if(this[i] != arr[i])
                return false;
        return true;
    }

    @property size_t length()const{ return limit; }

private:
    auto roundUp()(size_t val){ return (val+factor-1)/factor*factor; }
    auto roundDown()(size_t val){ return val/factor*factor; }
    // factor - number of elements in one machine word
    enum factor = size_t.sizeof*8/bits;
    PackedPtr!(T) ptr;
    size_t ofs, limit;
}


private struct SliceOverIndexed(T)
{
    enum assignableIndex = is(typeof((){ T.init[0] = Item.init; }));
    enum assignableSlice = is(typeof((){ T.init[0..0] = Item.init; }));
    auto opIndex(size_t idx)const
    in
    {
        assert(idx < to - from);
    }
    body
    {
        return (*arr)[from+idx];
    }

    static if(assignableIndex)
    void opIndexAssign(Item val, size_t idx)
    in
    {
        assert(idx < to - from);
    }
    body
    {
       (*arr)[from+idx] = val;
    }

    auto opSlice(size_t a, size_t b)
    {
        return typeof(this)(from+a, from+b, arr);
    }

    // static if(assignableSlice)
    void opSliceAssign(T)(T val, size_t start, size_t end)
    {
        (*arr)[start+from .. end+from] = val;
    }

    auto opSlice()
    {
        return typeof(this)(from, to, arr);
    }

    @property size_t length()const { return to-from;}

    auto opDollar()const { return length; }

    @property bool empty()const { return from == to; }

    @property auto front()const { return (*arr)[from]; }

    static if(assignableIndex)
    @property void front(Item val) { (*arr)[from] = val; }

    @property auto back()const { return (*arr)[to-1]; }

    static if(assignableIndex)
    @property void back(Item val) { (*arr)[to-1] = val; }

    @property auto save() inout { return this; }

    void popFront() {   from++; }

    void popBack() {    to--; }

    bool opEquals(T)(auto ref T arr) const
    {
        if(arr.length != length)
            return false;
        for(size_t i=0; i <length; i++)
            if(this[i] != arr[i])
                return false;
        return true;
    }
private:
    alias Item = typeof(T.init[0]);
    size_t from, to;
    T* arr;
}

static assert(isRandomAccessRange!(SliceOverIndexed!(int[])));

// BUG? forward reference to return type of sliceOverIndexed!Grapheme
SliceOverIndexed!(const(T)) sliceOverIndexed(T)(size_t a, size_t b, const(T)* x)
    if(is(Unqual!T == T))
{
    return SliceOverIndexed!(const(T))(a, b, x);
}

// BUG? inout is out of reach
//...SliceOverIndexed.arr only parameters or stack based variables can be inout
SliceOverIndexed!T sliceOverIndexed(T)(size_t a, size_t b, T* x)
    if(is(Unqual!T == T))
{
    return SliceOverIndexed!T(a, b, x);
}

unittest
{
    int[] idxArray = [2, 3, 5, 8, 13];
    auto sliced = sliceOverIndexed(0, idxArray.length, &idxArray);

    assert(!sliced.empty);
    assert(sliced.front == 2);
    sliced.front = 1;
    assert(sliced.front == 1);
    assert(sliced.back == 13);
    sliced.popFront();
    assert(sliced.front == 3);
    assert(sliced.back == 13);
    sliced.back = 11;
    assert(sliced.back == 11);
    sliced.popBack();

    assert(sliced.front == 3);
    assert(sliced[$-1] == 8);
    sliced = sliced[];
    assert(sliced[0] == 3);
    assert(sliced.back == 8);
    sliced = sliced[1..$];
    assert(sliced.front == 5);
    sliced = sliced[0..$-1];
    assert(sliced[$-1] == 5);

    int[] other = [2, 5];
    assert(sliced[] == sliceOverIndexed(1, 2, &other));
    sliceOverIndexed(0, 2, &idxArray)[0..2] = -1;
    assert(idxArray[0..2] == [-1, -1]);
    uint[] nullArr = null;
    auto nullSlice = sliceOverIndexed(0, 0, &idxArray);
    assert(nullSlice.empty);
}

private auto packedArrayView(T)(inout(size_t)* ptr, size_t items) @trusted pure nothrow
{
    return inout(PackedArrayView!T)(ptr, 0, items);
}


//============================================================================
// Partially unrolled binary search using Shar's method
//============================================================================

string genUnrolledSwitchSearch(size_t size)
{
    import std.conv : to;
    import core.bitop : bsr;
    import std.array : replace;
    assert(isPowerOf2(size));
    string code = `
    import core.bitop : bsr;
    auto power = bsr(m)+1;
    switch(power){`;
    size_t i = bsr(size);
    foreach_reverse(val; 0..bsr(size))
    {
        auto v = 2^^val;
        code ~= `
        case pow:
            if(pred(range[idx+m], needle))
                idx +=  m;
            goto case;
        `.replace("m", to!string(v))
        .replace("pow", to!string(i));
        i--;
    }
    code ~= `
        case 0:
            if(pred(range[idx], needle))
                idx += 1;
            goto default;
        `;
    code ~= `
        default:
    }`;
    return code;
}

bool isPowerOf2(size_t sz) @safe pure nothrow @nogc
{
    return (sz & (sz-1)) == 0;
}

size_t uniformLowerBound(alias pred, Range, T)(Range range, T needle)
    if(is(T : ElementType!Range))
{
    assert(isPowerOf2(range.length));
    size_t idx = 0, m = range.length/2;
    while(m != 0)
    {
        if(pred(range[idx+m], needle))
            idx += m;
        m /= 2;
    }
    if(pred(range[idx], needle))
        idx += 1;
    return idx;
}

size_t switchUniformLowerBound(alias pred, Range, T)(Range range, T needle)
    if(is(T : ElementType!Range))
{
    assert(isPowerOf2(range.length));
    size_t idx = 0, m = range.length/2;
    enum max = 1<<10;
    while(m >= max)
    {
        if(pred(range[idx+m], needle))
            idx += m;
        m /= 2;
    }
    mixin(genUnrolledSwitchSearch(max));
    return idx;
}

//
size_t floorPowerOf2(size_t arg) @safe pure nothrow @nogc
{
    import core.bitop : bsr;
    assert(arg > 1); // else bsr is undefined
    return 1<<bsr(arg-1);
}

size_t ceilPowerOf2(size_t arg) @safe pure nothrow @nogc
{
    import core.bitop : bsr;
    assert(arg > 1); // else bsr is undefined
    return 1<<bsr(arg-1)+1;
}

template sharMethod(alias uniLowerBound)
{
    size_t sharMethod(alias _pred="a<b", Range, T)(Range range, T needle)
        if(is(T : ElementType!Range))
    {
        import std.functional;
        alias pred = binaryFun!_pred;
        if(range.length == 0)
            return 0;
        if(isPowerOf2(range.length))
            return uniLowerBound!pred(range, needle);
        size_t n = floorPowerOf2(range.length);
        if(pred(range[n-1], needle))
        {// search in another 2^^k area that fully covers the tail of range
            size_t k = ceilPowerOf2(range.length - n + 1);
            return range.length - k + uniLowerBound!pred(range[$-k..$], needle);
        }
        else
            return uniLowerBound!pred(range[0..n], needle);
    }
}

alias sharLowerBound = sharMethod!uniformLowerBound;
alias sharSwitchLowerBound = sharMethod!switchUniformLowerBound;

unittest
{
    import std.range;
    auto stdLowerBound(T)(T[] range, T needle)
    {
        return assumeSorted(range).lowerBound(needle).length;
    }
    immutable MAX = 5*1173;
    auto arr = array(iota(5, MAX, 5));
    assert(arr.length == MAX/5-1);
    foreach(i; 0..MAX+5)
    {
        auto st = stdLowerBound(arr, i);
        assert(st == sharLowerBound(arr, i));
        assert(st == sharSwitchLowerBound(arr, i));
    }
    arr = [];
    auto st = stdLowerBound(arr, 33);
    assert(st == sharLowerBound(arr, 33));
    assert(st == sharSwitchLowerBound(arr, 33));
}
//============================================================================

@safe
{
// hope to see simillar stuff in public interface... once Allocators are out
//@@@BUG moveFront and friends? dunno, for now it's POD-only

@trusted size_t genericReplace(Policy=void, T, Range)
    (ref T dest, size_t from, size_t to, Range stuff)
{
    import std.algorithm : copy;
    size_t delta = to - from;
    size_t stuff_end = from+stuff.length;
    if(stuff.length > delta)
    {// replace increases length
        delta = stuff.length - delta;// now, new is > old  by delta
        static if(is(Policy == void))
            dest.length = dest.length+delta;//@@@BUG lame @property
        else
            dest = Policy.realloc(dest, dest.length+delta);
        copyBackwards(dest[to..dest.length-delta],
            dest[to+delta..dest.length]);
        copyForward(stuff, dest[from..stuff_end]);
    }
    else if(stuff.length == delta)
    {
        copy(stuff, dest[from..to]);
    }
    else
    {// replace decreases length by delta
        delta = delta - stuff.length;
        copy(stuff, dest[from..stuff_end]);
        copyForward(dest[to..dest.length],
            dest[stuff_end..dest.length-delta]);
        static if(is(Policy == void))
            dest.length = dest.length - delta;//@@@BUG lame @property
        else
            dest = Policy.realloc(dest, dest.length-delta);
    }
    return stuff_end;
}


// Simple storage manipulation policy
@trusted public struct GcPolicy
{
    static T[] dup(T)(const T[] arr)
    {
        return arr.dup;
    }

    static T[] alloc(T)(size_t size)
    {
        return new T[size];
    }

    static T[] realloc(T)(T[] arr, size_t sz)
    {
        arr.length = sz;
        return arr;
    }

    static void replaceImpl(T, Range)(ref T[] dest, size_t from, size_t to, Range stuff)
    {
        replaceInPlace(dest, from, to, stuff);
    }

    static void append(T, V)(ref T[] arr, V value)
        if(!isInputRange!V)
    {
        arr ~= force!T(value);
    }

    static void append(T, V)(ref T[] arr, V value)
        if(isInputRange!V)
    {
        insertInPlace(arr, arr.length, value);
    }

    static void destroy(T)(ref T arr)
        if(isDynamicArray!T && is(Unqual!T == T))
    {
        debug
        {
            arr[] = cast(typeof(T.init[0]))(0xdead_beef);
        }
        arr = null;
    }

    static void destroy(T)(ref T arr)
        if(isDynamicArray!T && !is(Unqual!T == T))
    {
        arr = null;
    }
}

// ditto
@trusted struct ReallocPolicy
{
    static T[] dup(T)(const T[] arr)
    {
        auto result = alloc!T(arr.length);
        result[] = arr[];
        return result;
    }

    static T[] alloc(T)(size_t size)
    {
        import std.exception : enforce;
        auto ptr = cast(T*)enforce(malloc(T.sizeof*size), "out of memory on C heap");
        return ptr[0..size];
    }

    static T[] realloc(T)(T[] arr, size_t size)
    {
        import std.exception : enforce;
        if(!size)
        {
            destroy(arr);
            return null;
        }
        auto ptr = cast(T*)enforce(core.stdc.stdlib.realloc(
                             arr.ptr, T.sizeof*size), "out of memory on C heap");
        return ptr[0..size];
    }

    static void replaceImpl(T, Range)(ref T[] dest, size_t from, size_t to, Range stuff)
    {
        genericReplace!(ReallocPolicy)(dest, from, to, stuff);
    }

    static void append(T, V)(ref T[] arr, V value)
        if(!isInputRange!V)
    {
        arr = realloc(arr, arr.length+1);
        arr[$-1] = force!T(value);
    }

    static void append(T, V)(ref T[] arr, V value)
        if(isInputRange!V && hasLength!V)
    {
        arr = realloc(arr, arr.length+value.length);
        copy(value, arr[$-value.length..$]);
    }

    static void destroy(T)(ref T[] arr)
    {
        if(arr.ptr)
            free(arr.ptr);
        arr = null;
    }
}

//build hack
alias _RealArray = CowArray!ReallocPolicy;

unittest
{
    with(ReallocPolicy)
    {
        bool test(T, U, V)(T orig, size_t from, size_t to, U toReplace, V result,
                   string file = __FILE__, size_t line = __LINE__)
        {
            {
                replaceImpl(orig, from, to, toReplace);
                scope(exit) destroy(orig);
                if(!equalS(orig, result))
                    return false;
            }
            return true;
        }
        static T[] arr(T)(T[] args... )
        {
            return dup(args);
        }

        assert(test(arr([1, 2, 3, 4]), 0, 0, [5, 6, 7], [5, 6, 7, 1, 2, 3, 4]));
        assert(test(arr([1, 2, 3, 4]), 0, 2, cast(int[])[], [3, 4]));
        assert(test(arr([1, 2, 3, 4]), 0, 4, [5, 6, 7], [5, 6, 7]));
        assert(test(arr([1, 2, 3, 4]), 0, 2, [5, 6, 7], [5, 6, 7, 3, 4]));
        assert(test(arr([1, 2, 3, 4]), 2, 3, [5, 6, 7], [1, 2, 5, 6, 7, 4]));
    }
}

/**
    Tests if T is some kind a set of code points. Intended for template constraints.
*/
public template isCodepointSet(T)
{
    static if(is(T dummy == InversionList!(Args), Args...))
        enum isCodepointSet = true;
    else
        enum isCodepointSet = false;
}

/**
    Tests if $(D T) is a pair of integers that implicitly convert to $(D V).
    The following code must compile for any pair $(D T):
    ---
    (T x){ V a = x[0]; V b = x[1];}
    ---
    The following must not compile:
     ---
    (T x){ V c = x[2];}
    ---
*/
public template isIntegralPair(T, V=uint)
{
    enum isIntegralPair = is(typeof((T x){ V a = x[0]; V b = x[1];}))
        && !is(typeof((T x){ V c = x[2]; }));
}


/**
    The recommended default type for set of $(CODEPOINTS).
    For details, see the current implementation: $(LREF InversionList).
*/
public alias CodepointSet = InversionList!GcPolicy;


//@@@BUG: std.typecons tuples depend on std.format to produce fields mixin
// which relies on std.uni.isGraphical and this chain blows up with Forward reference error
// hence below doesn't seem to work
// public alias CodepointInterval = Tuple!(uint, "a", uint, "b");

/**
    The recommended type of $(XREF _typecons, Tuple)
    to represent [a, b$(RPAREN) intervals of $(CODEPOINTS). As used in $(LREF InversionList).
    Any interval type should pass $(LREF isIntegralPair) trait.
*/
public struct CodepointInterval
{
pure:
    uint[2] _tuple;
    alias _tuple this;

@safe pure nothrow @nogc:

    this(uint low, uint high)
    {
        _tuple[0] = low;
        _tuple[1] = high;
    }
    bool opEquals(T)(T val) const
    {
        return this[0] == val[0] && this[1] == val[1];
    }
    @property ref inout(uint) a() inout { return _tuple[0]; }
    @property ref inout(uint) b() inout { return _tuple[1]; }
}

//@@@BUG another forward reference workaround
@trusted bool equalS(R1, R2)(R1 lhs, R2 rhs)
{
    for(;;){
        if(lhs.empty)
            return rhs.empty;
        if(rhs.empty)
            return false;
        if(lhs.front != rhs.front)
            return false;
        lhs.popFront();
        rhs.popFront();
    }
}



/**
    $(P
    $(D InversionList) is a set of $(CODEPOINTS)
    represented as an array of open-right [a, b$(RPAREN)
    intervals (see $(LREF CodepointInterval) above).
    The name comes from the way the representation reads left to right.
    For instance a set of all values [10, 50$(RPAREN), [80, 90$(RPAREN),
    plus a singular value 60 looks like this:
    )
    ---
    10, 50, 60, 61, 80, 90
    ---
    $(P
    The way to read this is: start with negative meaning that all numbers
    smaller then the next one are not present in this set (and positive
    - the contrary). Then switch positive/negative after each
    number passed from left to right.
    )
    $(P This way negative spans until 10, then positive until 50,
    then negative until 60, then positive until 61, and so on.
    As seen this provides a space-efficient storage of highly redundant data
    that comes in long runs. A description which Unicode $(CHARACTER)
    properties fit nicely. The technique itself could be seen as a variation
    on $(LUCKY RLE encoding).
    )

    $(P Sets are value types (just like $(D int) is) thus they
        are never aliased.
    )
        Example:
        ---
        auto a = CodepointSet('a', 'z'+1);
        auto b = CodepointSet('A', 'Z'+1);
        auto c = a;
        a = a | b;
        assert(a == CodepointSet('A', 'Z'+1, 'a', 'z'+1));
        assert(a != c);
        ---
    $(P See also $(LREF unicode) for simpler construction of sets
        from predefined ones.
    )

    $(P Memory usage is 8 bytes per each contiguous interval in a set.
    The value semantics are achieved by using the
    $(WEB en.wikipedia.org/wiki/Copy-on-write, COW) technique
    and thus it's $(RED not) safe to cast this type to $(D_KEYWORD shared).
    )

    Note:
    $(P It's not recommended to rely on the template parameters
    or the exact type of a current $(CODEPOINT) set in $(D std.uni).
    The type and parameters may change when the standard
    allocators design is finalized.
    Use $(LREF isCodepointSet) with templates or just stick with the default
    alias $(LREF CodepointSet) throughout the whole code base.
    )
*/
@trusted public struct InversionList(SP=GcPolicy)
{
    import std.range : assumeSorted;

public:

    /**
        Construct from another code point set of any type.
    */
    this(Set)(Set set) pure
        if(isCodepointSet!Set)
    {
        uint[] arr;
        foreach(v; set.byInterval)
        {
            arr ~= v.a;
            arr ~= v.b;
        }
        data = CowArray!(SP).reuse(arr);
    }

    /**
        Construct a set from a forward range of code point intervals.
    */
    this(Range)(Range intervals) pure
        if(isForwardRange!Range && isIntegralPair!(ElementType!Range))
    {
        uint[] arr;
        foreach(v; intervals)
        {
            SP.append(arr, v.a);
            SP.append(arr, v.b);
        }
        data = CowArray!(SP).reuse(arr);
        sanitize(); //enforce invariant: sort intervals etc.
    }

    //helper function that avoids sanity check to be CTFE-friendly
    private static fromIntervals(Range)(Range intervals) pure
    {
        import std.algorithm : map;
        import std.range : roundRobin;
        auto flattened = roundRobin(intervals.save.map!"a[0]"(),
            intervals.save.map!"a[1]"());
        InversionList set;
        set.data = CowArray!(SP)(flattened);
        return set;
    }
    //ditto untill sort is CTFE-able
    private static fromIntervals()(uint[] intervals...) pure
    in
    {
        import std.conv : text;
        assert(intervals.length % 2 == 0, "Odd number of interval bounds [a, b)!");
        for (uint i = 0; i < intervals.length; i += 2)
        {
            auto a = intervals[i], b = intervals[i+1];
            assert(a < b, text("illegal interval [a, b): ", a, " > ", b));
        }
    }
    body
    {
        InversionList set;
        set.data = CowArray!(SP)(intervals);
        return set;
    }

    /**
        Construct a set from plain values of code point intervals.
        Example:
        ---
        import std.algorithm;
        auto set = CodepointSet('a', 'z'+1, 'а', 'я'+1);
        foreach(v; 'a'..'z'+1)
            assert(set[v]);
        // Cyrillic lowercase interval
        foreach(v; 'а'..'я'+1)
            assert(set[v]);
        //specific order is not required, intervals may interesect
        auto set2 = CodepointSet('а', 'я'+1, 'a', 'd', 'b', 'z'+1);
        //the same end result
        assert(set2.byInterval.equal(set.byInterval));
        ---
    */
    this()(uint[] intervals...)
    in
    {
        import std.conv : text;
        assert(intervals.length % 2 == 0, "Odd number of interval bounds [a, b)!");
        for (uint i = 0; i < intervals.length; i += 2)
        {
            auto a = intervals[i], b = intervals[i+1];
            assert(a < b, text("illegal interval [a, b): ", a, " > ", b));
        }
    }
    body
    {
        data = CowArray!(SP)(intervals);
        sanitize(); //enforce invariant: sort intervals etc.
    }

    /**
        Get range that spans all of the $(CODEPOINT) intervals in this $(LREF InversionList).

        Example:
        ---
        import std.algorithm, std.typecons;
        auto set = CodepointSet('A', 'D'+1, 'a', 'd'+1);
        set.byInterval.equal([tuple('A', 'E'), tuple('a', 'e')]);
        ---
    */
    @property auto byInterval()
    {
        return Intervals!(typeof(data))(data);
    }

    /**
        Tests the presence of code point $(D val) in this set.

        Example:
        ---
        auto gothic = unicode.Gothic;
        // Gothic letter ahsa
        assert(gothic['\U00010330']);
        // no ascii in Gothic obviously
        assert(!gothic['$']);
        ---
    */
    bool opIndex(uint val) const
    {
        // the <= ensures that searching in  interval of [a, b) for 'a' you get .length == 1
        // return assumeSorted!((a,b) => a<=b)(data[]).lowerBound(val).length & 1;
        return sharSwitchLowerBound!"a<=b"(data[], val) & 1;
    }

    // Linear scan for $(D ch). Useful only for small sets.
    // TODO:
    // used internally in std.regex
    // should be properly exposed in a public API ?
    package auto scanFor()(dchar ch) const
    {
        immutable len = data.length;
        for(size_t i = 0; i < len; i++)
            if(ch < data[i])
                return i & 1;
        return 0;
    }

    /// Number of $(CODEPOINTS) in this set
    @property size_t length()
    {
        size_t sum = 0;
        foreach(iv; byInterval)
        {
            sum += iv.b - iv.a;
        }
        return sum;
    }

// bootstrap full set operations from 4 primitives (suitable as a template mixin):
// addInterval, skipUpTo, dropUpTo & byInterval iteration
//============================================================================
public:
    /**
        $(P Sets support natural syntax for set algebra, namely: )
        $(BOOKTABLE ,
            $(TR $(TH Operator) $(TH Math notation) $(TH Description) )
            $(TR $(TD &) $(TD a ∩ b) $(TD intersection) )
            $(TR $(TD |) $(TD a ∪ b) $(TD union) )
            $(TR $(TD -) $(TD a ∖ b) $(TD subtraction) )
            $(TR $(TD ~) $(TD a ~ b) $(TD symmetric set difference i.e. (a ∪ b) \ (a ∩ b)) )
        )

        Example:
        ---
        auto lower = unicode.LowerCase;
        auto upper = unicode.UpperCase;
        auto ascii = unicode.ASCII;

        assert((lower & upper).empty); // no intersection
        auto lowerASCII = lower & ascii;
        assert(lowerASCII.byCodepoint.equal(iota('a', 'z'+1)));
        // throw away all of the lowercase ASCII
        assert((ascii - lower).length == 128 - 26);

        auto onlyOneOf = lower ~ ascii;
        assert(!onlyOneOf['Δ']); // not ASCII and not lowercase
        assert(onlyOneOf['$']); // ASCII and not lowercase
        assert(!onlyOneOf['a']); // ASCII and lowercase
        assert(onlyOneOf['я']); // not ASCII but lowercase

        // throw away all cased letters from ASCII
        auto noLetters = ascii - (lower | upper);
        assert(noLetters.length == 128 - 26*2);
        ---
    */
    This opBinary(string op, U)(U rhs)
        if(isCodepointSet!U || is(U:dchar))
    {
        static if(op == "&" || op == "|" || op == "~")
        {// symmetric ops thus can swap arguments to reuse r-value
            static if(is(U:dchar))
            {
                auto tmp = this;
                mixin("tmp "~op~"= rhs; ");
                return tmp;
            }
            else
            {
                static if(is(Unqual!U == U))
                {
                    // try hard to reuse r-value
                    mixin("rhs "~op~"= this;");
                    return rhs;
                }
                else
                {
                    auto tmp = this;
                    mixin("tmp "~op~"= rhs;");
                    return tmp;
                }
            }
        }
        else static if(op == "-") // anti-symmetric
        {
            auto tmp = this;
            tmp -= rhs;
            return tmp;
        }
        else
            static assert(0, "no operator "~op~" defined for Set");
    }

    /// The 'op=' versions of the above overloaded operators.
    ref This opOpAssign(string op, U)(U rhs)
        if(isCodepointSet!U || is(U:dchar))
    {
        static if(op == "|")    // union
        {
            static if(is(U:dchar))
            {
                this.addInterval(rhs, rhs+1);
                return this;
            }
            else
                return this.add(rhs);
        }
        else static if(op == "&")   // intersection
                return this.intersect(rhs);// overloaded
        else static if(op == "-")   // set difference
                return this.sub(rhs);// overloaded
        else static if(op == "~")   // symmetric set difference
        {
            auto copy = this & rhs;
            this |= rhs;
            this -= copy;
            return this;
        }
        else
            static assert(0, "no operator "~op~" defined for Set");
    }

    /**
        Tests the presence of codepoint $(D ch) in this set,
        the same as $(LREF opIndex).
    */
    bool opBinaryRight(string op: "in", U)(U ch) const
        if(is(U : dchar))
    {
        return this[ch];
    }

    ///
    unittest
    {
        assert('я' in unicode.Cyrillic);
        assert(!('z' in unicode.Cyrillic));
    }



    /**
     * Obtains a set that is the inversion of this set.
     *
     * See_Also: $(LREF inverted)
     */
    auto opUnary(string op: "!")()
    {
        return this.inverted;
    }

    /**
        A range that spans each $(CODEPOINT) in this set.

        Example:
        ---
        import std.algorithm;
        auto set = unicode.ASCII;
        set.byCodepoint.equal(iota(0, 0x80));
        ---
    */
    @property auto byCodepoint()
    {
        @trusted static struct CodepointRange
        {
            this(This set)
            {
                r = set.byInterval;
                if(!r.empty)
                    cur = r.front.a;
            }

            @property dchar front() const
            {
                return cast(dchar)cur;
            }

            @property bool empty() const
            {
                return r.empty;
            }

            void popFront()
            {
                cur++;
                while(cur >= r.front.b)
                {
                    r.popFront();
                    if(r.empty)
                        break;
                    cur = r.front.a;
                }
            }
        private:
            uint cur;
            typeof(This.init.byInterval) r;
        }

        return CodepointRange(this);
    }

    /**
        $(P Obtain textual representation of this set in from of
        open-right intervals and feed it to $(D sink).
        )
        $(P Used by various standard formatting facilities such as
         $(XREF _format, formattedWrite), $(XREF _stdio, write),
         $(XREF _stdio, writef), $(XREF _conv, to) and others.
        )
        Example:
        ---
        import std.conv;
        assert(unicode.ASCII.to!string == "[0..128$(RPAREN)");
        ---
    */

    private import std.format : FormatException, FormatSpec;

    /***************************************
     * Obtain a textual representation of this InversionList
     * in form of open-right intervals.
     *
     * The formatting flag is applied individually to each value, for example:
     * $(LI $(B %s) and $(B %d) format the intervals as a [low..high$(RPAREN) range of integrals)
     * $(LI $(B %x) formats the intervals as a [low..high$(RPAREN) range of lowercase hex characters)
     * $(LI $(B %X) formats the intervals as a [low..high$(RPAREN) range of uppercase hex characters)
     */
    void toString(scope void delegate(const(char)[]) sink,
                  FormatSpec!char fmt) /* const */
    {
        import std.format;
        auto range = byInterval;
        if(range.empty)
            return;

        while (1)
        {
            auto i = range.front;
            range.popFront();

            put(sink, "[");
            formatValue(sink, i.a, fmt);
            put(sink, "..");
            formatValue(sink, i.b, fmt);
            put(sink, ")");
            if (range.empty) return;
            put(sink, " ");
        }
    }

    ///
    unittest
    {
        import std.conv : to;
        import std.format : format;
        import std.uni : unicode;

        assert(unicode.Cyrillic.to!string ==
            "[1024..1157) [1159..1320) [7467..7468) [7544..7545) [11744..11776) [42560..42648) [42655..42656)");

        // The specs '%s' and '%d' are equivalent to the to!string call above.
        assert(format("%d", unicode.Cyrillic) == unicode.Cyrillic.to!string);

        assert(format("%#x", unicode.Cyrillic) ==
            "[0x400..0x485) [0x487..0x528) [0x1d2b..0x1d2c) [0x1d78..0x1d79) [0x2de0..0x2e00) [0xa640..0xa698) [0xa69f..0xa6a0)");

        assert(format("%#X", unicode.Cyrillic) ==
            "[0X400..0X485) [0X487..0X528) [0X1D2B..0X1D2C) [0X1D78..0X1D79) [0X2DE0..0X2E00) [0XA640..0XA698) [0XA69F..0XA6A0)");
    }

    unittest
    {
        import std.exception : assertThrown;
        import std.format : format;
        assertThrown!FormatException(format("%a", unicode.ASCII));
    }


    /**
        Add an interval [a, b$(RPAREN) to this set.

        Example:
        ---
        CodepointSet someSet;
        someSet.add('0', '5').add('A','Z'+1);
        someSet.add('5', '9'+1);
        assert(someSet['0']);
        assert(someSet['5']);
        assert(someSet['9']);
        assert(someSet['Z']);
        ---
    */
    ref add()(uint a, uint b)
    {
        addInterval(a, b);
        return this;
    }

private:

  package(std)  // used from: std.regex.internal.parser
    ref intersect(U)(U rhs)
        if(isCodepointSet!U)
    {
        Marker mark;
        foreach( i; rhs.byInterval)
        {
            mark = this.dropUpTo(i.a, mark);
            mark = this.skipUpTo(i.b, mark);
        }
        this.dropUpTo(uint.max, mark);
        return this;
    }

    ref intersect()(dchar ch)
    {
        foreach(i; byInterval)
            if(i.a <= ch && ch < i.b)
                return this = This.init.add(ch, ch+1);
        this = This.init;
        return this;
    }

    ///
    unittest
    {
        assert(unicode.Cyrillic.intersect('-').byInterval.empty);
    }

    ref sub()(dchar ch)
    {
        return subChar(ch);
    }

    // same as the above except that skip & drop parts are swapped
  package(std)  // used from: std.regex.internal.parser
    ref sub(U)(U rhs)
        if(isCodepointSet!U)
    {
        uint top;
        Marker mark;
        foreach(i; rhs.byInterval)
        {
            mark = this.skipUpTo(i.a, mark);
            mark = this.dropUpTo(i.b, mark);
        }
        return this;
    }

  package(std)  // used from: std.regex.internal.parse
    ref add(U)(U rhs)
        if(isCodepointSet!U)
    {
        Marker start;
        foreach(i; rhs.byInterval)
        {
            start = addInterval(i.a, i.b, start);
        }
        return this;
    }

// end of mixin-able part
//============================================================================
public:
    /**
        Obtains a set that is the inversion of this set.

