summaryrefslogtreecommitdiffstats
path: root/intl/icu/source/i18n/regexcmp.cpp
blob: 2657cf37cbecca17b4886c5960287e9fa10f2a99 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
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
// Copyright (C) 2016 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html
//
//  file:  regexcmp.cpp
//
//  Copyright (C) 2002-2016 International Business Machines Corporation and others.
//  All Rights Reserved.
//
//  This file contains the ICU regular expression compiler, which is responsible
//  for processing a regular expression pattern into the compiled form that
//  is used by the match finding engine.
//

#include "unicode/utypes.h"

#if !UCONFIG_NO_REGULAR_EXPRESSIONS

#include "unicode/ustring.h"
#include "unicode/unistr.h"
#include "unicode/uniset.h"
#include "unicode/uchar.h"
#include "unicode/uchriter.h"
#include "unicode/parsepos.h"
#include "unicode/parseerr.h"
#include "unicode/regex.h"
#include "unicode/utf.h"
#include "unicode/utf16.h"
#include "patternprops.h"
#include "putilimp.h"
#include "cmemory.h"
#include "cstring.h"
#include "uvectr32.h"
#include "uvectr64.h"
#include "uassert.h"
#include "uinvchar.h"

#include "regeximp.h"
#include "regexcst.h"   // Contains state table for the regex pattern parser.
                        //   generated by a Perl script.
#include "regexcmp.h"
#include "regexst.h"
#include "regextxt.h"



U_NAMESPACE_BEGIN


//------------------------------------------------------------------------------
//
//  Constructor.
//
//------------------------------------------------------------------------------
RegexCompile::RegexCompile(RegexPattern *rxp, UErrorCode &status) :
   fParenStack(status), fSetStack(status), fSetOpStack(status)
{
    // Lazy init of all shared global sets (needed for init()'s empty text)
    RegexStaticSets::initGlobals(&status);

    fStatus           = &status;

    fRXPat            = rxp;
    fScanIndex        = 0;
    fLastChar         = -1;
    fPeekChar         = -1;
    fLineNum          = 1;
    fCharNum          = 0;
    fQuoteMode        = FALSE;
    fInBackslashQuote = FALSE;
    fModeFlags        = fRXPat->fFlags | 0x80000000;
    fEOLComments      = TRUE;

    fMatchOpenParen   = -1;
    fMatchCloseParen  = -1;
    fCaptureName      = NULL;
    fLastSetLiteral   = U_SENTINEL;

    if (U_SUCCESS(status) && U_FAILURE(rxp->fDeferredStatus)) {
        status = rxp->fDeferredStatus;
    }
}

static const UChar      chAmp       = 0x26;      // '&'
static const UChar      chDash      = 0x2d;      // '-'


//------------------------------------------------------------------------------
//
//  Destructor
//
//------------------------------------------------------------------------------
RegexCompile::~RegexCompile() {
    delete fCaptureName;         // Normally will be NULL, but can exist if pattern
                                 //   compilation stops with a syntax error.
}

static inline void addCategory(UnicodeSet *set, int32_t value, UErrorCode& ec) {
    set->addAll(UnicodeSet().applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, value, ec));
}

//------------------------------------------------------------------------------
//
//  Compile regex pattern.   The state machine for rexexp pattern parsing is here.
//                           The state tables are hand-written in the file regexcst.txt,
//                           and converted to the form used here by a perl
//                           script regexcst.pl
//
//------------------------------------------------------------------------------
void    RegexCompile::compile(
                         const UnicodeString &pat,   // Source pat to be compiled.
                         UParseError &pp,            // Error position info
                         UErrorCode &e)              // Error Code
{
    fRXPat->fPatternString = new UnicodeString(pat);
    UText patternText = UTEXT_INITIALIZER;
    utext_openConstUnicodeString(&patternText, fRXPat->fPatternString, &e);

    if (U_SUCCESS(e)) {
        compile(&patternText, pp, e);
        utext_close(&patternText);
    }
}

//
//   compile, UText mode
//     All the work is actually done here.
//
void    RegexCompile::compile(
                         UText *pat,                 // Source pat to be compiled.
                         UParseError &pp,            // Error position info
                         UErrorCode &e)              // Error Code
{
    fStatus             = &e;
    fParseErr           = &pp;
    fStackPtr           = 0;
    fStack[fStackPtr]   = 0;

    if (U_FAILURE(*fStatus)) {
        return;
    }

    // There should be no pattern stuff in the RegexPattern object.  They can not be reused.
    U_ASSERT(fRXPat->fPattern == NULL || utext_nativeLength(fRXPat->fPattern) == 0);

    // Prepare the RegexPattern object to receive the compiled pattern.
    fRXPat->fPattern        = utext_clone(fRXPat->fPattern, pat, FALSE, TRUE, fStatus);
    if (U_FAILURE(*fStatus)) {
        return;
    }
    fRXPat->fStaticSets     = RegexStaticSets::gStaticSets->fPropSets;
    fRXPat->fStaticSets8    = RegexStaticSets::gStaticSets->fPropSets8;


    // Initialize the pattern scanning state machine
    fPatternLength = utext_nativeLength(pat);
    uint16_t                state = 1;
    const RegexTableEl      *tableEl;

    // UREGEX_LITERAL force entire pattern to be treated as a literal string.
    if (fModeFlags & UREGEX_LITERAL) {
        fQuoteMode = TRUE;
    }

    nextChar(fC);                        // Fetch the first char from the pattern string.

    //
    // Main loop for the regex pattern parsing state machine.
    //   Runs once per state transition.
    //   Each time through optionally performs, depending on the state table,
    //      - an advance to the the next pattern char
    //      - an action to be performed.
    //      - pushing or popping a state to/from the local state return stack.
    //   file regexcst.txt is the source for the state table.  The logic behind
    //     recongizing the pattern syntax is there, not here.
    //
    for (;;) {
        //  Bail out if anything has gone wrong.
        //  Regex pattern parsing stops on the first error encountered.
        if (U_FAILURE(*fStatus)) {
            break;
        }

        U_ASSERT(state != 0);

        // Find the state table element that matches the input char from the pattern, or the
        //    class of the input character.  Start with the first table row for this
        //    state, then linearly scan forward until we find a row that matches the
        //    character.  The last row for each state always matches all characters, so
        //    the search will stop there, if not before.
        //
        tableEl = &gRuleParseStateTable[state];
        REGEX_SCAN_DEBUG_PRINTF(("char, line, col = (\'%c\', %d, %d)    state=%s ",
            fC.fChar, fLineNum, fCharNum, RegexStateNames[state]));

        for (;;) {    // loop through table rows belonging to this state, looking for one
                      //   that matches the current input char.
            REGEX_SCAN_DEBUG_PRINTF(("."));
            if (tableEl->fCharClass < 127 && fC.fQuoted == FALSE &&   tableEl->fCharClass == fC.fChar) {
                // Table row specified an individual character, not a set, and
                //   the input character is not quoted, and
                //   the input character matched it.
                break;
            }
            if (tableEl->fCharClass == 255) {
                // Table row specified default, match anything character class.
                break;
            }
            if (tableEl->fCharClass == 254 && fC.fQuoted)  {
                // Table row specified "quoted" and the char was quoted.
                break;
            }
            if (tableEl->fCharClass == 253 && fC.fChar == (UChar32)-1)  {
                // Table row specified eof and we hit eof on the input.
                break;
            }

            if (tableEl->fCharClass >= 128 && tableEl->fCharClass < 240 &&   // Table specs a char class &&
                fC.fQuoted == FALSE &&                                       //   char is not escaped &&
                fC.fChar != (UChar32)-1) {                                   //   char is not EOF
                U_ASSERT(tableEl->fCharClass <= 137);
                if (RegexStaticSets::gStaticSets->fRuleSets[tableEl->fCharClass-128].contains(fC.fChar)) {
                    // Table row specified a character class, or set of characters,
                    //   and the current char matches it.
                    break;
                }
            }

            // No match on this row, advance to the next  row for this state,
            tableEl++;
        }
        REGEX_SCAN_DEBUG_PRINTF(("\n"));

        //
        // We've found the row of the state table that matches the current input
        //   character from the rules string.
        // Perform any action specified  by this row in the state table.
        if (doParseActions(tableEl->fAction) == FALSE) {
            // Break out of the state machine loop if the
            //   the action signalled some kind of error, or
            //   the action was to exit, occurs on normal end-of-rules-input.
            break;
        }

        if (tableEl->fPushState != 0) {
            fStackPtr++;
            if (fStackPtr >= kStackSize) {
                error(U_REGEX_INTERNAL_ERROR);
                REGEX_SCAN_DEBUG_PRINTF(("RegexCompile::parse() - state stack overflow.\n"));
                fStackPtr--;
            }
            fStack[fStackPtr] = tableEl->fPushState;
        }

        //
        //  NextChar.  This is where characters are actually fetched from the pattern.
        //             Happens under control of the 'n' tag in the state table.
        //
        if (tableEl->fNextChar) {
            nextChar(fC);
        }

        // Get the next state from the table entry, or from the
        //   state stack if the next state was specified as "pop".
        if (tableEl->fNextState != 255) {
            state = tableEl->fNextState;
        } else {
            state = fStack[fStackPtr];
            fStackPtr--;
            if (fStackPtr < 0) {
                // state stack underflow
                // This will occur if the user pattern has mis-matched parentheses,
                //   with extra close parens.
                //
                fStackPtr++;
                error(U_REGEX_MISMATCHED_PAREN);
            }
        }

    }

    if (U_FAILURE(*fStatus)) {
        // Bail out if the pattern had errors.
        //   Set stack cleanup:  a successful compile would have left it empty,
        //   but errors can leave temporary sets hanging around.
        while (!fSetStack.empty()) {
            delete (UnicodeSet *)fSetStack.pop();
        }
        return;
    }

    //
    // The pattern has now been read and processed, and the compiled code generated.
    //

    //
    // The pattern's fFrameSize so far has accumulated the requirements for
    //   storage for capture parentheses, counters, etc. that are encountered
    //   in the pattern.  Add space for the two variables that are always
    //   present in the saved state:  the input string position (int64_t) and
    //   the position in the compiled pattern.
    //
    allocateStackData(RESTACKFRAME_HDRCOUNT);

    //
    // Optimization pass 1: NOPs, back-references, and case-folding
    //
    stripNOPs();

    //
    // Get bounds for the minimum and maximum length of a string that this
    //   pattern can match.  Used to avoid looking for matches in strings that
    //   are too short.
    //
    fRXPat->fMinMatchLen = minMatchLength(3, fRXPat->fCompiledPat->size()-1);

    //
    // Optimization pass 2: match start type
    //
    matchStartType();

    //
    // Set up fast latin-1 range sets
    //
    int32_t numSets = fRXPat->fSets->size();
    fRXPat->fSets8 = new Regex8BitSet[numSets];
    // Null pointer check.
    if (fRXPat->fSets8 == NULL) {
        e = *fStatus = U_MEMORY_ALLOCATION_ERROR;
        return;
    }
    int32_t i;
    for (i=0; i<numSets; i++) {
        UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(i);
        fRXPat->fSets8[i].init(s);
    }

}





//------------------------------------------------------------------------------
//
//  doParseAction        Do some action during regex pattern parsing.
//                       Called by the parse state machine.
//
//                       Generation of the match engine PCode happens here, or
//                       in functions called from the parse actions defined here.
//
//
//------------------------------------------------------------------------------
UBool RegexCompile::doParseActions(int32_t action)
{
    UBool   returnVal = TRUE;

    switch ((Regex_PatternParseAction)action) {

    case doPatStart:
        // Start of pattern compiles to:
        //0   SAVE   2        Fall back to position of FAIL
        //1   jmp    3
        //2   FAIL            Stop if we ever reach here.
        //3   NOP             Dummy, so start of pattern looks the same as
        //                    the start of an ( grouping.
        //4   NOP             Resreved, will be replaced by a save if there are
        //                    OR | operators at the top level
        appendOp(URX_STATE_SAVE, 2);
        appendOp(URX_JMP,  3);
        appendOp(URX_FAIL, 0);

        // Standard open nonCapture paren action emits the two NOPs and
        //   sets up the paren stack frame.
        doParseActions(doOpenNonCaptureParen);
        break;

    case doPatFinish:
        // We've scanned to the end of the pattern
        //  The end of pattern compiles to:
        //        URX_END
        //    which will stop the runtime match engine.
        //  Encountering end of pattern also behaves like a close paren,
        //   and forces fixups of the State Save at the beginning of the compiled pattern
        //   and of any OR operations at the top level.
        //
        handleCloseParen();
        if (fParenStack.size() > 0) {
            // Missing close paren in pattern.
            error(U_REGEX_MISMATCHED_PAREN);
        }

        // add the END operation to the compiled pattern.
        appendOp(URX_END, 0);

        // Terminate the pattern compilation state machine.
        returnVal = FALSE;
        break;



    case doOrOperator:
        // Scanning a '|', as in (A|B)
        {
            // Generate code for any pending literals preceding the '|'
            fixLiterals(FALSE);

            // Insert a SAVE operation at the start of the pattern section preceding
            //   this OR at this level.  This SAVE will branch the match forward
            //   to the right hand side of the OR in the event that the left hand
            //   side fails to match and backtracks.  Locate the position for the
            //   save from the location on the top of the parentheses stack.
            int32_t savePosition = fParenStack.popi();
            int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(savePosition);
            U_ASSERT(URX_TYPE(op) == URX_NOP);  // original contents of reserved location
            op = buildOp(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+1);
            fRXPat->fCompiledPat->setElementAt(op, savePosition);

            // Append an JMP operation into the compiled pattern.  The operand for
            //  the JMP will eventually be the location following the ')' for the
            //  group.  This will be patched in later, when the ')' is encountered.
            appendOp(URX_JMP, 0);

            // Push the position of the newly added JMP op onto the parentheses stack.
            // This registers if for fixup when this block's close paren is encountered.
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);

            // Append a NOP to the compiled pattern.  This is the slot reserved
            //   for a SAVE in the event that there is yet another '|' following
            //   this one.
            appendOp(URX_NOP, 0);
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);
        }
        break;


    case doBeginNamedCapture:
        // Scanning (?<letter.
        //   The first letter of the name will come through again under doConinueNamedCapture.
        fCaptureName = new UnicodeString();
        if (fCaptureName == NULL) {
            error(U_MEMORY_ALLOCATION_ERROR);
        }
        break;

    case  doContinueNamedCapture:
        fCaptureName->append(fC.fChar);
        break;

    case doBadNamedCapture:
        error(U_REGEX_INVALID_CAPTURE_GROUP_NAME);
        break;
        
    case doOpenCaptureParen:
        // Open Capturing Paren, possibly named.
        //   Compile to a
        //      - NOP, which later may be replaced by a save-state if the
        //         parenthesized group gets a * quantifier, followed by
        //      - START_CAPTURE  n    where n is stack frame offset to the capture group variables.
        //      - NOP, which may later be replaced by a save-state if there
        //             is an '|' alternation within the parens.
        //
        //    Each capture group gets three slots in the save stack frame:
        //         0: Capture Group start position (in input string being matched.)
        //         1: Capture Group end position.
        //         2: Start of Match-in-progress.
        //    The first two locations are for a completed capture group, and are
        //     referred to by back references and the like.
        //    The third location stores the capture start position when an START_CAPTURE is
        //      encountered.  This will be promoted to a completed capture when (and if) the corresponding
        //      END_CAPTURE is encountered.
        {
            fixLiterals();
            appendOp(URX_NOP, 0);
            int32_t  varsLoc = allocateStackData(3);    // Reserve three slots in match stack frame.
            appendOp(URX_START_CAPTURE, varsLoc);
            appendOp(URX_NOP, 0);