        See the '!' $(LREF opUnary) for the same but using operators.

        Example:
        ---
        set = unicode.ASCII;
        // union with the inverse gets all of the code points in the Unicode
        assert((set | set.inverted).length == 0x110000);
        // no intersection with the inverse
        assert((set & set.inverted).empty);
        ---
    */
    @property auto inverted()
    {
        InversionList inversion = this;
        if(inversion.data.length == 0)
        {
            inversion.addInterval(0, lastDchar+1);
            return inversion;
        }
        if(inversion.data[0] != 0)
            genericReplace(inversion.data, 0, 0, [0]);
        else
            genericReplace(inversion.data, 0, 1, cast(uint[])null);
        if(data[data.length-1] != lastDchar+1)
            genericReplace(inversion.data,
                inversion.data.length, inversion.data.length, [lastDchar+1]);
        else
            genericReplace(inversion.data,
                inversion.data.length-1, inversion.data.length, cast(uint[])null);

        return inversion;
    }

    /**
        Generates string with D source code of unary function with name of
        $(D funcName) taking a single $(D dchar) argument. If $(D funcName) is empty
        the code is adjusted to be a lambda function.

        The function generated tests if the $(CODEPOINT) passed
        belongs to this set or not. The result is to be used with string mixin.
        The intended usage area is aggressive optimization via meta programming
        in parser generators and the like.

        Note: Use with care for relatively small or regular sets. It
        could end up being slower then just using multi-staged tables.

        Example:
        ---
        import std.stdio;

        // construct set directly from [a, b$RPAREN intervals
        auto set = CodepointSet(10, 12, 45, 65, 100, 200);
        writeln(set);
        writeln(set.toSourceCode("func"));
        ---

        The above outputs something along the lines of:
        ---
        bool func(dchar ch)  @safe pure nothrow @nogc
        {
            if(ch < 45)
            {
                if(ch == 10 || ch == 11) return true;
                return false;
            }
            else if (ch < 65) return true;
            else
            {
                if(ch < 100) return false;
                if(ch < 200) return true;
                return false;
            }
        }
        ---
    */
    string toSourceCode(string funcName="")
    {
        import std.array : array;
        import std.format : format;
        import std.algorithm : countUntil;
        enum maxBinary = 3;
        static string linearScope(R)(R ivals, string indent)
        {
            string result = indent~"{\n";
            string deeper = indent~"    ";
            foreach(ival; ivals)
            {
                auto span = ival[1] - ival[0];
                assert(span != 0);
                if(span == 1)
                {
                    result ~= format("%sif(ch == %s) return true;\n", deeper, ival[0]);
                }
                else if(span == 2)
                {
                    result ~= format("%sif(ch == %s || ch == %s) return true;\n",
                        deeper, ival[0], ival[0]+1);
                }
                else
                {
                    if(ival[0] != 0) // dchar is unsigned and  < 0 is useless
                        result ~= format("%sif(ch < %s) return false;\n", deeper, ival[0]);
                    result ~= format("%sif(ch < %s) return true;\n", deeper, ival[1]);
                }
            }
            result ~= format("%sreturn false;\n%s}\n", deeper, indent); // including empty range of intervals
            return result;
        }

        static string binaryScope(R)(R ivals, string indent)
        {
            // time to do unrolled comparisons?
            if(ivals.length < maxBinary)
                return linearScope(ivals, indent);
            else
                return bisect(ivals, ivals.length/2, indent);
        }

        // not used yet if/elsebinary search is far better with DMD  as of 2.061
        // and GDC is doing fine job either way
        static string switchScope(R)(R ivals, string indent)
        {
            string result = indent~"switch(ch){\n";
            string deeper = indent~"    ";
            foreach(ival; ivals)
            {
                if(ival[0]+1 == ival[1])
                {
                    result ~= format("%scase %s: return true;\n",
                        deeper, ival[0]);
                }
                else
                {
                    result ~= format("%scase %s: .. case %s: return true;\n",
                         deeper, ival[0], ival[1]-1);
                }
            }
            result ~= deeper~"default: return false;\n"~indent~"}\n";
            return result;
        }

        static string bisect(R)(R range, size_t idx, string indent)
        {
            string deeper = indent ~ "    ";
            // bisect on one [a, b) interval at idx
            string result = indent~"{\n";
            // less branch, < a
            result ~= format("%sif(ch < %s)\n%s",
                deeper, range[idx][0], binaryScope(range[0..idx], deeper));
            // middle point,  >= a && < b
            result ~= format("%selse if (ch < %s) return true;\n",
                deeper, range[idx][1]);
            // greater or equal branch,  >= b
            result ~= format("%selse\n%s",
                deeper, binaryScope(range[idx+1..$], deeper));
            return result~indent~"}\n";
        }

        string code = format("bool %s(dchar ch) @safe pure nothrow @nogc\n",
            funcName.empty ? "function" : funcName);
        auto range = byInterval.array();
        // special case first bisection to be on ASCII vs beyond
        auto tillAscii = countUntil!"a[0] > 0x80"(range);
        if(tillAscii <= 0) // everything is ASCII or nothing is ascii (-1 & 0)
            code ~= binaryScope(range, "");
        else
            code ~= bisect(range, tillAscii, "");
        return code;
    }

    /**
        True if this set doesn't contain any $(CODEPOINTS).
        Example:
        ---
        CodepointSet emptySet;
        assert(emptySet.length == 0);
        assert(emptySet.empty);
        ---
    */
    @property bool empty() const
    {
        return data.length == 0;
    }

private:
    alias This = typeof(this);
    alias Marker = size_t;

    // a random-access range of integral pairs
    static struct Intervals(Range)
    {
        this(Range sp)
        {
            slice = sp;
            start = 0;
            end = sp.length;
        }

        this(Range sp, size_t s, size_t e)
        {
            slice = sp;
            start = s;
            end = e;
        }

        @property auto front()const
        {
            uint a = slice[start];
            uint b = slice[start+1];
            return CodepointInterval(a, b);
        }

        //may break sorted property - but we need std.sort to access it
        //hence package protection attribute
        package @property auto front(CodepointInterval val)
        {
            slice[start] = val.a;
            slice[start+1] = val.b;
        }

        @property auto back()const
        {
            uint a = slice[end-2];
            uint b = slice[end-1];
            return CodepointInterval(a, b);
        }

        //ditto about package
        package @property auto back(CodepointInterval val)
        {
            slice[end-2] = val.a;
            slice[end-1] = val.b;
        }

        void popFront()
        {
            start += 2;
        }

        void popBack()
        {
            end -= 2;
        }

        auto opIndex(size_t idx) const
        {
            uint a = slice[start+idx*2];
            uint b = slice[start+idx*2+1];
            return CodepointInterval(a, b);
        }

        //ditto about package
        package auto opIndexAssign(CodepointInterval val, size_t idx)
        {
            slice[start+idx*2] = val.a;
            slice[start+idx*2+1] = val.b;
        }

        auto opSlice(size_t s, size_t e)
        {
            return Intervals(slice, s*2+start, e*2+start);
        }

        @property size_t length()const {  return slice.length/2; }

        @property bool empty()const { return start == end; }

        @property auto save(){ return this; }
    private:
        size_t start, end;
        Range slice;
    }

    // called after construction from intervals
    // to make sure invariants hold
    void sanitize()
    {
        import std.algorithm : sort, SwapStrategy, max;
        if (data.length == 0)
            return;
        alias Ival = CodepointInterval;
        //intervals wrapper for a _range_ over packed array
        auto ivals = Intervals!(typeof(data[]))(data[]);
        //@@@BUG@@@ can't use "a.a < b.a" see issue 12265
        sort!((a,b) => a.a < b.a, SwapStrategy.stable)(ivals);
        // what follows is a variation on stable remove
        // differences:
        // - predicate is binary, and is tested against
        //   the last kept element (at 'i').
        // - predicate mutates lhs (merges rhs into lhs)
        size_t len = ivals.length;
        size_t i = 0;
        size_t j = 1;
        while (j < len)
        {
            if (ivals[i].b >= ivals[j].a)
            {
                ivals[i] = Ival(ivals[i].a, max(ivals[i].b, ivals[j].b));
                j++;
            }
            else //unmergable
            {
                // check if there is a hole after merges
                // (in the best case we do 0 writes to ivals)
                if (j != i+1)
                    ivals[i+1] = ivals[j]; //copy over
                i++;
                j++;
            }
        }
        len = i + 1;
        for (size_t k=0; k + 1 < len; k++)
        {
            assert(ivals[k].a < ivals[k].b);
            assert(ivals[k].b < ivals[k+1].a);
        }
        data.length = len * 2;
    }

    // special case for normal InversionList
    ref subChar(dchar ch)
    {
        auto mark = skipUpTo(ch);
        if(mark != data.length
            && data[mark] == ch && data[mark-1] == ch)
        {
            // it has split, meaning that ch happens to be in one of intervals
            data[mark] = data[mark]+1;
        }
        return this;
    }

    //
    Marker addInterval(int a, int b, Marker hint=Marker.init)
    in
    {
        assert(a <= b);
    }
    body
    {
        import std.range : assumeSorted, SearchPolicy;
        auto range = assumeSorted(data[]);
        size_t pos;
        size_t a_idx = hint + range[hint..$].lowerBound!(SearchPolicy.gallop)(a).length;
        if(a_idx == range.length)
        {
            //  [---+++----++++----++++++]
            //  [                         a  b]
            data.append(a, b);
            return data.length-1;
        }
        size_t b_idx = range[a_idx..range.length].lowerBound!(SearchPolicy.gallop)(b).length+a_idx;
        uint[3] buf = void;
        uint to_insert;
        debug(std_uni)
        {
            writefln("a_idx=%d; b_idx=%d;", a_idx, b_idx);
        }
        if(b_idx == range.length)
        {
            //  [-------++++++++----++++++-]
            //  [      s     a                 b]
            if(a_idx & 1)// a in positive
            {
                buf[0] = b;
                to_insert = 1;
            }
            else// a in negative
            {
                buf[0] = a;
                buf[1] = b;
                to_insert = 2;
            }
            pos = genericReplace(data, a_idx, b_idx, buf[0..to_insert]);
            return pos - 1;
        }

        uint top = data[b_idx];

        debug(std_uni)
        {
            writefln("a_idx=%d; b_idx=%d;", a_idx, b_idx);
            writefln("a=%s; b=%s; top=%s;", a, b, top);
        }
        if(a_idx & 1)
        {// a in positive
            if(b_idx & 1)// b in positive
            {
                //  [-------++++++++----++++++-]
                //  [       s    a        b    ]
                buf[0] = top;
                to_insert = 1;
            }
            else // b in negative
            {
                //  [-------++++++++----++++++-]
                //  [       s    a   b         ]
                if(top == b)
                {
                    assert(b_idx+1 < data.length);
                    buf[0] = data[b_idx+1];
                    pos = genericReplace(data, a_idx, b_idx+2, buf[0..1]);
                    return pos - 1;
                }
                buf[0] = b;
                buf[1] = top;
                to_insert = 2;
            }
        }
        else
        { // a in negative
            if(b_idx & 1) // b in positive
            {
                //  [----------+++++----++++++-]
                //  [     a     b              ]
                buf[0] = a;
                buf[1] = top;
                to_insert = 2;
            }
            else// b in negative
            {
                //  [----------+++++----++++++-]
                //  [  a       s      b        ]
                if(top == b)
                {
                    assert(b_idx+1 < data.length);
                    buf[0] = a;
                    buf[1] = data[b_idx+1];
                    pos = genericReplace(data, a_idx, b_idx+2, buf[0..2]);
                    return pos - 1;
                }
                buf[0] = a;
                buf[1] = b;
                buf[2] = top;
                to_insert = 3;
            }
        }
        pos = genericReplace(data, a_idx, b_idx+1, buf[0..to_insert]);
        debug(std_uni)
        {
            writefln("marker idx: %d; length=%d", pos, data[pos], data.length);
            writeln("inserting ", buf[0..to_insert]);
        }
        return pos - 1;
    }

    //
    Marker dropUpTo(uint a, Marker pos=Marker.init)
    in
    {
        assert(pos % 2 == 0); // at start of interval
    }
    body
    {
        auto range = assumeSorted!"a<=b"(data[pos..data.length]);
        if(range.empty)
            return pos;
        size_t idx = pos;
        idx += range.lowerBound(a).length;

        debug(std_uni)
        {
            writeln("dropUpTo full length=", data.length);
            writeln(pos,"~~~", idx);
        }
        if(idx == data.length)
            return genericReplace(data, pos, idx, cast(uint[])[]);
        if(idx & 1)
        {   // a in positive
            //[--+++----++++++----+++++++------...]
            //      |<---si       s  a  t
            genericReplace(data, pos, idx, [a]);
        }
        else
        {   // a in negative
            //[--+++----++++++----+++++++-------+++...]
            //      |<---si              s  a  t
            genericReplace(data, pos, idx, cast(uint[])[]);
        }
        return pos;
    }

    //
    Marker skipUpTo(uint a, Marker pos=Marker.init)
    out(result)
    {
        assert(result % 2 == 0);// always start of interval
        //(may be  0-width after-split)
    }
    body
    {
        assert(data.length % 2 == 0);
        auto range = assumeSorted!"a<=b"(data[pos..data.length]);
        size_t idx = pos+range.lowerBound(a).length;

        if(idx >= data.length) // could have Marker point to recently removed stuff
            return data.length;

        if(idx & 1)// inside of interval, check for split
        {

            uint top = data[idx];
            if(top == a)// no need to split, it's end
                return idx+1;
            uint start = data[idx-1];
            if(a == start)
                return idx-1;
            // split it up
            genericReplace(data, idx, idx+1, [a, a, top]);
            return idx+1;        // avoid odd index
        }
        return idx;
    }

    CowArray!SP data;
}

@system unittest
{
    // test examples
    import std.algorithm, std.typecons;
    import std.range : iota;
    auto set = CodepointSet('A', 'D'+1, 'a', 'd'+1);
    set.byInterval.equalS([tuple('A', 'E'), tuple('a', 'e')]);
    set = unicode.ASCII;
    assert(set.byCodepoint.equalS(iota(0, 0x80)));
    set = CodepointSet('a', 'z'+1, 'а', 'я'+1);
    foreach(v; 'a'..'z'+1)
        assert(set[v]);
    // Cyrillic lowercase interval
    foreach(v; 'а'..'я'+1)
        assert(set[v]);
    //specific order is not required, intervals may interesect
    auto set2 = CodepointSet('а', 'я'+1, 'a', 'd', 'b', 'z'+1);
    assert(set2.byInterval.equal(set.byInterval));

    auto gothic = unicode.Gothic;
    // Gothic letter ahsa
    assert(gothic['\U00010330']);
    // no ascii in Gothic obviously
    assert(!gothic['$']);

    CodepointSet emptySet;
    assert(emptySet.length == 0);
    assert(emptySet.empty);

    set = unicode.ASCII;
    // union with the inverse gets all of code points in the Unicode
    assert((set | set.inverted).length == 0x110000);
    // no intersection with inverse
    assert((set & set.inverted).empty);

    CodepointSet someSet;
    someSet.add('0', '5').add('A','Z'+1);
    someSet.add('5', '9'+1);
    assert(someSet['0']);
    assert(someSet['5']);
    assert(someSet['9']);
    assert(someSet['Z']);

    auto lower = unicode.LowerCase;
    auto upper = unicode.UpperCase;
    auto ascii = unicode.ASCII;
    assert((lower & upper).empty); // no intersection
    auto lowerASCII = lower & ascii;
    assert(lowerASCII.byCodepoint.equalS(iota('a', 'z'+1)));
    // throw away all of the lowercase ASCII
    assert((ascii - lower).length == 128 - 26);
    auto onlyOneOf = lower ~ ascii;
    assert(!onlyOneOf['Δ']); // not ASCII and not lowercase
    assert(onlyOneOf['$']); // ASCII and not lowercase
    assert(!onlyOneOf['a']); // ASCII and lowercase
    assert(onlyOneOf['я']); // not ASCII but lowercase

    auto noLetters = ascii - (lower | upper);
    assert(noLetters.length == 128 - 26*2);
    import std.conv;
    assert(unicode.ASCII.to!string() == "[0..128)");
}

// pedantic version for ctfe, and aligned-access only architectures
@trusted uint safeRead24(const ubyte* ptr, size_t idx) pure nothrow @nogc
{
    idx *= 3;
    version(LittleEndian)
        return ptr[idx] + (cast(uint)ptr[idx+1]<<8)
             + (cast(uint)ptr[idx+2]<<16);
    else
        return (cast(uint)ptr[idx]<<16) + (cast(uint)ptr[idx+1]<<8)
             + ptr[idx+2];
}

// ditto
@trusted void safeWrite24(ubyte* ptr, uint val, size_t idx) pure nothrow @nogc
{
    idx *= 3;
    version(LittleEndian)
    {
        ptr[idx] = val & 0xFF;
        ptr[idx+1] = (val>>8) & 0xFF;
        ptr[idx+2] = (val>>16) & 0xFF;
    }
    else
    {
        ptr[idx] = (val>>16) & 0xFF;
        ptr[idx+1] = (val>>8) & 0xFF;
        ptr[idx+2] = val & 0xFF;
    }
}

// unaligned x86-like read/write functions
@trusted uint unalignedRead24(const ubyte* ptr, size_t idx) pure nothrow @nogc
{
    uint* src = cast(uint*)(ptr+3*idx);
    version(LittleEndian)
        return *src & 0xFF_FFFF;
    else
        return *src >> 8;
}

// ditto
@trusted void unalignedWrite24(ubyte* ptr, uint val, size_t idx) pure nothrow @nogc
{
    uint* dest = cast(uint*)(cast(ubyte*)ptr + 3*idx);
    version(LittleEndian)
        *dest = val | (*dest & 0xFF00_0000);
    else
        *dest = (val<<8) | (*dest & 0xFF);
}

uint read24(const ubyte* ptr, size_t idx) @safe pure nothrow @nogc
{
    static if(hasUnalignedReads)
        return __ctfe ? safeRead24(ptr, idx) : unalignedRead24(ptr, idx);
    else
        return safeRead24(ptr, idx);
}

void write24(ubyte* ptr, uint val, size_t idx) @safe pure nothrow @nogc
{
    static if(hasUnalignedReads)
        return __ctfe ? safeWrite24(ptr, val, idx) : unalignedWrite24(ptr, val, idx);
    else
        return safeWrite24(ptr, val, idx);
}
@trusted struct CowArray(SP=GcPolicy)
{
    static auto reuse(uint[] arr)
    {
        CowArray cow;
        cow.data = arr;
        SP.append(cow.data, 1);
        assert(cow.refCount == 1);
        assert(cow.length == arr.length);
        return cow;
    }

    this(Range)(Range range)
        if(isInputRange!Range && hasLength!Range)
    {
        import std.algorithm : copy;
        length = range.length;
        copy(range, data[0..$-1]);
    }

    this(Range)(Range range)
        if(isForwardRange!Range && !hasLength!Range)
    {
        import std.algorithm : copy;
        auto len = walkLength(range.save);
        length = len;
        copy(range, data[0..$-1]);
    }

    this(this)
    {
        if(!empty)
        {
            refCount = refCount + 1;
        }
    }

    ~this()
    {
        if(!empty)
        {
            auto cnt = refCount;
            if(cnt == 1)
                SP.destroy(data);
            else
                refCount = cnt - 1;
        }
    }

    // no ref-count for empty U24 array
    @property bool empty() const { return data.length == 0; }

    // report one less then actual size
    @property size_t length() const
    {
        return data.length ? data.length - 1 : 0;
    }

    //+ an extra slot for ref-count
    @property void length(size_t len)
    {
        import std.algorithm : min, copy;
        if(len == 0)
        {
            if(!empty)
                freeThisReference();
            return;
        }
        immutable total = len + 1; // including ref-count
        if(empty)
        {
            data = SP.alloc!uint(total);
            refCount = 1;
            return;
        }
        auto cur_cnt = refCount;
        if(cur_cnt != 1) // have more references to this memory
        {
            refCount = cur_cnt - 1;
            auto new_data = SP.alloc!uint(total);
            // take shrinking into account
            auto to_copy = min(total, data.length) - 1;
            copy(data[0..to_copy], new_data[0..to_copy]);
            data = new_data; // before setting refCount!
            refCount = 1;
        }
        else // 'this' is the only reference
        {
            // use the realloc (hopefully in-place operation)
            data = SP.realloc(data, total);
            refCount = 1; // setup a ref-count in the new end of the array
        }
    }

    alias opDollar = length;

    uint opIndex()(size_t idx)const
    {
        return data[idx];
    }

    void opIndexAssign(uint val, size_t idx)
    {
        auto cnt = refCount;
        if(cnt != 1)
            dupThisReference(cnt);
        data[idx] = val;
    }

    //
    auto opSlice(size_t from, size_t to)
    {
        if(!empty)
        {
            auto cnt = refCount;
            if(cnt != 1)
                dupThisReference(cnt);
        }
        return data[from .. to];

    }

    //
    auto opSlice(size_t from, size_t to) const
    {
        return data[from .. to];
    }

    // length slices before the ref count
    auto opSlice()
    {
        return opSlice(0, length);
    }

    // ditto
    auto opSlice() const
    {
        return opSlice(0, length);
    }

    void append(Range)(Range range)
        if(isInputRange!Range && hasLength!Range && is(ElementType!Range : uint))
    {
        size_t nl = length + range.length;
        length = nl;
        copy(range, this[nl-range.length..nl]);
    }

    void append()(uint[] val...)
    {
        length = length + val.length;
        data[$-val.length-1 .. $-1] = val[];
    }

    bool opEquals()(auto const ref CowArray rhs)const
    {
        if(empty ^ rhs.empty)
            return false; // one is empty and the other isn't
        return empty || data[0..$-1] == rhs.data[0..$-1];
    }

private:
    // ref-count is right after the data
    @property uint refCount() const
    {
        return data[$-1];
    }

    @property void refCount(uint cnt)
    {
        data[$-1] = cnt;
    }

    void freeThisReference()
    {
        auto count = refCount;
        if(count != 1) // have more references to this memory
        {
            // dec shared ref-count
            refCount = count - 1;
            data = [];
        }
        else
            SP.destroy(data);
        assert(!data.ptr);
    }

    void dupThisReference(uint count)
    in
    {
        assert(!empty && count != 1 && count == refCount);
    }
    body
    {
        import std.algorithm : copy;
        // dec shared ref-count
        refCount = count - 1;
        // copy to the new chunk of RAM
        auto new_data = SP.alloc!uint(data.length);
        // bit-blit old stuff except the counter
        copy(data[0..$-1], new_data[0..$-1]);
        data = new_data; // before setting refCount!
        refCount = 1; // so that this updates the right one
    }

    uint[] data;
}

@safe unittest// Uint24 tests
{
    import std.algorithm;
    import std.conv;
    import std.range;
    void funcRef(T)(ref T u24)
    {
        u24.length = 2;
        u24[1] = 1024;
        T u24_c = u24;
        assert(u24[1] == 1024);
        u24.length = 0;
        assert(u24.empty);
        u24.append([1, 2]);
        assert(equalS(u24[], [1, 2]));
        u24.append(111);
        assert(equalS(u24[], [1, 2, 111]));
        assert(!u24_c.empty && u24_c[1] == 1024);
        u24.length = 3;
        copy(iota(0, 3), u24[]);
        assert(equalS(u24[], iota(0, 3)));
        assert(u24_c[1] == 1024);
    }

    void func2(T)(T u24)
    {
        T u24_2 = u24;
        T u24_3;
        u24_3 = u24_2;
        assert(u24_2 == u24_3);
        assert(equalS(u24[], u24_2[]));
        assert(equalS(u24_2[], u24_3[]));
        funcRef(u24_3);

        assert(equalS(u24_3[], iota(0, 3)));
        assert(!equalS(u24_2[], u24_3[]));
        assert(equalS(u24_2[], u24[]));
        u24_2 = u24_3;
        assert(equalS(u24_2[], iota(0, 3)));
        // to test that passed arg is intact outside
        // plus try out opEquals
        u24 = u24_3;
        u24 = T.init;
        u24_3 = T.init;
        assert(u24.empty);
        assert(u24 == u24_3);
        assert(u24 != u24_2);
    }

    foreach(Policy; TypeTuple!(GcPolicy, ReallocPolicy))
    {
        alias Range = typeof(CowArray!Policy.init[]);
        alias U24A = CowArray!Policy;
        static assert(isForwardRange!Range);
        static assert(isBidirectionalRange!Range);
        static assert(isOutputRange!(Range, uint));
        static assert(isRandomAccessRange!(Range));

        auto arr = U24A([42u, 36, 100]);
        assert(arr[0] == 42);
        assert(arr[1] == 36);
        arr[0] = 72;
        arr[1] = 0xFE_FEFE;
        assert(arr[0] == 72);
        assert(arr[1] == 0xFE_FEFE);
        assert(arr[2] == 100);
        U24A arr2 = arr;
        assert(arr2[0] == 72);
        arr2[0] = 11;
        // test COW-ness
        assert(arr[0] == 72);
        assert(arr2[0] == 11);
        // set this to about 100M to stress-test COW memory management
        foreach(v; 0..10_000)
            func2(arr);
        assert(equalS(arr[], [72, 0xFE_FEFE, 100]));

        auto r2 = U24A(iota(0, 100));
        assert(equalS(r2[], iota(0, 100)), text(r2[]));
        copy(iota(10, 170, 2), r2[10..90]);
        assert(equalS(r2[], chain(iota(0, 10), iota(10, 170, 2), iota(90, 100)))
               , text(r2[]));
    }
}

version(unittest)
{
    private alias AllSets = TypeTuple!(InversionList!GcPolicy, InversionList!ReallocPolicy);
}

@safe unittest// core set primitives test
{
    import std.conv;
    foreach(CodeList; AllSets)
    {
        CodeList a;
        //"plug a hole" test
        a.add(10, 20).add(25, 30).add(15, 27);
        assert(a == CodeList(10, 30), text(a));

        auto x = CodeList.init;
        x.add(10, 20).add(30, 40).add(50, 60);

        a = x;
        a.add(20, 49);//[10, 49) [50, 60)
        assert(a == CodeList(10, 49, 50 ,60));

        a = x;
        a.add(20, 50);
        assert(a == CodeList(10, 60), text(a));

        // simple unions, mostly edge effects
        x = CodeList.init;
        x.add(10, 20).add(40, 60);

        a = x;
        a.add(10, 25); //[10, 25) [40, 60)
        assert(a == CodeList(10, 25, 40, 60));

        a = x;
        a.add(5, 15); //[5, 20) [40, 60)
        assert(a == CodeList(5, 20, 40, 60));

        a = x;
        a.add(0, 10); // [0, 20) [40, 60)
        assert(a == CodeList(0, 20, 40, 60));

        a = x;
        a.add(0, 5); // prepand
        assert(a == CodeList(0, 5, 10, 20, 40, 60), text(a));

        a = x;
        a.add(5, 20);
        assert(a == CodeList(5, 20, 40, 60));

        a = x;
        a.add(3, 37);
        assert(a == CodeList(3, 37, 40, 60));

        a = x;
        a.add(37, 65);
        assert(a == CodeList(10, 20, 37, 65));

        // some tests on helpers for set intersection
        x = CodeList.init.add(10, 20).add(40, 60).add(100, 120);
        a = x;

        auto m = a.skipUpTo(60);
        a.dropUpTo(110, m);
        assert(a == CodeList(10, 20, 40, 60, 110, 120), text(a.data[]));

        a = x;
        a.dropUpTo(100);
        assert(a == CodeList(100, 120), text(a.data[]));

        a = x;
        m = a.skipUpTo(50);
        a.dropUpTo(140, m);
        assert(a == CodeList(10, 20, 40, 50), text(a.data[]));
        a = x;
        a.dropUpTo(60);
        assert(a == CodeList(100, 120), text(a.data[]));
    }
}


//test constructor to work with any order of intervals
@safe unittest
{
    import std.conv, std.range, std.algorithm;
    import std.typecons;
    //ensure constructor handles bad ordering and overlap
    auto c1 = CodepointSet('а', 'я'+1, 'А','Я'+1);
    foreach(ch; chain(iota('а', 'я'+1), iota('А','Я'+1)))
        assert(ch in c1, to!string(ch));

    //contiguos
    assert(CodepointSet(1000, 1006, 1006, 1009)
        .byInterval.equal([tuple(1000, 1009)]));
    //contains
    assert(CodepointSet(900, 1200, 1000, 1100)
        .byInterval.equal([tuple(900, 1200)]));
    //intersect left
    assert(CodepointSet(900, 1100, 1000, 1200)
        .byInterval.equal([tuple(900, 1200)]));
    //intersect right
    assert(CodepointSet(1000, 1200, 900, 1100)
        .byInterval.equal([tuple(900, 1200)]));

    //ditto with extra items at end
    assert(CodepointSet(1000, 1200, 900, 1100, 800, 850)
        .byInterval.equal([tuple(800, 850), tuple(900, 1200)]));
    assert(CodepointSet(900, 1100, 1000, 1200, 800, 850)
        .byInterval.equal([tuple(800, 850), tuple(900, 1200)]));

    //"plug a hole" test
    auto c2 = CodepointSet(20, 40,
        60, 80, 100, 140, 150, 200,
        40, 60, 80, 100, 140, 150
    );
    assert(c2.byInterval.equal([tuple(20, 200)]));

    auto c3 = CodepointSet(
        20, 40, 60, 80, 100, 140, 150, 200,
        0, 10, 15, 100, 10, 20, 200, 220);
    assert(c3.byInterval.equal([tuple(0, 140), tuple(150, 220)]));
}


@safe unittest
{   // full set operations
    import std.conv;
    foreach(CodeList; AllSets)
    {
        CodeList a, b, c, d;

        //"plug a hole"
        a.add(20, 40).add(60, 80).add(100, 140).add(150, 200);
        b.add(40, 60).add(80, 100).add(140, 150);
        c = a | b;
        d = b | a;
        assert(c == CodeList(20, 200), text(CodeList.stringof," ", c));
        assert(c == d, text(c," vs ", d));

        b = CodeList.init.add(25, 45).add(65, 85).add(95,110).add(150, 210);
        c = a | b; //[20,45) [60, 85) [95, 140) [150, 210)
        d = b | a;
        assert(c == CodeList(20, 45, 60, 85, 95, 140, 150, 210), text(c));
        assert(c == d, text(c," vs ", d));

        b = CodeList.init.add(10, 20).add(30,100).add(145,200);
        c = a | b;//[10, 140) [145, 200)
        d = b | a;
        assert(c == CodeList(10, 140, 145, 200));
        assert(c == d, text(c," vs ", d));

        b = CodeList.init.add(0, 10).add(15, 100).add(10, 20).add(200, 220);
        c = a | b;//[0, 140) [150, 220)
        d = b | a;
        assert(c == CodeList(0, 140, 150, 220));
        assert(c == d, text(c," vs ", d));


        a = CodeList.init.add(20, 40).add(60, 80);
        b = CodeList.init.add(25, 35).add(65, 75);
        c = a & b;
        d = b & a;
        assert(c == CodeList(25, 35, 65, 75), text(c));
        assert(c == d, text(c," vs ", d));

        a = CodeList.init.add(20, 40).add(60, 80).add(100, 140).add(150, 200);
        b = CodeList.init.add(25, 35).add(65, 75).add(110, 130).add(160, 180);
        c = a & b;
        d = b & a;
        assert(c == CodeList(25, 35, 65, 75, 110, 130, 160, 180), text(c));
        assert(c == d, text(c," vs ", d));

        a = CodeList.init.add(20, 40).add(60, 80).add(100, 140).add(150, 200);
        b = CodeList.init.add(10, 30).add(60, 120).add(135, 160);
        c = a & b;//[20, 30)[60, 80) [100, 120) [135, 140) [150, 160)
        d = b & a;

        assert(c == CodeList(20, 30, 60, 80, 100, 120, 135, 140, 150, 160),text(c));
        assert(c == d, text(c, " vs ",d));
        assert((c & a) == c);
        assert((d & b) == d);
        assert((c & d) == d);

        b = CodeList.init.add(40, 60).add(80, 100).add(140, 200);
        c = a & b;
        d = b & a;
        assert(c == CodeList(150, 200), text(c));
        assert(c == d, text(c, " vs ",d));
        assert((c & a) == c);
        assert((d & b) == d);
        assert((c & d) == d);

        assert((a & a) == a);
        assert((b & b) == b);

        a = CodeList.init.add(20, 40).add(60, 80).add(100, 140).add(150, 200);
        b = CodeList.init.add(30, 60).add(75, 120).add(190, 300);
        c = a - b;// [30, 40) [60, 75) [120, 140) [150, 190)
        d = b - a;// [40, 60) [80, 100) [200, 300)
        assert(c == CodeList(20, 30, 60, 75, 120, 140, 150, 190), text(c));
        assert(d == CodeList(40, 60, 80, 100, 200, 300), text(d));
        assert(c - d == c, text(c-d, " vs ", c));
        assert(d - c == d, text(d-c, " vs ", d));
        assert(c - c == CodeList.init);
        assert(d - d == CodeList.init);

        a = CodeList.init.add(20, 40).add( 60, 80).add(100, 140).add(150,            200);
        b = CodeList.init.add(10,  50).add(60,                           160).add(190, 300);
        c = a - b;// [160, 190)
        d = b - a;// [10, 20) [40, 50) [80, 100) [140, 150) [200, 300)
        assert(c == CodeList(160, 190), text(c));
        assert(d == CodeList(10, 20, 40, 50, 80, 100, 140, 150, 200, 300), text(d));
        assert(c - d == c, text(c-d, " vs ", c));
        assert(d - c == d, text(d-c, " vs ", d));
        assert(c - c == CodeList.init);
        assert(d - d == CodeList.init);

        a = CodeList.init.add(20,    40).add(60, 80).add(100,      140).add(150,  200);
        b = CodeList.init.add(10, 30).add(45,         100).add(130,             190);
        c = a ~ b; // [10, 20) [30, 40) [45, 60) [80, 130) [140, 150) [190, 200)
        d = b ~ a;
        assert(c == CodeList(10, 20, 30, 40, 45, 60, 80, 130, 140, 150, 190, 200),
               text(c));
        assert(c == d, text(c, " vs ", d));
    }
}

}

@safe unittest// vs single dchar
{
    import std.conv;
    CodepointSet a = CodepointSet(10, 100, 120, 200);
    assert(a - 'A' == CodepointSet(10, 65, 66, 100, 120, 200), text(a - 'A'));
    assert((a & 'B') == CodepointSet(66, 67));
}

@safe unittest// iteration & opIndex
{
    import std.conv;
    import std.typecons;
    foreach(CodeList; TypeTuple!(InversionList!(ReallocPolicy)))
    {
        auto arr = "ABCDEFGHIJKLMabcdefghijklm"d;
        auto a = CodeList('A','N','a', 'n');
        assert(equalS(a.byInterval,
                [tuple(cast(uint)'A', cast(uint)'N'), tuple(cast(uint)'a', cast(uint)'n')]
            ), text(a.byInterval));

        // same @@@BUG as in issue 8949 ?
        version(bug8949)
        {
            assert(equalS(retro(a.byInterval),
                [tuple(cast(uint)'a', cast(uint)'n'), tuple(cast(uint)'A', cast(uint)'N')]
            ), text(retro(a.byInterval)));
        }
        auto achr = a.byCodepoint;
        assert(equalS(achr, arr), text(a.byCodepoint));
        foreach(ch; a.byCodepoint)
            assert(a[ch]);
        auto x = CodeList(100, 500, 600, 900, 1200, 1500);
        assert(equalS(x.byInterval, [ tuple(100, 500), tuple(600, 900), tuple(1200, 1500)]), text(x.byInterval));
        foreach(ch; x.byCodepoint)
            assert(x[ch]);
        static if(is(CodeList == CodepointSet))
        {
            auto y = CodeList(x.byInterval);
            assert(equalS(x.byInterval, y.byInterval));
        }
        assert(equalS(CodepointSet.init.byInterval, cast(Tuple!(uint, uint)[])[]));
        assert(equalS(CodepointSet.init.byCodepoint, cast(dchar[])[]));
    }
}

//============================================================================
// Generic Trie template and various ways to build it
//============================================================================

// debug helper to get a shortened array dump
auto arrayRepr(T)(T x)
{
    import std.conv : text;
    if(x.length > 32)
    {
        return text(x[0..16],"~...~", x[x.length-16..x.length]);
    }
    else
        return text(x);
}

/**
    Maps $(D Key) to a suitable integer index within the range of $(D size_t).
    The mapping is constructed by applying predicates from $(D Prefix) left to right
    and concatenating the resulting bits.

    The first (leftmost) predicate defines the most significant bits of
    the resulting index.
 */
template mapTrieIndex(Prefix...)
{
    size_t mapTrieIndex(Key)(Key key)
        if(isValidPrefixForTrie!(Key, Prefix))
    {
        alias p = Prefix;
        size_t idx;
        foreach(i, v; p[0..$-1])
        {
            idx |= p[i](key);
            idx <<= p[i+1].bitSize;
        }
        idx |= p[$-1](key);
        return idx;
    }
}

/*
    $(D TrieBuilder) is a type used for incremental construction
    of $(LREF Trie)s.

    See $(LREF buildTrie) for generic helpers built on top of it.
*/
@trusted struct TrieBuilder(Value, Key, Args...)
    if(isBitPackableType!Value && isValidArgsForTrie!(Key, Args))
{
    import std.exception : enforce;

private:
    // last index is not stored in table, it is used as an offset to values in a block.
    static if(is(Value == bool))// always pack bool
        alias V = BitPacked!(Value, 1);
    else
        alias V = Value;
    static auto deduceMaxIndex(Preds...)()
    {
        size_t idx = 1;
        foreach(v; Preds)
            idx *= 2^^v.bitSize;
        return idx;
    }

    static if(is(typeof(Args[0]) : Key)) // Args start with upper bound on Key
    {
        alias Prefix = Args[1..$];
        enum lastPageSize = 2^^Prefix[$-1].bitSize;
        enum translatedMaxIndex = mapTrieIndex!(Prefix)(Args[0]);
        enum roughedMaxIndex =
            (translatedMaxIndex + lastPageSize-1)/lastPageSize*lastPageSize;
        // check warp around - if wrapped, use the default deduction rule
        enum maxIndex = roughedMaxIndex < translatedMaxIndex ?
            deduceMaxIndex!(Prefix)() : roughedMaxIndex;
    }
    else
    {
        alias Prefix = Args;
        enum maxIndex = deduceMaxIndex!(Prefix)();
    }

    alias getIndex = mapTrieIndex!(Prefix);

    enum lastLevel = Prefix.length-1;
    struct ConstructState
    {
        size_t idx_zeros, idx_ones;
    }
    // iteration over levels of Trie, each indexes its own level and thus a shortened domain
    size_t[Prefix.length] indices;
    // default filler value to use
    Value defValue;
    // this is a full-width index of next item
    size_t curIndex;
    // all-zeros page index, all-ones page index (+ indicator if there is such a page)
    ConstructState[Prefix.length] state;
    // the table being constructed
    MultiArray!(idxTypes!(Key, fullBitSize!(Prefix), Prefix[0..$]), V) table;

    @disable this();

    //shortcut for index variable at level 'level'
    @property ref idx(size_t level)(){ return indices[level]; }

    // this function assumes no holes in the input so
    // indices are going one by one
    void addValue(size_t level, T)(T val, size_t numVals)
    {
        alias j = idx!level;
        enum pageSize = 1<<Prefix[level].bitSize;
        if(numVals == 0)
            return;
        auto ptr = table.slice!(level);
        if(numVals == 1)
        {
            static if(level == Prefix.length-1)
                ptr[j] = val;
            else
            {// can incur narrowing conversion
                assert(j < ptr.length);
                ptr[j] = force!(typeof(ptr[j]))(val);
            }
            j++;
            if(j % pageSize == 0)
                spillToNextPage!level(ptr);
            return;
        }
        // longer row of values
        // get to the next page boundary
        size_t nextPB = (j + pageSize) & ~(pageSize-1);
        size_t n =  nextPB - j;// can fill right in this page
        if(numVals < n) //fits in current page
        {
            ptr[j..j+numVals]  = val;
            j += numVals;
            return;
        }
        static if(level != 0)//on the first level it always fits
        {
            numVals -= n;
            //write till the end of current page
            ptr[j..j+n]  = val;
            j += n;
            //spill to the next page
            spillToNextPage!level(ptr);
            // page at once loop
            if(state[level].idx_zeros != size_t.max && val == T.init)
            {
                alias NextIdx = typeof(table.slice!(level-1)[0]);
                addValue!(level-1)(force!NextIdx(state[level].idx_zeros),
                    numVals/pageSize);
                ptr = table.slice!level; //table structure might have changed
                numVals %= pageSize;
            }
            else
            {
                while(numVals >= pageSize)
                {
                    numVals -= pageSize;
                    ptr[j..j+pageSize]  = val;
                    j += pageSize;
                    spillToNextPage!level(ptr);
                }
            }
            if(numVals)
            {
                // the leftovers, an incomplete page
                ptr[j..j+numVals]  = val;
                j += numVals;
            }
        }
    }

    void spillToNextPage(size_t level, Slice)(ref Slice ptr)
    {
        // last level (i.e. topmost) has 1 "page"
        // thus it need not to add a new page on upper level
        static if(level != 0)
            spillToNextPageImpl!(level)(ptr);
    }

    // this can re-use the current page if duplicate or allocate a new one
    // it also makes sure that previous levels point to the correct page in this level
    void spillToNextPageImpl(size_t level, Slice)(ref Slice ptr)
    {
        alias NextIdx = typeof(table.slice!(level-1)[0]);
        NextIdx next_lvl_index;
        enum pageSize = 1<<Prefix[level].bitSize;
        assert(idx!level % pageSize == 0);
        auto last = idx!level-pageSize;
        auto slice = ptr[idx!level - pageSize..idx!level];
        size_t j;
        for(j=0; j<last; j+=pageSize)
        {
            if(ptr[j..j+pageSize] == slice)
            {
                // get index to it, reuse ptr space for the next block
                next_lvl_index = force!NextIdx(j/pageSize);
                version(none)
                {
                writefln("LEVEL(%s) page maped idx: %s: 0..%s  ---> [%s..%s]"
                        ,level
                        ,indices[level-1], pageSize, j, j+pageSize);
                writeln("LEVEL(", level
                        , ") mapped page is: ", slice, ": ", arrayRepr(ptr[j..j+pageSize]));
                writeln("LEVEL(", level
                        , ") src page is :", ptr, ": ", arrayRepr(slice[0..pageSize]));
                }
                idx!level -= pageSize; // reuse this page, it is duplicate
                break;
            }
        }
        if(j == last)
        {
    L_allocate_page:
            next_lvl_index = force!NextIdx(idx!level/pageSize - 1);
            if(state[level].idx_zeros == size_t.max && ptr.zeros(j, j+pageSize))
            {
                state[level].idx_zeros = next_lvl_index;
            }
            // allocate next page
            version(none)
            {
            writefln("LEVEL(%s) page allocated: %s"
                     , level, arrayRepr(slice[0..pageSize]));
            writefln("LEVEL(%s) index: %s ; page at this index %s"
                     , level
                     , next_lvl_index
                     , arrayRepr(
                         table.slice!(level)
                          [pageSize*next_lvl_index..(next_lvl_index+1)*pageSize]
                        ));
            }
            table.length!level = table.length!level + pageSize;
        }
    L_know_index:
        // for the previous level, values are indices to the pages in the current level
        addValue!(level-1)(next_lvl_index, 1);
        ptr = table.slice!level; //re-load the slice after moves
    }

    // idx - full-width index to fill with v (full-width index != key)
    // fills everything in the range of [curIndex, idx) with filler
    void putAt(size_t idx, Value v)
    {
        assert(idx >= curIndex);
        size_t numFillers = idx - curIndex;
        addValue!lastLevel(defValue, numFillers);
        addValue!lastLevel(v, 1);
        curIndex = idx + 1;
    }

    // ditto, but sets the range of [idxA, idxB) to v
    void putRangeAt(size_t idxA, size_t idxB, Value v)
    {
        assert(idxA >= curIndex);
        assert(idxB >= idxA);
        size_t numFillers = idxA - curIndex;
        addValue!lastLevel(defValue, numFillers);
        addValue!lastLevel(v, idxB - idxA);
        curIndex = idxB; // open-right
    }

    enum errMsg = "non-monotonic prefix function(s), an unsorted range or "~
        "duplicate key->value mapping";

public:
    /**
        Construct a builder, where $(D filler) is a value
        to indicate empty slots (or "not found" condition).
    */
    this(Value filler)
    {
        curIndex = 0;
        defValue = filler;
        // zeros-page index, ones-page index
        foreach(ref v; state)
            v = ConstructState(size_t.max, size_t.max);
        table = typeof(table)(indices);
        // one page per level is a bootstrap minimum
        foreach(i, Pred; Prefix)
            table.length!i = (1<<Pred.bitSize);
    }

    /**
        Put a value $(D v) into interval as
        mapped by keys from $(D a) to $(D b).
        All slots prior to $(D a) are filled with
        the default filler.
    */
    void putRange(Key a, Key b, Value v)
    {
        auto idxA = getIndex(a), idxB = getIndex(b);
        // indexes of key should always grow
        enforce(idxB >= idxA && idxA >= curIndex, errMsg);
        putRangeAt(idxA, idxB, v);
    }

    /**
        Put a value $(D v) into slot mapped by $(D key).
        All slots prior to $(D key) are filled with the
        default filler.
    */
    void putValue(Key key, Value v)
    {
        import std.conv : text;
        auto idx = getIndex(key);
        enforce(idx >= curIndex, text(errMsg, " ", idx));
        putAt(idx, v);
    }

    /// Finishes construction of Trie, yielding an immutable Trie instance.
    auto build()
    {
        static if(maxIndex != 0) // doesn't cover full range of size_t
        {
            assert(curIndex <= maxIndex);
            addValue!lastLevel(defValue, maxIndex - curIndex);
        }
        else
        {
            if(curIndex != 0 // couldn't wrap around
                || (Prefix.length != 1 && indices[lastLevel] == 0)) // can be just empty
            {
                addValue!lastLevel(defValue, size_t.max - curIndex);
                addValue!lastLevel(defValue, 1);
            }
            // else curIndex already completed the full range of size_t by wrapping around
        }
        return Trie!(V, Key, maxIndex, Prefix)(table);
    }
}

/*
    $(P A generic Trie data-structure for a fixed number of stages.
    The design goal is optimal speed with smallest footprint size.
    )
    $(P It's intentionally read-only and doesn't provide constructors.
     To construct one use a special builder,
     see $(LREF TrieBuilder) and $(LREF buildTrie).
    )

*/
@trusted public struct Trie(Value, Key, Args...)
    if(isValidPrefixForTrie!(Key, Args)
        || (isValidPrefixForTrie!(Key, Args[1..$])
            && is(typeof(Args[0]) : size_t)))
{
    static if(is(typeof(Args[0]) : size_t))
    {
        enum maxIndex = Args[0];
        enum hasBoundsCheck = true;
        alias Prefix = Args[1..$];
    }
    else
    {
        enum hasBoundsCheck = false;
        alias Prefix = Args;
    }

    private this()(typeof(_table) table)
    {
        _table = table;
    }

    // only for constant Tries constructed from precompiled tables
    private this()(const(size_t)[] offsets, const(size_t)[] sizes,
        const(size_t)[] data) const
    {
        _table = typeof(_table)(offsets, sizes, data);
    }

    /*
        $(P Lookup the $(D key) in this $(D Trie). )

        $(P The lookup always succeeds if key fits the domain
        provided during construction. The whole domain defined
        is covered so instead of not found condition
        the sentinel (filler) value could be used. )

        $(P See $(LREF buildTrie), $(LREF TrieBuilder) for how to
        define a domain of $(D Trie) keys and the sentinel value. )

        Note:
        Domain range-checking is only enabled in debug builds
        and results in assertion failure.
    */
    // templated to auto-detect pure, @safe and nothrow
    TypeOfBitPacked!Value opIndex()(Key key) const
    {
        static if(hasBoundsCheck)
            assert(mapTrieIndex!Prefix(key) < maxIndex);
        size_t idx;
        alias p = Prefix;
        idx = cast(size_t)p[0](key);
        foreach(i, v; p[0..$-1])
            idx = cast(size_t)((_table.ptr!i[idx]<<p[i+1].bitSize) + p[i+1](key));
        return _table.ptr!(p.length-1)[idx];
    }

    @property size_t bytes(size_t n=size_t.max)() const
    {
        return _table.bytes!n;
    }

    @property size_t pages(size_t n)() const
    {
        return (bytes!n+2^^(Prefix[n].bitSize-1))
                /2^^Prefix[n].bitSize;
    }

    void store(OutRange)(scope OutRange sink) const
        if(isOutputRange!(OutRange, char))
    {
        _table.store(sink);
    }

private:
    MultiArray!(idxTypes!(Key, fullBitSize!(Prefix), Prefix[0..$]), Value) _table;
}

// create a tuple of 'sliceBits' that slice the 'top' of bits into pieces of sizes 'sizes'
// left-to-right, the most significant bits first
template GetBitSlicing(size_t top, sizes...)
{
    static if(sizes.length > 0)
        alias GetBitSlicing =
            TypeTuple!(sliceBits!(top - sizes[0], top),
                       GetBitSlicing!(top - sizes[0], sizes[1..$]));
    else
        alias GetBitSlicing = TypeTuple!();
}

template callableWith(T)
{
    template callableWith(alias Pred)
    {
        static if(!is(typeof(Pred(T.init))))
            enum callableWith = false;
        else
        {
            alias Result = typeof(Pred(T.init));
            enum callableWith = isBitPackableType!(TypeOfBitPacked!(Result));
        }
    }
}

/*
    Check if $(D Prefix) is a valid set of predicates
    for $(D Trie) template having $(D Key) as the type of keys.
    This requires all predicates to be callable, take
    single argument of type $(D Key) and return unsigned value.
*/
template isValidPrefixForTrie(Key, Prefix...)
{
    enum isValidPrefixForTrie = allSatisfy!(callableWith!Key, Prefix); // TODO: tighten the screws
}

/*
    Check if $(D Args) is a set of maximum key value followed by valid predicates
    for $(D Trie) template having $(D Key) as the type of keys.
*/
template isValidArgsForTrie(Key, Args...)
{
    static if(Args.length > 1)
    {
        enum isValidArgsForTrie = isValidPrefixForTrie!(Key, Args)
            || (isValidPrefixForTrie!(Key, Args[1..$]) && is(typeof(Args[0]) : Key));
    }
    else
        enum isValidArgsForTrie = isValidPrefixForTrie!Args;
}

@property size_t sumOfIntegerTuple(ints...)()
{
    size_t count=0;
    foreach(v; ints)
        count += v;
    return count;
}

/**
    A shorthand for creating a custom multi-level fixed Trie
    from a $(D CodepointSet). $(D sizes) are numbers of bits per level,
    with the most significant bits used first.

    Note: The sum of $(D sizes) must be equal 21.

    See_Also: $(LREF toTrie), which is even simpler.

    Example:
    ---
    {
        import std.stdio;
        auto set = unicode("Number");
        auto trie = codepointSetTrie!(8, 5, 8)(set);
        writeln("Input code points to test:");
        foreach(line; stdin.byLine)
        {
            int count=0;
            foreach(dchar ch; line)
                if(trie[ch])// is number
                    count++;
            writefln("Contains %d number code points.", count);
        }
    }
    ---
*/
public template codepointSetTrie(sizes...)
    if(sumOfIntegerTuple!sizes == 21)
{
    auto codepointSetTrie(Set)(Set set)
        if(isCodepointSet!Set)
    {
        auto builder = TrieBuilder!(bool, dchar, lastDchar+1, GetBitSlicing!(21, sizes))(false);
        foreach(ival; set.byInterval)
            builder.putRange(ival[0], ival[1], true);
        return builder.build();
    }
}

/// Type of Trie generated by codepointSetTrie function.
public template CodepointSetTrie(sizes...)
    if(sumOfIntegerTuple!sizes == 21)
{
    alias Prefix = GetBitSlicing!(21, sizes);
    alias CodepointSetTrie = typeof(TrieBuilder!(bool, dchar, lastDchar+1, Prefix)(false).build());
}

/**
    A slightly more general tool for building fixed $(D Trie)
    for the Unicode data.

    Specifically unlike $(D codepointSetTrie) it's allows creating mappings
    of $(D dchar) to an arbitrary type $(D T).

    Note: Overload taking $(D CodepointSet)s will naturally convert
    only to bool mapping $(D Trie)s.

    Example:
    ---
    // pick characters from the Greek script
    auto set = unicode.Greek;

    // a user-defined property (or an expensive function)
    // that we want to look up
    static uint luckFactor(dchar ch)
    {
        // here we consider a character lucky
        // if its code point has a lot of identical hex-digits
        // e.g. arabic letter DDAL (\u0688) has a "luck factor" of 2
        ubyte[6] nibbles; // 6 4-bit chunks of code point
        uint value = ch;
        foreach(i; 0..6)
        {
            nibbles[i] = value & 0xF;
            value >>= 4;
        }
        uint luck;
        foreach(n; nibbles)
            luck = cast(uint)max(luck, count(nibbles[], n));
        return luck;
    }

    // only unsigned built-ins are supported at the moment
    alias LuckFactor = BitPacked!(uint, 3);

    // create a temporary associative array (AA)
    LuckFactor[dchar] map;
    foreach(ch; set.byCodepoint)
        map[ch] = luckFactor(ch);

    // bits per stage are chosen randomly, fell free to optimize
    auto trie = codepointTrie!(LuckFactor, 8, 5, 8)(map);

    // from now on the AA is not needed
    foreach(ch; set.byCodepoint)
        assert(trie[ch] == luckFactor(ch)); // verify
    // CJK is not Greek, thus it has the default value
    assert(trie['\u4444'] == 0);
    // and here is a couple of quite lucky Greek characters:
    // Greek small letter epsilon with dasia
    assert(trie['\u1F11'] == 3);
    // Ancient Greek metretes sign
    assert(trie['\U00010181'] == 3);
    ---
*/
public template codepointTrie(T, sizes...)
    if(sumOfIntegerTuple!sizes == 21)
{
    alias Prefix = GetBitSlicing!(21, sizes);

    static if(is(TypeOfBitPacked!T == bool))
    {
        auto codepointTrie(Set)(in Set set)
            if(isCodepointSet!Set)
        {
            return codepointSetTrie(set);
        }
    }

    auto codepointTrie()(T[dchar] map, T defValue=T.init)
    {
        return buildTrie!(T, dchar, Prefix)(map, defValue);
    }

    // unsorted range of pairs
    auto codepointTrie(R)(R range, T defValue=T.init)
        if(isInputRange!R
            && is(typeof(ElementType!R.init[0]) : T)
            && is(typeof(ElementType!R.init[1]) : dchar))
    {
        // build from unsorted array of pairs
        // TODO: expose index sorting functions for Trie
        return buildTrie!(T, dchar, Prefix)(range, defValue, true);
    }
}

pure unittest // codepointTrie example
{
    import std.algorithm;
    // pick characters from the Greek script
    auto set = unicode.Greek;

    // a user-defined property (or an expensive function)
    // that we want to look up
    static uint luckFactor(dchar ch)
    {
        // here we consider a character lucky
        // if its code point has a lot of identical hex-digits
        // e.g. arabic letter DDAL (\u0688) has a "luck factor" of 2
        ubyte[6] nibbles; // 6 4-bit chunks of code point
        uint value = ch;
        foreach(i; 0..6)
        {
            nibbles[i] = value & 0xF;
            value >>= 4;
        }
        uint luck;
        foreach(n; nibbles)
            luck = cast(uint)max(luck, count(nibbles[], n));
        return luck;
    }

    // only unsigned built-ins are supported at the moment
    alias LuckFactor = BitPacked!(uint, 3);

    // create a temporary associative array (AA)
    LuckFactor[dchar] map;
    foreach(ch; set.byCodepoint)
        map[ch] = LuckFactor(luckFactor(ch));

    // bits per stage are chosen randomly, fell free to optimize
    auto trie = codepointTrie!(LuckFactor, 8, 5, 8)(map);

    // from now on the AA is not needed
    foreach(ch; set.byCodepoint)
        assert(trie[ch] == luckFactor(ch)); // verify
    // CJK is not Greek, thus it has the default value
    assert(trie['\u4444'] == 0);
    // and here is a couple of quite lucky Greek characters:
    // Greek small letter epsilon with dasia
    assert(trie['\u1F11'] == 3);
    // Ancient Greek metretes sign
    assert(trie['\U00010181'] == 3);

}

/// Type of Trie as generated by codepointTrie function.
public template CodepointTrie(T, sizes...)
    if(sumOfIntegerTuple!sizes == 21)
{
    alias Prefix = GetBitSlicing!(21, sizes);
    alias CodepointTrie = typeof(TrieBuilder!(T, dchar, lastDchar+1, Prefix)(T.init).build());
}

// @@@BUG multiSort can's access private symbols from uni
public template cmpK0(alias Pred)
{
    import std.typecons;
    static bool cmpK0(Value, Key)
        (Tuple!(Value, Key) a, Tuple!(Value, Key) b)
    {
        return Pred(a[1]) < Pred(b[1]);
    }
}

/*
    The most general utility for construction of $(D Trie)s
    short of using $(D TrieBuilder) directly.

    Provides a number of convenience overloads.
    $(D Args) is tuple of maximum key value followed by
    predicates to construct index from key.

    Alternatively if the first argument is not a value convertible to $(D Key)
    then the whole tuple of $(D Args) is treated as predicates
    and the maximum Key is deduced from predicates.
*/
public template buildTrie(Value, Key, Args...)
    if(isValidArgsForTrie!(Key, Args))
{
    static if(is(typeof(Args[0]) : Key)) // prefix starts with upper bound on Key
    {
        alias Prefix = Args[1..$];
    }
    else
        alias Prefix = Args;

    alias getIndex = mapTrieIndex!(Prefix);

    // for multi-sort
    template GetComparators(size_t n)
    {
        static if(n > 0)
            alias GetComparators =
                TypeTuple!(GetComparators!(n-1), cmpK0!(Prefix[n-1]));
        else
            alias GetComparators = TypeTuple!();
    }

    /*
        Build $(D Trie) from a range of a Key-Value pairs,
        assuming it is sorted by Key as defined by the following lambda:
        ------
        (a, b) => mapTrieIndex!(Prefix)(a) < mapTrieIndex!(Prefix)(b)
        ------
        Exception is thrown if it's detected that the above order doesn't hold.

        In other words $(LREF mapTrieIndex) should be a
        monotonically increasing function that maps $(D Key) to an integer.

        See_Also: $(XREF _algorithm, sort),
        $(XREF _range, SortedRange),
        $(XREF _algorithm, setUnion).
    */
    auto buildTrie(Range)(Range range, Value filler=Value.init)
        if(isInputRange!Range && is(typeof(Range.init.front[0]) : Value)
            && is(typeof(Range.init.front[1]) : Key))
    {
        auto builder = TrieBuilder!(Value, Key, Prefix)(filler);
        foreach(v; range)
            builder.putValue(v[1], v[0]);
        return builder.build();
    }

    /*
        If $(D Value) is bool (or BitPacked!(bool, x)) then it's possible
        to build $(D Trie) from a range of open-right intervals of $(D Key)s.
        The requirement  on the ordering of keys (and the behavior on the
        violation of it) is the same as for Key-Value range overload.

        Intervals denote ranges of !$(D filler) i.e. the opposite of filler.
        If no filler provided keys inside of the intervals map to true,
        and $(D filler) is false.
    */
    auto buildTrie(Range)(Range range, Value filler=Value.init)
        if(is(TypeOfBitPacked!Value ==  bool)
            && isInputRange!Range && is(typeof(Range.init.front[0]) : Key)
            && is(typeof(Range.init.front[1]) : Key))
    {
        auto builder = TrieBuilder!(Value, Key, Prefix)(filler);
        foreach(ival; range)
            builder.putRange(ival[0], ival[1], !filler);
        return builder.build();
    }

    auto buildTrie(Range)(Range range, Value filler, bool unsorted)
        if(isInputRange!Range
            && is(typeof(Range.init.front[0]) : Value)
            && is(typeof(Range.init.front[1]) : Key))
    {
        import std.algorithm : multiSort;
        alias Comps = GetComparators!(Prefix.length);
        if(unsorted)
            multiSort!(Comps)(range);
        return buildTrie(range, filler);
    }

    /*
        If $(D Value) is bool (or BitPacked!(bool, x)) then it's possible
        to build $(D Trie) simply from an input range of $(D Key)s.
        The requirement  on the ordering of keys (and the behavior on the
        violation of it) is the same as for Key-Value range overload.