            // On the Parentheses stack, start a new frame and add the postions
            //   of the two NOPs.  Depending on what follows in the pattern, the
            //   NOPs may be changed to SAVE_STATE or JMP ops, with a target
            //   address of the end of the parenthesized group.
            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
            fParenStack.push(capturing, *fStatus);                        // Frame type.
            fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus);   // The first  NOP location
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP loc

            // Save the mapping from group number to stack frame variable position.
            fRXPat->fGroupMap->addElement(varsLoc, *fStatus);

            // If this is a named capture group, add the name->group number mapping.
            if (fCaptureName != NULL) {
                int32_t groupNumber = fRXPat->fGroupMap->size();
                int32_t previousMapping = uhash_puti(fRXPat->fNamedCaptureMap, fCaptureName, groupNumber, fStatus);
                fCaptureName = NULL;    // hash table takes ownership of the name (key) string.
                if (previousMapping > 0 && U_SUCCESS(*fStatus)) {
                    error(U_REGEX_INVALID_CAPTURE_GROUP_NAME);
                }
            }
        }
        break;

    case doOpenNonCaptureParen:
        // Open non-caputuring (grouping only) Paren.
        //   Compile to a
        //      - NOP, which later may be replaced by a save-state if the
        //         parenthesized group gets a * quantifier, followed by
        //      - NOP, which may later be replaced by a save-state if there
        //             is an '|' alternation within the parens.
        {
            fixLiterals();
            appendOp(URX_NOP, 0);
            appendOp(URX_NOP, 0);

            // On the Parentheses stack, start a new frame and add the postions
            //   of the two NOPs.
            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
            fParenStack.push(plain,      *fStatus);                       // Begin a new frame.
            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first  NOP location
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP loc
        }
         break;


    case doOpenAtomicParen:
        // Open Atomic Paren.  (?>
        //   Compile to a
        //      - NOP, which later may be replaced if the parenthesized group
        //         has a quantifier, followed by
        //      - STO_SP  save state stack position, so it can be restored at the ")"
        //      - NOP, which may later be replaced by a save-state if there
        //             is an '|' alternation within the parens.
        {
            fixLiterals();
            appendOp(URX_NOP, 0);
            int32_t  varLoc = allocateData(1);    // Reserve a data location for saving the state stack ptr.
            appendOp(URX_STO_SP, varLoc);
            appendOp(URX_NOP, 0);

            // On the Parentheses stack, start a new frame and add the postions
            //   of the two NOPs.  Depending on what follows in the pattern, the
            //   NOPs may be changed to SAVE_STATE or JMP ops, with a target
            //   address of the end of the parenthesized group.
            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
            fParenStack.push(atomic, *fStatus);                           // Frame type.
            fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus);   // The first NOP
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP
        }
        break;


    case doOpenLookAhead:
        // Positive Look-ahead   (?=  stuff  )
        //
        //   Note:   Addition of transparent input regions, with the need to
        //           restore the original regions when failing out of a lookahead
        //           block, complicated this sequence.  Some conbined opcodes
        //           might make sense - or might not, lookahead aren't that common.
        //
        //      Caution:  min match length optimization knows about this
        //               sequence; don't change without making updates there too.
        //
        // Compiles to
        //    1    START_LA     dataLoc     Saves SP, Input Pos
        //    2.   STATE_SAVE   4            on failure of lookahead, goto 4
        //    3    JMP          6           continue ...
        //
        //    4.   LA_END                   Look Ahead failed.  Restore regions.
        //    5.   BACKTRACK                and back track again.
        //
        //    6.   NOP              reserved for use by quantifiers on the block.
        //                          Look-ahead can't have quantifiers, but paren stack
        //                             compile time conventions require the slot anyhow.
        //    7.   NOP              may be replaced if there is are '|' ops in the block.
        //    8.     code for parenthesized stuff.
        //    9.   LA_END
        //
        //  Two data slots are reserved, for saving the stack ptr and the input position.
        {
            fixLiterals();
            int32_t dataLoc = allocateData(2);
            appendOp(URX_LA_START, dataLoc);
            appendOp(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+ 2);
            appendOp(URX_JMP, fRXPat->fCompiledPat->size()+ 3);
            appendOp(URX_LA_END, dataLoc);
            appendOp(URX_BACKTRACK, 0);
            appendOp(URX_NOP, 0);
            appendOp(URX_NOP, 0);

            // On the Parentheses stack, start a new frame and add the postions
            //   of the NOPs.
            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
            fParenStack.push(lookAhead, *fStatus);                        // Frame type.
            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first  NOP location
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP location
        }
        break;

    case doOpenLookAheadNeg:
        // Negated Lookahead.   (?! stuff )
        // Compiles to
        //    1.    START_LA    dataloc
        //    2.    SAVE_STATE  7         // Fail within look-ahead block restores to this state,
        //                                //   which continues with the match.
        //    3.    NOP                   // Std. Open Paren sequence, for possible '|'
        //    4.       code for parenthesized stuff.
        //    5.    END_LA                // Cut back stack, remove saved state from step 2.
        //    6.    BACKTRACK             // code in block succeeded, so neg. lookahead fails.
        //    7.    END_LA                // Restore match region, in case look-ahead was using
        //                                        an alternate (transparent) region.
        {
            fixLiterals();
            int32_t dataLoc = allocateData(2);
            appendOp(URX_LA_START, dataLoc);
            appendOp(URX_STATE_SAVE, 0);    // dest address will be patched later.
            appendOp(URX_NOP, 0);

            // On the Parentheses stack, start a new frame and add the postions
            //   of the StateSave and NOP.
            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
            fParenStack.push(negLookAhead, *fStatus);                    // Frame type
            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The STATE_SAVE location
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP location

            // Instructions #5 - #7 will be added when the ')' is encountered.
        }
        break;

    case doOpenLookBehind:
        {
            //   Compile a (?<= look-behind open paren.
            //
            //          Compiles to
            //              0       URX_LB_START     dataLoc
            //              1       URX_LB_CONT      dataLoc
            //              2                        MinMatchLen
            //              3                        MaxMatchLen
            //              4       URX_NOP          Standard '(' boilerplate.
            //              5       URX_NOP          Reserved slot for use with '|' ops within (block).
            //              6         <code for LookBehind expression>
            //              7       URX_LB_END       dataLoc    # Check match len, restore input  len
            //              8       URX_LA_END       dataLoc    # Restore stack, input pos
            //
            //          Allocate a block of matcher data, to contain (when running a match)
            //              0:    Stack ptr on entry
            //              1:    Input Index on entry
            //              2:    Start index of match current match attempt.
            //              3:    Original Input String len.

            // Generate match code for any pending literals.
            fixLiterals();

            // Allocate data space
            int32_t dataLoc = allocateData(4);

            // Emit URX_LB_START
            appendOp(URX_LB_START, dataLoc);

            // Emit URX_LB_CONT
            appendOp(URX_LB_CONT, dataLoc);
            appendOp(URX_RESERVED_OP, 0);    // MinMatchLength.  To be filled later.
            appendOp(URX_RESERVED_OP, 0);    // MaxMatchLength.  To be filled later.

            // Emit the NOPs
            appendOp(URX_NOP, 0);
            appendOp(URX_NOP, 0);

            // On the Parentheses stack, start a new frame and add the postions
            //   of the URX_LB_CONT and the NOP.
            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
            fParenStack.push(lookBehind, *fStatus);                       // Frame type
            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first NOP location
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The 2nd   NOP location

            // The final two instructions will be added when the ')' is encountered.
        }

        break;

    case doOpenLookBehindNeg:
        {
            //   Compile a (?<! negated look-behind open paren.
            //
            //          Compiles to
            //              0       URX_LB_START     dataLoc    # Save entry stack, input len
            //              1       URX_LBN_CONT     dataLoc    # Iterate possible match positions
            //              2                        MinMatchLen
            //              3                        MaxMatchLen
            //              4                        continueLoc (9)
            //              5       URX_NOP          Standard '(' boilerplate.
            //              6       URX_NOP          Reserved slot for use with '|' ops within (block).
            //              7         <code for LookBehind expression>
            //              8       URX_LBN_END      dataLoc    # Check match len, cause a FAIL
            //              9       ...
            //
            //          Allocate a block of matcher data, to contain (when running a match)
            //              0:    Stack ptr on entry
            //              1:    Input Index on entry
            //              2:    Start index of match current match attempt.
            //              3:    Original Input String len.

            // Generate match code for any pending literals.
            fixLiterals();

            // Allocate data space
            int32_t dataLoc = allocateData(4);

            // Emit URX_LB_START
            appendOp(URX_LB_START, dataLoc);

            // Emit URX_LBN_CONT
            appendOp(URX_LBN_CONT, dataLoc);
            appendOp(URX_RESERVED_OP, 0);    // MinMatchLength.  To be filled later.
            appendOp(URX_RESERVED_OP, 0);    // MaxMatchLength.  To be filled later.
            appendOp(URX_RESERVED_OP, 0);    // Continue Loc.    To be filled later.

            // Emit the NOPs
            appendOp(URX_NOP, 0);
            appendOp(URX_NOP, 0);

            // On the Parentheses stack, start a new frame and add the postions
            //   of the URX_LB_CONT and the NOP.
            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
            fParenStack.push(lookBehindN, *fStatus);                      // Frame type
            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first NOP location
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The 2nd   NOP location

            // The final two instructions will be added when the ')' is encountered.
        }
        break;

    case doConditionalExpr:
        // Conditionals such as (?(1)a:b)
    case doPerlInline:
        // Perl inline-condtionals.  (?{perl code}a|b) We're not perl, no way to do them.
        error(U_REGEX_UNIMPLEMENTED);
        break;


    case doCloseParen:
        handleCloseParen();
        if (fParenStack.size() <= 0) {
            //  Extra close paren, or missing open paren.
            error(U_REGEX_MISMATCHED_PAREN);
        }
        break;

    case doNOP:
        break;


    case doBadOpenParenType:
    case doRuleError:
        error(U_REGEX_RULE_SYNTAX);
        break;


    case doMismatchedParenErr:
        error(U_REGEX_MISMATCHED_PAREN);
        break;

    case doPlus:
        //  Normal '+'  compiles to
        //     1.   stuff to be repeated  (already built)
        //     2.   jmp-sav 1
        //     3.   ...
        //
        //  Or, if the item to be repeated can match a zero length string,
        //     1.   STO_INP_LOC  data-loc
        //     2.      body of stuff to be repeated
        //     3.   JMP_SAV_X    2
        //     4.   ...

        //
        //  Or, if the item to be repeated is simple
        //     1.   Item to be repeated.
        //     2.   LOOP_SR_I    set number  (assuming repeated item is a set ref)
        //     3.   LOOP_C       stack location
        {
            int32_t  topLoc = blockTopLoc(FALSE);        // location of item #1
            int32_t  frameLoc;

            // Check for simple constructs, which may get special optimized code.
            if (topLoc == fRXPat->fCompiledPat->size() - 1) {
                int32_t repeatedOp = (int32_t)fRXPat->fCompiledPat->elementAti(topLoc);

                if (URX_TYPE(repeatedOp) == URX_SETREF) {
                    // Emit optimized code for [char set]+
                    appendOp(URX_LOOP_SR_I, URX_VAL(repeatedOp));
                    frameLoc = allocateStackData(1);
                    appendOp(URX_LOOP_C, frameLoc);
                    break;
                }

                if (URX_TYPE(repeatedOp) == URX_DOTANY ||
                    URX_TYPE(repeatedOp) == URX_DOTANY_ALL ||
                    URX_TYPE(repeatedOp) == URX_DOTANY_UNIX) {
                    // Emit Optimized code for .+ operations.
                    int32_t loopOpI = buildOp(URX_LOOP_DOT_I, 0);
                    if (URX_TYPE(repeatedOp) == URX_DOTANY_ALL) {
                        // URX_LOOP_DOT_I operand is a flag indicating ". matches any" mode.
                        loopOpI |= 1;
                    }
                    if (fModeFlags & UREGEX_UNIX_LINES) {
                        loopOpI |= 2;
                    }
                    appendOp(loopOpI);
                    frameLoc = allocateStackData(1);
                    appendOp(URX_LOOP_C, frameLoc);
                    break;
                }

            }

            // General case.

            // Check for minimum match length of zero, which requires
            //    extra loop-breaking code.
            if (minMatchLength(topLoc, fRXPat->fCompiledPat->size()-1) == 0) {
                // Zero length match is possible.
                // Emit the code sequence that can handle it.
                insertOp(topLoc);
                frameLoc = allocateStackData(1);

                int32_t op = buildOp(URX_STO_INP_LOC, frameLoc);
                fRXPat->fCompiledPat->setElementAt(op, topLoc);

                appendOp(URX_JMP_SAV_X, topLoc+1);
            } else {
                // Simpler code when the repeated body must match something non-empty
                appendOp(URX_JMP_SAV, topLoc);
            }
        }
        break;

    case doNGPlus:
        //  Non-greedy '+?'  compiles to
        //     1.   stuff to be repeated  (already built)
        //     2.   state-save  1
        //     3.   ...
        {
            int32_t topLoc      = blockTopLoc(FALSE);
            appendOp(URX_STATE_SAVE, topLoc);
        }
        break;


    case doOpt:
        // Normal (greedy) ? quantifier.
        //  Compiles to
        //     1. state save 3
        //     2.    body of optional block
        //     3. ...
        // Insert the state save into the compiled pattern, and we're done.
        {
            int32_t   saveStateLoc = blockTopLoc(TRUE);
            int32_t   saveStateOp  = buildOp(URX_STATE_SAVE, fRXPat->fCompiledPat->size());
            fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc);
        }
        break;

    case doNGOpt:
        // Non-greedy ?? quantifier
        //   compiles to
        //    1.  jmp   4
        //    2.     body of optional block
        //    3   jmp   5
        //    4.  state save 2
        //    5    ...
        //  This code is less than ideal, with two jmps instead of one, because we can only
        //  insert one instruction at the top of the block being iterated.
        {
            int32_t  jmp1_loc = blockTopLoc(TRUE);
            int32_t  jmp2_loc = fRXPat->fCompiledPat->size();

            int32_t  jmp1_op  = buildOp(URX_JMP, jmp2_loc+1);
            fRXPat->fCompiledPat->setElementAt(jmp1_op, jmp1_loc);

            appendOp(URX_JMP, jmp2_loc+2);

            appendOp(URX_STATE_SAVE, jmp1_loc+1);
        }
        break;


    case doStar:
        // Normal (greedy) * quantifier.
        // Compiles to
        //       1.   STATE_SAVE   4
        //       2.      body of stuff being iterated over
        //       3.   JMP_SAV      2
        //       4.   ...
        //
        // Or, if the body is a simple [Set],
        //       1.   LOOP_SR_I    set number
        //       2.   LOOP_C       stack location
        //       ...
        //
        // Or if this is a .*
        //       1.   LOOP_DOT_I    (. matches all mode flag)
        //       2.   LOOP_C        stack location
        //
        // Or, if the body can match a zero-length string, to inhibit infinite loops,
        //       1.   STATE_SAVE   5
        //       2.   STO_INP_LOC  data-loc
        //       3.      body of stuff
        //       4.   JMP_SAV_X    2
        //       5.   ...
        {
            // location of item #1, the STATE_SAVE
            int32_t   topLoc = blockTopLoc(FALSE);
            int32_t   dataLoc = -1;