        Keys found in range denote !$(D filler) i.e. the opposite of filler.
        If no filler provided keys map to true, and $(D filler) is false.
    */
    auto buildTrie(Range)(Range range, Value filler=Value.init)
        if(is(TypeOfBitPacked!Value ==  bool)
            && isInputRange!Range && is(typeof(Range.init.front) : Key))
    {
        auto builder = TrieBuilder!(Value, Key, Prefix)(filler);
        foreach(v; range)
            builder.putValue(v, !filler);
        return builder.build();
    }

    /*
        If $(D Key) is unsigned integer $(D Trie) could be constructed from array
        of values where array index serves as key.
    */
    auto buildTrie()(Value[] array, Value filler=Value.init)
        if(isUnsigned!Key)
    {
        auto builder = TrieBuilder!(Value, Key, Prefix)(filler);
        foreach(idx, v; array)
            builder.putValue(idx, v);
        return builder.build();
    }

    /*
        Builds $(D Trie) from associative array.
    */
    auto buildTrie(Key, Value)(Value[Key] map, Value filler=Value.init)
    {
        import std.range : zip, array;
        auto range = array(zip(map.values, map.keys));
        return buildTrie(range, filler, true); // sort it
    }
}

// helper in place of assumeSize to
//reduce mangled name & help DMD inline Trie functors
struct clamp(size_t bits)
{
    static size_t opCall(T)(T arg){ return arg; }
    enum bitSize = bits;
}

struct clampIdx(size_t idx, size_t bits)
{
    static size_t opCall(T)(T arg){ return arg[idx]; }
    enum bitSize = bits;
}

/**
    Conceptual type that outlines the common properties of all UTF Matchers.

    Note: For illustration purposes only, every method
    call results in assertion failure.
    Use $(LREF utfMatcher) to obtain a concrete matcher
    for UTF-8 or UTF-16 encodings.
*/
public struct MatcherConcept
{
    /**
        $(P Perform a semantic equivalent 2 operations:
        decoding a $(CODEPOINT) at front of $(D inp) and testing if
        it belongs to the set of $(CODEPOINTS) of this matcher. )

        $(P The effect on $(D inp) depends on the kind of function called:)

        $(P Match. If the codepoint is found in the set then range $(D inp)
        is advanced by its size in $(S_LINK Code unit, code units),
        otherwise the range is not modifed.)

        $(P Skip. The range is always advanced by the size
        of the tested $(CODEPOINT) regardless of the result of test.)

        $(P Test. The range is left unaffected regardless
        of the result of test.)
    */
    public bool match(Range)(ref Range inp)
        if(isRandomAccessRange!Range && is(ElementType!Range : char))
    {
       assert(false);
    }

    ///ditto
    public bool skip(Range)(ref Range inp)
        if(isRandomAccessRange!Range && is(ElementType!Range : char))
    {
        assert(false);
    }

    ///ditto
    public bool test(Range)(ref Range inp)
        if(isRandomAccessRange!Range && is(ElementType!Range : char))
    {
        assert(false);
    }
    ///
    @safe unittest
    {
        string truth = "2² = 4";
        auto m = utfMatcher!char(unicode.Number);
        assert(m.match(truth)); // '2' is a number all right
        assert(truth == "² = 4"); // skips on match
        assert(m.match(truth)); // so is the superscript '2'
        assert(!m.match(truth)); // space is not a number
        assert(truth == " = 4"); // unaffected on no match
        assert(!m.skip(truth)); // same test ...
        assert(truth == "= 4"); // but skips a codepoint regardless
        assert(!m.test(truth)); // '=' is not a number
        assert(truth == "= 4"); // test never affects argument
    }

    /*
        Advanced feature - provide direct access to a subset of matcher based a
        set of known encoding lengths. Lengths are provided in
        $(S_LINK Code unit, code units). The sub-matcher then may do less
        operations per any $(D test)/$(D match).

        Use with care as the sub-matcher won't match
        any $(CODEPOINTS) that have encoded length that doesn't belong
        to the selected set of lengths. Also the sub-matcher object references
        the parent matcher and must not be used past the liftetime
        of the latter.

        Another caveat of using sub-matcher is that skip is not available
        preciesly because sub-matcher doesn't detect all lengths.
    */
    @property auto subMatcher(Lengths...)()
    {
        assert(0);
        return this;
    }

    ///
    @safe unittest
    {
        auto m = utfMatcher!char(unicode.Number);
        string square = "2²";
        // about sub-matchers
        assert(!m.subMatcher!(2,3,4).test(square)); // ASCII no covered
        assert(m.subMatcher!1.match(square)); // ASCII-only, works
        assert(!m.subMatcher!1.test(square)); // unicode '²'
        assert(m.subMatcher!(2,3,4).match(square));  //
        assert(square == "");
        wstring wsquare = "2²";
        auto m16 = utfMatcher!wchar(unicode.Number);
        // may keep ref, but the orignal (m16) must be kept alive
        auto bmp = m16.subMatcher!1;
        assert(bmp.match(wsquare)); // Okay, in basic multilingual plan
        assert(bmp.match(wsquare)); // And '²' too
    }
}

/**
    Test if $(D M) is an UTF Matcher for ranges of $(D Char).
*/
public enum isUtfMatcher(M, C) = __traits(compiles, (){
    C[] s;
    auto d = s.decoder;
    M m;
    assert(is(typeof(m.match(d)) == bool));
    assert(is(typeof(m.test(d)) == bool));
    static if(is(typeof(m.skip(d))))
    {
        assert(is(typeof(m.skip(d)) == bool));
        assert(is(typeof(m.skip(s)) == bool));
    }
    assert(is(typeof(m.match(s)) == bool));
    assert(is(typeof(m.test(s)) == bool));
});

unittest
{
    alias CharMatcher = typeof(utfMatcher!char(CodepointSet.init));
    alias WcharMatcher = typeof(utfMatcher!wchar(CodepointSet.init));
    static assert(isUtfMatcher!(CharMatcher, char));
    static assert(isUtfMatcher!(CharMatcher, immutable(char)));
    static assert(isUtfMatcher!(WcharMatcher, wchar));
    static assert(isUtfMatcher!(WcharMatcher, immutable(wchar)));
}

enum Mode {
    alwaysSkip,
    neverSkip,
    skipOnMatch
}

mixin template ForwardStrings()
{
    private bool fwdStr(string fn, C)(ref C[] str) const pure
    {
        alias type = typeof(units(str));
        return mixin(fn~"(*cast(type*)&str)");
    }
}

template Utf8Matcher()
{
    enum validSize(int sz) = sz >= 1 && sz <=4;

    void badEncoding() pure @safe
    {
        import std.utf;
        throw new UTFException("Invalid UTF-8 sequence");
    }

    //for 1-stage ASCII
    alias AsciiSpec = TypeTuple!(bool, char, clamp!7);
    //for 2-stage lookup of 2 byte UTF-8 sequences
    alias Utf8Spec2 = TypeTuple!(bool, char[2],
        clampIdx!(0, 5), clampIdx!(1, 6));
    //ditto for 3 byte
    alias Utf8Spec3 = TypeTuple!(bool, char[3],
        clampIdx!(0, 4),
        clampIdx!(1, 6),
        clampIdx!(2, 6)
    );
    //ditto for 4 byte
    alias Utf8Spec4 = TypeTuple!(bool, char[4],
        clampIdx!(0, 3), clampIdx!(1, 6),
        clampIdx!(2, 6), clampIdx!(3, 6)
    );
    alias Tables = TypeTuple!(
        typeof(TrieBuilder!(AsciiSpec)(false).build()),
        typeof(TrieBuilder!(Utf8Spec2)(false).build()),
        typeof(TrieBuilder!(Utf8Spec3)(false).build()),
        typeof(TrieBuilder!(Utf8Spec4)(false).build())
    );
    alias Table(int size) = Tables[size-1];

    enum leadMask(size_t size) = (cast(size_t)1<<(7 - size))-1;
    enum encMask(size_t size) = ((1<<size)-1)<<(8-size);

    char truncate()(char ch) pure @safe
    {
        ch -= 0x80;
        if (ch < 0x40)
        {
            return ch;
        }
        else
        {
            badEncoding();
            return cast(char)0;
        }
    }

    static auto encode(size_t sz)(dchar ch)
        if(sz > 1)
    {
        import std.utf : encode;
        char[4] buf;
        std.utf.encode(buf, ch);
        char[sz] ret;
        buf[0] &= leadMask!sz;
        foreach(n; 1..sz)
            buf[n] = buf[n] & 0x3f; //keep 6 lower bits
        ret[] = buf[0..sz];
        return ret;
    }

    auto build(Set)(Set set)
    {
        import std.algorithm : map;
        auto ascii = set & unicode.ASCII;
        auto utf8_2 = set & CodepointSet(0x80, 0x800);
        auto utf8_3 = set & CodepointSet(0x800, 0x1_0000);
        auto utf8_4 = set & CodepointSet(0x1_0000, lastDchar+1);
        auto asciiT = ascii.byCodepoint.map!(x=>cast(char)x).buildTrie!(AsciiSpec);
        auto utf8_2T = utf8_2.byCodepoint.map!(x=>encode!2(x)).buildTrie!(Utf8Spec2);
        auto utf8_3T = utf8_3.byCodepoint.map!(x=>encode!3(x)).buildTrie!(Utf8Spec3);
        auto utf8_4T = utf8_4.byCodepoint.map!(x=>encode!4(x)).buildTrie!(Utf8Spec4);
        alias Ret = Impl!(1,2,3,4);
        return Ret(asciiT, utf8_2T, utf8_3T, utf8_4T);
    }

    // Bootstrap UTF-8 static matcher interface
    // from 3 primitives: tab!(size), lookup and Sizes
    mixin template DefMatcher()
    {
        import std.format : format;
        enum hasASCII = staticIndexOf!(1, Sizes) >= 0;
        alias UniSizes = Erase!(1, Sizes);

        //generate dispatch code sequence for unicode parts
        static auto genDispatch()
        {
            string code;
            foreach(size; UniSizes)
                code ~= format(q{
                    if ((ch & ~leadMask!%d) == encMask!(%d))
                        return lookup!(%d, mode)(inp);
                    else
                }, size, size, size);
            static if (Sizes.length == 4) //covers all code unit cases
                code ~= "{ badEncoding(); return false; }";
            else
                code ~= "return false;"; //may be just fine but not covered
            return code;
        }
        enum dispatch = genDispatch();

        public bool match(Range)(ref Range inp) const pure @trusted
            if(isRandomAccessRange!Range && is(ElementType!Range : char))
        {
            enum mode = Mode.skipOnMatch;
            assert(!inp.empty);
            auto ch = inp[0];
            static if(hasASCII)
            {
                if (ch < 0x80)
                {
                    bool r = tab!1[ch];
                    if(r)
                        inp.popFront();
                    return r;
                }
                else
                    mixin(dispatch);
            }
            else
                mixin(dispatch);
        }

        static if(Sizes.length == 4) // can skip iff can detect all encodings
        {
            public bool skip(Range)(ref Range inp) const pure @trusted
                if(isRandomAccessRange!Range && is(ElementType!Range : char))
            {
                enum mode = Mode.alwaysSkip;
                assert(!inp.empty);
                auto ch = inp[0];
                static if(hasASCII)
                {
                    if (ch < 0x80)
                    {
                        inp.popFront();
                        return tab!1[ch];
                    }
                    else
                        mixin(dispatch);
                }
                else
                    mixin(dispatch);
            }
        }

        public bool test(Range)(ref Range inp) const pure @trusted
            if(isRandomAccessRange!Range && is(ElementType!Range : char))
        {
            enum mode = Mode.neverSkip;
            assert(!inp.empty);
            auto ch = inp[0];
            static if(hasASCII)
            {
                if (ch < 0x80)
                    return tab!1[ch];
                else
                    mixin(dispatch);
            }
            else
                mixin(dispatch);
        }

        bool match(C)(ref C[] str) const pure @trusted
            if(isSomeChar!C)
        {
            return fwdStr!"match"(str);
        }

        bool skip(C)(ref C[] str) const pure @trusted
            if(isSomeChar!C)
        {
            return fwdStr!"skip"(str);
        }

        bool test(C)(ref C[] str) const pure @trusted
            if(isSomeChar!C)
        {
            return fwdStr!"test"(str);
        }

        mixin ForwardStrings;
    }

    struct Impl(Sizes...)
    {
        static assert(allSatisfy!(validSize, Sizes),
            "Only lengths of 1, 2, 3 and 4 code unit are possible for UTF-8");
    private:
        //pick tables for chosen sizes
        alias OurTabs = staticMap!(Table, Sizes);
        OurTabs tables;
        mixin DefMatcher;
        //static disptach helper UTF size ==> table
        alias tab(int i) = tables[i - 1];

        package @property auto subMatcher(SizesToPick...)() @trusted
        {
            return CherryPick!(Impl, SizesToPick)(&this);
        }

        bool lookup(int size, Mode mode, Range)(ref Range inp) const pure @trusted
        {
            import std.typecons;
            if(inp.length < size)
            {
                badEncoding();
                return false;
            }
            char[size] needle = void;
            needle[0] = leadMask!size & inp[0];
            foreach(i; staticIota!(1, size))
            {
                needle[i] = truncate(inp[i]);
            }
            //overlong encoding checks
            static if(size == 2)
            {
                //0x80-0x7FF
                //got 6 bits in needle[1], must use at least 8 bits
                //must use at least 2 bits in needle[1]
                if(needle[0] < 2) badEncoding();
            }
            else static if(size == 3)
            {
                //0x800-0xFFFF
                //got 6 bits in needle[2], must use at least 12bits
                //must use 6 bits in needle[1] or anything in needle[0]
                if(needle[0] == 0 && needle[1] < 0x20) badEncoding();
            }
            else static if(size == 4)
            {
                //0x800-0xFFFF
                //got 2x6=12 bits in needle[2..3] must use at least 17bits
                //must use 5 bits (or above) in needle[1] or anything in needle[0]
                if(needle[0] == 0 && needle[1] < 0x10) badEncoding();
            }
            static if(mode == Mode.alwaysSkip)
            {
                inp.popFrontN(size);
                return tab!size[needle];
            }
            else static if(mode == Mode.neverSkip)
            {
                return tab!size[needle];
            }
            else
            {
                static assert(mode == Mode.skipOnMatch);
                if (tab!size[needle])
                {
                    inp.popFrontN(size);
                    return true;
                }
                else
                    return false;
            }
        }
    }

    struct CherryPick(I, Sizes...)
    {
        static assert(allSatisfy!(validSize, Sizes),
            "Only lengths of 1, 2, 3 and 4 code unit are possible for UTF-8");
    private:
        I* m;
        @property ref tab(int i)() const pure { return m.tables[i - 1]; }
        bool lookup(int size, Mode mode, Range)(ref Range inp) const pure
        {
            return m.lookup!(size, mode)(inp);
        }
        mixin DefMatcher;
    }
}

template Utf16Matcher()
{
    enum validSize(int sz) = sz >= 1 && sz <=2;

    void badEncoding() pure
    {
        import std.utf;
        throw new UTFException("Invalid UTF-16 sequence");
    }

    alias Seq = TypeTuple;
    // 1-stage ASCII
    alias AsciiSpec = Seq!(bool, wchar, clamp!7);
    //2-stage BMP
    alias BmpSpec = Seq!(bool, wchar, sliceBits!(7, 16), sliceBits!(0, 7));
    //4-stage - full Unicode
    //assume that 0xD800 & 0xDC00 bits are cleared
    //thus leaving 10 bit per wchar to worry about
    alias UniSpec = Seq!(bool, wchar[2],
        assumeSize!(x=>x[0]>>4, 6), assumeSize!(x=>x[0]&0xf, 4),
        assumeSize!(x=>x[1]>>6, 4), assumeSize!(x=>x[1]&0x3f, 6),
    );
    alias Ascii = typeof(TrieBuilder!(AsciiSpec)(false).build());
    alias Bmp = typeof(TrieBuilder!(BmpSpec)(false).build());
    alias Uni = typeof(TrieBuilder!(UniSpec)(false).build());

    auto encode2(dchar ch)
    {
        ch -= 0x1_0000;
        assert(ch <= 0xF_FFFF);
        wchar[2] ret;
        //do not put surrogate bits, they are sliced off
        ret[0] = (ch>>10);
        ret[1] = (ch & 0xFFF);
        return ret;
    }

    auto build(Set)(Set set)
    {
        import std.algorithm : map;
        auto ascii = set & unicode.ASCII;
        auto bmp = (set & CodepointSet.fromIntervals(0x80, 0xFFFF+1))
            - CodepointSet.fromIntervals(0xD800, 0xDFFF+1);
        auto other = set - (bmp | ascii);
        auto asciiT = ascii.byCodepoint.map!(x=>cast(char)x).buildTrie!(AsciiSpec);
        auto bmpT = bmp.byCodepoint.map!(x=>cast(wchar)x).buildTrie!(BmpSpec);
        auto otherT = other.byCodepoint.map!(x=>encode2(x)).buildTrie!(UniSpec);
        alias Ret = Impl!(1,2);
        return Ret(asciiT, bmpT, otherT);
    }

    //bootstrap full UTF-16 matcher interace from
    //sizeFlags, lookupUni and ascii
    mixin template DefMatcher()
    {
        public bool match(Range)(ref Range inp) const pure @trusted
            if(isRandomAccessRange!Range && is(ElementType!Range : wchar))
        {
            enum mode = Mode.skipOnMatch;
            assert(!inp.empty);
            auto ch = inp[0];
            static if(sizeFlags & 1)
            {
                if (ch < 0x80)
                {
                  if (ascii[ch])
                  {
                      inp.popFront();
                      return true;
                  }
                  else
                      return false;
                }
                return lookupUni!mode(inp);
            }
            else
                return lookupUni!mode(inp);
        }

        static if(Sizes.length == 2)
        {
            public bool skip(Range)(ref Range inp) const pure @trusted
                if(isRandomAccessRange!Range && is(ElementType!Range : wchar))
            {
                enum mode = Mode.alwaysSkip;
                assert(!inp.empty);
                auto ch = inp[0];
                static if(sizeFlags & 1)
                {
                    if (ch < 0x80)
                    {
                        inp.popFront();
                        return ascii[ch];
                    }
                    else
                        return lookupUni!mode(inp);
                }
                else
                    return lookupUni!mode(inp);
            }
        }

        public bool test(Range)(ref Range inp) const pure @trusted
            if(isRandomAccessRange!Range && is(ElementType!Range : wchar))
        {
            enum mode = Mode.neverSkip;
            assert(!inp.empty);
            auto ch = inp[0];
            static if(sizeFlags & 1)
                return ch < 0x80 ? ascii[ch] : lookupUni!mode(inp);
            else
                return lookupUni!mode(inp);
        }

        bool match(C)(ref C[] str) const pure @trusted
            if(isSomeChar!C)
        {
            return fwdStr!"match"(str);
        }

        bool skip(C)(ref C[] str) const pure @trusted
            if(isSomeChar!C)
        {
            return fwdStr!"skip"(str);
        }

        bool test(C)(ref C[] str) const pure @trusted
            if(isSomeChar!C)
        {
            return fwdStr!"test"(str);
        }

        mixin ForwardStrings; //dispatch strings to range versions
    }

    struct Impl(Sizes...)
        if(Sizes.length >= 1 && Sizes.length <= 2)
    {
    private:
        static assert(allSatisfy!(validSize, Sizes),
            "Only lengths of 1 and 2 code units are possible in UTF-16");
        static if(Sizes.length > 1)
            enum sizeFlags = Sizes[0] | Sizes[1];
        else
            enum sizeFlags = Sizes[0];

        static if(sizeFlags & 1)
        {
            Ascii ascii;
            Bmp bmp;
        }
        static if(sizeFlags & 2)
        {
            Uni uni;
        }
        mixin DefMatcher;

        package @property auto subMatcher(SizesToPick...)() @trusted
        {
            return CherryPick!(Impl, SizesToPick)(&this);
        }

        bool lookupUni(Mode mode, Range)(ref Range inp) const pure
        {
            wchar x = cast(wchar)(inp[0] - 0xD800);
            //not a high surrogate
            if(x > 0x3FF)
            {
                //low surrogate
                if(x <= 0x7FF) badEncoding();
                static if(sizeFlags & 1)
                {
                    auto ch = inp[0];
                    static if(mode == Mode.alwaysSkip)
                        inp.popFront();
                    static if(mode == Mode.skipOnMatch)
                    {
                        if (bmp[ch])
                        {
                            inp.popFront();
                            return true;
                        }
                        else
                            return false;
                    }
                    else
                        return bmp[ch];
                }
                else //skip is not available for sub-matchers, so just false
                    return false;
            }
            else
            {
                static if(sizeFlags & 2)
                {
                    if(inp.length < 2)
                        badEncoding();
                    wchar y = cast(wchar)(inp[1] - 0xDC00);
                    //not a low surrogate
                    if(y > 0x3FF)
                        badEncoding();
                    wchar[2] needle = [inp[0] & 0x3ff, inp[1] & 0x3ff];
                    static if(mode == Mode.alwaysSkip)
                        inp.popFrontN(2);
                    static if(mode == Mode.skipOnMatch)
                    {
                        if (uni[needle])
                        {
                            inp.popFrontN(2);
                            return true;
                        }
                        else
                            return false;
                    }
                    else
                        return uni[needle];
                }
                else //ditto
                    return false;
            }
        }
    }

    struct CherryPick(I, Sizes...)
        if(Sizes.length >= 1 && Sizes.length <= 2)
    {
    private:
        I* m;
        enum sizeFlags = I.sizeFlags;

        static if(sizeFlags & 1)
        {
            @property ref ascii()() const pure{ return m.ascii; }
        }

        bool lookupUni(Mode mode, Range)(ref Range inp) const pure
        {
            return m.lookupUni!mode(inp);
        }
        mixin DefMatcher;
        static assert(allSatisfy!(validSize, Sizes),
            "Only lengths of 1 and 2 code units are possible in UTF-16");
    }
}

private auto utf8Matcher(Set)(Set set) @trusted
{
    return Utf8Matcher!().build(set);
}

private auto utf16Matcher(Set)(Set set) @trusted
{
    return Utf16Matcher!().build(set);
}

/**
    Constructs a matcher object
    to classify $(CODEPOINTS) from the $(D set) for encoding
    that has $(D Char) as code unit.

    See $(LREF MatcherConcept) for API outline.
*/
public auto utfMatcher(Char, Set)(Set set) @trusted
    if(isCodepointSet!Set)
{
    static if(is(Char : char))
        return utf8Matcher(set);
    else static if(is(Char : wchar))
        return utf16Matcher(set);
    else static if(is(Char : dchar))
        static assert(false, "UTF-32 needs no decoding,
            and thus not supported by utfMatcher");
    else
        static assert(false, "Only character types 'char' and 'wchar' are allowed");
}


//a range of code units, packed with index to speed up forward iteration
package auto decoder(C)(C[] s, size_t offset=0) @safe pure nothrow @nogc
    if(is(C : wchar) || is(C : char))
{
    static struct Decoder
    {
    pure nothrow:
        C[] str;
        size_t idx;
        @property C front(){ return str[idx]; }
        @property C back(){ return str[$-1]; }
        void popFront(){ idx++; }
        void popBack(){ str = str[0..$-1]; }
        void popFrontN(size_t n){ idx += n; }
        @property bool empty(){ return idx == str.length; }
        @property auto save(){ return this; }
        auto opIndex(size_t i){ return str[idx+i]; }
        @property size_t length(){ return str.length - idx; }
        alias opDollar = length;
        auto opSlice(size_t a, size_t b){ return Decoder(str[0..idx+b], idx+a); }
    }
    static assert(isRandomAccessRange!Decoder);
    static assert(is(ElementType!Decoder : C));
    return Decoder(s, offset);
}

/*
    Expose UTF string $(D s) as a random-access
    range of $(S_LINK Code unit, code units).
*/
package auto units(C)(C[] s) @safe pure nothrow @nogc
    if(is(C : wchar) || is(C : char))
{
    static struct Units
    {
    pure nothrow:
        C[] str;
        @property C front(){ return str[0]; }
        @property C back(){ return str[$-1]; }
        void popFront(){ str = str[1..$]; }
        void popBack(){ str = str[0..$-1]; }
        void popFrontN(size_t n){ str = str[n..$]; }
        @property bool empty(){ return 0 == str.length; }
        @property auto save(){ return this; }
        auto opIndex(size_t i){ return str[i]; }
        @property size_t length(){ return str.length; }
        alias opDollar = length;
        auto opSlice(size_t a, size_t b){ return Units(str[a..b]); }
    }
    static assert(isRandomAccessRange!Units);
    static assert(is(ElementType!Units : C));
    return Units(s);
}

@safe unittest
{
    import std.range;
    string rs = "hi! ネемног砀 текста";
    auto codec = rs.decoder;
    auto utf8 =  utf8Matcher(unicode.Letter);
    auto asc = utf8.subMatcher!(1);
    auto uni = utf8.subMatcher!(2,3,4);
    assert(asc.test(codec));
    assert(!uni.match(codec));
    assert(utf8.skip(codec));
    assert(codec.idx == 1);

    assert(!uni.match(codec));
    assert(asc.test(codec));
    assert(utf8.skip(codec));
    assert(codec.idx == 2);
    assert(!asc.match(codec));

    assert(!utf8.test(codec));
    assert(!utf8.skip(codec));

    assert(!asc.test(codec));
    assert(!utf8.test(codec));
    assert(!utf8.skip(codec));
    assert(utf8.test(codec));
    foreach(i; 0..7)
    {
        assert(!asc.test(codec));
        assert(uni.test(codec));
        assert(utf8.skip(codec));
    }
    assert(!utf8.test(codec));
    assert(!utf8.skip(codec));
    //the same with match where applicable
    codec = rs.decoder;
    assert(utf8.match(codec));
    assert(codec.idx == 1);
    assert(utf8.match(codec));
    assert(codec.idx == 2);
    assert(!utf8.match(codec));
    assert(codec.idx == 2);
    assert(!utf8.skip(codec));
    assert(!utf8.skip(codec));

    foreach(i; 0..7)
    {
        assert(!asc.test(codec));
        assert(utf8.test(codec));
        assert(utf8.match(codec));
    }
    auto i = codec.idx;
    assert(!utf8.match(codec));
    assert(codec.idx == i);
}

@safe unittest
{
    import std.range;
    static bool testAll(Matcher, Range)(ref Matcher m, ref Range r)
    {
        bool t = m.test(r);
        auto save = r.idx;
        assert(t == m.match(r));
        assert(r.idx == save || t); //ether no change or was match
        r.idx = save;
        static if(is(typeof(m.skip(r))))
        {
            assert(t == m.skip(r));
            assert(r.idx != save); //always changed
            r.idx = save;
        }
        return t;
    }
    auto utf16 = utfMatcher!wchar(unicode.L);
    auto bmp = utf16.subMatcher!1;
    auto nonBmp = utf16.subMatcher!1;
    auto utf8 = utfMatcher!char(unicode.L);
    auto ascii = utf8.subMatcher!1;
    auto uni2 = utf8.subMatcher!2;
    auto uni3 = utf8.subMatcher!3;
    auto uni24 = utf8.subMatcher!(2,4);
    foreach(ch; unicode.L.byCodepoint.stride(3))
    {
        import std.utf : encode;
        char[4] buf;
        wchar[2] buf16;
        auto len = std.utf.encode(buf, ch);
        auto len16 = std.utf.encode(buf16, ch);
        auto c8 = buf[0..len].decoder;
        auto c16 = buf16[0..len16].decoder;
        assert(testAll(utf16, c16));
        assert(testAll(bmp, c16) || len16 != 1);
        assert(testAll(nonBmp, c16) || len16 != 2);

        assert(testAll(utf8, c8));

        //submatchers return false on out of their domain
        assert(testAll(ascii, c8) || len != 1);
        assert(testAll(uni2, c8) || len != 2);
        assert(testAll(uni3, c8) || len != 3);
        assert(testAll(uni24, c8) || (len != 2 && len !=4));
    }
}

// cover decode fail cases of Matcher
unittest
{
    import std.exception : collectException;
    import std.format : format;
    import std.algorithm;
    auto utf16 = utfMatcher!wchar(unicode.L);
    auto utf8 = utfMatcher!char(unicode.L);
    //decode failure cases UTF-8
    alias fails8 = TypeTuple!("\xC1", "\x80\x00","\xC0\x00", "\xCF\x79",
        "\xFF\x00\0x00\0x00\x00", "\xC0\0x80\0x80\x80", "\x80\0x00\0x00\x00",
        "\xCF\x00\0x00\0x00\x00");
    foreach(msg; fails8){
        assert(collectException((){
            auto s = msg;
            import std.utf;
            size_t idx = 0;
            //decode(s, idx);
            utf8.test(s);
        }()), format("%( %2x %)", cast(ubyte[])msg));
    }
    //decode failure cases UTF-16
    alias fails16 = TypeTuple!([0xD811], [0xDC02]);
    foreach(msg; fails16){
        assert(collectException((){
            auto s = msg.map!(x => cast(wchar)x);
            utf16.test(s);
        }()));
    }
}

/++
    Convenience function to construct optimal configurations for
    packed Trie from any $(D set) of $(CODEPOINTS).

    The parameter $(D level) indicates the number of trie levels to use,
    allowed values are: 1, 2, 3 or 4. Levels represent different trade-offs
    speed-size wise.

    $(P Level 1 is fastest and the most memory hungry (a bit array). )
    $(P Level 4 is the slowest and has the smallest footprint. )

    See the $(S_LINK Synopsis, Synopsis) section for example.

    Note:
    Level 4 stays very practical (being faster and more predictable)
    compared to using direct lookup on the $(D set) itself.


+/
public auto toTrie(size_t level, Set)(Set set)
    if(isCodepointSet!Set)
{
    static if(level == 1)
        return codepointSetTrie!(21)(set);
    else static if(level == 2)
        return codepointSetTrie!(10, 11)(set);
    else static if(level == 3)
        return codepointSetTrie!(8, 5, 8)(set);
    else static if(level == 4)
         return codepointSetTrie!(6, 4, 4, 7)(set);
    else
        static assert(false,
            "Sorry, toTrie doesn't support levels > 4, use codepointSetTrie directly");
}

/**
    $(P Builds a $(D Trie) with typically optimal speed-size trade-off
    and wraps it into a delegate of the following type:
    $(D bool delegate(dchar ch)). )

    $(P Effectively this creates a 'tester' lambda suitable
    for algorithms like std.algorithm.find that take unary predicates. )

    See the $(S_LINK Synopsis, Synopsis) section for example.
*/
public auto toDelegate(Set)(Set set)
    if(isCodepointSet!Set)
{
    // 3 is very small and is almost as fast as 2-level (due to CPU caches?)
    auto t = toTrie!3(set);
    return (dchar ch) => t[ch];
}

/**
    $(P Opaque wrapper around unsigned built-in integers and
    code unit (char/wchar/dchar) types.
    Parameter $(D sz) indicates that the value is confined
    to the range of [0, 2^^sz$(RPAREN). With this knowledge it can be
    packed more tightly when stored in certain
    data-structures like trie. )

    Note:
    $(P The $(D BitPacked!(T, sz)) is implicitly convertible to $(D T)
    but not vise-versa. Users have to ensure the value fits in
    the range required and use the $(D cast)
    operator to perform the conversion.)
*/
struct BitPacked(T, size_t sz)
    if(isIntegral!T || is(T:dchar))
{
    enum bitSize = sz;
    T _value;
    alias _value this;
}

/*
    Depending on the form of the passed argument $(D bitSizeOf) returns
    the amount of bits required to represent a given type
    or a return type of a given functor.
*/
template bitSizeOf(Args...)
    if(Args.length == 1)
{
    alias T = Args[0];
    static if(__traits(compiles, { size_t val = T.bitSize; })) //(is(typeof(T.bitSize) : size_t))
    {
        enum bitSizeOf = T.bitSize;
    }
    else static if(is(ReturnType!T dummy == BitPacked!(U, bits), U, size_t bits))
    {
        enum bitSizeOf = bitSizeOf!(ReturnType!T);
    }
    else
    {
        enum bitSizeOf = T.sizeof*8;
    }
}

/**
    Tests if $(D T) is some instantiation of $(LREF BitPacked)!(U, x)
    and thus suitable for packing.
*/
template isBitPacked(T)
{
    static if(is(T dummy == BitPacked!(U, bits), U, size_t bits))
        enum isBitPacked = true;
    else
        enum isBitPacked = false;
}

/**
    Gives the type $(D U) from $(LREF BitPacked)!(U, x)
    or $(D T) itself for every other type.
*/
template TypeOfBitPacked(T)
{
    static if(is(T dummy == BitPacked!(U, bits), U, size_t bits))
        alias TypeOfBitPacked = U;
    else
        alias TypeOfBitPacked = T;
}

/*
    Wrapper, used in definition of custom data structures from $(D Trie) template.
    Applying it to a unary lambda function indicates that the returned value always
    fits within $(D bits) of bits.
*/
struct assumeSize(alias Fn, size_t bits)
{
    enum bitSize = bits;
    static auto ref opCall(T)(auto ref T arg)
    {
        return Fn(arg);
    }
}

/*
    A helper for defining lambda function that yields a slice
    of certain bits from an unsigned integral value.
    The resulting lambda is wrapped in assumeSize and can be used directly
    with $(D Trie) template.
*/
struct sliceBits(size_t from, size_t to)
{
    //for now bypass assumeSize, DMD has trouble inlining it
    enum bitSize = to-from;
    static auto opCall(T)(T x)
    out(result)
    {
        assert(result < (1<<to-from));
    }
    body
    {
        static assert(from < to);
        static if(from == 0)
            return x & ((1<<to)-1);
        else
        return (x >> from) & ((1<<(to-from))-1);
    }
}

@safe pure nothrow @nogc uint low_8(uint x) { return x&0xFF; }
@safe pure nothrow @nogc uint midlow_8(uint x){ return (x&0xFF00)>>8; }
alias lo8 = assumeSize!(low_8, 8);
alias mlo8 = assumeSize!(midlow_8, 8);

static assert(bitSizeOf!lo8 == 8);
static assert(bitSizeOf!(sliceBits!(4, 7)) == 3);
static assert(bitSizeOf!(BitPacked!(uint, 2)) == 2);

template Sequence(size_t start, size_t end)
{
    static if(start < end)
        alias Sequence = TypeTuple!(start, Sequence!(start+1, end));
    else
        alias Sequence = TypeTuple!();
}

//---- TRIE TESTS ----
unittest
{
    import std.conv;
    import std.algorithm;
    import std.range;
    static trieStats(TRIE)(TRIE t)
    {
        version(std_uni_stats)
        {
            import std.stdio;
            writeln("---TRIE FOOTPRINT STATS---");
            foreach(i; staticIota!(0, t.table.dim) )
            {
                writefln("lvl%s = %s bytes;  %s pages"
                         , i, t.bytes!i, t.pages!i);
            }
            writefln("TOTAL: %s bytes", t.bytes);
            version(none)
            {
                writeln("INDEX (excluding value level):");
                foreach(i; staticIota!(0, t.table.dim-1) )
                    writeln(t.table.slice!(i)[0..t.table.length!i]);
            }
            writeln("---------------------------");
        }
    }
    //@@@BUG link failure, lambdas not found by linker somehow (in case of trie2)
    // alias lo8   = assumeSize!(8, function (uint x) { return x&0xFF; });
    // alias next8 = assumeSize!(7, function (uint x) { return (x&0x7F00)>>8; });
    alias Set = CodepointSet;
    auto set = Set('A','Z','a','z');
    auto trie = buildTrie!(bool, uint, 256, lo8)(set.byInterval);// simple bool array
    for(int a='a'; a<'z';a++)
        assert(trie[a]);
    for(int a='A'; a<'Z';a++)
        assert(trie[a]);
    for(int a=0; a<'A'; a++)
        assert(!trie[a]);
    for(int a ='Z'; a<'a'; a++)
        assert(!trie[a]);
    trieStats(trie);

    auto redundant2 = Set(
        1, 18, 256+2, 256+111, 512+1, 512+18, 768+2, 768+111);
    auto trie2 = buildTrie!(bool, uint, 1024, mlo8, lo8)(redundant2.byInterval);
    trieStats(trie2);
    foreach(e; redundant2.byCodepoint)
        assert(trie2[e], text(cast(uint)e, " - ", trie2[e]));
    foreach(i; 0..1024)
    {
        assert(trie2[i] == (i in redundant2));
    }


    auto redundant3 = Set(
          2,    4,    6,    8,    16,
       2+16, 4+16, 16+6, 16+8, 16+16,
       2+32, 4+32, 32+6, 32+8,
      );

    enum max3 = 256;
    // sliceBits
    auto trie3 = buildTrie!(bool, uint, max3,
            sliceBits!(6,8), sliceBits!(4,6), sliceBits!(0,4)
        )(redundant3.byInterval);
    trieStats(trie3);
    foreach(i; 0..max3)
        assert(trie3[i] == (i in redundant3), text(cast(uint)i));

    auto redundant4 = Set(
            10, 64, 64+10, 128, 128+10, 256, 256+10, 512,
            1000, 2000, 3000, 4000, 5000, 6000
        );
    enum max4 = 2^^16;
    auto trie4 = buildTrie!(bool, size_t, max4,
            sliceBits!(13, 16), sliceBits!(9, 13), sliceBits!(6, 9) , sliceBits!(0, 6)
        )(redundant4.byInterval);
    foreach(i; 0..max4){
        if(i in redundant4)
            assert(trie4[i], text(cast(uint)i));
    }
    trieStats(trie4);

        alias mapToS = mapTrieIndex!(useItemAt!(0, char));
        string[] redundantS = ["tea", "start", "orange"];
        redundantS.sort!((a,b) => mapToS(a) < mapToS(b))();
        auto strie = buildTrie!(bool, string, useItemAt!(0, char))(redundantS);
        // using first char only
        assert(redundantS == ["orange", "start", "tea"]);
        assert(strie["test"], text(strie["test"]));
        assert(!strie["aea"]);
        assert(strie["s"]);

    // a bit size test
    auto a = array(map!(x => to!ubyte(x))(iota(0, 256)));
    auto bt = buildTrie!(bool, ubyte, sliceBits!(7, 8), sliceBits!(5, 7), sliceBits!(0, 5))(a);
    trieStats(bt);
    foreach(i; 0..256)
        assert(bt[cast(ubyte)i]);
}

template useItemAt(size_t idx, T)
    if(isIntegral!T || is(T: dchar))
{
    size_t impl(in T[] arr){ return arr[idx]; }
    alias useItemAt = assumeSize!(impl, 8*T.sizeof);
}

template useLastItem(T)
{
    size_t impl(in T[] arr){ return arr[$-1]; }
    alias useLastItem = assumeSize!(impl, 8*T.sizeof);
}

template fullBitSize(Prefix...)
{
    static if(Prefix.length > 0)
        enum fullBitSize = bitSizeOf!(Prefix[0])+fullBitSize!(Prefix[1..$]);
    else
        enum fullBitSize = 0;
}

template idxTypes(Key, size_t fullBits, Prefix...)
{
    static if(Prefix.length == 1)
    {// the last level is value level, so no index once reduced to 1-level
        alias idxTypes = TypeTuple!();
    }
    else
    {
        // Important note on bit packing
        // Each level has to hold enough of bits to address the next one
        // The bottom level is known to hold full bit width
        // thus it's size in pages is full_bit_width - size_of_last_prefix
        // Recourse on this notion
        alias idxTypes =
            TypeTuple!(
                idxTypes!(Key, fullBits - bitSizeOf!(Prefix[$-1]), Prefix[0..$-1]),
                BitPacked!(typeof(Prefix[$-2](Key.init)), fullBits - bitSizeOf!(Prefix[$-1]))
            );
    }
}

//============================================================================

@safe pure int comparePropertyName(Char1, Char2)(const(Char1)[] a, const(Char2)[] b) if (is(Char1 : dchar) && is(Char2 : dchar))
{
    import std.ascii : toLower;
    import std.algorithm : cmp, map, filter;
    static bool pred(dchar c) {return !c.isWhite && c != '-' && c != '_';}
    return cmp(
        a.map!toLower.filter!pred,
        b.map!toLower.filter!pred);
}

@safe pure unittest
{
    assert(!comparePropertyName("foo-bar", "fooBar"));
}

bool propertyNameLess(Char1, Char2)(const(Char1)[] a, const(Char2)[] b) @safe pure if (is(Char1 : dchar) && is(Char2 : dchar))
{
    return comparePropertyName(a, b) < 0;
}

//============================================================================
// Utilities for compression of Unicode code point sets
//============================================================================

@safe void compressTo(uint val, ref ubyte[] arr) pure nothrow
{
    // not optimized as usually done 1 time (and not public interface)
    if(val < 128)
        arr ~= cast(ubyte)val;
    else if(val < (1<<13))
    {
        arr ~= (0b1_00<<5) | cast(ubyte)(val>>8);
        arr ~= val & 0xFF;
    }
    else
    {
        assert(val < (1<<21));
        arr ~= (0b1_01<<5) | cast(ubyte)(val>>16);
        arr ~= (val >> 8) & 0xFF;
        arr ~= val  & 0xFF;
    }
}

@safe uint decompressFrom(const(ubyte)[] arr, ref size_t idx) pure
{
    import std.exception : enforce;
    uint first = arr[idx++];
    if(!(first & 0x80)) // no top bit -> [0..127]
        return first;
    uint extra = ((first>>5) & 1) + 1; // [1, 2]
    uint val = (first & 0x1F);
    enforce(idx + extra <= arr.length, "bad code point interval encoding");
    foreach(j; 0..extra)
        val = (val<<8) | arr[idx+j];
    idx += extra;
    return val;
}


package ubyte[] compressIntervals(Range)(Range intervals)
    if(isInputRange!Range && isIntegralPair!(ElementType!Range))
{
    ubyte[] storage;
    uint base = 0;
    // RLE encode
    foreach(val; intervals)
    {
        compressTo(val[0]-base, storage);
        base = val[0];
        if(val[1] != lastDchar+1) // till the end of the domain so don't store it
        {
            compressTo(val[1]-base, storage);
            base = val[1];
        }
    }
    return storage;
}

@safe pure unittest
{
    import std.typecons;
    auto run = [tuple(80, 127), tuple(128, (1<<10)+128)];
    ubyte[] enc = [cast(ubyte)80, 47, 1, (0b1_00<<5) | (1<<2), 0];
    assert(compressIntervals(run) == enc);
    auto run2 = [tuple(0, (1<<20)+512+1), tuple((1<<20)+512+4, lastDchar+1)];
    ubyte[] enc2 = [cast(ubyte)0, (0b1_01<<5) | (1<<4), 2, 1, 3]; // odd length-ed
    assert(compressIntervals(run2) == enc2);
    size_t  idx = 0;
    assert(decompressFrom(enc, idx) == 80);
    assert(decompressFrom(enc, idx) == 47);
    assert(decompressFrom(enc, idx) == 1);
    assert(decompressFrom(enc, idx) == (1<<10));
    idx = 0;
    assert(decompressFrom(enc2, idx) == 0);
    assert(decompressFrom(enc2, idx) == (1<<20)+512+1);
    assert(equalS(decompressIntervals(compressIntervals(run)), run));
    assert(equalS(decompressIntervals(compressIntervals(run2)), run2));
}

// Creates a range of $(D CodepointInterval) that lazily decodes compressed data.
@safe package auto decompressIntervals(const(ubyte)[] data) pure
{
    return DecompressedIntervals(data);
}

@safe struct DecompressedIntervals
{
pure:
    const(ubyte)[] _stream;
    size_t _idx;
    CodepointInterval _front;

    this(const(ubyte)[] stream)
    {
        _stream = stream;
        popFront();
    }

    @property CodepointInterval front()
    {
        assert(!empty);
        return _front;
    }

    void popFront()
    {
        if(_idx == _stream.length)
        {
            _idx = size_t.max;
            return;
        }
        uint base = _front[1];
        _front[0] = base + decompressFrom(_stream, _idx);
        if(_idx == _stream.length)// odd length ---> till the end
            _front[1] = lastDchar+1;
        else
        {
            base = _front[0];
            _front[1] = base + decompressFrom(_stream, _idx);
        }
    }

    @property bool empty() const
    {
        return _idx == size_t.max;
    }

    @property DecompressedIntervals save() { return this; }
}

static assert(isInputRange!DecompressedIntervals);
static assert(isForwardRange!DecompressedIntervals);
//============================================================================

version(std_uni_bootstrap){}
else
{

// helper for looking up code point sets
@trusted ptrdiff_t findUnicodeSet(alias table, C)(in C[] name) pure
{
    import std.range : assumeSorted;
    import std.algorithm : map;
    auto range = assumeSorted!((a,b) => propertyNameLess(a,b))
        (table.map!"a.name"());
    size_t idx = range.lowerBound(name).length;
    if(idx < range.length && comparePropertyName(range[idx], name) == 0)
        return idx;
    return -1;
}

// another one that loads it
@trusted bool loadUnicodeSet(alias table, Set, C)(in C[] name, ref Set dest) pure
{
    auto idx = findUnicodeSet!table(name);
    if(idx >= 0)
    {
        dest = Set(asSet(table[idx].compressed));
        return true;
    }
    return false;
}

@trusted bool loadProperty(Set=CodepointSet, C)
    (in C[] name, ref Set target) pure
{
    alias ucmp = comparePropertyName;
    // conjure cumulative properties by hand
    if(ucmp(name, "L") == 0 || ucmp(name, "Letter") == 0)
    {
        target = asSet(uniProps.Lu);
        target |= asSet(uniProps.Ll);
        target |= asSet(uniProps.Lt);
        target |= asSet(uniProps.Lo);
        target |= asSet(uniProps.Lm);
    }
    else if(ucmp(name,"LC") == 0 || ucmp(name,"Cased Letter")==0)
    {
        target = asSet(uniProps.Ll);
        target |= asSet(uniProps.Lu);
        target |= asSet(uniProps.Lt);// Title case
    }
    else if(ucmp(name, "M") == 0 || ucmp(name, "Mark") == 0)
    {
        target = asSet(uniProps.Mn);
        target |= asSet(uniProps.Mc);
        target |= asSet(uniProps.Me);
    }
    else if(ucmp(name, "N") == 0 || ucmp(name, "Number") == 0)
    {
        target = asSet(uniProps.Nd);
        target |= asSet(uniProps.Nl);
        target |= asSet(uniProps.No);
    }
    else if(ucmp(name, "P") == 0 || ucmp(name, "Punctuation") == 0)
    {
        target = asSet(uniProps.Pc);
        target |= asSet(uniProps.Pd);
        target |= asSet(uniProps.Ps);
        target |= asSet(uniProps.Pe);
        target |= asSet(uniProps.Pi);
        target |= asSet(uniProps.Pf);
        target |= asSet(uniProps.Po);
    }
    else if(ucmp(name, "S") == 0 || ucmp(name, "Symbol") == 0)
    {
        target = asSet(uniProps.Sm);
        target |= asSet(uniProps.Sc);
        target |= asSet(uniProps.Sk);
        target |= asSet(uniProps.So);
    }
    else if(ucmp(name, "Z") == 0 || ucmp(name, "Separator") == 0)
    {
        target = asSet(uniProps.Zs);
        target |= asSet(uniProps.Zl);
        target |= asSet(uniProps.Zp);
    }
    else if(ucmp(name, "C") == 0 || ucmp(name, "Other") == 0)
    {
        target = asSet(uniProps.Co);
        target |= asSet(uniProps.Lo);
        target |= asSet(uniProps.No);
        target |= asSet(uniProps.So);
        target |= asSet(uniProps.Po);
    }
    else if(ucmp(name, "graphical") == 0){
        target = asSet(uniProps.Alphabetic);

        target |= asSet(uniProps.Mn);
        target |= asSet(uniProps.Mc);
        target |= asSet(uniProps.Me);

        target |= asSet(uniProps.Nd);
        target |= asSet(uniProps.Nl);
        target |= asSet(uniProps.No);

        target |= asSet(uniProps.Pc);
        target |= asSet(uniProps.Pd);
        target |= asSet(uniProps.Ps);
        target |= asSet(uniProps.Pe);
        target |= asSet(uniProps.Pi);
        target |= asSet(uniProps.Pf);
        target |= asSet(uniProps.Po);

        target |= asSet(uniProps.Zs);

        target |= asSet(uniProps.Sm);
        target |= asSet(uniProps.Sc);
        target |= asSet(uniProps.Sk);
        target |= asSet(uniProps.So);
    }
    else if(ucmp(name, "any") == 0)
        target = Set.fromIntervals(0, 0x110000);
    else if(ucmp(name, "ascii") == 0)
        target = Set.fromIntervals(0, 0x80);
    else
        return loadUnicodeSet!(uniProps.tab)(name, target);
    return true;
}

// CTFE-only helper for checking property names at compile-time
@safe bool isPrettyPropertyName(C)(in C[] name)
{
    import std.algorithm : find;
    auto names = [
        "L", "Letter",
        "LC", "Cased Letter",
        "M", "Mark",
        "N", "Number",
        "P", "Punctuation",
        "S", "Symbol",
        "Z", "Separator",
        "Graphical",
        "any",
        "ascii"
    ];
    auto x = find!(x => comparePropertyName(x, name) == 0)(names);
    return !x.empty;
}

// ditto, CTFE-only, not optimized
@safe private static bool findSetName(alias table, C)(in C[] name)
{
    return findUnicodeSet!table(name) >= 0;
}

template SetSearcher(alias table, string kind)
{
    /// Run-time checked search.
    static auto opCall(C)(in C[] name)
        if(is(C : dchar))
    {
        import std.conv : to;
        CodepointSet set;
        if(loadUnicodeSet!table(name, set))
            return set;
        throw new Exception("No unicode set for "~kind~" by name "
            ~name.to!string()~" was found.");
    }
    /// Compile-time checked search.
    static @property auto opDispatch(string name)()
    {
        static if(findSetName!table(name))
        {
            CodepointSet set;
            loadUnicodeSet!table(name, set);
            return set;
        }
        else
            static assert(false, "No unicode set for "~kind~" by name "
                ~name~" was found.");
    }
}

/**
    A single entry point to lookup Unicode $(CODEPOINT) sets by name or alias of
    a block, script or general category.

    It uses well defined standard rules of property name lookup.
    This includes fuzzy matching of names, so that
    'White_Space', 'white-SpAce' and 'whitespace' are all considered equal
    and yield the same set of white space $(CHARACTERS).
*/
@safe public struct unicode
{
    /**
        Performs the lookup of set of $(CODEPOINTS)
        with compile-time correctness checking.
        This short-cut version combines 3 searches:
        across blocks, scripts, and common binary properties.

        Note that since scripts and blocks overlap the
        usual trick to disambiguate is used - to get a block use
        $(D unicode.InBlockName), to search a script
        use $(D unicode.ScriptName).

        See_Also: $(LREF block), $(LREF script)
        and (not included in this search) $(LREF hangulSyllableType).

        Example:
        ---
        auto ascii = unicode.ASCII;
        assert(ascii['A']);
        assert(ascii['~']);
        assert(!ascii['\u00e0']);
        // matching is case-insensitive
        assert(ascii == unicode.ascII);
        assert(!ascii['à']);
        // underscores, '-' and whitespace in names are ignored too
        auto latin = unicode.in_latin1_Supplement;
        assert(latin['à']);
        assert(!latin['$']);
        // BTW Latin 1 Supplement is a block, hence "In" prefix
        assert(latin == unicode("In Latin 1 Supplement"));
        import std.exception;
        // run-time look up throws if no such set is found
        assert(collectException(unicode("InCyrilliac")));
        ---
    */

    static @property auto opDispatch(string name)() pure
    {
        static if(findAny(name))
            return loadAny(name);
        else
            static assert(false, "No unicode set by name "~name~" was found.");
    }

    /**
        The same lookup across blocks, scripts, or binary properties,
        but performed at run-time.
        This version is provided for cases where $(D name)
        is not known beforehand; otherwise compile-time
        checked $(LREF opDispatch) is typically a better choice.

        See the $(S_LINK Unicode properties, table of properties) for available
        sets.
    */
    static auto opCall(C)(in C[] name)
        if(is(C : dchar))
    {
        return loadAny(name);
    }

    /**
        Narrows down the search for sets of $(CODEPOINTS) to all Unicode blocks.

        Note:
        Here block names are unambiguous as no scripts are searched
        and thus to search use simply $(D unicode.block.BlockName) notation.

        See $(S_LINK Unicode properties, table of properties) for available sets.

        Example:
        ---
        // use .block for explicitness
        assert(unicode.block.Greek_and_Coptic == unicode.InGreek_and_Coptic);
        ---

        See_Also: $(S_LINK Unicode properties, table of properties).
    */
    struct block
    {
        mixin SetSearcher!(blocks.tab, "block");
    }

    /**
        Narrows down the search for sets of $(CODEPOINTS) to all Unicode scripts.

        See the $(S_LINK Unicode properties, table of properties) for available
        sets.

        Example:
        ---
        auto arabicScript = unicode.script.arabic;
        auto arabicBlock = unicode.block.arabic;
        // there is an intersection between script and block
        assert(arabicBlock['؁']);
        assert(arabicScript['؁']);
        // but they are different
        assert(arabicBlock != arabicScript);
        assert(arabicBlock == unicode.inArabic);
        assert(arabicScript == unicode.arabic);
        ---
    */
    struct script
    {
        mixin SetSearcher!(scripts.tab, "script");
    }

    /**
        Fetch a set of $(CODEPOINTS) that have the given hangul syllable type.

        Other non-binary properties (once supported) follow the same
        notation - $(D unicode.propertyName.propertyValue) for compile-time
        checked access and $(D unicode.propertyName(propertyValue))
        for run-time checked one.

        See the $(S_LINK Unicode properties, table of properties) for available
        sets.

        Example:
        ---
        // L here is syllable type not Letter as in unicode.L short-cut
        auto leadingVowel = unicode.hangulSyllableType("L");
        // check that some leading vowels are present
        foreach(vowel; '\u1110'..'\u115F')
            assert(leadingVowel[vowel]);
        assert(leadingVowel == unicode.hangulSyllableType.L);
        ---
    */
    struct hangulSyllableType
    {
        mixin SetSearcher!(hangul.tab, "hangul syllable type");
    }

private:
    alias ucmp = comparePropertyName;

    static bool findAny(string name)
    {
        return isPrettyPropertyName(name)
            || findSetName!(uniProps.tab)(name) || findSetName!(scripts.tab)(name)
            || (ucmp(name[0..2],"In") == 0 && findSetName!(blocks.tab)(name[2..$]));
    }

    static auto loadAny(Set=CodepointSet, C)(in C[] name) pure
    {
        import std.conv : to;
        Set set;
        bool loaded = loadProperty(name, set) || loadUnicodeSet!(scripts.tab)(name, set)
            || (name.length > 2 && ucmp(name[0..2],"In") == 0
                && loadUnicodeSet!(blocks.tab)(name[2..$], set));
        if(loaded)
            return set;
        throw new Exception("No unicode set by name "~name.to!string()~" was found.");
    }

    // FIXME: re-disable once the compiler is fixed
    // Disabled to prevent the mistake of creating instances of this pseudo-struct.
    //@disable ~this();
}

unittest
{
    import std.exception : collectException;
    auto ascii = unicode.ASCII;
    assert(ascii['A']);
    assert(ascii['~']);
    assert(!ascii['\u00e0']);
    // matching is case-insensitive
    assert(ascii == unicode.ascII);
    assert(!ascii['à']);
    // underscores, '-' and whitespace in names are ignored too
    auto latin = unicode.Inlatin1_Supplement;
    assert(latin['à']);
    assert(!latin['$']);
    // BTW Latin 1 Supplement is a block, hence "In" prefix
    assert(latin == unicode("In Latin 1 Supplement"));
    import std.exception;
    // R-T look up throws if no such set is found
    assert(collectException(unicode("InCyrilliac")));
    assert(collectException(unicode("X")));

    assert(unicode.block.Greek_and_Coptic == unicode.InGreek_and_Coptic);

    // L here is explicitly syllable type not "Letter" as in unicode.L
    auto leadingVowel = unicode.hangulSyllableType("L");
    // check that some leading vowels are present
    foreach(vowel; '\u1110'..'\u115F'+1)
        assert(leadingVowel[vowel]);
    assert(leadingVowel == unicode.hangulSyllableType.L);

    auto arabicScript = unicode.script.arabic;
    auto arabicBlock = unicode.block.arabic;
    // there is an intersection between script and block
    assert(arabicBlock['؁']);
    assert(arabicScript['؁']);
    // but they are different
    assert(arabicBlock != arabicScript);
    assert(arabicBlock == unicode.inArabic);
    assert(arabicScript == unicode.arabic);
}

unittest
{
    assert(unicode("InHebrew") == asSet(blocks.Hebrew));
    assert(unicode("separator") == (asSet(uniProps.Zs) | asSet(uniProps.Zl) | asSet(uniProps.Zp)));
    assert(unicode("In-Kharoshthi") == asSet(blocks.Kharoshthi));
}

enum EMPTY_CASE_TRIE = ushort.max;// from what gen_uni uses internally

// control - '\r'
enum controlSwitch = `
    case '\u0000':..case '\u0008':case '\u000E':..case '\u001F':case '\u007F':..case '\u0084':case '\u0086':..case '\u009F': case '\u0009':..case '\u000C': case '\u0085':
`;
// TODO: redo the most of hangul stuff algorithmically in case of Graphemes too
// kill unrolled switches

private static bool isRegionalIndicator(dchar ch) @safe
{
    return ch >= '\U0001F1E6' && ch <= '\U0001F1FF';
}

template genericDecodeGrapheme(bool getValue)
{
    alias graphemeExtend = graphemeExtendTrie;
    alias spacingMark = mcTrie;
    static if(getValue)
        alias Value = Grapheme;
    else
        alias Value = void;

    Value genericDecodeGrapheme(Input)(ref Input range)
    {
        enum GraphemeState {
            Start,
            CR,
            RI,
            L,
            V,
            LVT
        }
        static if(getValue)
            Grapheme grapheme;
        auto state = GraphemeState.Start;
        enum eat = q{
            static if(getValue)
                grapheme ~= ch;
            range.popFront();
        };

        dchar ch;
        assert(!range.empty, "Attempting to decode grapheme from an empty " ~ Input.stringof);
        while(!range.empty)
        {
            ch = range.front;
            final switch(state) with(GraphemeState)
            {
            case Start:
                mixin(eat);
                if(ch == '\r')
                    state = CR;
                else if(isRegionalIndicator(ch))
                    state = RI;
                else if(isHangL(ch))
                    state = L;
                else if(hangLV[ch] || isHangV(ch))
                    state = V;
                else if(hangLVT[ch])
                    state = LVT;
                else if(isHangT(ch))
                    state = LVT;
                else
                {
                    switch(ch)
                    {
                    mixin(controlSwitch);
                        goto L_End;
                    default:
                        goto L_End_Extend;
                    }
                }
            break;
            case CR:
                if(ch == '\n')
                    mixin(eat);
                goto L_End_Extend;
            case RI:
                if(isRegionalIndicator(ch))
                    mixin(eat);
                else
                    goto L_End_Extend;
            break;
            case L:
                if(isHangL(ch))
                    mixin(eat);
                else if(isHangV(ch) || hangLV[ch])
                {
                    state = V;
                    mixin(eat);
                }
                else if(hangLVT[ch])
                {
                    state = LVT;
                    mixin(eat);
                }
                else
                    goto L_End_Extend;
            break;
            case V:
                if(isHangV(ch))
                    mixin(eat);
                else if(isHangT(ch))
                {
                    state = LVT;
                    mixin(eat);
                }
                else
                    goto L_End_Extend;
            break;
            case LVT:
                if(isHangT(ch))
                {
                    mixin(eat);
                }
                else
                    goto L_End_Extend;
            break;
            }
        }
    L_End_Extend:
        while(!range.empty)
        {
            ch = range.front;
            // extend & spacing marks
            if(!graphemeExtend[ch] && !spacingMark[ch])
                break;
            mixin(eat);
        }
    L_End:
        static if(getValue)
            return grapheme;
    }

}

public: // Public API continues

/++
    Computes the length of grapheme cluster starting at $(D index).
    Both the resulting length and the $(D index) are measured
    in $(S_LINK Code unit, code units).