            // Check for simple *, where the construct being repeated
            //   compiled to single opcode, and might be optimizable.
            if (topLoc == fRXPat->fCompiledPat->size() - 1) {
                int32_t repeatedOp = (int32_t)fRXPat->fCompiledPat->elementAti(topLoc);

                if (URX_TYPE(repeatedOp) == URX_SETREF) {
                    // Emit optimized code for a [char set]*
                    int32_t loopOpI = buildOp(URX_LOOP_SR_I, URX_VAL(repeatedOp));
                    fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc);
                    dataLoc = allocateStackData(1);
                    appendOp(URX_LOOP_C, dataLoc);
                    break;
                }

                if (URX_TYPE(repeatedOp) == URX_DOTANY ||
                    URX_TYPE(repeatedOp) == URX_DOTANY_ALL ||
                    URX_TYPE(repeatedOp) == URX_DOTANY_UNIX) {
                    // Emit Optimized code for .* operations.
                    int32_t loopOpI = buildOp(URX_LOOP_DOT_I, 0);
                    if (URX_TYPE(repeatedOp) == URX_DOTANY_ALL) {
                        // URX_LOOP_DOT_I operand is a flag indicating . matches any mode.
                        loopOpI |= 1;
                    }
                    if ((fModeFlags & UREGEX_UNIX_LINES) != 0) {
                        loopOpI |= 2;
                    }
                    fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc);
                    dataLoc = allocateStackData(1);
                    appendOp(URX_LOOP_C, dataLoc);
                    break;
                }
            }

            // Emit general case code for this *
            // The optimizations did not apply.

            int32_t   saveStateLoc = blockTopLoc(TRUE);
            int32_t   jmpOp        = buildOp(URX_JMP_SAV, saveStateLoc+1);

            // Check for minimum match length of zero, which requires
            //    extra loop-breaking code.
            if (minMatchLength(saveStateLoc, fRXPat->fCompiledPat->size()-1) == 0) {
                insertOp(saveStateLoc);
                dataLoc = allocateStackData(1);

                int32_t op = buildOp(URX_STO_INP_LOC, dataLoc);
                fRXPat->fCompiledPat->setElementAt(op, saveStateLoc+1);
                jmpOp      = buildOp(URX_JMP_SAV_X, saveStateLoc+2);
            }

            // Locate the position in the compiled pattern where the match will continue
            //   after completing the *.   (4 or 5 in the comment above)
            int32_t continueLoc = fRXPat->fCompiledPat->size()+1;

            // Put together the save state op and store it into the compiled code.
            int32_t saveStateOp = buildOp(URX_STATE_SAVE, continueLoc);
            fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc);

            // Append the URX_JMP_SAV or URX_JMPX operation to the compiled pattern.
            appendOp(jmpOp);
        }
        break;

    case doNGStar:
        // Non-greedy *? quantifier
        // compiles to
        //     1.   JMP    3
        //     2.      body of stuff being iterated over
        //     3.   STATE_SAVE  2
        //     4    ...
        {
            int32_t     jmpLoc  = blockTopLoc(TRUE);                   // loc  1.
            int32_t     saveLoc = fRXPat->fCompiledPat->size();        // loc  3.
            int32_t     jmpOp   = buildOp(URX_JMP, saveLoc);
            fRXPat->fCompiledPat->setElementAt(jmpOp, jmpLoc);
            appendOp(URX_STATE_SAVE, jmpLoc+1);
        }
        break;


    case doIntervalInit:
        // The '{' opening an interval quantifier was just scanned.
        // Init the counter varaiables that will accumulate the values as the digits
        //    are scanned.
        fIntervalLow = 0;
        fIntervalUpper = -1;
        break;

    case doIntevalLowerDigit:
        // Scanned a digit from the lower value of an {lower,upper} interval
        {
            int32_t digitValue = u_charDigitValue(fC.fChar);
            U_ASSERT(digitValue >= 0);
            int64_t val = (int64_t)fIntervalLow*10 + digitValue;
            if (val > INT32_MAX) {
                error(U_REGEX_NUMBER_TOO_BIG);
            } else {
                fIntervalLow = (int32_t)val;
            }
        }
        break;

    case doIntervalUpperDigit:
        // Scanned a digit from the upper value of an {lower,upper} interval
        {
            if (fIntervalUpper < 0) {
                fIntervalUpper = 0;
            }
            int32_t digitValue = u_charDigitValue(fC.fChar);
            U_ASSERT(digitValue >= 0);
            int64_t val = (int64_t)fIntervalUpper*10 + digitValue;
            if (val > INT32_MAX) {
                error(U_REGEX_NUMBER_TOO_BIG);
            } else {
                fIntervalUpper = (int32_t)val;
            }
        }
        break;

    case doIntervalSame:
        // Scanned a single value interval like {27}.  Upper = Lower.
        fIntervalUpper = fIntervalLow;
        break;

    case doInterval:
        // Finished scanning a normal {lower,upper} interval.  Generate the code for it.
        if (compileInlineInterval() == FALSE) {
            compileInterval(URX_CTR_INIT, URX_CTR_LOOP);
        }
        break;

    case doPossessiveInterval:
        // Finished scanning a Possessive {lower,upper}+ interval.  Generate the code for it.
        {
            // Remember the loc for the top of the block being looped over.
            //   (Can not reserve a slot in the compiled pattern at this time, because
            //    compileInterval needs to reserve also, and blockTopLoc can only reserve
            //    once per block.)
            int32_t topLoc = blockTopLoc(FALSE);

            // Produce normal looping code.
            compileInterval(URX_CTR_INIT, URX_CTR_LOOP);

            // Surround the just-emitted normal looping code with a STO_SP ... LD_SP
            //  just as if the loop was inclosed in atomic parentheses.

            // First the STO_SP before the start of the loop
            insertOp(topLoc);

            int32_t  varLoc = allocateData(1);   // Reserve a data location for saving the
            int32_t  op     = buildOp(URX_STO_SP, varLoc);
            fRXPat->fCompiledPat->setElementAt(op, topLoc);

            int32_t loopOp = (int32_t)fRXPat->fCompiledPat->popi();
            U_ASSERT(URX_TYPE(loopOp) == URX_CTR_LOOP && URX_VAL(loopOp) == topLoc);
            loopOp++;     // point LoopOp after the just-inserted STO_SP
            fRXPat->fCompiledPat->push(loopOp, *fStatus);

            // Then the LD_SP after the end of the loop
            appendOp(URX_LD_SP, varLoc);
        }

        break;

    case doNGInterval:
        // Finished scanning a non-greedy {lower,upper}? interval.  Generate the code for it.
        compileInterval(URX_CTR_INIT_NG, URX_CTR_LOOP_NG);
        break;

    case doIntervalError:
        error(U_REGEX_BAD_INTERVAL);
        break;

    case doLiteralChar:
        // We've just scanned a "normal" character from the pattern,
        literalChar(fC.fChar);
        break;


    case doEscapedLiteralChar:
        // We've just scanned an backslashed escaped character with  no
        //   special meaning.  It represents itself.
        if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 &&
            ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) ||     // in [A-Z]
            (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) {   // in [a-z]
               error(U_REGEX_BAD_ESCAPE_SEQUENCE);
             }
        literalChar(fC.fChar);
        break;


    case doDotAny:
        // scanned a ".",  match any single character.
        {
            fixLiterals(FALSE);
            if (fModeFlags & UREGEX_DOTALL) {
                appendOp(URX_DOTANY_ALL, 0);
            } else if (fModeFlags & UREGEX_UNIX_LINES) {
                appendOp(URX_DOTANY_UNIX, 0);
            } else {
                appendOp(URX_DOTANY, 0);
            }
        }
        break;

    case doCaret:
        {
            fixLiterals(FALSE);
            if (       (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) {
                appendOp(URX_CARET, 0);
            } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) {
                appendOp(URX_CARET_M, 0);
            } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) {
                appendOp(URX_CARET, 0);   // Only testing true start of input.
            } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) {
                appendOp(URX_CARET_M_UNIX, 0);
            }
        }
        break;

    case doDollar:
        {
            fixLiterals(FALSE);
            if (       (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) {
                appendOp(URX_DOLLAR, 0);
            } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) {
                appendOp(URX_DOLLAR_M, 0);
            } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) {
                appendOp(URX_DOLLAR_D, 0);
            } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) {
                appendOp(URX_DOLLAR_MD, 0);
            }
        }
        break;

    case doBackslashA:
        fixLiterals(FALSE);
        appendOp(URX_CARET, 0);
        break;

    case doBackslashB:
        {
            #if  UCONFIG_NO_BREAK_ITERATION==1
            if (fModeFlags & UREGEX_UWORD) {
                error(U_UNSUPPORTED_ERROR);
            }
            #endif
            fixLiterals(FALSE);
            int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B;
            appendOp(op, 1);
        }
        break;

    case doBackslashb:
        {
            #if  UCONFIG_NO_BREAK_ITERATION==1
            if (fModeFlags & UREGEX_UWORD) {
                error(U_UNSUPPORTED_ERROR);
            }
            #endif
            fixLiterals(FALSE);
            int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B;
            appendOp(op, 0);
        }
        break;

    case doBackslashD:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_D, 1);
        break;

    case doBackslashd:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_D, 0);
        break;

    case doBackslashG:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_G, 0);
        break;

    case doBackslashH:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_H, 1);
        break;

    case doBackslashh:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_H, 0);
        break;

    case doBackslashR:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_R, 0);
        break;

    case doBackslashS:
        fixLiterals(FALSE);
        appendOp(URX_STAT_SETREF_N, URX_ISSPACE_SET);
        break;

    case doBackslashs:
        fixLiterals(FALSE);
        appendOp(URX_STATIC_SETREF, URX_ISSPACE_SET);
        break;

    case doBackslashV:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_V, 1);
        break;

    case doBackslashv:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_V, 0);
        break;

    case doBackslashW:
        fixLiterals(FALSE);
        appendOp(URX_STAT_SETREF_N, URX_ISWORD_SET);
        break;

    case doBackslashw:
        fixLiterals(FALSE);
        appendOp(URX_STATIC_SETREF, URX_ISWORD_SET);
        break;

    case doBackslashX:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_X, 0);
        break;


    case doBackslashZ:
        fixLiterals(FALSE);
        appendOp(URX_DOLLAR, 0);
        break;

    case doBackslashz:
        fixLiterals(FALSE);
        appendOp(URX_BACKSLASH_Z, 0);
        break;

    case doEscapeError:
        error(U_REGEX_BAD_ESCAPE_SEQUENCE);
        break;

    case doExit:
        fixLiterals(FALSE);
        returnVal = FALSE;
        break;

    case doProperty:
        {
            fixLiterals(FALSE);
            UnicodeSet *theSet = scanProp();
            compileSet(theSet);
        }
        break;

    case doNamedChar:
        {
            UChar32 c = scanNamedChar();
            literalChar(c);
        }
        break;


    case doBackRef:
        // BackReference.  Somewhat unusual in that the front-end can not completely parse
        //                 the regular expression, because the number of digits to be consumed
        //                 depends on the number of capture groups that have been defined.  So
        //                 we have to do it here instead.
        {
            int32_t  numCaptureGroups = fRXPat->fGroupMap->size();
            int32_t  groupNum = 0;
            UChar32  c        = fC.fChar;

            for (;;) {
                // Loop once per digit, for max allowed number of digits in a back reference.
                int32_t digit = u_charDigitValue(c);
                groupNum = groupNum * 10 + digit;
                if (groupNum >= numCaptureGroups) {
                    break;
                }
                c = peekCharLL();
                if (RegexStaticSets::gStaticSets->fRuleDigitsAlias->contains(c) == FALSE) {
                    break;
                }
                nextCharLL();
            }

            // Scan of the back reference in the source regexp is complete.  Now generate
            //  the compiled code for it.
            // Because capture groups can be forward-referenced by back-references,
            //  we fill the operand with the capture group number.  At the end
            //  of compilation, it will be changed to the variable's location.
            U_ASSERT(groupNum > 0);  // Shouldn't happen.  '\0' begins an octal escape sequence,
                                     //    and shouldn't enter this code path at all.
            fixLiterals(FALSE);
            if (fModeFlags & UREGEX_CASE_INSENSITIVE) {
                appendOp(URX_BACKREF_I, groupNum);
            } else {
                appendOp(URX_BACKREF, groupNum);
            }
        }
        break;

    case doBeginNamedBackRef:
        U_ASSERT(fCaptureName == NULL);
        fCaptureName = new UnicodeString;
        if (fCaptureName == NULL) {
            error(U_MEMORY_ALLOCATION_ERROR);
        }
        break;
            
    case doContinueNamedBackRef:
        fCaptureName->append(fC.fChar);
        break;

    case doCompleteNamedBackRef:
        {
        int32_t groupNumber = uhash_geti(fRXPat->fNamedCaptureMap, fCaptureName);
        if (groupNumber == 0) {
            // Group name has not been defined.
            //   Could be a forward reference. If we choose to support them at some
            //   future time, extra mechanism will be required at this point.
            error(U_REGEX_INVALID_CAPTURE_GROUP_NAME);
        } else {
            // Given the number, handle identically to a \n numbered back reference.
            // See comments above, under doBackRef
            fixLiterals(FALSE);
            if (fModeFlags & UREGEX_CASE_INSENSITIVE) {
                appendOp(URX_BACKREF_I, groupNumber);
            } else {
                appendOp(URX_BACKREF, groupNumber);
            }
        }
        delete fCaptureName;
        fCaptureName = NULL;
        break;
        }
       
    case doPossessivePlus:
        // Possessive ++ quantifier.
        // Compiles to
        //       1.   STO_SP
        //       2.      body of stuff being iterated over
        //       3.   STATE_SAVE 5
        //       4.   JMP        2
        //       5.   LD_SP
        //       6.   ...
        //
        //  Note:  TODO:  This is pretty inefficient.  A mass of saved state is built up
        //                then unconditionally discarded.  Perhaps introduce a new opcode.  Ticket 6056
        //
        {
            // Emit the STO_SP
            int32_t   topLoc = blockTopLoc(TRUE);
            int32_t   stoLoc = allocateData(1);  // Reserve the data location for storing save stack ptr.
            int32_t   op     = buildOp(URX_STO_SP, stoLoc);
            fRXPat->fCompiledPat->setElementAt(op, topLoc);

            // Emit the STATE_SAVE
            appendOp(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+2);

            // Emit the JMP
            appendOp(URX_JMP, topLoc+1);

            // Emit the LD_SP
            appendOp(URX_LD_SP, stoLoc);
        }
        break;

    case doPossessiveStar:
        // Possessive *+ quantifier.
        // Compiles to
        //       1.   STO_SP       loc
        //       2.   STATE_SAVE   5
        //       3.      body of stuff being iterated over
        //       4.   JMP          2
        //       5.   LD_SP        loc
        //       6    ...
        // TODO:  do something to cut back the state stack each time through the loop.
        {
            // Reserve two slots at the top of the block.
            int32_t   topLoc = blockTopLoc(TRUE);
            insertOp(topLoc);

            // emit   STO_SP     loc
            int32_t   stoLoc = allocateData(1);    // Reserve the data location for storing save stack ptr.
            int32_t   op     = buildOp(URX_STO_SP, stoLoc);
            fRXPat->fCompiledPat->setElementAt(op, topLoc);

            // Emit the SAVE_STATE   5
            int32_t L7 = fRXPat->fCompiledPat->size()+1;
            op = buildOp(URX_STATE_SAVE, L7);
            fRXPat->fCompiledPat->setElementAt(op, topLoc+1);

            // Append the JMP operation.
            appendOp(URX_JMP, topLoc+1);

            // Emit the LD_SP       loc
            appendOp(URX_LD_SP, stoLoc);
        }
        break;

    case doPossessiveOpt:
        // Possessive  ?+ quantifier.
        //  Compiles to
        //     1. STO_SP      loc
        //     2. SAVE_STATE  5
        //     3.    body of optional block
        //     4. LD_SP       loc
        //     5. ...
        //
        {
            // Reserve two slots at the top of the block.
            int32_t   topLoc = blockTopLoc(TRUE);
            insertOp(topLoc);