    Params:
        C = type that is implicitly convertible to $(D dchars)
        input = array of grapheme clusters
        index = starting index into $(D input[])

    Returns:
        length of grapheme cluster

    Example:
    ---
    // ASCII as usual is 1 code unit, 1 code point etc.
    assert(graphemeStride("  ", 1) == 1);
    // A + combing ring above
    string city = "A\u030Arhus";
    size_t first = graphemeStride(city, 0);
    assert(first == 3); //\u030A has 2 UTF-8 code units
    assert(city[0..first] == "A\u030A");
    assert(city[first..$] == "rhus");
    ---
+/
size_t graphemeStride(C)(in C[] input, size_t index)
    if(is(C : dchar))
{
    auto src = input[index..$];
    auto n = src.length;
    genericDecodeGrapheme!(false)(src);
    return n - src.length;
}

// for now tested separately see test_grapheme.d
@safe unittest
{
    assert(graphemeStride("  ", 1) == 1);
    // A + combing ring above
    string city = "A\u030Arhus";
    size_t first = graphemeStride(city, 0);
    assert(first == 3); //\u030A has 2 UTF-8 code units
    assert(city[0..first] == "A\u030A");
    assert(city[first..$] == "rhus");
}

/++
    Reads one full grapheme cluster from an input range of dchar $(D inp).

    For examples see the $(LREF Grapheme) below.

    Note:
    This function modifies $(D inp) and thus $(D inp)
    must be an L-value.
+/
Grapheme decodeGrapheme(Input)(ref Input inp)
    if(isInputRange!Input && is(Unqual!(ElementType!Input) == dchar))
{
    return genericDecodeGrapheme!true(inp);
}

unittest
{
    Grapheme gr;
    string s = " \u0020\u0308 ";
    gr = decodeGrapheme(s);
    assert(gr.length == 1 && gr[0] == ' ');
    gr = decodeGrapheme(s);
    assert(gr.length == 2 && equalS(gr[0..2], " \u0308"));
    s = "\u0300\u0308\u1100";
    assert(equalS(decodeGrapheme(s)[], "\u0300\u0308"));
    assert(equalS(decodeGrapheme(s)[], "\u1100"));
    s = "\u11A8\u0308\uAC01";
    assert(equalS(decodeGrapheme(s)[], "\u11A8\u0308"));
    assert(equalS(decodeGrapheme(s)[], "\uAC01"));
}

/++
    $(P Iterate a string by grapheme.)

    $(P Useful for doing string manipulation that needs to be aware
    of graphemes.)

    See_Also:
        $(LREF byCodePoint)
+/
auto byGrapheme(Range)(Range range)
    if(isInputRange!Range && is(Unqual!(ElementType!Range) == dchar))
{
    // TODO: Bidirectional access
    static struct Result
    {
        private Range _range;
        private Grapheme _front;

        bool empty() @property
        {
            return _front.length == 0;
        }

        Grapheme front() @property
        {
            return _front;
        }

        void popFront()
        {
            _front = _range.empty ? Grapheme.init : _range.decodeGrapheme();
        }

        static if(isForwardRange!Range)
        {
            Result save() @property
            {
                return Result(_range.save, _front);
            }
        }
    }

    auto result = Result(range);
    result.popFront();
    return result;
}

///
unittest
{
    import std.conv;
    import std.range;
    import std.algorithm;
    auto text = "noe\u0308l"; // noël using e + combining diaeresis
    assert(text.walkLength == 5); // 5 code points

    auto gText = text.byGrapheme;
    assert(gText.walkLength == 4); // 4 graphemes

    assert(gText.take(3).equal("noe\u0308".byGrapheme));
    assert(gText.drop(3).equal("l".byGrapheme));
}

// For testing non-forward-range input ranges
version(unittest)
private static struct InputRangeString
{
    private string s;

    bool empty() @property { return s.empty; }
    dchar front() @property { return s.front; }
    void popFront() { s.popFront(); }
}

unittest
{
    import std.conv;
    import std.range;
    import std.algorithm;
    assert("".byGrapheme.walkLength == 0);

    auto reverse = "le\u0308on";
    assert(reverse.walkLength == 5);

    auto gReverse = reverse.byGrapheme;
    assert(gReverse.walkLength == 4);

    foreach(text; TypeTuple!("noe\u0308l"c, "noe\u0308l"w, "noe\u0308l"d))
    {
        assert(text.walkLength == 5);
        static assert(isForwardRange!(typeof(text)));

        auto gText = text.byGrapheme;
        static assert(isForwardRange!(typeof(gText)));
        assert(gText.walkLength == 4);
        assert(gText.array.retro.equal(gReverse));
    }

    auto nonForwardRange = InputRangeString("noe\u0308l").byGrapheme;
    static assert(!isForwardRange!(typeof(nonForwardRange)));
    assert(nonForwardRange.walkLength == 4);
}

/++
    $(P Lazily transform a range of $(LREF Grapheme)s to a range of code points.)

    $(P Useful for converting the result to a string after doing operations
    on graphemes.)

    $(P Acts as the identity function when given a range of code points.)
+/
auto byCodePoint(Range)(Range range)
    if(isInputRange!Range && is(Unqual!(ElementType!Range) == Grapheme))
{
    // TODO: Propagate bidirectional access
    static struct Result
    {
        private Range _range;
        private size_t i = 0;

        bool empty() @property
        {
            return _range.empty;
        }

        dchar front() @property
        {
            return _range.front[i];
        }

        void popFront()
        {
            ++i;

            if(i >= _range.front.length)
            {
                _range.popFront();
                i = 0;
            }
        }

        static if(isForwardRange!Range)
        {
            Result save() @property
            {
                return Result(_range.save, i);
            }
        }
    }

    return Result(range);
}

/// Ditto
Range byCodePoint(Range)(Range range)
    if(isInputRange!Range && is(Unqual!(ElementType!Range) == dchar))
{
    return range;
}

///
unittest
{
    import std.conv : text;
    import std.range;

    string s = "noe\u0308l"; // noël

    // reverse it and convert the result to a string
    string reverse = s.byGrapheme
        .array
        .retro
        .byCodePoint
        .text;

    assert(reverse == "le\u0308on"); // lëon
}

unittest
{
    import std.conv;
    import std.algorithm;
    assert("".byGrapheme.byCodePoint.equal(""));

    string text = "noe\u0308l";
    static assert(is(typeof(text.byCodePoint) == string));

    auto gText = InputRangeString(text).byGrapheme;
    static assert(!isForwardRange!(typeof(gText)));

    auto cpText = gText.byCodePoint;
    static assert(!isForwardRange!(typeof(cpText)));

    assert(cpText.walkLength == text.walkLength);
}

@trusted:

/++
    $(P A structure designed to effectively pack $(CHARACTERS)
    of a $(CLUSTER).
    )

    $(P $(D Grapheme) has value semantics so 2 copies of a $(D Grapheme)
    always refer to distinct objects. In most actual scenarios a $(D Grapheme)
    fits on the stack and avoids memory allocation overhead for all but quite
    long clusters.
    )

    Example:
    ---
    import std.algorithm;
    string bold = "ku\u0308hn";

    // note that decodeGrapheme takes parameter by ref
    // slicing a grapheme yields a range of dchar
    assert(decodeGrapheme(bold)[].equal("k"));

    // the next grapheme is 2 characters long
    auto wideOne = decodeGrapheme(bold);
    assert(wideOne.length == 2);
    assert(wideOne[].equal("u\u0308"));

    // the usual range manipulation is possible
    assert(wideOne[].filter!isMark.equal("\u0308"));
    ---

    See_Also: $(LREF decodeGrapheme), $(LREF graphemeStride)
+/
@trusted struct Grapheme
{
    import std.exception : enforce;

public:
    this(C)(in C[] chars...)
        if(is(C : dchar))
    {
        this ~= chars;
    }

    this(Input)(Input seq)
        if(!isDynamicArray!Input
            && isInputRange!Input && is(ElementType!Input : dchar))
    {
        this ~= seq;
    }

    /// Gets a $(CODEPOINT) at the given index in this cluster.
    dchar opIndex(size_t index) const pure nothrow @nogc
    {
        assert(index < length);
        return read24(isBig ? ptr_ : small_.ptr, index);
    }

    /++
        Writes a $(CODEPOINT) $(D ch) at given index in this cluster.

        Warning:
        Use of this facility may invalidate grapheme cluster,
        see also $(LREF Grapheme.valid).

        Example:
        ---
        auto g = Grapheme("A\u0302");
        assert(g[0] == 'A');
        assert(g.valid);
        g[1] = '~'; // ASCII tilda is not a combining mark
        assert(g[1] == '~');
        assert(!g.valid);
        ---
    +/
    void opIndexAssign(dchar ch, size_t index) pure nothrow @nogc
    {
        assert(index < length);
        write24(isBig ? ptr_ : small_.ptr, ch, index);
    }

    /++
        Random-access range over Grapheme's $(CHARACTERS).

        Warning: Invalidates when this Grapheme leaves the scope,
        attempts to use it then would lead to memory corruption.
    +/
    @system auto opSlice(size_t a, size_t b) pure nothrow @nogc
    {
        return sliceOverIndexed(a, b, &this);
    }

    /// ditto
    @system auto opSlice() pure nothrow @nogc
    {
        return sliceOverIndexed(0, length, &this);
    }

    /// Grapheme cluster length in $(CODEPOINTS).
    @property size_t length() const pure nothrow @nogc
    {
        return isBig ? len_ : slen_ & 0x7F;
    }

    /++
        Append $(CHARACTER) $(D ch) to this grapheme.
        Warning:
        Use of this facility may invalidate grapheme cluster,
        see also $(D valid).

        Example:
        ---
        auto g = Grapheme("A");
        assert(g.valid);
        g ~= '\u0301';
        assert(g[].equal("A\u0301"));
        assert(g.valid);
        g ~= "B";
        // not a valid grapheme cluster anymore
        assert(!g.valid);
        // still could be useful though
        assert(g[].equal("A\u0301B"));
        ---
        See_Also: $(LREF Grapheme.valid)
    +/
    ref opOpAssign(string op)(dchar ch)
    {
        static if(op == "~")
        {
            if(!isBig)
            {
                if(slen_ + 1 > small_cap)
                    convertToBig();// & fallthrough to "big" branch
                else
                {
                    write24(small_.ptr, ch, smallLength);
                    slen_++;
                    return this;
                }
            }

            assert(isBig);
            if(len_ + 1 > cap_)
            {
                cap_ += grow;
                ptr_ = cast(ubyte*)enforce(realloc(ptr_, 3*(cap_+1)));
            }
            write24(ptr_, ch, len_++);
            return this;
        }
        else
            static assert(false, "No operation "~op~" defined for Grapheme");
    }

    /// Append all $(CHARACTERS) from the input range $(D inp) to this Grapheme.
    ref opOpAssign(string op, Input)(Input inp)
        if(isInputRange!Input && is(ElementType!Input : dchar))
    {
        static if(op == "~")
        {
            foreach(dchar ch; inp)
                this ~= ch;
            return this;
        }
        else
            static assert(false, "No operation "~op~" defined for Grapheme");
    }

    /++
        True if this object contains valid extended grapheme cluster.
        Decoding primitives of this module always return a valid $(D Grapheme).

        Appending to and direct manipulation of grapheme's $(CHARACTERS) may
        render it no longer valid. Certain applications may chose to use
        Grapheme as a "small string" of any $(CODEPOINTS) and ignore this property
        entirely.
    +/
    @property bool valid()() /*const*/
    {
        auto r = this[];
        genericDecodeGrapheme!false(r);
        return r.length == 0;
    }

    this(this)
    {
        if(isBig)
        {// dup it
            auto raw_cap = 3*(cap_+1);
            auto p = cast(ubyte*)enforce(malloc(raw_cap));
            p[0..raw_cap] = ptr_[0..raw_cap];
            ptr_ = p;
        }
    }

    ~this()
    {
        if(isBig)
        {
            free(ptr_);
        }
    }


private:
    enum small_bytes = ((ubyte*).sizeof+3*size_t.sizeof-1);
    // "out of the blue" grow rate, needs testing
    // (though graphemes are typically small < 9)
    enum grow = 20;
    enum small_cap = small_bytes/3;
    enum small_flag = 0x80, small_mask = 0x7F;
    // 16 bytes in 32bits, should be enough for the majority of cases
    union
    {
        struct
        {
            ubyte* ptr_;
            size_t cap_;
            size_t len_;
            size_t padding_;
        }
        struct
        {
            ubyte[small_bytes] small_;
            ubyte slen_;
        }
    }

    void convertToBig()
    {
        size_t k = smallLength;
        ubyte* p = cast(ubyte*)enforce(malloc(3*(grow+1)));
        for(int i=0; i<k; i++)
            write24(p, read24(small_.ptr, i), i);
        // now we can overwrite small array data
        ptr_ = p;
        len_ = slen_;
        assert(grow > len_);
        cap_ = grow;
        setBig();
    }

    void setBig() pure nothrow @nogc { slen_ |= small_flag; }

    @property size_t smallLength() const pure nothrow @nogc
    {
        return slen_ & small_mask;
    }
    @property ubyte isBig() const pure nothrow @nogc
    {
        return slen_ & small_flag;
    }
}

static assert(Grapheme.sizeof == size_t.sizeof*4);

// verify the example
unittest
{
    import std.algorithm;
    string bold = "ku\u0308hn";

    // note that decodeGrapheme takes parameter by ref
    auto first = decodeGrapheme(bold);

    assert(first.length == 1);
    assert(first[0] == 'k');

    // the next grapheme is 2 characters long
    auto wideOne = decodeGrapheme(bold);
    // slicing a grapheme yields a random-access range of dchar
    assert(wideOne[].equalS("u\u0308"));
    assert(wideOne.length == 2);
    static assert(isRandomAccessRange!(typeof(wideOne[])));

    // all of the usual range manipulation is possible
    assert(wideOne[].filter!isMark().equalS("\u0308"));

    auto g = Grapheme("A");
    assert(g.valid);
    g ~= '\u0301';
    assert(g[].equalS("A\u0301"));
    assert(g.valid);
    g ~= "B";
    // not a valid grapheme cluster anymore
    assert(!g.valid);
    // still could be useful though
    assert(g[].equalS("A\u0301B"));
}

unittest
{
    auto g = Grapheme("A\u0302");
    assert(g[0] == 'A');
    assert(g.valid);
    g[1] = '~'; // ASCII tilda is not a combining mark
    assert(g[1] == '~');
    assert(!g.valid);
}

unittest
{
    import std.conv;
    import std.algorithm;
    import std.range;

    // not valid clusters (but it just a test)
    auto g  = Grapheme('a', 'b', 'c', 'd', 'e');
    assert(g[0] == 'a');
    assert(g[1] == 'b');
    assert(g[2] == 'c');
    assert(g[3] == 'd');
    assert(g[4] == 'e');
    g[3] = 'Й';
    assert(g[2] == 'c');
    assert(g[3] == 'Й', text(g[3], " vs ", 'Й'));
    assert(g[4] == 'e');
    assert(!g.valid);

    g ~= 'ц';
    g ~= '~';
    assert(g[0] == 'a');
    assert(g[1] == 'b');
    assert(g[2] == 'c');
    assert(g[3] == 'Й');
    assert(g[4] == 'e');
    assert(g[5] == 'ц');
    assert(g[6] == '~');
    assert(!g.valid);

    Grapheme copy = g;
    copy[0] = 'X';
    copy[1] = '-';
    assert(g[0] == 'a' && copy[0] == 'X');
    assert(g[1] == 'b' && copy[1] == '-');
    assert(equalS(g[2..g.length], copy[2..copy.length]));
    copy = Grapheme("АБВГДЕЁЖЗИКЛМ");
    assert(equalS(copy[0..8], "АБВГДЕЁЖ"), text(copy[0..8]));
    copy ~= "xyz";
    assert(equalS(copy[13..15], "xy"), text(copy[13..15]));
    assert(!copy.valid);

    Grapheme h;
    foreach(dchar v; iota(cast(int)'A', cast(int)'Z'+1).map!"cast(dchar)a"())
        h ~= v;
    assert(equalS(h[], iota(cast(int)'A', cast(int)'Z'+1)));
}

/++
    $(P Does basic case-insensitive comparison of strings $(D str1) and $(D str2).
    This function uses simpler comparison rule thus achieving better performance
    than $(LREF icmp). However keep in mind the warning below.)

    Params:
        str1 = a string or a $(D ForwardRange) of $(D dchar)s
        str2 = a string or a $(D ForwardRange) of $(D dchar)s

    Returns:
        An $(D int) that is 0 if the strings match,
        &lt;0 if $(D str1) is lexicographically "less" than $(D str2),
        &gt;0 if $(D str1) is lexicographically "greater" than $(D str2)

    Warning:
    This function only handles 1:1 $(CODEPOINT) mapping
    and thus is not sufficient for certain alphabets
    like German, Greek and few others.

    Example:
    ---
    assert(sicmp("Август", "авгусТ") == 0);
    // Greek also works as long as there is no 1:M mapping in sight
    assert(sicmp("ΌΎ", "όύ") == 0);
    // things like the following won't get matched as equal
    // Greek small letter iota with dialytika and tonos
    assert(sicmp("ΐ", "\u03B9\u0308\u0301") != 0);

    // while icmp has no problem with that
    assert(icmp("ΐ", "\u03B9\u0308\u0301") == 0);
    assert(icmp("ΌΎ", "όύ") == 0);
    ---

    See_Also:
        $(LREF icmp)
        $(XREF_PACK algorithm,comparison,cmp)
+/
int sicmp(S1, S2)(S1 str1, S2 str2)
    if(isForwardRange!S1 && is(Unqual!(ElementType!S1) == dchar)
    && isForwardRange!S2 && is(Unqual!(ElementType!S2) == dchar))
{
    import std.utf : decode;

    alias sTable = simpleCaseTable;
    size_t ridx=0;
    foreach(dchar lhs; str1)
    {
        if(ridx == str2.length)
            return 1;
        dchar rhs = std.utf.decode(str2, ridx);
        int diff = lhs - rhs;
        if(!diff)
            continue;
        size_t idx = simpleCaseTrie[lhs];
        size_t idx2 = simpleCaseTrie[rhs];
        // simpleCaseTrie is packed index table
        if(idx != EMPTY_CASE_TRIE)
        {
            if(idx2 != EMPTY_CASE_TRIE)
            {// both cased chars
                // adjust idx --> start of bucket
                idx = idx - sTable[idx].n;
                idx2 = idx2 - sTable[idx2].n;
                if(idx == idx2)// one bucket, equivalent chars
                    continue;
                else//  not the same bucket
                    diff = sTable[idx].ch - sTable[idx2].ch;
            }
            else
                diff = sTable[idx - sTable[idx].n].ch - rhs;
        }
        else if(idx2 != EMPTY_CASE_TRIE)
        {
            diff = lhs - sTable[idx2 - sTable[idx2].n].ch;
        }
        // one of chars is not cased at all
        return diff;
    }
    return ridx == str2.length ? 0 : -1;
}
// overloads for the most common cases to reduce compile time
@safe pure /*TODO nothrow*/
{
    int sicmp(const(char)[] str1, const(char)[] str2)
    { return sicmp!(const(char)[], const(char)[])(str1, str2); }
    int sicmp(const(wchar)[] str1, const(wchar)[] str2)
    { return sicmp!(const(wchar)[], const(wchar)[])(str1, str2); }
    int sicmp(const(dchar)[] str1, const(dchar)[] str2)
    { return sicmp!(const(dchar)[], const(dchar)[])(str1, str2); }
}

private int fullCasedCmp(Range)(dchar lhs, dchar rhs, ref Range rtail)
    @trusted pure /*TODO nothrow*/
{
    import std.algorithm : skipOver;
    alias fTable = fullCaseTable;
    size_t idx = fullCaseTrie[lhs];
    // fullCaseTrie is packed index table
    if(idx == EMPTY_CASE_TRIE)
        return lhs;
    size_t start = idx - fTable[idx].n;
    size_t end = fTable[idx].size + start;
    assert(fTable[start].entry_len == 1);
    for(idx=start; idx<end; idx++)
    {
        auto entryLen = fTable[idx].entry_len;
        if(entryLen == 1)
        {
            if(fTable[idx].seq[0] == rhs)
            {
                return 0;
            }
        }
        else
        {// OK it's a long chunk, like 'ss' for German
            dstring seq = fTable[idx].seq[0..entryLen];
            if(rhs == seq[0]
                && rtail.skipOver(seq[1..$]))
            {
                // note that this path modifies rtail
                // iff we managed to get there
                return 0;
            }
        }
    }
    return fTable[start].seq[0]; // new remapped character for accurate diffs
}

/++
    $(P Does case insensitive comparison of $(D str1) and $(D str2).
    Follows the rules of full case-folding mapping.
    This includes matching as equal german ß with "ss" and
    other 1:M $(CODEPOINT) mappings unlike $(LREF sicmp).
    The cost of $(D icmp) being pedantically correct is
    slightly worse performance.
    )

    Example:
    ---
    assert(icmp("Rußland", "Russland") == 0);
    assert(icmp("ᾩ -> \u1F70\u03B9", "\u1F61\u03B9 -> ᾲ") == 0);
    ---
+/
int icmp(S1, S2)(S1 str1, S2 str2)
    if(isForwardRange!S1 && is(Unqual!(ElementType!S1) == dchar)
    && isForwardRange!S2 && is(Unqual!(ElementType!S2) == dchar))
{
    for(;;)
    {
        if(str1.empty)
            return str2.empty ? 0 : -1;
        dchar lhs = str1.front;
        if(str2.empty)
            return 1;
        dchar rhs = str2.front;
        str1.popFront();
        str2.popFront();
        int diff = lhs - rhs;
        if(!diff)
            continue;
        // first try to match lhs to <rhs,right-tail> sequence
        int cmpLR = fullCasedCmp(lhs, rhs, str2);
        if(!cmpLR)
            continue;
        // then rhs to <lhs,left-tail> sequence
        int cmpRL = fullCasedCmp(rhs, lhs, str1);
        if(!cmpRL)
            continue;
        // cmpXX contain remapped codepoints
        // to obtain stable ordering of icmp
        diff = cmpLR - cmpRL;
        return diff;
    }
}
// overloads for the most common cases to reduce compile time
@safe pure /*TODO nothrow*/
{
    int icmp(const(char)[] str1, const(char)[] str2)
    { return icmp!(const(char)[], const(char)[])(str1, str2); }
    int icmp(const(wchar)[] str1, const(wchar)[] str2)
    { return icmp!(const(wchar)[], const(wchar)[])(str1, str2); }
    int icmp(const(dchar)[] str1, const(dchar)[] str2)
    { return icmp!(const(dchar)[], const(dchar)[])(str1, str2); }
}

unittest
{
    import std.conv;
    import std.exception : assertCTFEable;
    import std.algorithm;
    assertCTFEable!(
    {
    foreach(cfunc; TypeTuple!(icmp, sicmp))
    {
        foreach(S1; TypeTuple!(string, wstring, dstring))
        foreach(S2; TypeTuple!(string, wstring, dstring))
        (){ // avoid slow optimizations for large functions @@@BUG@@@ 2396
            assert(cfunc("".to!S1(), "".to!S2()) == 0);
            assert(cfunc("A".to!S1(), "".to!S2()) > 0);
            assert(cfunc("".to!S1(), "0".to!S2()) < 0);
            assert(cfunc("abc".to!S1(), "abc".to!S2()) == 0);
            assert(cfunc("abcd".to!S1(), "abc".to!S2()) > 0);
            assert(cfunc("abc".to!S1(), "abcd".to!S2()) < 0);
            assert(cfunc("Abc".to!S1(), "aBc".to!S2()) == 0);
            assert(cfunc("авГуст".to!S1(), "АВгУСТ".to!S2()) == 0);
            // Check example:
            assert(cfunc("Август".to!S1(), "авгусТ".to!S2()) == 0);
            assert(cfunc("ΌΎ".to!S1(), "όύ".to!S2()) == 0);
        }();
        // check that the order is properly agnostic to the case
        auto strs = [ "Apple", "ORANGE",  "orAcle", "amp", "banana"];
        sort!((a,b) => cfunc(a,b) < 0)(strs);
        assert(strs == ["amp", "Apple",  "banana", "orAcle", "ORANGE"]);
    }
    assert(icmp("ßb", "ssa") > 0);
    // Check example:
    assert(icmp("Russland", "Rußland") == 0);
    assert(icmp("ᾩ -> \u1F70\u03B9", "\u1F61\u03B9 -> ᾲ") == 0);
    assert(icmp("ΐ"w, "\u03B9\u0308\u0301") == 0);
    assert(sicmp("ΐ", "\u03B9\u0308\u0301") != 0);
    //bugzilla 11057
    assert( icmp("K", "L") < 0 );
    });
}

// This is package for the moment to be used as a support tool for std.regex
// It needs a better API
/*
    Return a range of all $(CODEPOINTS) that casefold to
    and from this $(D ch).
*/
package auto simpleCaseFoldings(dchar ch)
{
    alias sTable = simpleCaseTable;
    static struct Range
    {
    pure nothrow:
        uint idx; //if == uint.max, then read c.
        union
        {
            dchar c; // == 0 - empty range
            uint len;
        }
        @property bool isSmall() const { return idx == uint.max; }

        this(dchar ch)
        {
            idx = uint.max;
            c = ch;
        }

        this(uint start, uint size)
        {
            idx = start;
            len = size;
        }

        @property dchar front() const
        {
            assert(!empty);
            if(isSmall)
            {
                return c;
            }
            auto ch = sTable[idx].ch;
            return ch;
        }

        @property bool empty() const
        {
            if(isSmall)
            {
                return c == 0;
            }
            return len == 0;
        }

        @property uint length() const
        {
            if(isSmall)
            {
                return c == 0 ? 0 : 1;
            }
            return len;
        }

        void popFront()
        {
            if(isSmall)
                c = 0;
            else
            {
                idx++;
                len--;
            }
        }
    }
    immutable idx = simpleCaseTrie[ch];
    if (idx == EMPTY_CASE_TRIE)
        return Range(ch);
    auto entry = sTable[idx];
    immutable start = idx - entry.n;
    return Range(start, entry.size);
}

unittest
{
    import std.exception : assertCTFEable;
    import std.algorithm : canFind;
    import std.array;
    assertCTFEable!((){
        auto r = simpleCaseFoldings('Э').array;
        assert(r.length == 2);
        assert(r.canFind('э') && r.canFind('Э'));
        auto sr = simpleCaseFoldings('~');
        assert(sr.equalS("~"));
        //A with ring above - casefolds to the same bucket as Angstrom sign
        sr = simpleCaseFoldings('Å');
        assert(sr.length == 3);
        assert(sr.canFind('å') && sr.canFind('Å') && sr.canFind('\u212B'));
    });
}

/++
    $(P Returns the $(S_LINK Combining class, combining class) of $(D ch).)

    Example:
    ---
    // shorten the code
    alias CC = combiningClass;

    // combining tilda
    assert(CC('\u0303') == 230);
    // combining ring below
    assert(CC('\u0325') == 220);
    // the simple consequence is that  "tilda" should be
    // placed after a "ring below" in a sequence
    ---
+/
ubyte combiningClass(dchar ch) @safe pure nothrow @nogc
{
    return combiningClassTrie[ch];
}

@safe pure nothrow @nogc unittest
{
    foreach(ch; 0..0x80)
        assert(combiningClass(ch) == 0);
    assert(combiningClass('\u05BD') == 22);
    assert(combiningClass('\u0300') == 230);
    assert(combiningClass('\u0317') == 220);
    assert(combiningClass('\u1939') == 222);
}

/// Unicode character decomposition type.
enum UnicodeDecomposition {
    /// Canonical decomposition. The result is canonically equivalent sequence.
    Canonical,
    /**
         Compatibility decomposition. The result is compatibility equivalent sequence.
         Note: Compatibility decomposition is a $(B lossy) conversion,
         typically suitable only for fuzzy matching and internal processing.
    */
    Compatibility
}

/**
    Shorthand aliases for character decomposition type, passed as a
    template parameter to $(LREF decompose).
*/
enum {
    Canonical = UnicodeDecomposition.Canonical,
    Compatibility = UnicodeDecomposition.Compatibility
};

/++
    Try to canonically compose 2 $(CHARACTERS).
    Returns the composed $(CHARACTER) if they do compose and dchar.init otherwise.

    The assumption is that $(D first) comes before $(D second) in the original text,
    usually meaning that the first is a starter.

    Note: Hangul syllables are not covered by this function.
    See $(D composeJamo) below.

    Example:
    ---
    assert(compose('A','\u0308') == '\u00C4');
    assert(compose('A', 'B') == dchar.init);
    assert(compose('C', '\u0301') == '\u0106');
    // note that the starter is the first one
    // thus the following doesn't compose
    assert(compose('\u0308', 'A') == dchar.init);
    ---
+/
public dchar compose(dchar first, dchar second) pure nothrow
{
    import std.internal.unicode_comp;
    import std.algorithm : map;
    import std.range : assumeSorted;
    size_t packed = compositionJumpTrie[first];
    if(packed == ushort.max)
        return dchar.init;
    // unpack offset and length
    size_t idx = packed & composeIdxMask, cnt = packed >> composeCntShift;
    // TODO: optimize this micro binary search (no more then 4-5 steps)
    auto r = compositionTable[idx..idx+cnt].map!"a.rhs"().assumeSorted();
    auto target = r.lowerBound(second).length;
    if(target == cnt)
        return dchar.init;
    auto entry = compositionTable[idx+target];
    if(entry.rhs != second)
        return dchar.init;
    return entry.composed;
}

/++
    Returns a full $(S_LINK Canonical decomposition, Canonical)
    (by default) or $(S_LINK Compatibility decomposition, Compatibility)
    decomposition of $(CHARACTER) $(D ch).
    If no decomposition is available returns a $(LREF Grapheme)
    with the $(D ch) itself.

    Note:
    This function also decomposes hangul syllables
    as prescribed by the standard.

    See_Also: $(LREF decomposeHangul) for a restricted version
    that takes into account only hangul syllables  but
    no other decompositions.