            // Emit the STO_SP
            int32_t   stoLoc = allocateData(1);   // Reserve the data location for storing save stack ptr.
            int32_t   op     = buildOp(URX_STO_SP, stoLoc);
            fRXPat->fCompiledPat->setElementAt(op, topLoc);

            // Emit the SAVE_STATE
            int32_t   continueLoc = fRXPat->fCompiledPat->size()+1;
            op = buildOp(URX_STATE_SAVE, continueLoc);
            fRXPat->fCompiledPat->setElementAt(op, topLoc+1);

            // Emit the LD_SP
            appendOp(URX_LD_SP, stoLoc);
        }
        break;


    case doBeginMatchMode:
        fNewModeFlags = fModeFlags;
        fSetModeFlag  = TRUE;
        break;

    case doMatchMode:   //  (?i)    and similar
        {
            int32_t  bit = 0;
            switch (fC.fChar) {
            case 0x69: /* 'i' */   bit = UREGEX_CASE_INSENSITIVE; break;
            case 0x64: /* 'd' */   bit = UREGEX_UNIX_LINES;       break;
            case 0x6d: /* 'm' */   bit = UREGEX_MULTILINE;        break;
            case 0x73: /* 's' */   bit = UREGEX_DOTALL;           break;
            case 0x75: /* 'u' */   bit = 0; /* Unicode casing */  break;
            case 0x77: /* 'w' */   bit = UREGEX_UWORD;            break;
            case 0x78: /* 'x' */   bit = UREGEX_COMMENTS;         break;
            case 0x2d: /* '-' */   fSetModeFlag = FALSE;          break;
            default:
                U_ASSERT(FALSE);   // Should never happen.  Other chars are filtered out
                                   // by the scanner.
            }
            if (fSetModeFlag) {
                fNewModeFlags |= bit;
            } else {
                fNewModeFlags &= ~bit;
            }
        }
        break;

    case doSetMatchMode:
        // Emit code to match any pending literals, using the not-yet changed match mode.
        fixLiterals();

        // We've got a (?i) or similar.  The match mode is being changed, but
        //   the change is not scoped to a parenthesized block.
        U_ASSERT(fNewModeFlags < 0);
        fModeFlags = fNewModeFlags;

        break;


    case doMatchModeParen:
        // We've got a (?i: or similar.  Begin a parenthesized block, save old
        //   mode flags so they can be restored at the close of the block.
        //
        //   Compile to a
        //      - NOP, which later may be replaced by a save-state if the
        //         parenthesized group gets a * quantifier, followed by
        //      - NOP, which may later be replaced by a save-state if there
        //             is an '|' alternation within the parens.
        {
            fixLiterals(FALSE);
            appendOp(URX_NOP, 0);
            appendOp(URX_NOP, 0);

            // On the Parentheses stack, start a new frame and add the postions
            //   of the two NOPs (a normal non-capturing () frame, except for the
            //   saving of the orignal mode flags.)
            fParenStack.push(fModeFlags, *fStatus);
            fParenStack.push(flags, *fStatus);                            // Frame Marker
            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first NOP
            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP

            // Set the current mode flags to the new values.
            U_ASSERT(fNewModeFlags < 0);
            fModeFlags = fNewModeFlags;
        }
        break;

    case doBadModeFlag:
        error(U_REGEX_INVALID_FLAG);
        break;

    case doSuppressComments:
        // We have just scanned a '(?'.  We now need to prevent the character scanner from
        // treating a '#' as a to-the-end-of-line comment.
        //   (This Perl compatibility just gets uglier and uglier to do...)
        fEOLComments = FALSE;
        break;


    case doSetAddAmp:
        {
          UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
          set->add(chAmp);
        }
        break;

    case doSetAddDash:
        {
          UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
          set->add(chDash);
        }
        break;

     case doSetBackslash_s:
        {
         UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
         set->addAll(*RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]);
         break;
        }

     case doSetBackslash_S:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            UnicodeSet SSet(*RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]);
            SSet.complement();
            set->addAll(SSet);
            break;
        }

    case doSetBackslash_d:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            // TODO - make a static set, ticket 6058.
            addCategory(set, U_GC_ND_MASK, *fStatus);
            break;
        }

    case doSetBackslash_D:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            UnicodeSet digits;
            // TODO - make a static set, ticket 6058.
            digits.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK, *fStatus);
            digits.complement();
            set->addAll(digits);
            break;
        }

    case doSetBackslash_h:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            UnicodeSet h;
            h.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ZS_MASK, *fStatus);
            h.add((UChar32)9);   // Tab
            set->addAll(h);
            break;
        }

    case doSetBackslash_H:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            UnicodeSet h;
            h.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ZS_MASK, *fStatus);
            h.add((UChar32)9);   // Tab
            h.complement();
            set->addAll(h);
            break;
        }

    case doSetBackslash_v:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            set->add((UChar32)0x0a, (UChar32)0x0d);  // add range
            set->add((UChar32)0x85);
            set->add((UChar32)0x2028, (UChar32)0x2029);
            break;
        }

    case doSetBackslash_V:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            UnicodeSet v;
            v.add((UChar32)0x0a, (UChar32)0x0d);  // add range
            v.add((UChar32)0x85);
            v.add((UChar32)0x2028, (UChar32)0x2029);
            v.complement();
            set->addAll(v);
            break;
        }

    case doSetBackslash_w:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            set->addAll(*RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]);
            break;
        }

    case doSetBackslash_W:
        {
            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
            UnicodeSet SSet(*RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]);
            SSet.complement();
            set->addAll(SSet);
            break;
        }

    case doSetBegin:
        fixLiterals(FALSE);
        fSetStack.push(new UnicodeSet(), *fStatus);
        fSetOpStack.push(setStart, *fStatus);
        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) {
            fSetOpStack.push(setCaseClose, *fStatus);
        }
        break;

    case doSetBeginDifference1:
        //  We have scanned something like [[abc]-[
        //  Set up a new UnicodeSet for the set beginning with the just-scanned '['
        //  Push a Difference operator, which will cause the new set to be subtracted from what
        //    went before once it is created.
        setPushOp(setDifference1);
        fSetOpStack.push(setStart, *fStatus);
        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) {
            fSetOpStack.push(setCaseClose, *fStatus);
        }
        break;

    case doSetBeginIntersection1:
        //  We have scanned something like  [[abc]&[
        //   Need both the '&' operator and the open '[' operator.
        setPushOp(setIntersection1);
        fSetOpStack.push(setStart, *fStatus);
        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) {
            fSetOpStack.push(setCaseClose, *fStatus);
        }
        break;

    case doSetBeginUnion:
        //  We have scanned something like  [[abc][
        //     Need to handle the union operation explicitly [[abc] | [
        setPushOp(setUnion);
        fSetOpStack.push(setStart, *fStatus);
        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) {
            fSetOpStack.push(setCaseClose, *fStatus);
        }
        break;

    case doSetDifference2:
        // We have scanned something like [abc--
        //   Consider this to unambiguously be a set difference operator.
        setPushOp(setDifference2);
        break;

    case doSetEnd:
        // Have encountered the ']' that closes a set.
        //    Force the evaluation of any pending operations within this set,
        //    leave the completed set on the top of the set stack.
        setEval(setEnd);
        U_ASSERT(fSetOpStack.peeki()==setStart);
        fSetOpStack.popi();
        break;

    case doSetFinish:
        {
        // Finished a complete set expression, including all nested sets.
        //   The close bracket has already triggered clearing out pending set operators,
        //    the operator stack should be empty and the operand stack should have just
        //    one entry, the result set.
        U_ASSERT(fSetOpStack.empty());
        UnicodeSet *theSet = (UnicodeSet *)fSetStack.pop();
        U_ASSERT(fSetStack.empty());
        compileSet(theSet);
        break;
        }

    case doSetIntersection2:
        // Have scanned something like [abc&&
        setPushOp(setIntersection2);
        break;

    case doSetLiteral:
        // Union the just-scanned literal character into the set being built.
        //    This operation is the highest precedence set operation, so we can always do
        //    it immediately, without waiting to see what follows.  It is necessary to perform
        //    any pending '-' or '&' operation first, because these have the same precedence
        //    as union-ing in a literal'
        {
            setEval(setUnion);
            UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
            s->add(fC.fChar);
            fLastSetLiteral = fC.fChar;
            break;
        }

    case doSetLiteralEscaped:
        // A back-slash escaped literal character was encountered.
        // Processing is the same as with setLiteral, above, with the addition of
        //  the optional check for errors on escaped ASCII letters.
        {
            if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 &&
                ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) ||     // in [A-Z]
                 (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) {   // in [a-z]
                error(U_REGEX_BAD_ESCAPE_SEQUENCE);
            }
            setEval(setUnion);
            UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
            s->add(fC.fChar);
            fLastSetLiteral = fC.fChar;
            break;
        }

        case doSetNamedChar:
        // Scanning a \N{UNICODE CHARACTER NAME}
        //  Aside from the source of the character, the processing is identical to doSetLiteral,
        //    above.
        {
            UChar32  c = scanNamedChar();
            setEval(setUnion);
            UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
            s->add(c);
            fLastSetLiteral = c;
            break;
        }

    case doSetNamedRange:
        // We have scanned literal-\N{CHAR NAME}.  Add the range to the set.
        // The left character is already in the set, and is saved in fLastSetLiteral.
        // The right side needs to be picked up, the scan is at the 'N'.
        // Lower Limit > Upper limit being an error matches both Java
        //        and ICU UnicodeSet behavior.
        {
            UChar32  c = scanNamedChar();
            if (U_SUCCESS(*fStatus) && (fLastSetLiteral == U_SENTINEL || fLastSetLiteral > c)) {
                error(U_REGEX_INVALID_RANGE);
            }
            UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
            s->add(fLastSetLiteral, c);
            fLastSetLiteral = c;
            break;
        }


    case  doSetNegate:
        // Scanned a '^' at the start of a set.
        // Push the negation operator onto the set op stack.
        // A twist for case-insensitive matching:
        //   the case closure operation must happen _before_ negation.
        //   But the case closure operation will already be on the stack if it's required.
        //   This requires checking for case closure, and swapping the stack order
        //    if it is present.
        {
            int32_t  tosOp = fSetOpStack.peeki();
            if (tosOp == setCaseClose) {
                fSetOpStack.popi();
                fSetOpStack.push(setNegation, *fStatus);
                fSetOpStack.push(setCaseClose, *fStatus);
            } else {
                fSetOpStack.push(setNegation, *fStatus);
            }
        }
        break;

    case doSetNoCloseError:
        error(U_REGEX_MISSING_CLOSE_BRACKET);
        break;

    case doSetOpError:
        error(U_REGEX_RULE_SYNTAX);   //  -- or && at the end of a set.  Illegal.
        break;

    case doSetPosixProp:
        {
            UnicodeSet *s = scanPosixProp();
            if (s != NULL) {
                UnicodeSet *tos = (UnicodeSet *)fSetStack.peek();
                tos->addAll(*s);
                delete s;
            }  // else error.  scanProp() reported the error status already.
        }
        break;

    case doSetProp:
        //  Scanned a \p \P within [brackets].
        {
            UnicodeSet *s = scanProp();
            if (s != NULL) {
                UnicodeSet *tos = (UnicodeSet *)fSetStack.peek();
                tos->addAll(*s);
                delete s;
            }  // else error.  scanProp() reported the error status already.
        }
        break;


    case doSetRange:
        // We have scanned literal-literal.  Add the range to the set.
        // The left character is already in the set, and is saved in fLastSetLiteral.
        // The right side is the current character.
        // Lower Limit > Upper limit being an error matches both Java
        //        and ICU UnicodeSet behavior.
        {

        if (fLastSetLiteral == U_SENTINEL || fLastSetLiteral > fC.fChar) {
            error(U_REGEX_INVALID_RANGE);
        }
        UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
        s->add(fLastSetLiteral, fC.fChar);
        break;
        }

    default:
        U_ASSERT(FALSE);
        error(U_REGEX_INTERNAL_ERROR);
        break;
    }

    if (U_FAILURE(*fStatus)) {
        returnVal = FALSE;
    }

    return returnVal;
}



//------------------------------------------------------------------------------
//
//   literalChar           We've encountered a literal character from the pattern,
//                             or an escape sequence that reduces to a character.
//                         Add it to the string containing all literal chars/strings from
//                             the pattern.
//
//------------------------------------------------------------------------------
void RegexCompile::literalChar(UChar32 c)  {
    fLiteralChars.append(c);
}


//------------------------------------------------------------------------------
//
//    fixLiterals           When compiling something that can follow a literal
//                          string in a pattern, emit the code to match the
//                          accumulated literal string.
//
//                          Optionally, split the last char of the string off into
//                          a single "ONE_CHAR" operation, so that quantifiers can
//                          apply to that char alone.  Example:   abc*
//                          The * must apply to the 'c' only.
//
//------------------------------------------------------------------------------
void    RegexCompile::fixLiterals(UBool split) {

    // If no literal characters have been scanned but not yet had code generated
    //   for them, nothing needs to be done.
    if (fLiteralChars.length() == 0) {
        return;
    }

    int32_t indexOfLastCodePoint = fLiteralChars.moveIndex32(fLiteralChars.length(), -1);
    UChar32 lastCodePoint = fLiteralChars.char32At(indexOfLastCodePoint);

    // Split:  We need to  ensure that the last item in the compiled pattern
    //     refers only to the last literal scanned in the pattern, so that
    //     quantifiers (*, +, etc.) affect only it, and not a longer string.
    //     Split before case folding for case insensitive matches.

    if (split) {
        fLiteralChars.truncate(indexOfLastCodePoint);
        fixLiterals(FALSE);   // Recursive call, emit code to match the first part of the string.
                              //  Note that the truncated literal string may be empty, in which case
                              //  nothing will be emitted.

        literalChar(lastCodePoint);  // Re-add the last code point as if it were a new literal.
        fixLiterals(FALSE);          // Second recursive call, code for the final code point.
        return;
    }

    // If we are doing case-insensitive matching, case fold the string.  This may expand
    //   the string, e.g. the German sharp-s turns into "ss"
    if (fModeFlags & UREGEX_CASE_INSENSITIVE) {
        fLiteralChars.foldCase();
        indexOfLastCodePoint = fLiteralChars.moveIndex32(fLiteralChars.length(), -1);
        lastCodePoint = fLiteralChars.char32At(indexOfLastCodePoint);
    }

    if (indexOfLastCodePoint == 0) {
        // Single character, emit a URX_ONECHAR op to match it.
        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) &&
                 u_hasBinaryProperty(lastCodePoint, UCHAR_CASE_SENSITIVE)) {
            appendOp(URX_ONECHAR_I, lastCodePoint);
        } else {
            appendOp(URX_ONECHAR, lastCodePoint);
        }
    } else {
        // Two or more chars, emit a URX_STRING to match them.
        if (fLiteralChars.length() > 0x00ffffff || fRXPat->fLiteralText.length() > 0x00ffffff) {
            error(U_REGEX_PATTERN_TOO_BIG);
        }
        if (fModeFlags & UREGEX_CASE_INSENSITIVE) {
            appendOp(URX_STRING_I, fRXPat->fLiteralText.length());
        } else {
            // TODO here:  add optimization to split case sensitive strings of length two
            //             into two single char ops, for efficiency.
            appendOp(URX_STRING, fRXPat->fLiteralText.length());
        }
        appendOp(URX_STRING_LEN, fLiteralChars.length());