    Example:
    ---
    import std.algorithm;
    assert(decompose('Ĉ')[].equal("C\u0302"));
    assert(decompose('D')[].equal("D"));
    assert(decompose('\uD4DC')[].equal("\u1111\u1171\u11B7"));
    assert(decompose!Compatibility('¹').equal("1"));
    ---
+/
public Grapheme decompose(UnicodeDecomposition decompType=Canonical)(dchar ch)
{
    import std.internal.unicode_decomp;
    import std.algorithm : until;
    static if(decompType == Canonical)
    {
        alias table = decompCanonTable;
        alias mapping = canonMappingTrie;
    }
    else static if(decompType == Compatibility)
    {
        alias table = decompCompatTable;
        alias mapping = compatMappingTrie;
    }
    ushort idx = mapping[ch];
    if(!idx) // not found, check hangul arithmetic decomposition
        return decomposeHangul(ch);
    auto decomp = table[idx..$].until(0);
    return Grapheme(decomp);
}

unittest
{
    // verify examples
    assert(compose('A','\u0308') == '\u00C4');
    assert(compose('A', 'B') == dchar.init);
    assert(compose('C', '\u0301') == '\u0106');
    // note that the starter is the first one
    // thus the following doesn't compose
    assert(compose('\u0308', 'A') == dchar.init);

    import std.algorithm;
    assert(decompose('Ĉ')[].equalS("C\u0302"));
    assert(decompose('D')[].equalS("D"));
    assert(decompose('\uD4DC')[].equalS("\u1111\u1171\u11B7"));
    assert(decompose!Compatibility('¹')[].equalS("1"));
}

//----------------------------------------------------------------------------
// Hangul specific composition/decomposition
enum jamoSBase = 0xAC00;
enum jamoLBase = 0x1100;
enum jamoVBase = 0x1161;
enum jamoTBase = 0x11A7;
enum jamoLCount = 19, jamoVCount = 21, jamoTCount = 28;
enum jamoNCount = jamoVCount * jamoTCount;
enum jamoSCount = jamoLCount * jamoNCount;

// Tests if $(D ch) is a Hangul leading consonant jamo.
bool isJamoL(dchar ch) pure nothrow @nogc
{
    // first cmp rejects ~ 1M code points above leading jamo range
    return ch < jamoLBase+jamoLCount && ch >= jamoLBase;
}

// Tests if $(D ch) is a Hangul vowel jamo.
bool isJamoT(dchar ch) pure nothrow @nogc
{
    // first cmp rejects ~ 1M code points above trailing jamo range
    // Note: ch == jamoTBase doesn't indicate trailing jamo (TIndex must be > 0)
    return ch < jamoTBase+jamoTCount && ch > jamoTBase;
}

// Tests if $(D ch) is a Hangul trailnig consonant jamo.
bool isJamoV(dchar ch) pure nothrow @nogc
{
    // first cmp rejects ~ 1M code points above vowel range
    return  ch < jamoVBase+jamoVCount && ch >= jamoVBase;
}

int hangulSyllableIndex(dchar ch) pure nothrow @nogc
{
    int idxS = cast(int)ch - jamoSBase;
    return idxS >= 0 && idxS < jamoSCount ? idxS : -1;
}

// internal helper: compose hangul syllables leaving dchar.init in holes
void hangulRecompose(dchar[] seq) pure nothrow @nogc
{
    for(size_t idx = 0; idx + 1 < seq.length; )
    {
        if(isJamoL(seq[idx]) && isJamoV(seq[idx+1]))
        {
            int indexL = seq[idx] - jamoLBase;
            int indexV = seq[idx+1] - jamoVBase;
            int indexLV = indexL * jamoNCount + indexV * jamoTCount;
            if(idx + 2 < seq.length && isJamoT(seq[idx+2]))
            {
                seq[idx] = jamoSBase + indexLV + seq[idx+2] - jamoTBase;
                seq[idx+1] = dchar.init;
                seq[idx+2] = dchar.init;
                idx += 3;
            }
            else
            {
                seq[idx] = jamoSBase + indexLV;
                seq[idx+1] = dchar.init;
                idx += 2;
            }
        }
        else
            idx++;
    }
}

//----------------------------------------------------------------------------
public:

/**
    Decomposes a Hangul syllable. If $(D ch) is not a composed syllable
    then this function returns $(LREF Grapheme) containing only $(D ch) as is.
*/
Grapheme decomposeHangul(dchar ch)
{
    int idxS = cast(int)ch - jamoSBase;
    if(idxS < 0 || idxS >= jamoSCount) return Grapheme(ch);
    int idxL = idxS / jamoNCount;
    int idxV = (idxS % jamoNCount) / jamoTCount;
    int idxT = idxS % jamoTCount;

    int partL = jamoLBase + idxL;
    int partV = jamoVBase + idxV;
    if(idxT > 0) // there is a trailling consonant (T); <L,V,T> decomposition
        return Grapheme(partL, partV, jamoTBase + idxT);
    else // <L, V> decomposition
        return Grapheme(partL, partV);
}

///
unittest
{
    import std.algorithm;
    assert(decomposeHangul('\uD4DB')[].equal("\u1111\u1171\u11B6"));
}

/++
    Try to compose hangul syllable out of a leading consonant ($(D lead)),
    a $(D vowel) and optional $(D trailing) consonant jamos.

    On success returns the composed LV or LVT hangul syllable.

    If any of $(D lead) and $(D vowel) are not a valid hangul jamo
    of the respective $(CHARACTER) class returns dchar.init.
+/
dchar composeJamo(dchar lead, dchar vowel, dchar trailing=dchar.init) pure nothrow @nogc
{
    if(!isJamoL(lead))
        return dchar.init;
    int indexL = lead - jamoLBase;
    if(!isJamoV(vowel))
        return dchar.init;
    int indexV = vowel - jamoVBase;
    int indexLV = indexL * jamoNCount + indexV * jamoTCount;
    dchar syllable = jamoSBase + indexLV;
    return isJamoT(trailing) ? syllable + (trailing - jamoTBase) : syllable;
}

///
unittest
{
    assert(composeJamo('\u1111', '\u1171', '\u11B6') == '\uD4DB');
    // leaving out T-vowel, or passing any codepoint
    // that is not trailing consonant composes an LV-syllable
    assert(composeJamo('\u1111', '\u1171') == '\uD4CC');
    assert(composeJamo('\u1111', '\u1171', ' ') == '\uD4CC');
    assert(composeJamo('\u1111', 'A') == dchar.init);
    assert(composeJamo('A', '\u1171') == dchar.init);
}

unittest
{
    import std.conv;

    static void testDecomp(UnicodeDecomposition T)(dchar ch, string r)
    {
        Grapheme g = decompose!T(ch);
        assert(equalS(g[], r), text(g[], " vs ", r));
    }
    testDecomp!Canonical('\u1FF4', "\u03C9\u0301\u0345");
    testDecomp!Canonical('\uF907', "\u9F9C");
    testDecomp!Compatibility('\u33FF', "\u0067\u0061\u006C");
    testDecomp!Compatibility('\uA7F9', "\u0153");

    // check examples
    assert(decomposeHangul('\uD4DB')[].equalS("\u1111\u1171\u11B6"));
    assert(composeJamo('\u1111', '\u1171', '\u11B6') == '\uD4DB');
    assert(composeJamo('\u1111', '\u1171') == '\uD4CC'); // leave out T-vowel
    assert(composeJamo('\u1111', '\u1171', ' ') == '\uD4CC');
    assert(composeJamo('\u1111', 'A') == dchar.init);
    assert(composeJamo('A', '\u1171') == dchar.init);
}

/**
    Enumeration type for normalization forms,
    passed as template parameter for functions like $(LREF normalize).
*/
enum NormalizationForm {
    NFC,
    NFD,
    NFKC,
    NFKD
}


enum {
    /**
        Shorthand aliases from values indicating normalization forms.
    */
    NFC = NormalizationForm.NFC,
    ///ditto
    NFD = NormalizationForm.NFD,
    ///ditto
    NFKC = NormalizationForm.NFKC,
    ///ditto
    NFKD = NormalizationForm.NFKD
};

/++
    Returns $(D input) string normalized to the chosen form.
    Form C is used by default.

    For more information on normalization forms see
    the $(S_LINK Normalization, normalization section).

    Note:
    In cases where the string in question is already normalized,
    it is returned unmodified and no memory allocation happens.
+/
inout(C)[] normalize(NormalizationForm norm=NFC, C)(inout(C)[] input)
{
    import std.algorithm : sort, SwapStrategy;
    import std.range : zip;
    import std.array : appender;

    auto anchors = splitNormalized!norm(input);
    if(anchors[0] == input.length && anchors[1] == input.length)
        return input;
    dchar[] decomposed;
    decomposed.reserve(31);
    ubyte[] ccc;
    ccc.reserve(31);
    auto app = appender!(C[])();
    do
    {
        app.put(input[0..anchors[0]]);
        foreach(dchar ch; input[anchors[0]..anchors[1]])
            static if(norm == NFD || norm == NFC)
            {
                foreach(dchar c; decompose!Canonical(ch)[])
                    decomposed ~= c;
            }
            else // NFKD & NFKC
            {
                foreach(dchar c; decompose!Compatibility(ch)[])
                    decomposed ~= c;
            }
        ccc.length = decomposed.length;
        size_t firstNonStable = 0;
        ubyte lastClazz = 0;

        foreach(idx, dchar ch; decomposed)
        {
            auto clazz = combiningClass(ch);
            ccc[idx] = clazz;
            if(clazz == 0 && lastClazz != 0)
            {
                // found a stable code point after unstable ones
                sort!("a[0] < b[0]", SwapStrategy.stable)
                    (zip(ccc[firstNonStable..idx], decomposed[firstNonStable..idx]));
                firstNonStable = decomposed.length;
            }
            else if(clazz != 0 && lastClazz == 0)
            {
                // found first unstable code point after stable ones
                firstNonStable = idx;
            }
            lastClazz = clazz;
        }
        sort!("a[0] < b[0]", SwapStrategy.stable)
            (zip(ccc[firstNonStable..$], decomposed[firstNonStable..$]));
        static if(norm == NFC || norm == NFKC)
        {
            import std.algorithm : countUntil;
            size_t idx = 0;
            auto first = countUntil(ccc, 0);
            if(first >= 0) // no starters?? no recomposition
            {
                for(;;)
                {
                    auto second = recompose(first, decomposed, ccc);
                    if(second == decomposed.length)
                        break;
                    first = second;
                }
                // 2nd pass for hangul syllables
                hangulRecompose(decomposed);
            }
        }
        static if(norm == NFD || norm == NFKD)
            app.put(decomposed);
        else
        {
            import std.algorithm : remove;
            auto clean = remove!("a == dchar.init", SwapStrategy.stable)(decomposed);
            app.put(decomposed[0 .. clean.length]);
        }
        // reset variables
        decomposed.length = 0;
        decomposed.assumeSafeAppend();
        ccc.length = 0;
        ccc.assumeSafeAppend();
        input = input[anchors[1]..$];
        // and move on
        anchors = splitNormalized!norm(input);
    }while(anchors[0] != input.length);
    app.put(input[0..anchors[0]]);
    return cast(inout(C)[])app.data;
}

///
unittest
{
    // any encoding works
    wstring greet = "Hello world";
    assert(normalize(greet) is greet); // the same exact slice

    // An example of a character with all 4 forms being different:
    // Greek upsilon with acute and hook symbol (code point 0x03D3)
    assert(normalize!NFC("ϓ") == "\u03D3");
    assert(normalize!NFD("ϓ") == "\u03D2\u0301");
    assert(normalize!NFKC("ϓ") == "\u038E");
    assert(normalize!NFKD("ϓ") == "\u03A5\u0301");
}

unittest
{
    import std.conv;

    assert(normalize!NFD("abc\uF904def") == "abc\u6ED1def", text(normalize!NFD("abc\uF904def")));
    assert(normalize!NFKD("2¹⁰") == "210", normalize!NFKD("2¹⁰"));
    assert(normalize!NFD("Äffin") == "A\u0308ffin");

    // check example

    // any encoding works
    wstring greet = "Hello world";
    assert(normalize(greet) is greet); // the same exact slice

    // An example of a character with all 4 forms being different:
    // Greek upsilon with acute and hook symbol (code point 0x03D3)
    assert(normalize!NFC("ϓ") == "\u03D3");
    assert(normalize!NFD("ϓ") == "\u03D2\u0301");
    assert(normalize!NFKC("ϓ") == "\u038E");
    assert(normalize!NFKD("ϓ") == "\u03A5\u0301");
}

// canonically recompose given slice of code points, works in-place and mutates data
private size_t recompose(size_t start, dchar[] input, ubyte[] ccc) pure nothrow
{
    assert(input.length == ccc.length);
    int accumCC = -1;// so that it's out of 0..255 range
    bool foundSolidStarter = false;
    // writefln("recomposing %( %04x %)", input);
    // first one is always a starter thus we start at i == 1
    size_t i = start+1;
    for(; ; )
    {
        if(i == input.length)
            break;
        int curCC = ccc[i];
        // In any character sequence beginning with a starter S
        // a character C is blocked from S if and only if there
        // is some character B between S and C, and either B
        // is a starter or it has the same or higher combining class as C.
        //------------------------
        // Applying to our case:
        // S is input[0]
        // accumCC is the maximum CCC of characters between C and S,
        //     as ccc are sorted
        // C is input[i]

        if(curCC > accumCC)
        {
            dchar comp = compose(input[start], input[i]);
            if(comp != dchar.init)
            {
                input[start] = comp;
                input[i] = dchar.init;// put a sentinel
                // current was merged so its CCC shouldn't affect
                // composing with the next one
            }
            else {
                // if it was a starter then accumCC is now 0, end of loop
                accumCC = curCC;
                if(accumCC == 0)
                    break;
            }
        }
        else{
            // ditto here
            accumCC = curCC;
            if(accumCC == 0)
                break;
        }
        i++;
    }
    return i;
}

// returns tuple of 2 indexes that delimit:
// normalized text, piece that needs normalization and
// the rest of input starting with stable code point
private auto splitNormalized(NormalizationForm norm, C)(const(C)[] input)
{
    import std.typecons : tuple;
    auto result = input;
    ubyte lastCC = 0;

    foreach(idx, dchar ch; input)
    {
        static if(norm == NFC)
            if(ch < 0x0300)
            {
                lastCC = 0;
                continue;
            }
        ubyte CC = combiningClass(ch);
        if(lastCC > CC && CC != 0)
        {
            return seekStable!norm(idx, input);
        }

        if(notAllowedIn!norm(ch))
        {
           return seekStable!norm(idx, input);
        }
        lastCC = CC;
    }
    return tuple(input.length, input.length);
}

private auto seekStable(NormalizationForm norm, C)(size_t idx, in C[] input)
{
    import std.utf : codeLength;
    import std.typecons : tuple;

    auto br = input[0..idx];
    size_t region_start = 0;// default
    for(;;)
    {
        if(br.empty)// start is 0
            break;
        dchar ch = br.back;
        if(combiningClass(ch) == 0 && allowedIn!norm(ch))
        {
            region_start = br.length - std.utf.codeLength!C(ch);
            break;
        }
        br.popFront();
    }
    ///@@@BUG@@@ can't use find: " find is a nested function and can't be used..."
    size_t region_end=input.length;// end is $ by default
    foreach(i, dchar ch; input[idx..$])
    {
        if(combiningClass(ch) == 0 && allowedIn!norm(ch))
        {
            region_end = i+idx;
            break;
        }
    }
    // writeln("Region to normalize: ", input[region_start..region_end]);
    return tuple(region_start, region_end);
}

/**
    Tests if dchar $(D ch) is always allowed (Quick_Check=YES) in normalization
    form $(D norm).
*/
public bool allowedIn(NormalizationForm norm)(dchar ch)
{
    return !notAllowedIn!norm(ch);
}

///
unittest
{
    // e.g. Cyrillic is always allowed, so is ASCII
    assert(allowedIn!NFC('я'));
    assert(allowedIn!NFD('я'));
    assert(allowedIn!NFKC('я'));
    assert(allowedIn!NFKD('я'));
    assert(allowedIn!NFC('Z'));
}

// not user friendly name but more direct
private bool notAllowedIn(NormalizationForm norm)(dchar ch)
{
    static if(norm == NFC)
        alias qcTrie = nfcQCTrie;
    else static if(norm == NFD)
        alias qcTrie = nfdQCTrie;
    else static if(norm == NFKC)
        alias qcTrie = nfkcQCTrie;
    else static if(norm == NFKD)
        alias qcTrie = nfkdQCTrie;
    else
        static assert("Unknown normalization form "~norm);
    return qcTrie[ch];
}

unittest
{
    assert(allowedIn!NFC('я'));
    assert(allowedIn!NFD('я'));
    assert(allowedIn!NFKC('я'));
    assert(allowedIn!NFKD('я'));
    assert(allowedIn!NFC('Z'));
}

}

version(std_uni_bootstrap)
{
    // old version used for bootstrapping of gen_uni.d that generates
    // up to date optimal versions of all of isXXX functions
    @safe pure nothrow @nogc public bool isWhite(dchar c)
    {
        import std.ascii : isWhite;
        return isWhite(c) ||
               c == lineSep || c == paraSep ||
               c == '\u0085' || c == '\u00A0' || c == '\u1680' || c == '\u180E' ||
               (c >= '\u2000' && c <= '\u200A') ||
               c == '\u202F' || c == '\u205F' || c == '\u3000';
    }
}
else
{

// trusted -> avoid bounds check
@trusted pure nothrow @nogc
{
    // hide template instances behind functions (Bugzilla 13232)
    ushort toLowerIndex(dchar c) { return toLowerIndexTrie[c]; }
    ushort toLowerSimpleIndex(dchar c) { return toLowerSimpleIndexTrie[c]; }
    dchar toLowerTab(size_t idx) { return toLowerTable[idx]; }

    ushort toTitleIndex(dchar c) { return toTitleIndexTrie[c]; }
    ushort toTitleSimpleIndex(dchar c) { return toTitleSimpleIndexTrie[c]; }
    dchar toTitleTab(size_t idx) { return toTitleTable[idx]; }

    ushort toUpperIndex(dchar c) { return toUpperIndexTrie[c]; }
    ushort toUpperSimpleIndex(dchar c) { return toUpperSimpleIndexTrie[c]; }
    dchar toUpperTab(size_t idx) { return toUpperTable[idx]; }
}

public:

/++
    Whether or not $(D c) is a Unicode whitespace $(CHARACTER).
    (general Unicode category: Part of C0(tab, vertical tab, form feed,
    carriage return, and linefeed characters), Zs, Zl, Zp, and NEL(U+0085))
+/
@safe pure nothrow @nogc
public bool isWhite(dchar c)
{
    return isWhiteGen(c); // call pregenerated binary search
}

/++
    Return whether $(D c) is a Unicode lowercase $(CHARACTER).
+/
@safe pure nothrow @nogc
bool isLower(dchar c)
{
    import std.ascii : isLower, isASCII;
    if(isASCII(c))
        return isLower(c);
    return lowerCaseTrie[c];
}

@safe unittest
{
    import std.ascii : isLower;
    foreach(v; 0..0x80)
        assert(isLower(v) == .isLower(v));
    assert(.isLower('я'));
    assert(.isLower('й'));
    assert(!.isLower('Ж'));
    // Greek HETA
    assert(!.isLower('\u0370'));
    assert(.isLower('\u0371'));
    assert(!.isLower('\u039C')); // capital MU
    assert(.isLower('\u03B2')); // beta
    // from extended Greek
    assert(!.isLower('\u1F18'));
    assert(.isLower('\u1F00'));
    foreach(v; unicode.lowerCase.byCodepoint)
        assert(.isLower(v) && !isUpper(v));
}


/++
    Return whether $(D c) is a Unicode uppercase $(CHARACTER).
+/
@safe pure nothrow @nogc
bool isUpper(dchar c)
{
    import std.ascii : isUpper, isASCII;
    if(isASCII(c))
        return isUpper(c);
    return upperCaseTrie[c];
}

@safe unittest
{
    import std.ascii : isLower;
    foreach(v; 0..0x80)
        assert(isLower(v) == .isLower(v));
    assert(!isUpper('й'));
    assert(isUpper('Ж'));
    // Greek HETA
    assert(isUpper('\u0370'));
    assert(!isUpper('\u0371'));
    assert(isUpper('\u039C')); // capital MU
    assert(!isUpper('\u03B2')); // beta
    // from extended Greek
    assert(!isUpper('\u1F00'));
    assert(isUpper('\u1F18'));
    foreach(v; unicode.upperCase.byCodepoint)
        assert(isUpper(v) && !.isLower(v));
}


/++
    If $(D c) is a Unicode uppercase $(CHARACTER), then its lowercase equivalent
    is returned. Otherwise $(D c) is returned.

    Warning: certain alphabets like German and Greek have no 1:1
    upper-lower mapping. Use overload of toLower which takes full string instead.
+/
@safe pure nothrow @nogc
dchar toLower(dchar c)
{
     // optimize ASCII case
    if(c < 0xAA)
    {
        if(c < 'A')
            return c;
        if(c <= 'Z')
            return c + 32;
        return c;
    }
    size_t idx = toLowerSimpleIndex(c);
    if(idx != ushort.max)
    {
        return toLowerTab(idx);
    }
    return c;
}

//TODO: Hidden for now, needs better API.
//Other transforms could use better API as well, but this one is a new primitive.
@safe pure nothrow @nogc
private dchar toTitlecase(dchar c)
{
    // optimize ASCII case
    if(c < 0xAA)
    {
        if(c < 'a')
            return c;
        if(c <= 'z')
            return c - 32;
        return c;
    }
    size_t idx = toTitleSimpleIndex(c);
    if(idx != ushort.max)
    {
        return toTitleTab(idx);
    }
    return c;
}

private alias UpperTriple = TypeTuple!(toUpperIndex, MAX_SIMPLE_UPPER, toUpperTab);
private alias LowerTriple = TypeTuple!(toLowerIndex, MAX_SIMPLE_LOWER, toLowerTab);

// generic toUpper/toLower on whole string, creates new or returns as is
private S toCase(alias indexFn, uint maxIdx, alias tableFn, S)(S s) @trusted pure
    if(isSomeString!S)
{
    import std.array : appender;

    foreach(i, dchar cOuter; s)
    {
        ushort idx = indexFn(cOuter);
        if(idx == ushort.max)
            continue;
        auto result = appender!S(s[0..i]);
        result.reserve(s.length);
        foreach(dchar c; s[i .. $])
        {
            idx = indexFn(c);
            if(idx == ushort.max)
                result.put(c);
            else if(idx < maxIdx)
            {
                c = tableFn(idx);
                result.put(c);
            }
            else
            {
                auto val = tableFn(idx);
                // unpack length + codepoint
                uint len = val>>24;
                result.put(cast(dchar)(val & 0xFF_FFFF));
                foreach(j; idx+1..idx+len)
                    result.put(tableFn(j));
            }
        }
        return result.data;
    }
    return s;
}

unittest //12428
{
    import std.array;
    auto s = "abcdefghij".replicate(300);
    s = s[0..10];

    toUpper(s);

    assert(s == "abcdefghij");
}


// generic toUpper/toLower on whole range, returns range
private auto toCaser(alias indexFn, uint maxIdx, alias tableFn, Range)(Range str)
    // Accept range of dchar's
    if (isInputRange!Range &&
        isSomeChar!(ElementEncodingType!Range) &&
        ElementEncodingType!Range.sizeof == dchar.sizeof)
{
    static struct ToCaserImpl
    {
        @property bool empty()
        {
            return !nLeft && r.empty;
        }

        @property auto front()
        {
            if (!nLeft)
            {
                dchar c = r.front;
                const idx = indexFn(c);
                if (idx == ushort.max)
                {
                    buf[0] = c;
                    nLeft = 1;
                }
                else if (idx < maxIdx)
                {
                    buf[0] = tableFn(idx);
                    nLeft = 1;
                }
                else
                {
                    auto val = tableFn(idx);
                    // unpack length + codepoint
                    nLeft = val >> 24;
                    if (nLeft == 0)
                        nLeft = 1;
                    assert(nLeft <= buf.length);
                    buf[nLeft - 1] = cast(dchar)(val & 0xFF_FFFF);
                    foreach (j; 1 .. nLeft)
                        buf[nLeft - j - 1] = tableFn(idx + j);
                }
            }
            return buf[nLeft - 1];
        }

        void popFront()
        {
            if (!nLeft)
                front();
            assert(nLeft);
            --nLeft;
            if (!nLeft)
                r.popFront();
        }

        static if (isForwardRange!Range)
        {
            @property auto save()
            {
                auto ret = this;
                ret.r = r.save;
                return ret;
            }
        }

      private:
        Range r;
        uint nLeft;
        dchar[3] buf = void;
    }

    return ToCaserImpl(str);
}

/*********************
 * Convert input range or string to upper or lower case.
 *
 * Does not allocate memory.
 * Characters in UTF-8 or UTF-16 format that cannot be decoded
 * are treated as $(XREF utf, replacementDchar).
 *
 * Params:
 *      str = string or range of characters
 *
 * Returns:
 *      an InputRange of dchars
 *
 * See_Also:
 *      $(LREF toUpper), $(LREF toLower)
 */

auto asLowerCase(Range)(Range str)
    if (isInputRange!Range && isSomeChar!(ElementEncodingType!Range))
{
    static if (ElementEncodingType!Range.sizeof < dchar.sizeof)
    {
        import std.utf : byDchar;

        // Decode first
        return toCaser!LowerTriple(str.byDchar);
    }
    else
    {
        return toCaser!LowerTriple(str);
    }
}

/// ditto
auto asUpperCase(Range)(Range str)
    if (isInputRange!Range && isSomeChar!(ElementEncodingType!Range))
{
    static if (ElementEncodingType!Range.sizeof < dchar.sizeof)
    {
        import std.utf : byDchar;

        // Decode first
        return toCaser!UpperTriple(str.byDchar);
    }
    else
    {
        return toCaser!UpperTriple(str);
    }
}

///
@safe pure unittest
{
    import std.algorithm: equal;

    assert("hEllo".asUpperCase.equal("HELLO"));
}

unittest
{
    import std.array;

    auto a = "HELLo".asLowerCase;
    auto savea = a.save;
    auto s = a.array;
    assert(s == "hello");
    s = savea.array;
    assert(s == "hello");

    string[] lower = ["123", "abcфеж", "\u0131\u023f\u03c9", "i\u0307\u1Fe2"];
    string[] upper = ["123", "ABCФЕЖ", "I\u2c7e\u2126", "\u0130\u03A5\u0308\u0300"];

    foreach (i, slwr; lower)
    {
        import std.utf : byChar;

        auto sx = slwr.asUpperCase.byChar.array;
        assert(sx == toUpper(slwr));
        auto sy = upper[i].asLowerCase.byChar.array;
        assert(sy == toLower(upper[i]));
    }

    // Not necessary to call r.front
    for (auto r = lower[3].asUpperCase; !r.empty; r.popFront())
    {
    }

    import std.algorithm : equal;

    "HELLo"w.asLowerCase.equal("hello"d);
    "HELLo"w.asUpperCase.equal("HELLO"d);
    "HELLo"d.asLowerCase.equal("hello"d);
    "HELLo"d.asUpperCase.equal("HELLO"d);

    import std.utf : byChar;
    assert(toLower("\u1Fe2") == asLowerCase("\u1Fe2").byChar.array);
}

import std.stdio;
// generic capitalizer on whole range, returns range
private auto toCapitalizer(alias indexFnUpper, uint maxIdxUpper, alias tableFnUpper,
                           Range)(Range str)
    // Accept range of dchar's
    if (isInputRange!Range &&
        isSomeChar!(ElementEncodingType!Range) &&
        ElementEncodingType!Range.sizeof == dchar.sizeof)
{
    static struct ToCapitalizerImpl
    {
        @property bool empty()
        {
            return lower ? lwr.empty : !nLeft && r.empty;
        }

        @property auto front()
        {
            if (lower)
                return lwr.front;

            if (!nLeft)
            {
                dchar c = r.front;
                const idx = indexFnUpper(c);
                if (idx == ushort.max)
                {
                    buf[0] = c;
                    nLeft = 1;
                }
                else if (idx < maxIdxUpper)
                {
                    buf[0] = tableFnUpper(idx);
                    nLeft = 1;
                }
                else
                {
                    auto val = tableFnUpper(idx);
                    // unpack length + codepoint
                    nLeft = val >> 24;
                    if (nLeft == 0)
                        nLeft = 1;
                    assert(nLeft <= buf.length);
                    buf[nLeft - 1] = cast(dchar)(val & 0xFF_FFFF);
                    foreach (j; 1 .. nLeft)
                        buf[nLeft - j - 1] = tableFnUpper(idx + j);
                }
            }
            return buf[nLeft - 1];
        }

        void popFront()
        {
            if (lower)
                lwr.popFront();
            else
            {
                if (!nLeft)
                    front();
                assert(nLeft);
                --nLeft;
                if (!nLeft)
                {
                    r.popFront();
                    lwr = r.asLowerCase();
                    lower = true;
                }
            }
        }

        static if (isForwardRange!Range)
        {
            @property auto save()
            {
                auto ret = this;
                ret.r = r.save;
                ret.lwr = lwr.save;
                return ret;
            }
        }

      private:
        Range r;
        typeof(r.asLowerCase) lwr; // range representing the lower case rest of string
        bool lower = false;     // false for first character, true for rest of string
        dchar[3] buf = void;
        uint nLeft = 0;
    }

    return ToCapitalizerImpl(str);
}

/*********************
 * Capitalize input range or string, meaning convert the first
 * character to upper case and subsequent characters to lower case.
 *
 * Does not allocate memory.
 * Characters in UTF-8 or UTF-16 format that cannot be decoded
 * are treated as $(XREF utf, replacementDchar).
 *
 * Params:
 *      str = string or range of characters
 *
 * Returns:
 *      an InputRange of dchars
 *
 * See_Also:
 *      $(LREF toUpper), $(LREF toLower)
 *      $(LREF asUpperCase), $(LREF asLowerCase)
 */

auto asCapitalized(Range)(Range str)
    if (isInputRange!Range && isSomeChar!(ElementEncodingType!Range))
{
    static if (ElementEncodingType!Range.sizeof < dchar.sizeof)
    {
        import std.utf : byDchar;

        // Decode first
        return toCapitalizer!UpperTriple(str.byDchar);
    }
    else
    {
        return toCapitalizer!UpperTriple(str);
    }
}

///
@safe pure unittest
{
    import std.algorithm: equal;

    assert("hEllo".asCapitalized.equal("Hello"));
}

@safe pure nothrow @nogc unittest
{
    auto r = "hEllo".asCapitalized();
    assert(r.front == 'H');
}

unittest
{
    import std.array;

    auto a = "hELLo".asCapitalized;
    auto savea = a.save;
    auto s = a.array;
    assert(s == "Hello");
    s = savea.array;
    assert(s == "Hello");

    string[2][] cases =
    [
        ["", ""],
        ["h", "H"],
        ["H", "H"],
        ["3", "3"],
        ["123", "123"],
        ["h123A", "H123a"],
        ["феж", "Феж"],
        ["\u1Fe2", "\u03a5\u0308\u0300"],
    ];

    foreach (i; 0 .. cases.length)
    {
        import std.utf : byChar;

        auto r = cases[i][0].asCapitalized.byChar.array;
        auto result = cases[i][1];
        assert(r == result);
    }