        // Add this string into the accumulated strings of the compiled pattern.
        fRXPat->fLiteralText.append(fLiteralChars);
    }

    fLiteralChars.remove();
}


int32_t RegexCompile::buildOp(int32_t type, int32_t val) {
    if (U_FAILURE(*fStatus)) {
        return 0;
    }
    if (type < 0 || type > 255) {
        U_ASSERT(FALSE);
        error(U_REGEX_INTERNAL_ERROR);
        type = URX_RESERVED_OP;
    }
    if (val > 0x00ffffff) {
        U_ASSERT(FALSE);
        error(U_REGEX_INTERNAL_ERROR);
        val = 0;
    }
    if (val < 0) {
        if (!(type == URX_RESERVED_OP_N || type == URX_RESERVED_OP)) {
            U_ASSERT(FALSE);
            error(U_REGEX_INTERNAL_ERROR);
            return -1;
        }
        if (URX_TYPE(val) != 0xff) {
            U_ASSERT(FALSE);
            error(U_REGEX_INTERNAL_ERROR);
            return -1;
        }
        type = URX_RESERVED_OP_N;
    }
    return (type << 24) | val;
}


//------------------------------------------------------------------------------
//
//   appendOp()             Append a new instruction onto the compiled pattern
//                          Includes error checking, limiting the size of the
//                          pattern to lengths that can be represented in the
//                          24 bit operand field of an instruction.
//
//------------------------------------------------------------------------------
void RegexCompile::appendOp(int32_t op) {
    if (U_FAILURE(*fStatus)) {
        return;
    }
    fRXPat->fCompiledPat->addElement(op, *fStatus);
    if ((fRXPat->fCompiledPat->size() > 0x00fffff0) && U_SUCCESS(*fStatus)) {
        error(U_REGEX_PATTERN_TOO_BIG);
    }
}

void RegexCompile::appendOp(int32_t type, int32_t val) {
    appendOp(buildOp(type, val));
}


//------------------------------------------------------------------------------
//
//   insertOp()             Insert a slot for a new opcode into the already
//                          compiled pattern code.
//
//                          Fill the slot with a NOP.  Our caller will replace it
//                          with what they really wanted.
//
//------------------------------------------------------------------------------
void   RegexCompile::insertOp(int32_t where) {
    UVector64 *code = fRXPat->fCompiledPat;
    U_ASSERT(where>0 && where < code->size());

    int32_t  nop = buildOp(URX_NOP, 0);
    code->insertElementAt(nop, where, *fStatus);

    // Walk through the pattern, looking for any ops with targets that
    //  were moved down by the insert.  Fix them.
    int32_t loc;
    for (loc=0; loc<code->size(); loc++) {
        int32_t op = (int32_t)code->elementAti(loc);
        int32_t opType = URX_TYPE(op);
        int32_t opValue = URX_VAL(op);
        if ((opType == URX_JMP         ||
            opType == URX_JMPX         ||
            opType == URX_STATE_SAVE   ||
            opType == URX_CTR_LOOP     ||
            opType == URX_CTR_LOOP_NG  ||
            opType == URX_JMP_SAV      ||
            opType == URX_JMP_SAV_X    ||
            opType == URX_RELOC_OPRND)    && opValue > where) {
            // Target location for this opcode is after the insertion point and
            //   needs to be incremented to adjust for the insertion.
            opValue++;
            op = buildOp(opType, opValue);
            code->setElementAt(op, loc);
        }
    }

    // Now fix up the parentheses stack.  All positive values in it are locations in
    //  the compiled pattern.   (Negative values are frame boundaries, and don't need fixing.)
    for (loc=0; loc<fParenStack.size(); loc++) {
        int32_t x = fParenStack.elementAti(loc);
        U_ASSERT(x < code->size());
        if (x>where) {
            x++;
            fParenStack.setElementAt(x, loc);
        }
    }

    if (fMatchCloseParen > where) {
        fMatchCloseParen++;
    }
    if (fMatchOpenParen > where) {
        fMatchOpenParen++;
    }
}


//------------------------------------------------------------------------------
//
//   allocateData()        Allocate storage in the matcher's static data area.
//                         Return the index for the newly allocated data.
//                         The storage won't actually exist until we are running a match
//                         operation, but the storage indexes are inserted into various
//                         opcodes while compiling the pattern.
//
//------------------------------------------------------------------------------
int32_t RegexCompile::allocateData(int32_t size) {
    if (U_FAILURE(*fStatus)) {
        return 0;
    }
    if (size <= 0 || size > 0x100 || fRXPat->fDataSize < 0) {
        error(U_REGEX_INTERNAL_ERROR);
        return 0;
    }
    int32_t dataIndex = fRXPat->fDataSize;
    fRXPat->fDataSize += size;
    if (fRXPat->fDataSize >= 0x00fffff0) {
        error(U_REGEX_INTERNAL_ERROR);
    }
    return dataIndex;
}


//------------------------------------------------------------------------------
//
//   allocateStackData()   Allocate space in the back-tracking stack frame.
//                         Return the index for the newly allocated data.
//                         The frame indexes are inserted into various
//                         opcodes while compiling the pattern, meaning that frame
//                         size must be restricted to the size that will fit
//                         as an operand (24 bits).
//
//------------------------------------------------------------------------------
int32_t RegexCompile::allocateStackData(int32_t size) {
    if (U_FAILURE(*fStatus)) {
        return 0;
    }
    if (size <= 0 || size > 0x100 || fRXPat->fFrameSize < 0) {
        error(U_REGEX_INTERNAL_ERROR);
        return 0;
    }
    int32_t dataIndex = fRXPat->fFrameSize;
    fRXPat->fFrameSize += size;
    if (fRXPat->fFrameSize >= 0x00fffff0) {
        error(U_REGEX_PATTERN_TOO_BIG);
    }
    return dataIndex;
}


//------------------------------------------------------------------------------
//
//   blockTopLoc()          Find or create a location in the compiled pattern
//                          at the start of the operation or block that has
//                          just been compiled.  Needed when a quantifier (* or
//                          whatever) appears, and we need to add an operation
//                          at the start of the thing being quantified.
//
//                          (Parenthesized Blocks) have a slot with a NOP that
//                          is reserved for this purpose.  .* or similar don't
//                          and a slot needs to be added.
//
//       parameter reserveLoc   :  TRUE -  ensure that there is space to add an opcode
//                                         at the returned location.
//                                 FALSE - just return the address,
//                                         do not reserve a location there.
//
//------------------------------------------------------------------------------
int32_t   RegexCompile::blockTopLoc(UBool reserveLoc) {
    int32_t   theLoc;
    fixLiterals(TRUE);  // Emit code for any pending literals.
                        //   If last item was a string, emit separate op for the its last char.
    if (fRXPat->fCompiledPat->size() == fMatchCloseParen)
    {
        // The item just processed is a parenthesized block.
        theLoc = fMatchOpenParen;   // A slot is already reserved for us.
        U_ASSERT(theLoc > 0);
        U_ASSERT(URX_TYPE(((uint32_t)fRXPat->fCompiledPat->elementAti(theLoc))) == URX_NOP);
    }
    else {
        // Item just compiled is a single thing, a ".", or a single char, a string or a set reference.
        // No slot for STATE_SAVE was pre-reserved in the compiled code.
        // We need to make space now.
        theLoc = fRXPat->fCompiledPat->size()-1;
        int32_t opAtTheLoc = (int32_t)fRXPat->fCompiledPat->elementAti(theLoc);
        if (URX_TYPE(opAtTheLoc) == URX_STRING_LEN) {
            // Strings take two opcode, we want the position of the first one.
            // We can have a string at this point if a single character case-folded to two.
            theLoc--;
        }
        if (reserveLoc) {
            int32_t  nop = buildOp(URX_NOP, 0);
            fRXPat->fCompiledPat->insertElementAt(nop, theLoc, *fStatus);
        }
    }
    return theLoc;
}



//------------------------------------------------------------------------------
//
//    handleCloseParen      When compiling a close paren, we need to go back
//                          and fix up any JMP or SAVE operations within the
//                          parenthesized block that need to target the end
//                          of the block.  The locations of these are kept on
//                          the paretheses stack.
//
//                          This function is called both when encountering a
//                          real ) and at the end of the pattern.
//
//------------------------------------------------------------------------------
void  RegexCompile::handleCloseParen() {
    int32_t   patIdx;
    int32_t   patOp;
    if (fParenStack.size() <= 0) {
        error(U_REGEX_MISMATCHED_PAREN);
        return;
    }

    // Emit code for any pending literals.
    fixLiterals(FALSE);

    // Fixup any operations within the just-closed parenthesized group
    //    that need to reference the end of the (block).
    //    (The first one popped from the stack is an unused slot for
    //     alternation (OR) state save, but applying the fixup to it does no harm.)
    for (;;) {
        patIdx = fParenStack.popi();
        if (patIdx < 0) {
            // value < 0 flags the start of the frame on the paren stack.
            break;
        }
        U_ASSERT(patIdx>0 && patIdx <= fRXPat->fCompiledPat->size());
        patOp = (int32_t)fRXPat->fCompiledPat->elementAti(patIdx);
        U_ASSERT(URX_VAL(patOp) == 0);          // Branch target for JMP should not be set.
        patOp |= fRXPat->fCompiledPat->size();  // Set it now.
        fRXPat->fCompiledPat->setElementAt(patOp, patIdx);
        fMatchOpenParen     = patIdx;
    }

    //  At the close of any parenthesized block, restore the match mode flags  to
    //  the value they had at the open paren.  Saved value is
    //  at the top of the paren stack.
    fModeFlags = fParenStack.popi();
    U_ASSERT(fModeFlags < 0);

    // DO any additional fixups, depending on the specific kind of
    // parentesized grouping this is

    switch (patIdx) {
    case plain:
    case flags:
        // No additional fixups required.
        //   (Grouping-only parentheses)
        break;
    case capturing:
        // Capturing Parentheses.
        //   Insert a End Capture op into the pattern.
        //   The frame offset of the variables for this cg is obtained from the
        //       start capture op and put it into the end-capture op.
        {
            int32_t   captureOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1);
            U_ASSERT(URX_TYPE(captureOp) == URX_START_CAPTURE);

            int32_t   frameVarLocation = URX_VAL(captureOp);
            appendOp(URX_END_CAPTURE, frameVarLocation);
        }
        break;
    case atomic:
        // Atomic Parenthesis.
        //   Insert a LD_SP operation to restore the state stack to the position
        //   it was when the atomic parens were entered.
        {
            int32_t   stoOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1);
            U_ASSERT(URX_TYPE(stoOp) == URX_STO_SP);
            int32_t   stoLoc = URX_VAL(stoOp);
            appendOp(URX_LD_SP, stoLoc);
        }
        break;

    case lookAhead:
        {
            int32_t  startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5);
            U_ASSERT(URX_TYPE(startOp) == URX_LA_START);
            int32_t dataLoc  = URX_VAL(startOp);
            appendOp(URX_LA_END, dataLoc);
        }
        break;

    case negLookAhead:
        {
            // See comment at doOpenLookAheadNeg
            int32_t  startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-1);
            U_ASSERT(URX_TYPE(startOp) == URX_LA_START);
            int32_t dataLoc  = URX_VAL(startOp);
            appendOp(URX_LA_END, dataLoc);
            appendOp(URX_BACKTRACK, 0);
            appendOp(URX_LA_END, dataLoc);

            // Patch the URX_SAVE near the top of the block.
            // The destination of the SAVE is the final LA_END that was just added.
            int32_t saveOp   = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen);
            U_ASSERT(URX_TYPE(saveOp) == URX_STATE_SAVE);
            int32_t dest     = fRXPat->fCompiledPat->size()-1;
            saveOp           = buildOp(URX_STATE_SAVE, dest);
            fRXPat->fCompiledPat->setElementAt(saveOp, fMatchOpenParen);
        }
        break;

    case lookBehind:
        {
            // See comment at doOpenLookBehind.

            // Append the URX_LB_END and URX_LA_END to the compiled pattern.
            int32_t  startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-4);
            U_ASSERT(URX_TYPE(startOp) == URX_LB_START);
            int32_t dataLoc  = URX_VAL(startOp);
            appendOp(URX_LB_END, dataLoc);
            appendOp(URX_LA_END, dataLoc);

            // Determine the min and max bounds for the length of the
            //  string that the pattern can match.
            //  An unbounded upper limit is an error.
            int32_t patEnd   = fRXPat->fCompiledPat->size() - 1;
            int32_t minML    = minMatchLength(fMatchOpenParen, patEnd);
            int32_t maxML    = maxMatchLength(fMatchOpenParen, patEnd);
            if (URX_TYPE(maxML) != 0) {
                error(U_REGEX_LOOK_BEHIND_LIMIT);
                break;
            }
            if (maxML == INT32_MAX) {
                error(U_REGEX_LOOK_BEHIND_LIMIT);
                break;
            }
            U_ASSERT(minML <= maxML);

            // Insert the min and max match len bounds into the URX_LB_CONT op that
            //  appears at the top of the look-behind block, at location fMatchOpenParen+1
            fRXPat->fCompiledPat->setElementAt(minML,  fMatchOpenParen-2);
            fRXPat->fCompiledPat->setElementAt(maxML,  fMatchOpenParen-1);

        }
        break;



    case lookBehindN:
        {
            // See comment at doOpenLookBehindNeg.

            // Append the URX_LBN_END to the compiled pattern.
            int32_t  startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5);
            U_ASSERT(URX_TYPE(startOp) == URX_LB_START);
            int32_t dataLoc  = URX_VAL(startOp);
            appendOp(URX_LBN_END, dataLoc);

            // Determine the min and max bounds for the length of the
            //  string that the pattern can match.
            //  An unbounded upper limit is an error.
            int32_t patEnd   = fRXPat->fCompiledPat->size() - 1;
            int32_t minML    = minMatchLength(fMatchOpenParen, patEnd);
            int32_t maxML    = maxMatchLength(fMatchOpenParen, patEnd);
            if (URX_TYPE(maxML) != 0) {
                error(U_REGEX_LOOK_BEHIND_LIMIT);
                break;
            }
            if (maxML == INT32_MAX) {
                error(U_REGEX_LOOK_BEHIND_LIMIT);
                break;
            }
            U_ASSERT(minML <= maxML);

            // Insert the min and max match len bounds into the URX_LB_CONT op that
            //  appears at the top of the look-behind block, at location fMatchOpenParen+1
            fRXPat->fCompiledPat->setElementAt(minML,  fMatchOpenParen-3);
            fRXPat->fCompiledPat->setElementAt(maxML,  fMatchOpenParen-2);

            // Insert the pattern location to continue at after a successful match
            //  as the last operand of the URX_LBN_CONT
            int32_t op = buildOp(URX_RELOC_OPRND, fRXPat->fCompiledPat->size());
            fRXPat->fCompiledPat->setElementAt(op,  fMatchOpenParen-1);
        }
        break;



    default:
        U_ASSERT(FALSE);
    }

    // remember the next location in the compiled pattern.
    // The compilation of Quantifiers will look at this to see whether its looping
    //   over a parenthesized block or a single item
    fMatchCloseParen = fRXPat->fCompiledPat->size();
}



//------------------------------------------------------------------------------
//
//   compileSet       Compile the pattern operations for a reference to a
//                    UnicodeSet.
//
//------------------------------------------------------------------------------
void        RegexCompile::compileSet(UnicodeSet *theSet)
{
    if (theSet == NULL) {
        return;
    }
    //  Remove any strings from the set.
    //  There shoudn't be any, but just in case.
    //     (Case Closure can add them; if we had a simple case closure avaialble that
    //      ignored strings, that would be better.)
    theSet->removeAllStrings();
    int32_t  setSize = theSet->size();

    switch (setSize) {
    case 0:
        {
            // Set of no elements.   Always fails to match.
            appendOp(URX_BACKTRACK, 0);
            delete theSet;
        }
        break;

    case 1:
        {
            // The set contains only a single code point.  Put it into
            //   the compiled pattern as a single char operation rather
            //   than a set, and discard the set itself.
            literalChar(theSet->charAt(0));
            delete theSet;
        }
        break;

    default:
        {
            //  The set contains two or more chars.  (the normal case)
            //  Put it into the compiled pattern as a set.
            int32_t setNumber = fRXPat->fSets->size();
            fRXPat->fSets->addElement(theSet, *fStatus);
            appendOp(URX_SETREF, setNumber);
        }
    }
}