    // Don't call r.front
    for (auto r = "\u1Fe2".asCapitalized; !r.empty; r.popFront())
    {
    }

    import std.algorithm : equal;

    "HELLo"w.asCapitalized.equal("Hello"d);
    "hElLO"w.asCapitalized.equal("Hello"d);
    "hello"d.asCapitalized.equal("Hello"d);
    "HELLO"d.asCapitalized.equal("Hello"d);

    import std.utf : byChar;
    assert(asCapitalized("\u0130").byChar.array == asUpperCase("\u0130").byChar.array);
}

// TODO: helper, I wish std.utf was more flexible (and stright)
private size_t encodeTo(char[] buf, size_t idx, dchar c) @trusted pure nothrow @nogc
{
    if (c <= 0x7F)
    {
        buf[idx] = cast(char)c;
        idx++;
    }
    else if (c <= 0x7FF)
    {
        buf[idx] = cast(char)(0xC0 | (c >> 6));
        buf[idx+1] = cast(char)(0x80 | (c & 0x3F));
        idx += 2;
    }
    else if (c <= 0xFFFF)
    {
        buf[idx] = cast(char)(0xE0 | (c >> 12));
        buf[idx+1] = cast(char)(0x80 | ((c >> 6) & 0x3F));
        buf[idx+2] = cast(char)(0x80 | (c & 0x3F));
        idx += 3;
    }
    else if (c <= 0x10FFFF)
    {
        buf[idx] = cast(char)(0xF0 | (c >> 18));
        buf[idx+1] = cast(char)(0x80 | ((c >> 12) & 0x3F));
        buf[idx+2] = cast(char)(0x80 | ((c >> 6) & 0x3F));
        buf[idx+3] = cast(char)(0x80 | (c & 0x3F));
        idx += 4;
    }
    else
        assert(0);
    return idx;
}

unittest
{
    char[] s = "abcd".dup;
    size_t i = 0;
    i = encodeTo(s, i, 'X');
    assert(s == "Xbcd");

    i = encodeTo(s, i, cast(dchar)'\u00A9');
    assert(s == "X\xC2\xA9d");
}

// TODO: helper, I wish std.utf was more flexible (and stright)
private size_t encodeTo(wchar[] buf, size_t idx, dchar c) @trusted pure
{
    import std.utf;
    if (c <= 0xFFFF)
    {
        if (0xD800 <= c && c <= 0xDFFF)
            throw (new UTFException("Encoding an isolated surrogate code point in UTF-16")).setSequence(c);
        buf[idx] = cast(wchar)c;
        idx++;
    }
    else if (c <= 0x10FFFF)
    {
        buf[idx] = cast(wchar)((((c - 0x10000) >> 10) & 0x3FF) + 0xD800);
        buf[idx+1] = cast(wchar)(((c - 0x10000) & 0x3FF) + 0xDC00);
        idx += 2;
    }
    else
        assert(0);
    return idx;
}

private size_t encodeTo(dchar[] buf, size_t idx, dchar c) @trusted pure nothrow @nogc
{
    buf[idx] = c;
    idx++;
    return idx;
}

private void toCaseInPlace(alias indexFn, uint maxIdx, alias tableFn, C)(ref C[] s) @trusted pure
    if (is(C == char) || is(C == wchar)  || is(C == dchar))
{
    import std.utf;
    size_t curIdx = 0;
    size_t destIdx = 0;
    alias slowToCase = toCaseInPlaceAlloc!(indexFn, maxIdx, tableFn);
    size_t lastUnchanged = 0;
    // in-buffer move of bytes to a new start index
    // the trick is that it may not need to copy at all
    static size_t moveTo(C[] str, size_t dest, size_t from, size_t to)
    {
        // Interestingly we may just bump pointer for a while
        // then have to copy if a re-cased char was smaller the original
        // later we may regain pace with char that got bigger
        // In the end it sometimes flip-flops between the 2 cases below
        if(dest == from)
            return to;
        // got to copy
        foreach(C c; str[from..to])
            str[dest++] = c;
        return dest;
    }
    while(curIdx != s.length)
    {
        size_t startIdx = curIdx;
        dchar ch = decode(s, curIdx);
        // TODO: special case for ASCII
        auto caseIndex = indexFn(ch);
        if(caseIndex == ushort.max) // unchanged, skip over
        {
            continue;
        }
        else if(caseIndex < maxIdx)  // 1:1 codepoint mapping
        {
            // previous cased chars had the same length as uncased ones
            // thus can just adjust pointer
            destIdx = moveTo(s, destIdx, lastUnchanged, startIdx);
            lastUnchanged = curIdx;
            dchar cased = tableFn(caseIndex);
            auto casedLen = codeLength!C(cased);
            if(casedLen + destIdx > curIdx) // no place to fit cased char
            {
                // switch to slow codepath, where we allocate
                return slowToCase(s, startIdx, destIdx);
            }
            else
            {
                destIdx = encodeTo(s, destIdx, cased);
            }
        }
        else  // 1:m codepoint mapping, slow codepath
        {
            destIdx = moveTo(s, destIdx, lastUnchanged, startIdx);
            lastUnchanged = curIdx;
            return slowToCase(s, startIdx, destIdx);
        }
        assert(destIdx <= curIdx);
    }
    if(lastUnchanged != s.length)
    {
        destIdx = moveTo(s, destIdx, lastUnchanged, s.length);
    }
    s = s[0..destIdx];
}

// helper to precalculate size of case-converted string
private template toCaseLength(alias indexFn, uint maxIdx, alias tableFn)
{
    size_t toCaseLength(C)(in C[] str)
    {
        import std.utf;
        size_t codeLen = 0;
        size_t lastNonTrivial = 0;
        size_t curIdx = 0;
        while(curIdx != str.length)
        {
            size_t startIdx = curIdx;
            dchar ch = decode(str, curIdx);
            ushort caseIndex = indexFn(ch);
            if(caseIndex == ushort.max)
                continue;
            else if(caseIndex < maxIdx)
            {
                codeLen += startIdx - lastNonTrivial;
                lastNonTrivial = curIdx;
                dchar cased = tableFn(caseIndex);
                codeLen += codeLength!C(cased);
            }
            else
            {
                codeLen += startIdx - lastNonTrivial;
                lastNonTrivial = curIdx;
                auto val = tableFn(caseIndex);
                auto len = val>>24;
                dchar cased = val & 0xFF_FFFF;
                codeLen += codeLength!C(cased);
                foreach(j; caseIndex+1..caseIndex+len)
                    codeLen += codeLength!C(tableFn(j));
            }
        }
        if(lastNonTrivial != str.length)
            codeLen += str.length - lastNonTrivial;
        return codeLen;
    }
}

unittest
{
    import std.conv;
    alias toLowerLength = toCaseLength!(LowerTriple);
    assert(toLowerLength("abcd") == 4);
    assert(toLowerLength("аБВгд456") == 10+3);
}

// slower code path that preallocates and then copies
// case-converted stuf to the new string
private template toCaseInPlaceAlloc(alias indexFn, uint maxIdx, alias tableFn)
{
    void toCaseInPlaceAlloc(C)(ref C[] s, size_t curIdx,
        size_t destIdx) @trusted pure
        if (is(C == char) || is(C == wchar) || is(C == dchar))
    {
        import std.utf : decode;
        alias caseLength = toCaseLength!(indexFn, maxIdx, tableFn);
        auto trueLength = destIdx + caseLength(s[curIdx..$]);
        C[] ns = new C[trueLength];
        ns[0..destIdx] = s[0..destIdx];
        size_t lastUnchanged = curIdx;
        while(curIdx != s.length)
        {
            size_t startIdx = curIdx; // start of current codepoint
            dchar ch = decode(s, curIdx);
            auto caseIndex = indexFn(ch);
            if(caseIndex == ushort.max) // skip over
            {
                continue;
            }
            else if(caseIndex < maxIdx)  // 1:1 codepoint mapping
            {
                dchar cased = tableFn(caseIndex);
                auto toCopy = startIdx - lastUnchanged;
                ns[destIdx .. destIdx+toCopy] = s[lastUnchanged .. startIdx];
                lastUnchanged = curIdx;
                destIdx += toCopy;
                destIdx = encodeTo(ns, destIdx, cased);
            }
            else  // 1:m codepoint mapping, slow codepath
            {
                auto toCopy = startIdx - lastUnchanged;
                ns[destIdx .. destIdx+toCopy] = s[lastUnchanged .. startIdx];
                lastUnchanged = curIdx;
                destIdx += toCopy;
                auto val = tableFn(caseIndex);
                // unpack length + codepoint
                uint len = val>>24;
                destIdx = encodeTo(ns, destIdx, cast(dchar)(val & 0xFF_FFFF));
                foreach(j; caseIndex+1..caseIndex+len)
                    destIdx = encodeTo(ns, destIdx, tableFn(j));
            }
        }
        if(lastUnchanged != s.length)
        {
            auto toCopy = s.length - lastUnchanged;
            ns[destIdx..destIdx+toCopy] = s[lastUnchanged..$];
            destIdx += toCopy;
        }
        assert(ns.length == destIdx);
        s = ns;
    }
}

/++
    Converts $(D s) to lowercase (by performing Unicode lowercase mapping) in place.
    For a few characters string length may increase after the transformation,
    in such a case the function reallocates exactly once.
    If $(D s) does not have any uppercase characters, then $(D s) is unaltered.
+/
void toLowerInPlace(C)(ref C[] s) @trusted pure
    if (is(C == char) || is(C == wchar) || is(C == dchar))
{
    toCaseInPlace!(LowerTriple)(s);
}
// overloads for the most common cases to reduce compile time
@safe pure /*TODO nothrow*/
{
    void toLowerInPlace(ref char[] s)
    { toLowerInPlace!char(s); }
    void toLowerInPlace(ref wchar[] s)
    { toLowerInPlace!wchar(s); }
    void toLowerInPlace(ref dchar[] s)
    { toLowerInPlace!dchar(s); }
}

/++
    Converts $(D s) to uppercase  (by performing Unicode uppercase mapping) in place.
    For a few characters string length may increase after the transformation,
    in such a case the function reallocates exactly once.
    If $(D s) does not have any lowercase characters, then $(D s) is unaltered.
+/
void toUpperInPlace(C)(ref C[] s) @trusted pure
    if (is(C == char) || is(C == wchar) || is(C == dchar))
{
    toCaseInPlace!(UpperTriple)(s);
}
// overloads for the most common cases to reduce compile time/code size
@safe pure /*TODO nothrow*/
{
    void toUpperInPlace(ref char[] s)
    { toUpperInPlace!char(s); }
    void toUpperInPlace(ref wchar[] s)
    { toUpperInPlace!wchar(s); }
    void toUpperInPlace(ref dchar[] s)
    { toUpperInPlace!dchar(s); }
}

/++
    Returns a string which is identical to $(D s) except that all of its
    characters are converted to lowercase (by preforming Unicode lowercase mapping).
    If none of $(D s) characters were affected, then $(D s) itself is returned.
+/
S toLower(S)(S s) @trusted pure
    if(isSomeString!S)
{
    return toCase!(LowerTriple)(s);
}
// overloads for the most common cases to reduce compile time
@safe pure /*TODO nothrow*/
{
    string toLower(string s)
    { return toLower!string(s); }
    wstring toLower(wstring s)
    { return toLower!wstring(s); }
    dstring toLower(dstring s)
    { return toLower!dstring(s); }
}


@trusted unittest //@@@BUG std.format is not @safe
{
    import std.format : format;
    foreach(ch; 0..0x80)
        assert(std.ascii.toLower(ch) == toLower(ch));
    assert(toLower('Я') == 'я');
    assert(toLower('Δ') == 'δ');
    foreach(ch; unicode.upperCase.byCodepoint)
    {
        dchar low = ch.toLower();
        assert(low == ch || isLower(low), format("%s -> %s", ch, low));
    }
    assert(toLower("АЯ") == "ая");

    assert("\u1E9E".toLower == "\u00df");
    assert("\u00df".toUpper == "SS");
}

//bugzilla 9629
unittest
{
    wchar[] test = "hello þ world"w.dup;
    auto piece = test[6..7];
    toUpperInPlace(piece);
    assert(test == "hello Þ world");
}


unittest
{
    import std.algorithm : cmp;
    string s1 = "FoL";
    string s2 = toLower(s1);
    assert(cmp(s2, "fol") == 0, s2);
    assert(s2 != s1);

    char[] s3 = s1.dup;
    toLowerInPlace(s3);
    assert(s3 == s2);

    s1 = "A\u0100B\u0101d";
    s2 = toLower(s1);
    s3 = s1.dup;
    assert(cmp(s2, "a\u0101b\u0101d") == 0);
    assert(s2 !is s1);
    toLowerInPlace(s3);
    assert(s3 == s2);

    s1 = "A\u0460B\u0461d";
    s2 = toLower(s1);
    s3 = s1.dup;
    assert(cmp(s2, "a\u0461b\u0461d") == 0);
    assert(s2 !is s1);
    toLowerInPlace(s3);
    assert(s3 == s2);

    s1 = "\u0130";
    s2 = toLower(s1);
    s3 = s1.dup;
    assert(s2 == "i\u0307");
    assert(s2 !is s1);
    toLowerInPlace(s3);
    assert(s3 == s2);

    // Test on wchar and dchar strings.
    assert(toLower("Some String"w) == "some string"w);
    assert(toLower("Some String"d) == "some string"d);

    // bugzilla 12455
    dchar c = 'İ'; // '\U0130' LATIN CAPITAL LETTER I WITH DOT ABOVE
    assert(isUpper(c));
    assert(toLower(c) == 'i');
    // extend on 12455 reprot - check simple-case toUpper too
    c = '\u1f87';
    assert(isLower(c));
    assert(toUpper(c) == '\u1F8F');
}


/++
    If $(D c) is a Unicode lowercase $(CHARACTER), then its uppercase equivalent
    is returned. Otherwise $(D c) is returned.

    Warning:
    Certain alphabets like German and Greek have no 1:1
    upper-lower mapping. Use overload of toUpper which takes full string instead.

    toUpper can be used as an argument to $(XREF_PACK algorithm,iteration,map)
    to produce an algorithm that can convert a range of characters to upper case
    without allocating memory.
    A string can then be produced by using $(XREF_PACK algorithm,mutation,copy)
    to send it to an $(XREF array, appender).
+/
@safe pure nothrow @nogc
dchar toUpper(dchar c)
{
    // optimize ASCII case
    if(c < 0xAA)
    {
        if(c < 'a')
            return c;
        if(c <= 'z')
            return c - 32;
        return c;
    }
    size_t idx = toUpperSimpleIndex(c);
    if(idx != ushort.max)
    {
        return toUpperTab(idx);
    }
    return c;
}

///
unittest
{
    import std.algorithm;
    import std.uni;
    import std.array;

    auto abuf = appender!(char[])();
    "hello".map!toUpper.copy(&abuf);
    assert(abuf.data == "HELLO");
}

@trusted unittest
{
    import std.format : format;
    foreach(ch; 0..0x80)
        assert(std.ascii.toUpper(ch) == toUpper(ch));
    assert(toUpper('я') == 'Я');
    assert(toUpper('δ') == 'Δ');
    auto title = unicode.Titlecase_Letter;
    foreach(ch; unicode.lowerCase.byCodepoint)
    {
        dchar up = ch.toUpper();
        assert(up == ch || isUpper(up) || title[up],
            format("%x -> %x", ch, up));
    }
}

/++
    Returns a string which is identical to $(D s) except that all of its
    characters are converted to uppercase (by preforming Unicode uppercase mapping).
    If none of $(D s) characters were affected, then $(D s) itself is returned.
+/
S toUpper(S)(S s) @trusted pure
    if(isSomeString!S)
{
    return toCase!(UpperTriple)(s);
}
// overloads for the most common cases to reduce compile time
@safe pure /*TODO nothrow*/
{
    string toUpper(string s)
    { return toUpper!string(s); }
    wstring toUpper(wstring s)
    { return toUpper!wstring(s); }
    dstring toUpper(dstring s)
    { return toUpper!dstring(s); }
}

unittest
{
    import std.algorithm : cmp;

    string s1 = "FoL";
    string s2;
    char[] s3;

    s2 = toUpper(s1);
    s3 = s1.dup; toUpperInPlace(s3);
    assert(s3 == s2, s3);
    assert(cmp(s2, "FOL") == 0);
    assert(s2 !is s1);

    s1 = "a\u0100B\u0101d";
    s2 = toUpper(s1);
    s3 = s1.dup; toUpperInPlace(s3);
    assert(s3 == s2);
    assert(cmp(s2, "A\u0100B\u0100D") == 0);
    assert(s2 !is s1);

    s1 = "a\u0460B\u0461d";
    s2 = toUpper(s1);
    s3 = s1.dup; toUpperInPlace(s3);
    assert(s3 == s2);
    assert(cmp(s2, "A\u0460B\u0460D") == 0);
    assert(s2 !is s1);
}

unittest
{
    static void doTest(C)(const(C)[] s, const(C)[] trueUp, const(C)[] trueLow)
    {
        import std.format : format;
        string diff = "src: %( %x %)\nres: %( %x %)\ntru: %( %x %)";
        auto low = s.toLower() , up = s.toUpper();
        auto lowInp = s.dup, upInp = s.dup;
        lowInp.toLowerInPlace();
        upInp.toUpperInPlace();
        assert(low == trueLow, format(diff, low, trueLow));
        assert(up == trueUp,  format(diff, up, trueUp));
        assert(lowInp == trueLow,
            format(diff, cast(ubyte[])s, cast(ubyte[])lowInp, cast(ubyte[])trueLow));
        assert(upInp == trueUp,
            format(diff, cast(ubyte[])s, cast(ubyte[])upInp, cast(ubyte[])trueUp));
    }
    foreach(S; TypeTuple!(dstring, wstring, string))
    {

        S easy = "123";
        S good = "abCФеж";
        S awful = "\u0131\u023f\u2126";
        S wicked = "\u0130\u1FE2";
        auto options = [easy, good, awful, wicked];
        S[] lower = ["123", "abcфеж", "\u0131\u023f\u03c9", "i\u0307\u1Fe2"];
        S[] upper = ["123", "ABCФЕЖ", "I\u2c7e\u2126", "\u0130\u03A5\u0308\u0300"];

        foreach(val; TypeTuple!(easy, good))
        {
            auto e = val.dup;
            auto g = e;
            e.toUpperInPlace();
            assert(e is g);
            e.toLowerInPlace();
            assert(e is g);
        }
        foreach(i, v; options)
        {
            doTest(v, upper[i], lower[i]);
        }

        // a few combinatorial runs
        foreach(i; 0..options.length)
        foreach(j; i..options.length)
        foreach(k; j..options.length)
        {
            auto sample = options[i] ~ options[j] ~ options[k];
            auto sample2 = options[k] ~ options[j] ~ options[i];
            doTest(sample, upper[i] ~ upper[j] ~ upper[k],
                lower[i] ~ lower[j] ~ lower[k]);
            doTest(sample2, upper[k] ~ upper[j] ~ upper[i],
                lower[k] ~ lower[j] ~ lower[i]);
        }
    }
}


/++
    Returns whether $(D c) is a Unicode alphabetic $(CHARACTER)
    (general Unicode category: Alphabetic).
+/
@safe pure nothrow @nogc
bool isAlpha(dchar c)
{
    // optimization
    if(c < 0xAA)
    {
        size_t x = c - 'A';
        if(x <= 'Z' - 'A')
            return true;
        else
        {
            x = c - 'a';
            if(x <= 'z'-'a')
                return true;
        }
        return false;
    }

    return alphaTrie[c];
}

@safe unittest
{
    auto alpha = unicode("Alphabetic");
    foreach(ch; alpha.byCodepoint)
        assert(isAlpha(ch));
    foreach(ch; 0..0x4000)
        assert((ch in alpha) == isAlpha(ch));
}


/++
    Returns whether $(D c) is a Unicode mark
    (general Unicode category: Mn, Me, Mc).
+/
@safe pure nothrow @nogc
bool isMark(dchar c)
{
    return markTrie[c];
}

@safe unittest
{
    auto mark = unicode("Mark");
    foreach(ch; mark.byCodepoint)
        assert(isMark(ch));
    foreach(ch; 0..0x4000)
        assert((ch in mark) == isMark(ch));
}

/++
    Returns whether $(D c) is a Unicode numerical $(CHARACTER)
    (general Unicode category: Nd, Nl, No).
+/
@safe pure nothrow @nogc
bool isNumber(dchar c)
{
    return numberTrie[c];
}

@safe unittest
{
    auto n = unicode("N");
    foreach(ch; n.byCodepoint)
        assert(isNumber(ch));
    foreach(ch; 0..0x4000)
        assert((ch in n) == isNumber(ch));
}


/++
    Returns whether $(D c) is a Unicode punctuation $(CHARACTER)
    (general Unicode category: Pd, Ps, Pe, Pc, Po, Pi, Pf).
+/
@safe pure nothrow @nogc
bool isPunctuation(dchar c)
{
    return punctuationTrie[c];
}

unittest
{
    assert(isPunctuation('\u0021'));
    assert(isPunctuation('\u0028'));
    assert(isPunctuation('\u0029'));
    assert(isPunctuation('\u002D'));
    assert(isPunctuation('\u005F'));
    assert(isPunctuation('\u00AB'));
    assert(isPunctuation('\u00BB'));
    foreach(ch; unicode("P").byCodepoint)
        assert(isPunctuation(ch));
}

/++
    Returns whether $(D c) is a Unicode symbol $(CHARACTER)
    (general Unicode category: Sm, Sc, Sk, So).
+/
@safe pure nothrow @nogc
bool isSymbol(dchar c)
{
   return symbolTrie[c];
}

unittest
{
    import std.format : format;
    assert(isSymbol('\u0024'));
    assert(isSymbol('\u002B'));
    assert(isSymbol('\u005E'));
    assert(isSymbol('\u00A6'));
    foreach(ch; unicode("S").byCodepoint)
        assert(isSymbol(ch), format("%04x", ch));
}

/++
    Returns whether $(D c) is a Unicode space $(CHARACTER)
    (general Unicode category: Zs)
    Note: This doesn't include '\n', '\r', \t' and other non-space $(CHARACTER).
    For commonly used less strict semantics see $(LREF isWhite).
+/
@safe pure nothrow @nogc
bool isSpace(dchar c)
{
    return isSpaceGen(c);
}

unittest
{
    assert(isSpace('\u0020'));
    auto space = unicode.Zs;
    foreach(ch; space.byCodepoint)
        assert(isSpace(ch));
    foreach(ch; 0..0x1000)
        assert(isSpace(ch) == space[ch]);
}


/++
    Returns whether $(D c) is a Unicode graphical $(CHARACTER)
    (general Unicode category: L, M, N, P, S, Zs).

+/
@safe pure nothrow @nogc
bool isGraphical(dchar c)
{
    return graphicalTrie[c];
}


unittest
{
    auto set = unicode("Graphical");
    import std.format : format;
    foreach(ch; set.byCodepoint)
        assert(isGraphical(ch), format("%4x", ch));
    foreach(ch; 0..0x4000)
        assert((ch in set) == isGraphical(ch));
}


/++
    Returns whether $(D c) is a Unicode control $(CHARACTER)
    (general Unicode category: Cc).
+/
@safe pure nothrow @nogc
bool isControl(dchar c)
{
    return isControlGen(c);
}

unittest
{
    assert(isControl('\u0000'));
    assert(isControl('\u0081'));
    assert(!isControl('\u0100'));
    auto cc = unicode.Cc;
    foreach(ch; cc.byCodepoint)
        assert(isControl(ch));
    foreach(ch; 0..0x1000)
        assert(isControl(ch) == cc[ch]);
}


/++
    Returns whether $(D c) is a Unicode formatting $(CHARACTER)
    (general Unicode category: Cf).
+/
@safe pure nothrow @nogc
bool isFormat(dchar c)
{
    return isFormatGen(c);
}


unittest
{
    assert(isFormat('\u00AD'));
    foreach(ch; unicode("Format").byCodepoint)
        assert(isFormat(ch));
}

// code points for private use, surrogates are not likely to change in near feature
// if need be they can be generated from unicode data as well

/++
    Returns whether $(D c) is a Unicode Private Use $(CODEPOINT)
    (general Unicode category: Co).
+/
@safe pure nothrow @nogc
bool isPrivateUse(dchar c)
{
    return (0x00_E000 <= c && c <= 0x00_F8FF)
        || (0x0F_0000 <= c && c <= 0x0F_FFFD)
        || (0x10_0000 <= c && c <= 0x10_FFFD);
}

/++
    Returns whether $(D c) is a Unicode surrogate $(CODEPOINT)
    (general Unicode category: Cs).
+/
@safe pure nothrow @nogc
bool isSurrogate(dchar c)
{
    return (0xD800 <= c && c <= 0xDFFF);
}

/++
    Returns whether $(D c) is a Unicode high surrogate (lead surrogate).
+/
@safe pure nothrow @nogc
bool isSurrogateHi(dchar c)
{
    return (0xD800 <= c && c <= 0xDBFF);
}

/++
    Returns whether $(D c) is a Unicode low surrogate (trail surrogate).
+/
@safe pure nothrow @nogc
bool isSurrogateLo(dchar c)
{
    return (0xDC00 <= c && c <= 0xDFFF);
}

/++
    Returns whether $(D c) is a Unicode non-character i.e.
    a $(CODEPOINT) with no assigned abstract character.
    (general Unicode category: Cn)
+/
@safe pure nothrow @nogc
bool isNonCharacter(dchar c)
{
    return nonCharacterTrie[c];
}

unittest
{
    auto set = unicode("Cn");
    foreach(ch; set.byCodepoint)
        assert(isNonCharacter(ch));
}

private:
// load static data from pre-generated tables into usable datastructures


@safe auto asSet(const (ubyte)[] compressed) pure
{
    return CodepointSet.fromIntervals(decompressIntervals(compressed));
}

@safe pure nothrow auto asTrie(T...)(in TrieEntry!T e)
{
    return const(CodepointTrie!T)(e.offsets, e.sizes, e.data);
}

@safe pure nothrow @nogc @property
{
    // It's important to use auto return here, so that the compiler
    // only runs semantic on the return type if the function gets
    // used. Also these are functions rather than templates to not
    // increase the object size of the caller.
    auto lowerCaseTrie() { static immutable res = asTrie(lowerCaseTrieEntries); return res; }
    auto upperCaseTrie() { static immutable res = asTrie(upperCaseTrieEntries); return res; }
    auto simpleCaseTrie() { static immutable res = asTrie(simpleCaseTrieEntries); return res; }
    auto fullCaseTrie() { static immutable res = asTrie(fullCaseTrieEntries); return res; }
    auto alphaTrie() { static immutable res = asTrie(alphaTrieEntries); return res; }
    auto markTrie() { static immutable res = asTrie(markTrieEntries); return res; }
    auto numberTrie() { static immutable res = asTrie(numberTrieEntries); return res; }
    auto punctuationTrie() { static immutable res = asTrie(punctuationTrieEntries); return res; }
    auto symbolTrie() { static immutable res = asTrie(symbolTrieEntries); return res; }
    auto graphicalTrie() { static immutable res = asTrie(graphicalTrieEntries); return res; }
    auto nonCharacterTrie() { static immutable res = asTrie(nonCharacterTrieEntries); return res; }

    //normalization quick-check tables
    auto nfcQCTrie()
    {
        import std.internal.unicode_norm;
        static immutable res = asTrie(nfcQCTrieEntries);
        return res;
    }

    auto nfdQCTrie()
    {
        import std.internal.unicode_norm;
        static immutable res = asTrie(nfdQCTrieEntries);
        return res;
    }

    auto nfkcQCTrie()
    {
        import std.internal.unicode_norm;
        static immutable res = asTrie(nfkcQCTrieEntries);
        return res;
    }

    auto nfkdQCTrie()
    {
        import std.internal.unicode_norm;
        static immutable res = asTrie(nfkdQCTrieEntries);
        return res;
    }

    //grapheme breaking algorithm tables
    auto mcTrie()
    {
        import std.internal.unicode_grapheme;
        static immutable res = asTrie(mcTrieEntries);
        return res;
    }

    auto graphemeExtendTrie()
    {
        import std.internal.unicode_grapheme;
        static immutable res = asTrie(graphemeExtendTrieEntries);
        return res;
    }

    auto hangLV()
    {
        import std.internal.unicode_grapheme;
        static immutable res = asTrie(hangulLVTrieEntries);
        return res;
    }

    auto hangLVT()
    {
        import std.internal.unicode_grapheme;
        static immutable res = asTrie(hangulLVTTrieEntries);
        return res;
    }

    // tables below are used for composition/decomposition
    auto combiningClassTrie()
    {
        import std.internal.unicode_comp;
        static immutable res = asTrie(combiningClassTrieEntries);
        return res;
    }

    auto compatMappingTrie()
    {
        import std.internal.unicode_decomp;
        static immutable res = asTrie(compatMappingTrieEntries);
        return res;
    }

    auto canonMappingTrie()
    {
        import std.internal.unicode_decomp;
        static immutable res = asTrie(canonMappingTrieEntries);
        return res;
    }

    auto compositionJumpTrie()
    {
        import std.internal.unicode_comp;
        static immutable res = asTrie(compositionJumpTrieEntries);
        return res;
    }

    //case conversion tables
    auto toUpperIndexTrie() { static immutable res = asTrie(toUpperIndexTrieEntries); return res; }
    auto toLowerIndexTrie() { static immutable res = asTrie(toLowerIndexTrieEntries); return res; }
    auto toTitleIndexTrie() { static immutable res = asTrie(toTitleIndexTrieEntries); return res; }
    //simple case conversion tables
    auto toUpperSimpleIndexTrie() { static immutable res = asTrie(toUpperSimpleIndexTrieEntries); return res; }
    auto toLowerSimpleIndexTrie() { static immutable res = asTrie(toLowerSimpleIndexTrieEntries); return res; }
    auto toTitleSimpleIndexTrie() { static immutable res = asTrie(toTitleSimpleIndexTrieEntries); return res; }

}

}// version(!std_uni_bootstrap)