//------------------------------------------------------------------------------
//
//   compileInterval    Generate the code for a {min, max} style interval quantifier.
//                      Except for the specific opcodes used, the code is the same
//                      for all three types (greedy, non-greedy, possessive) of
//                      intervals.  The opcodes are supplied as parameters.
//                      (There are two sets of opcodes - greedy & possessive use the
//                      same ones, while non-greedy has it's own.)
//
//                      The code for interval loops has this form:
//                         0  CTR_INIT   counter loc (in stack frame)
//                         1             5  patt address of CTR_LOOP at bottom of block
//                         2             min count
//                         3             max count   (-1 for unbounded)
//                         4  ...        block to be iterated over
//                         5  CTR_LOOP
//
//                       In
//------------------------------------------------------------------------------
void        RegexCompile::compileInterval(int32_t InitOp,  int32_t LoopOp)
{
    // The CTR_INIT op at the top of the block with the {n,m} quantifier takes
    //   four slots in the compiled code.  Reserve them.
    int32_t   topOfBlock = blockTopLoc(TRUE);
    insertOp(topOfBlock);
    insertOp(topOfBlock);
    insertOp(topOfBlock);

    // The operands for the CTR_INIT opcode include the index in the matcher data
    //   of the counter.  Allocate it now. There are two data items
    //        counterLoc   -->  Loop counter
    //               +1    -->  Input index (for breaking non-progressing loops)
    //                          (Only present if unbounded upper limit on loop)
    int32_t   dataSize = fIntervalUpper < 0 ? 2 : 1;
    int32_t   counterLoc = allocateStackData(dataSize);

    int32_t   op = buildOp(InitOp, counterLoc);
    fRXPat->fCompiledPat->setElementAt(op, topOfBlock);

    // The second operand of CTR_INIT is the location following the end of the loop.
    //   Must put in as a URX_RELOC_OPRND so that the value will be adjusted if the
    //   compilation of something later on causes the code to grow and the target
    //   position to move.
    int32_t loopEnd = fRXPat->fCompiledPat->size();
    op = buildOp(URX_RELOC_OPRND, loopEnd);
    fRXPat->fCompiledPat->setElementAt(op, topOfBlock+1);

    // Followed by the min and max counts.
    fRXPat->fCompiledPat->setElementAt(fIntervalLow, topOfBlock+2);
    fRXPat->fCompiledPat->setElementAt(fIntervalUpper, topOfBlock+3);

    // Apend the CTR_LOOP op.  The operand is the location of the CTR_INIT op.
    //   Goes at end of the block being looped over, so just append to the code so far.
    appendOp(LoopOp, topOfBlock);

    if ((fIntervalLow & 0xff000000) != 0 ||
        (fIntervalUpper > 0 && (fIntervalUpper & 0xff000000) != 0)) {
            error(U_REGEX_NUMBER_TOO_BIG);
        }

    if (fIntervalLow > fIntervalUpper && fIntervalUpper != -1) {
        error(U_REGEX_MAX_LT_MIN);
    }
}



UBool RegexCompile::compileInlineInterval() {
    if (fIntervalUpper > 10 || fIntervalUpper < fIntervalLow) {
        // Too big to inline.  Fail, which will cause looping code to be generated.
        //   (Upper < Lower picks up unbounded upper and errors, both.)
        return FALSE;
    }

    int32_t   topOfBlock = blockTopLoc(FALSE);
    if (fIntervalUpper == 0) {
        // Pathological case.  Attempt no matches, as if the block doesn't exist.
        // Discard the generated code for the block.
        // If the block included parens, discard the info pertaining to them as well.
        fRXPat->fCompiledPat->setSize(topOfBlock);
        if (fMatchOpenParen >= topOfBlock) {
            fMatchOpenParen = -1;
        }
        if (fMatchCloseParen >= topOfBlock) {
            fMatchCloseParen = -1;
        }
        return TRUE;
    }

    if (topOfBlock != fRXPat->fCompiledPat->size()-1 && fIntervalUpper != 1) {
        // The thing being repeated is not a single op, but some
        //   more complex block.  Do it as a loop, not inlines.
        //   Note that things "repeated" a max of once are handled as inline, because
        //     the one copy of the code already generated is just fine.
        return FALSE;
    }

    // Pick up the opcode that is to be repeated
    //
    int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(topOfBlock);

    // Compute the pattern location where the inline sequence
    //   will end, and set up the state save op that will be needed.
    //
    int32_t endOfSequenceLoc = fRXPat->fCompiledPat->size()-1
                                + fIntervalUpper + (fIntervalUpper-fIntervalLow);
    int32_t saveOp = buildOp(URX_STATE_SAVE, endOfSequenceLoc);
    if (fIntervalLow == 0) {
        insertOp(topOfBlock);
        fRXPat->fCompiledPat->setElementAt(saveOp, topOfBlock);
    }



    //  Loop, emitting the op for the thing being repeated each time.
    //    Loop starts at 1 because one instance of the op already exists in the pattern,
    //    it was put there when it was originally encountered.
    int32_t i;
    for (i=1; i<fIntervalUpper; i++ ) {
        if (i >= fIntervalLow) {
            appendOp(saveOp);
        }
        appendOp(op);
    }
    return TRUE;
}



//------------------------------------------------------------------------------
//
//   caseInsensitiveStart  given a single code point from a pattern string, determine the 
//                         set of characters that could potentially begin a case-insensitive 
//                         match of a string beginning with that character, using full Unicode
//                         case insensitive matching.
//
//          This is used in optimizing find().
//
//          closeOver(USET_CASE_INSENSITIVE) does most of what is needed, but
//          misses cases like this:
//             A string from the pattern begins with 'ss' (although all we know
//                 in this context is that it begins with 's')
//             The pattern could match a string beginning with a German sharp-s
//
//           To the ordinary case closure for a character c, we add all other
//           characters cx where the case closure of cx incudes a string form that begins
//           with the original character c.
//
//           This function could be made smarter. The full pattern string is available
//           and it would be possible to verify that the extra characters being added
//           to the starting set fully match, rather than having just a first-char of the
//           folded form match.
//
//------------------------------------------------------------------------------
void  RegexCompile::findCaseInsensitiveStarters(UChar32 c, UnicodeSet *starterChars) {

// Machine Generated below.
// It may need updating with new versions of Unicode.
// Intltest test RegexTest::TestCaseInsensitiveStarters will fail if an update is needed.
// The update tool is here: svn+ssh://source.icu-project.org/repos/icu/tools/trunk/unicode/c/genregexcasing

// Machine Generated Data. Do not hand edit.
    static const UChar32 RECaseFixCodePoints[] = {
        0x61, 0x66, 0x68, 0x69, 0x6a, 0x73, 0x74, 0x77, 0x79, 0x2bc, 
        0x3ac, 0x3ae, 0x3b1, 0x3b7, 0x3b9, 0x3c1, 0x3c5, 0x3c9, 0x3ce, 0x565, 
        0x574, 0x57e, 0x1f00, 0x1f01, 0x1f02, 0x1f03, 0x1f04, 0x1f05, 0x1f06, 0x1f07, 
        0x1f20, 0x1f21, 0x1f22, 0x1f23, 0x1f24, 0x1f25, 0x1f26, 0x1f27, 0x1f60, 0x1f61, 
        0x1f62, 0x1f63, 0x1f64, 0x1f65, 0x1f66, 0x1f67, 0x1f70, 0x1f74, 0x1f7c, 0x110000};

    static const int16_t RECaseFixStringOffsets[] = {
        0x0, 0x1, 0x6, 0x7, 0x8, 0x9, 0xd, 0xe, 0xf, 0x10, 
        0x11, 0x12, 0x13, 0x17, 0x1b, 0x20, 0x21, 0x2a, 0x2e, 0x2f, 
        0x30, 0x34, 0x35, 0x37, 0x39, 0x3b, 0x3d, 0x3f, 0x41, 0x43, 
        0x45, 0x47, 0x49, 0x4b, 0x4d, 0x4f, 0x51, 0x53, 0x55, 0x57, 
        0x59, 0x5b, 0x5d, 0x5f, 0x61, 0x63, 0x65, 0x66, 0x67, 0};

    static const int16_t RECaseFixCounts[] = {
        0x1, 0x5, 0x1, 0x1, 0x1, 0x4, 0x1, 0x1, 0x1, 0x1, 
        0x1, 0x1, 0x4, 0x4, 0x5, 0x1, 0x9, 0x4, 0x1, 0x1, 
        0x4, 0x1, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 
        0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 
        0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x1, 0x1, 0x1, 0};

    static const UChar RECaseFixData[] = {
        0x1e9a, 0xfb00, 0xfb01, 0xfb02, 0xfb03, 0xfb04, 0x1e96, 0x130, 0x1f0, 0xdf, 
        0x1e9e, 0xfb05, 0xfb06, 0x1e97, 0x1e98, 0x1e99, 0x149, 0x1fb4, 0x1fc4, 0x1fb3, 
        0x1fb6, 0x1fb7, 0x1fbc, 0x1fc3, 0x1fc6, 0x1fc7, 0x1fcc, 0x390, 0x1fd2, 0x1fd3, 
        0x1fd6, 0x1fd7, 0x1fe4, 0x3b0, 0x1f50, 0x1f52, 0x1f54, 0x1f56, 0x1fe2, 0x1fe3, 
        0x1fe6, 0x1fe7, 0x1ff3, 0x1ff6, 0x1ff7, 0x1ffc, 0x1ff4, 0x587, 0xfb13, 0xfb14, 
        0xfb15, 0xfb17, 0xfb16, 0x1f80, 0x1f88, 0x1f81, 0x1f89, 0x1f82, 0x1f8a, 0x1f83, 
        0x1f8b, 0x1f84, 0x1f8c, 0x1f85, 0x1f8d, 0x1f86, 0x1f8e, 0x1f87, 0x1f8f, 0x1f90, 
        0x1f98, 0x1f91, 0x1f99, 0x1f92, 0x1f9a, 0x1f93, 0x1f9b, 0x1f94, 0x1f9c, 0x1f95, 
        0x1f9d, 0x1f96, 0x1f9e, 0x1f97, 0x1f9f, 0x1fa0, 0x1fa8, 0x1fa1, 0x1fa9, 0x1fa2, 
        0x1faa, 0x1fa3, 0x1fab, 0x1fa4, 0x1fac, 0x1fa5, 0x1fad, 0x1fa6, 0x1fae, 0x1fa7, 
        0x1faf, 0x1fb2, 0x1fc2, 0x1ff2, 0};

// End of machine generated data.

    if (c < UCHAR_MIN_VALUE || c > UCHAR_MAX_VALUE) {
        // This function should never be called with an invalid input character.
        U_ASSERT(FALSE);
        starterChars->clear();
    } else if (u_hasBinaryProperty(c, UCHAR_CASE_SENSITIVE)) {
        UChar32 caseFoldedC  = u_foldCase(c, U_FOLD_CASE_DEFAULT);
        starterChars->set(caseFoldedC, caseFoldedC);

        int32_t i;
        for (i=0; RECaseFixCodePoints[i]<c ; i++) {
            // Simple linear search through the sorted list of interesting code points.
        }

        if (RECaseFixCodePoints[i] == c) {
            int32_t dataIndex = RECaseFixStringOffsets[i];
            int32_t numCharsToAdd = RECaseFixCounts[i];
            UChar32 cpToAdd = 0;
            for (int32_t j=0; j<numCharsToAdd; j++) {
                U16_NEXT_UNSAFE(RECaseFixData, dataIndex, cpToAdd);
                starterChars->add(cpToAdd);
            }
        }

        starterChars->closeOver(USET_CASE_INSENSITIVE);
        starterChars->removeAllStrings();
    } else {
        // Not a cased character. Just return it alone.
        starterChars->set(c, c);
    }
}




//------------------------------------------------------------------------------
//
//   matchStartType    Determine how a match can start.
//                     Used to optimize find() operations.
//
//                     Operation is very similar to minMatchLength().  Walk the compiled
//                     pattern, keeping an on-going minimum-match-length.  For any
//                     op where the min match coming in is zero, add that ops possible
//                     starting matches to the possible starts for the overall pattern.
//
//------------------------------------------------------------------------------
void   RegexCompile::matchStartType() {
    if (U_FAILURE(*fStatus)) {
        return;
    }


    int32_t    loc;                    // Location in the pattern of the current op being processed.
    int32_t    op;                     // The op being processed
    int32_t    opType;                 // The opcode type of the op
    int32_t    currentLen = 0;         // Minimum length of a match to this point (loc) in the pattern
    int32_t    numInitialStrings = 0;  // Number of strings encountered that could match at start.

    UBool      atStart = TRUE;         // True if no part of the pattern yet encountered
                                       //   could have advanced the position in a match.
                                       //   (Maximum match length so far == 0)

    // forwardedLength is a vector holding minimum-match-length values that
    //   are propagated forward in the pattern by JMP or STATE_SAVE operations.
    //   It must be one longer than the pattern being checked because some  ops
    //   will jmp to a end-of-block+1 location from within a block, and we must
    //   count those when checking the block.
    int32_t end = fRXPat->fCompiledPat->size();
    UVector32  forwardedLength(end+1, *fStatus);
    forwardedLength.setSize(end+1);
    for (loc=3; loc<end; loc++) {
        forwardedLength.setElementAt(INT32_MAX, loc);
    }

    for (loc = 3; loc<end; loc++) {
        op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
        opType = URX_TYPE(op);

        // The loop is advancing linearly through the pattern.
        // If the op we are now at was the destination of a branch in the pattern,
        // and that path has a shorter minimum length than the current accumulated value,
        // replace the current accumulated value.
        if (forwardedLength.elementAti(loc) < currentLen) {
            currentLen = forwardedLength.elementAti(loc);
            U_ASSERT(currentLen>=0 && currentLen < INT32_MAX);
        }

        switch (opType) {
            // Ops that don't change the total length matched
        case URX_RESERVED_OP:
        case URX_END:
        case URX_FAIL:
        case URX_STRING_LEN:
        case URX_NOP:
        case URX_START_CAPTURE:
        case URX_END_CAPTURE:
        case URX_BACKSLASH_B:
        case URX_BACKSLASH_BU:
        case URX_BACKSLASH_G:
        case URX_BACKSLASH_Z:
        case URX_DOLLAR:
        case URX_DOLLAR_M:
        case URX_DOLLAR_D:
        case URX_DOLLAR_MD:
        case URX_RELOC_OPRND:
        case URX_STO_INP_LOC:
        case URX_BACKREF:         // BackRef.  Must assume that it might be a zero length match
        case URX_BACKREF_I:

        case URX_STO_SP:          // Setup for atomic or possessive blocks.  Doesn't change what can match.
        case URX_LD_SP:
            break;

        case URX_CARET:
            if (atStart) {
                fRXPat->fStartType = START_START;
            }
            break;

        case URX_CARET_M:
        case URX_CARET_M_UNIX:
            if (atStart) {
                fRXPat->fStartType = START_LINE;
            }
            break;

        case URX_ONECHAR:
            if (currentLen == 0) {
                // This character could appear at the start of a match.
                //   Add it to the set of possible starting characters.
                fRXPat->fInitialChars->add(URX_VAL(op));
                numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;


        case URX_SETREF:
            if (currentLen == 0) {
                int32_t  sn = URX_VAL(op);
                U_ASSERT(sn > 0 && sn < fRXPat->fSets->size());
                const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn);
                fRXPat->fInitialChars->addAll(*s);
                numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;

        case URX_LOOP_SR_I:
            // [Set]*, like a SETREF, above, in what it can match,
            //  but may not match at all, so currentLen is not incremented.
            if (currentLen == 0) {
                int32_t  sn = URX_VAL(op);
                U_ASSERT(sn > 0 && sn < fRXPat->fSets->size());
                const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn);
                fRXPat->fInitialChars->addAll(*s);
                numInitialStrings += 2;
            }
            atStart = FALSE;
            break;

        case URX_LOOP_DOT_I:
            if (currentLen == 0) {
                // .* at the start of a pattern.
                //    Any character can begin the match.
                fRXPat->fInitialChars->clear();
                fRXPat->fInitialChars->complement();
                numInitialStrings += 2;
            }
            atStart = FALSE;
            break;


        case URX_STATIC_SETREF:
            if (currentLen == 0) {
                int32_t  sn = URX_VAL(op);
                U_ASSERT(sn>0 && sn<URX_LAST_SET);
                const UnicodeSet *s = fRXPat->fStaticSets[sn];
                fRXPat->fInitialChars->addAll(*s);
                numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;



        case URX_STAT_SETREF_N:
            if (currentLen == 0) {
                int32_t  sn = URX_VAL(op);
                const UnicodeSet *s = fRXPat->fStaticSets[sn];
                UnicodeSet sc(*s);
                sc.complement();
                fRXPat->fInitialChars->addAll(sc);
                numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;



        case URX_BACKSLASH_D:
            // Digit Char
             if (currentLen == 0) {
                 UnicodeSet s;
                 s.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK, *fStatus);
                 if (URX_VAL(op) != 0) {
                     s.complement();
                 }
                 fRXPat->fInitialChars->addAll(s);
                 numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;


        case URX_BACKSLASH_H:
            // Horiz white space
            if (currentLen == 0) {
                UnicodeSet s;
                s.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ZS_MASK, *fStatus);
                s.add((UChar32)9);   // Tab
                if (URX_VAL(op) != 0) {
                    s.complement();
                }
                fRXPat->fInitialChars->addAll(s);
                numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;


        case URX_BACKSLASH_R:       // Any line ending sequence
        case URX_BACKSLASH_V:       // Any line ending code point, with optional negation
            if (currentLen == 0) {
                UnicodeSet s;
                s.add((UChar32)0x0a, (UChar32)0x0d);  // add range
                s.add((UChar32)0x85);
                s.add((UChar32)0x2028, (UChar32)0x2029);
                if (URX_VAL(op) != 0) {
                     // Complement option applies to URX_BACKSLASH_V only.
                     s.complement();
                }
                fRXPat->fInitialChars->addAll(s);
                numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;



        case URX_ONECHAR_I:
            // Case Insensitive Single Character.
            if (currentLen == 0) {
                UChar32  c = URX_VAL(op);
                if (u_hasBinaryProperty(c, UCHAR_CASE_SENSITIVE)) {
                    UnicodeSet starters(c, c);
                    starters.closeOver(USET_CASE_INSENSITIVE);
                    // findCaseInsensitiveStarters(c, &starters);
                    //   For ONECHAR_I, no need to worry about text chars that expand on folding into strings.
                    //   The expanded folding can't match the pattern.
                    fRXPat->fInitialChars->addAll(starters);
                } else {
                    // Char has no case variants.  Just add it as-is to the
                    //   set of possible starting chars.
                    fRXPat->fInitialChars->add(c);
                }
                numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;


        case URX_BACKSLASH_X:   // Grahpeme Cluster.  Minimum is 1, max unbounded.
        case URX_DOTANY_ALL:    // . matches one or two.
        case URX_DOTANY:
        case URX_DOTANY_UNIX:
            if (currentLen == 0) {
                // These constructs are all bad news when they appear at the start
                //   of a match.  Any character can begin the match.
                fRXPat->fInitialChars->clear();
                fRXPat->fInitialChars->complement();
                numInitialStrings += 2;
            }
            currentLen++;
            atStart = FALSE;
            break;


        case URX_JMPX:
            loc++;             // Except for extra operand on URX_JMPX, same as URX_JMP.
            U_FALLTHROUGH;
        case URX_JMP:
            {
                int32_t  jmpDest = URX_VAL(op);
                if (jmpDest < loc) {
                    // Loop of some kind.  Can safely ignore, the worst that will happen
                    //  is that we understate the true minimum length
                    currentLen = forwardedLength.elementAti(loc+1);

                } else {
                    // Forward jump.  Propagate the current min length to the target loc of the jump.
                    U_ASSERT(jmpDest <= end+1);
                    if (forwardedLength.elementAti(jmpDest) > currentLen) {
                        forwardedLength.setElementAt(currentLen, jmpDest);
                    }
                }
            }
            atStart = FALSE;
            break;

        case URX_JMP_SAV:
        case URX_JMP_SAV_X:
            // Combo of state save to the next loc, + jmp backwards.
            //   Net effect on min. length computation is nothing.
            atStart = FALSE;
            break;

        case URX_BACKTRACK:
            // Fails are kind of like a branch, except that the min length was
            //   propagated already, by the state save.
            currentLen = forwardedLength.elementAti(loc+1);
            atStart = FALSE;
            break;


        case URX_STATE_SAVE:
            {
                // State Save, for forward jumps, propagate the current minimum.
                //             of the state save.
                int32_t  jmpDest = URX_VAL(op);
                if (jmpDest > loc) {
                    if (currentLen < forwardedLength.elementAti(jmpDest)) {
                        forwardedLength.setElementAt(currentLen, jmpDest);
                    }
                }
            }
            atStart = FALSE;
            break;




        case URX_STRING:
            {
                loc++;
                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
                int32_t stringLen   = URX_VAL(stringLenOp);
                U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN);
                U_ASSERT(stringLenOp >= 2);
                if (currentLen == 0) {
                    // Add the starting character of this string to the set of possible starting
                    //   characters for this pattern.
                    int32_t stringStartIdx = URX_VAL(op);
                    UChar32  c = fRXPat->fLiteralText.char32At(stringStartIdx);
                    fRXPat->fInitialChars->add(c);

                    // Remember this string.  After the entire pattern has been checked,
                    //  if nothing else is identified that can start a match, we'll use it.
                    numInitialStrings++;
                    fRXPat->fInitialStringIdx = stringStartIdx;
                    fRXPat->fInitialStringLen = stringLen;
                }

                currentLen += stringLen;
                atStart = FALSE;
            }
            break;

        case URX_STRING_I:
            {
                // Case-insensitive string.  Unlike exact-match strings, we won't
                //   attempt a string search for possible match positions.  But we
                //   do update the set of possible starting characters.
                loc++;
                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
                int32_t stringLen   = URX_VAL(stringLenOp);
                U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN);
                U_ASSERT(stringLenOp >= 2);
                if (currentLen == 0) {
                    // Add the starting character of this string to the set of possible starting
                    //   characters for this pattern.
                    int32_t stringStartIdx = URX_VAL(op);
                    UChar32  c = fRXPat->fLiteralText.char32At(stringStartIdx);
                    UnicodeSet s;
                    findCaseInsensitiveStarters(c, &s);
                    fRXPat->fInitialChars->addAll(s);
                    numInitialStrings += 2;  // Matching on an initial string not possible.
                }
                currentLen += stringLen;
                atStart = FALSE;
            }
            break;

        case URX_CTR_INIT:
        case URX_CTR_INIT_NG:
            {
                // Loop Init Ops.  These don't change the min length, but they are 4 word ops
                //   so location must be updated accordingly.
                // Loop Init Ops.
                //   If the min loop count == 0
                //      move loc forwards to the end of the loop, skipping over the body.
                //   If the min count is > 0,
                //      continue normal processing of the body of the loop.
                int32_t loopEndLoc   = (int32_t)fRXPat->fCompiledPat->elementAti(loc+1);
                        loopEndLoc   = URX_VAL(loopEndLoc);
                int32_t minLoopCount = (int32_t)fRXPat->fCompiledPat->elementAti(loc+2);
                if (minLoopCount == 0) {
                    // Min Loop Count of 0, treat like a forward branch and
                    //   move the current minimum length up to the target
                    //   (end of loop) location.
                    U_ASSERT(loopEndLoc <= end+1);
                    if (forwardedLength.elementAti(loopEndLoc) > currentLen) {
                        forwardedLength.setElementAt(currentLen, loopEndLoc);
                    }
                }
                loc+=3;  // Skips over operands of CTR_INIT
            }
            atStart = FALSE;
            break;


        case URX_CTR_LOOP:
        case URX_CTR_LOOP_NG:
            // Loop ops.
            //  The jump is conditional, backwards only.
            atStart = FALSE;
            break;

        case URX_LOOP_C:
            // More loop ops.  These state-save to themselves.
            //   don't change the minimum match
            atStart = FALSE;
            break;


        case URX_LA_START:
        case URX_LB_START:
            {
                // Look-around.  Scan forward until the matching look-ahead end,
                //   without processing the look-around block.  This is overly pessimistic.

                // Keep track of the nesting depth of look-around blocks.  Boilerplate code for
                //   lookahead contains two LA_END instructions, so count goes up by two
                //   for each LA_START.
                int32_t  depth = (opType == URX_LA_START? 2: 1);
                for (;;) {
                    loc++;
                    op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
                    if (URX_TYPE(op) == URX_LA_START) {
                        depth+=2;
                    }
                    if (URX_TYPE(op) == URX_LB_START) {
                        depth++;
                    }
                    if (URX_TYPE(op) == URX_LA_END || URX_TYPE(op)==URX_LBN_END) {
                        depth--;
                        if (depth == 0) {
                            break;
                        }
                    }
                    if (URX_TYPE(op) == URX_STATE_SAVE) {
                        // Need this because neg lookahead blocks will FAIL to outside
                        //   of the block.
                        int32_t  jmpDest = URX_VAL(op);
                        if (jmpDest > loc) {
                            if (currentLen < forwardedLength.elementAti(jmpDest)) {
                                forwardedLength.setElementAt(currentLen, jmpDest);
                            }
                        }
                    }
                    U_ASSERT(loc <= end);
                }
            }
            break;

        case URX_LA_END:
        case URX_LB_CONT:
        case URX_LB_END:
        case URX_LBN_CONT:
        case URX_LBN_END:
            U_ASSERT(FALSE);     // Shouldn't get here.  These ops should be
                                 //  consumed by the scan in URX_LA_START and LB_START

            break;

        default:
            U_ASSERT(FALSE);
            }

        }


    // We have finished walking through the ops.  Check whether some forward jump
    //   propagated a shorter length to location end+1.
    if (forwardedLength.elementAti(end+1) < currentLen) {
        currentLen = forwardedLength.elementAti(end+1);
    }


    fRXPat->fInitialChars8->init(fRXPat->fInitialChars);


    // Sort out what we should check for when looking for candidate match start positions.
    // In order of preference,
    //     1.   Start of input text buffer.
    //     2.   A literal string.
    //     3.   Start of line in multi-line mode.
    //     4.   A single literal character.
    //     5.   A character from a set of characters.
    //
    if (fRXPat->fStartType == START_START) {
        // Match only at the start of an input text string.
        //    start type is already set.  We're done.
    } else if (numInitialStrings == 1 && fRXPat->fMinMatchLen > 0) {
        // Match beginning only with a literal string.
        UChar32  c = fRXPat->fLiteralText.char32At(fRXPat->fInitialStringIdx);
        U_ASSERT(fRXPat->fInitialChars->contains(c));
        fRXPat->fStartType   = START_STRING;
        fRXPat->fInitialChar = c;
    } else if (fRXPat->fStartType == START_LINE) {
        // Match at start of line in Multi-Line mode.
        // Nothing to do here; everything is already set.
    } else if (fRXPat->fMinMatchLen == 0) {
        // Zero length match possible.  We could start anywhere.
        fRXPat->fStartType = START_NO_INFO;
    } else if (fRXPat->fInitialChars->size() == 1) {
        // All matches begin with the same char.
        fRXPat->fStartType   = START_CHAR;
        fRXPat->fInitialChar = fRXPat->fInitialChars->charAt(0);
        U_ASSERT(fRXPat->fInitialChar != (UChar32)-1);
    } else if (fRXPat->fInitialChars->contains((UChar32)0, (UChar32)0x10ffff) == FALSE &&
        fRXPat->fMinMatchLen > 0) {
        // Matches start with a set of character smaller than the set of all chars.
        fRXPat->fStartType = START_SET;
    } else {
        // Matches can start with anything
        fRXPat->fStartType = START_NO_INFO;
    }

    return;
}



//------------------------------------------------------------------------------
//
//   minMatchLength    Calculate the length of the shortest string that could
//                     match the specified pattern.
//                     Length is in 16 bit code units, not code points.
//
//                     The calculated length may not be exact.  The returned
//                     value may be shorter than the actual minimum; it must
//                     never be longer.
//
//                     start and end are the range of p-code operations to be
//                     examined.  The endpoints are included in the range.
//
//------------------------------------------------------------------------------
int32_t   RegexCompile::minMatchLength(int32_t start, int32_t end) {
    if (U_FAILURE(*fStatus)) {
        return 0;
    }

    U_ASSERT(start <= end);
    U_ASSERT(end < fRXPat->fCompiledPat->size());


    int32_t    loc;
    int32_t    op;
    int32_t    opType;
    int32_t    currentLen = 0;


    // forwardedLength is a vector holding minimum-match-length values that
    //   are propagated forward in the pattern by JMP or STATE_SAVE operations.
    //   It must be one longer than the pattern being checked because some  ops
    //   will jmp to a end-of-block+1 location from within a block, and we must
    //   count those when checking the block.
    UVector32  forwardedLength(end+2, *fStatus);
    forwardedLength.setSize(end+2);
    for (loc=start; loc<=end+1; loc++) {
        forwardedLength.setElementAt(INT32_MAX, loc);
    }

    for (loc = start; loc<=end; loc++) {
        op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
        opType = URX_TYPE(op);

        // The loop is advancing linearly through the pattern.
        // If the op we are now at was the destination of a branch in the pattern,
        // and that path has a shorter minimum length than the current accumulated value,
        // replace the current accumulated value.
        // U_ASSERT(currentLen>=0 && currentLen < INT32_MAX);  // MinLength == INT32_MAX for some
                                                               //   no-match-possible cases.
        if (forwardedLength.elementAti(loc) < currentLen) {
            currentLen = forwardedLength.elementAti(loc);
            U_ASSERT(currentLen>=0 && currentLen < INT32_MAX);
        }

        switch (opType) {
            // Ops that don't change the total length matched
        case URX_RESERVED_OP:
        case URX_END:
        case URX_STRING_LEN:
        case URX_NOP:
        case URX_START_CAPTURE:
        case URX_END_CAPTURE:
        case URX_BACKSLASH_B:
        case URX_BACKSLASH_BU:
        case URX_BACKSLASH_G:
        case URX_BACKSLASH_Z:
        case URX_CARET:
        case URX_DOLLAR:
        case URX_DOLLAR_M:
        case URX_DOLLAR_D:
        case URX_DOLLAR_MD:
        case URX_RELOC_OPRND:
        case URX_STO_INP_LOC:
        case URX_CARET_M:
        case URX_CARET_M_UNIX:
        case URX_BACKREF:         // BackRef.  Must assume that it might be a zero length match
        case URX_BACKREF_I:

        case URX_STO_SP:          // Setup for atomic or possessive blocks.  Doesn't change what can match.
        case URX_LD_SP:

        case URX_JMP_SAV:
        case URX_JMP_SAV_X:
            break;


            // Ops that match a minimum of one character (one or two 16 bit code units.)
            //
        case URX_ONECHAR:
        case URX_STATIC_SETREF:
        case URX_STAT_SETREF_N:
        case URX_SETREF:
        case URX_BACKSLASH_D:
        case URX_BACKSLASH_H:
        case URX_BACKSLASH_R:
        case URX_BACKSLASH_V:
        case URX_ONECHAR_I:
        case URX_BACKSLASH_X:   // Grahpeme Cluster.  Minimum is 1, max unbounded.
        case URX_DOTANY_ALL:    // . matches one or two.
        case URX_DOTANY:
        case URX_DOTANY_UNIX:
            currentLen++;
            break;


        case URX_JMPX:
            loc++;              // URX_JMPX has an extra operand, ignored here,
                                //   otherwise processed identically to URX_JMP.
            U_FALLTHROUGH;
        case URX_JMP:
            {
                int32_t  jmpDest = URX_VAL(op);
                if (jmpDest < loc) {
                    // Loop of some kind.  Can safely ignore, the worst that will happen
                    //  is that we understate the true minimum length
                    currentLen = forwardedLength.elementAti(loc+1);
                } else {
                    // Forward jump.  Propagate the current min length to the target loc of the jump.
                    U_ASSERT(jmpDest <= end+1);
                    if (forwardedLength.elementAti(jmpDest) > currentLen) {
                        forwardedLength.setElementAt(currentLen, jmpDest);
                    }
                }
            }
            break;

        case URX_BACKTRACK:
            {
                // Back-tracks are kind of like a branch, except that the min length was
                //   propagated already, by the state save.
                currentLen = forwardedLength.elementAti(loc+1);
            }
            break;


        case URX_STATE_SAVE:
            {
                // State Save, for forward jumps, propagate the current minimum.
                //             of the state save.
                int32_t  jmpDest = URX_VAL(op);
                if (jmpDest > loc) {
                    if (currentLen < forwardedLength.elementAti(jmpDest)) {
                        forwardedLength.setElementAt(currentLen, jmpDest);
                    }
                }
            }
            break;


        case URX_STRING:
            {
                loc++;
                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
                currentLen += URX_VAL(stringLenOp);
            }
            break;


        case URX_STRING_I:
            {
                loc++;
                // TODO: with full case folding, matching input text may be shorter than
                //       the string we have here.  More smarts could put some bounds on it.
                //       Assume a min length of one for now.  A min length of zero causes
                //        optimization failures for a pattern like "string"+
                // currentLen += URX_VAL(stringLenOp);
                currentLen += 1;
            }
            break;

        case URX_CTR_INIT:
        case URX_CTR_INIT_NG:
            {
                // Loop Init Ops.
                //   If the min loop count == 0
                //      move loc forwards to the end of the loop, skipping over the body.
                //   If the min count is > 0,
                //      continue normal processing of the body of the loop.
                int32_t loopEndLoc   = (int32_t)fRXPat->fCompiledPat->elementAti(loc+1);
                        loopEndLoc   = URX_VAL(loopEndLoc);
                int32_t minLoopCount = (int32_t)fRXPat->fCompiledPat->elementAti(loc+2);
                if (minLoopCount == 0) {
                    loc = loopEndLoc;
                } else {
                    loc+=3;  // Skips over operands of CTR_INIT
                }
            }
            break;


        case URX_CTR_LOOP:
        case URX_CTR_LOOP_NG:
            // Loop ops.
            //  The jump is conditional, backwards only.
            break;

        case URX_LOOP_SR_I:
        case URX_LOOP_DOT_I:
        case URX_LOOP_C:
            // More loop ops.  These state-save to themselves.
            //   don't change the minimum match - could match nothing at all.
            break;


        case URX_LA_START:
        case URX_LB_START:
            {
                // Look-around.  Scan forward until the matching look-ahead end,
                //   without processing the look-around block.  This is overly pessimistic for look-ahead,
                //   it assumes that the look-ahead match might be zero-length.
                //   TODO:  Positive lookahead could recursively do the block, then continue
                //          with the longer of the block or the value coming in.  Ticket 6060
                int32_t  depth = (opType == URX_LA_START? 2: 1);;
                for (;;) {
                    loc++;
                    op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
                    if (URX_TYPE(op) == URX_LA_START) {
                        // The boilerplate for look-ahead includes two LA_END insturctions,
                        //    Depth will be decremented by each one when it is seen.
                        depth += 2;
                    }
                    if (URX_TYPE(op) == URX_LB_START) {
                        depth++;
                    }
                    if (URX_TYPE(op) == URX_LA_END) {
                        depth--;
                        if (depth == 0) {
                            break;
                        }
                    }
                    if (URX_TYPE(op)==URX_LBN_END) {
                        depth--;
                        if (depth == 0) {
                            break;
                        }
                    }
                    if (URX_TYPE(op) == URX_STATE_SAVE) {
                        // Need this because neg lookahead blocks will FAIL to outside
                        //   of the block.
                        int32_t  jmpDest = URX_VAL(op);
                        if (jmpDest > loc) {
                            if (currentLen < forwardedLength.elementAti(jmpDest)) {
                                forwardedLength.setElementAt(currentLen, jmpDest);
                            }
                        }
                    }
                    U_ASSERT(loc <= end);
                }
            }
            break;

        case URX_LA_END:
        case URX_LB_CONT:
        case URX_LB_END:
        case URX_LBN_CONT:
        case URX_LBN_END:
            // Only come here if the matching URX_LA_START or URX_LB_START was not in the
            //   range being sized, which happens when measuring size of look-behind blocks.
            break;

        default:
            U_ASSERT(FALSE);
            }

        }

    // We have finished walking through the ops.  Check whether some forward jump
    //   propagated a shorter length to location end+1.
    if (forwardedLength.elementAti(end+1) < currentLen) {
        currentLen = forwardedLength.elementAti(end+1);
        U_ASSERT(currentLen>=0 && currentLen < INT32_MAX);
    }

    return currentLen;
}

// Increment with overflow check.
// val and delta will both be positive.

static int32_t safeIncrement(int32_t val, int32_t delta) {
    if (INT32_MAX - val > delta) {
        return val + delta;
    } else {
        return INT32_MAX;
    }
}


//------------------------------------------------------------------------------
//
//   maxMatchLength    Calculate the length of the longest string that could
//                     match the specified pattern.
//                     Length is in 16 bit code units, not code points.
//
//                     The calculated length may not be exact.  The returned
//                     value may be longer than the actual maximum; it must
//                     never be shorter.
//
//------------------------------------------------------------------------------
int32_t   RegexCompile::maxMatchLength(int32_t start, int32_t end) {
    if (U_FAILURE(*fStatus)) {
        return 0;
    }
    U_ASSERT(start <= end);
    U_ASSERT(end < fRXPat->fCompiledPat->size());


    int32_t    loc;
    int32_t    op;
    int32_t    opType;
    int32_t    currentLen = 0;
    UVector32  forwardedLength(end+1, *fStatus);
    forwardedLength.setSize(end+1);

    for (loc=start; loc<=end; loc++) {
        forwardedLength.setElementAt(0, loc);
    }

    for (loc = start; loc<=end; loc++) {
        op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
        opType = URX_TYPE(op);

        // The loop is advancing linearly through the pattern.
        // If the op we are now at was the destination of a branch in the pattern,
        // and that path has a longer maximum length than the current accumulated value,
        // replace the current accumulated value.
        if (forwardedLength.elementAti(loc) > currentLen) {
            currentLen = forwardedLength.elementAti(loc);
        }

        switch (opType) {
            // Ops that don't change the total length matched
        case URX_RESERVED_OP:
        case URX_END:
        case URX_STRING_LEN:
        case URX_NOP:
        case URX_START_CAPTURE:
        case URX_END_CAPTURE:
        case URX_BACKSLASH_B:
        case URX_BACKSLASH_BU:
        case URX_BACKSLASH_G:
        case URX_BACKSLASH_Z:
        case URX_CARET:
        case URX_DOLLAR:
        case URX_DOLLAR_M:
        case URX_DOLLAR_D:
        case URX_DOLLAR_MD:
        case URX_RELOC_OPRND:
        case URX_STO_INP_LOC:
        case URX_CARET_M:
        case URX_CARET_M_UNIX:

        case URX_STO_SP:          // Setup for atomic or possessive blocks.  Doesn't change what can match.
        case URX_LD_SP:

        case URX_LB_END:
        case URX_LB_CONT:
        case URX_LBN_CONT:
        case URX_LBN_END:
            break;


            // Ops that increase that cause an unbounded increase in the length
            //   of a matched string, or that increase it a hard to characterize way.
            //   Call the max length unbounded, and stop further checking.
        case URX_BACKREF:         // BackRef.  Must assume that it might be a zero length match
        case URX_BACKREF_I:
        case URX_BACKSLASH_X:   // Grahpeme Cluster.  Minimum is 1, max unbounded.
            currentLen = INT32_MAX;
            break;


            // Ops that match a max of one character (possibly two 16 bit code units.)
            //
        case URX_STATIC_SETREF:
        case URX_STAT_SETREF_N:
        case URX_SETREF:
        case URX_BACKSLASH_D:
        case URX_BACKSLASH_H:
        case URX_BACKSLASH_R:
        case URX_BACKSLASH_V:
        case URX_ONECHAR_I:
        case URX_DOTANY_ALL:
        case URX_DOTANY:
        case URX_DOTANY_UNIX:
            currentLen = safeIncrement(currentLen, 2);
            break;

            // Single literal character.  Increase current max length by one or two,
            //       depending on whether the char is in the supplementary range.
        case URX_ONECHAR:
            currentLen = safeIncrement(currentLen, 1);
            if (URX_VAL(op) > 0x10000) {
                currentLen = safeIncrement(currentLen, 1);
            }
            break;

            // Jumps.
            //
        case URX_JMP:
        case URX_JMPX:
        case URX_JMP_SAV:
        case URX_JMP_SAV_X:
            {
                int32_t  jmpDest = URX_VAL(op);
                if (jmpDest < loc) {
                    // Loop of some kind.  Max match length is unbounded.
                    currentLen = INT32_MAX;
                } else {
                    // Forward jump.  Propagate the current min length to the target loc of the jump.
                    if (forwardedLength.elementAti(jmpDest) < currentLen) {
                        forwardedLength.setElementAt(currentLen, jmpDest);
                    }
                    currentLen = 0;
                }
            }
            break;

        case URX_BACKTRACK:
            // back-tracks are kind of like a branch, except that the max length was
            //   propagated already, by the state save.
            currentLen = forwardedLength.elementAti(loc+1);
            break;


        case URX_STATE_SAVE:
            {
                // State Save, for forward jumps, propagate the current minimum.
                //               of the state save.
                //             For backwards jumps, they create a loop, maximum
                //               match length is unbounded.
                int32_t  jmpDest = URX_VAL(op);
                if (jmpDest > loc) {
                    if (currentLen > forwardedLength.elementAti(jmpDest)) {
                        forwardedLength.setElementAt(currentLen, jmpDest);
                    }
                } else {
                    currentLen = INT32_MAX;
                }
            }
            break;




        case URX_STRING:
            {
                loc++;
                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
                currentLen = safeIncrement(currentLen, URX_VAL(stringLenOp));
                break;
            }

        case URX_STRING_I:
            // TODO:  This code assumes that any user string that matches will be no longer
            //        than our compiled string, with case insensitive matching.
            //        Our compiled string has been case-folded already.
            //
            //        Any matching user string will have no more code points than our
            //        compiled (folded) string.  Folding may add code points, but
            //        not remove them.
            //
            //        There is a potential problem if a supplemental code point
            //        case-folds to a BMP code point.  In this case our compiled string
            //        could be shorter (in code units) than a matching user string.
            //
            //        At this time (Unicode 6.1) there are no such characters, and this case
            //        is not being handled.  A test, intltest regex/Bug9283, will fail if
            //        any problematic characters are added to Unicode.
            //
            //        If this happens, we can make a set of the BMP chars that the
            //        troublesome supplementals fold to, scan our string, and bump the
            //        currentLen one extra for each that is found.
            //
            {
                loc++;
                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
                currentLen = safeIncrement(currentLen, URX_VAL(stringLenOp));
            }
            break;

        case URX_CTR_INIT:
        case URX_CTR_INIT_NG:
            // For Loops, recursively call this function on the pattern for the loop body,
            //   then multiply the result by the maximum loop count.
            {
                int32_t  loopEndLoc = URX_VAL(fRXPat->fCompiledPat->elementAti(loc+1));
                if (loopEndLoc == loc+4) {
                    // Loop has an empty body. No affect on max match length.
                    // Continue processing with code after the loop end.
                    loc = loopEndLoc;
                    break;
                }

                int32_t maxLoopCount = static_cast<int32_t>(fRXPat->fCompiledPat->elementAti(loc+3));
                if (maxLoopCount == -1) {
                    // Unbounded Loop. No upper bound on match length.
                    currentLen = INT32_MAX;
                    break;
                }

                U_ASSERT(loopEndLoc >= loc+4);
                int64_t blockLen = maxMatchLength(loc+4, loopEndLoc-1);  // Recursive call.
                int64_t updatedLen = (int64_t)currentLen + blockLen * maxLoopCount; 
                if (updatedLen >= INT32_MAX) {
                    currentLen = INT32_MAX;
                    break;
                }
                currentLen = (int32_t)updatedLen;
                loc = loopEndLoc;
                break;
            }

        case URX_CTR_LOOP:
        case URX_CTR_LOOP_NG:
            // These opcodes will be skipped over by code for URX_CRT_INIT.
            // We shouldn't encounter them here.
            U_ASSERT(FALSE);
            break;

        case URX_LOOP_SR_I:
        case URX_LOOP_DOT_I:
        case URX_LOOP_C:
            // For anything to do with loops, make the match length unbounded.
            currentLen = INT32_MAX;
            break;



        case URX_LA_START:
        case URX_LA_END:
            // Look-ahead.  Just ignore, treat the look-ahead block as if
            // it were normal pattern.  Gives a too-long match length,
            //  but good enough for now.
            break;

            // End of look-ahead ops should always be consumed by the processing at
            //  the URX_LA_START op.
            // U_ASSERT(FALSE);
            // break;

        case URX_LB_START:
            {
                // Look-behind.  Scan forward until the matching look-around end,
                //   without processing the look-behind block.
                int32_t  depth = 0;
                for (;;) {
                    loc++;
                    op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
                    if (URX_TYPE(op) == URX_LA_START || URX_TYPE(op) == URX_LB_START) {
                        depth++;
                    }
                    if (URX_TYPE(op) == URX_LA_END || URX_TYPE(op)==URX_LBN_END) {
                        if (depth == 0) {
                            break;
                        }
                        depth--;
                    }
                    U_ASSERT(loc < end);
                }
            }
            break;

        default:
            U_ASSERT(FALSE);
        }


        if (currentLen == INT32_MAX) {
            //  The maximum length is unbounded.
            //  Stop further processing of the pattern.
            break;
        }

    }
    return currentLen;

}


//------------------------------------------------------------------------------
//
//   stripNOPs    Remove any NOP operations from the compiled pattern code.
//                Extra NOPs are inserted for some constructs during the initial
//                code generation to provide locations that may be patched later.
//                Many end up unneeded, and are removed by this function.
//
//                In order to minimize the number of passes through the pattern,
//                back-reference fixup is also performed here (adjusting
//                back-reference operands to point to the correct frame offsets).
//
//------------------------------------------------------------------------------
void RegexCompile::stripNOPs() {

    if (U_FAILURE(*fStatus)) {
        return;
    }

    int32_t    end = fRXPat->fCompiledPat->size();
    UVector32  deltas(end, *fStatus);

    // Make a first pass over the code, computing the amount that things
    //   will be offset at each location in the original code.
    int32_t   loc;
    int32_t   d = 0;
    for (loc=0; loc<end; loc++) {
        deltas.addElement(d, *fStatus);
        int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
        if (URX_TYPE(op) == URX_NOP) {
            d++;
        }
    }

    UnicodeString caseStringBuffer;

    // Make a second pass over the code, removing the NOPs by moving following
    //  code up, and patching operands that refer to code locations that
    //  are being moved.  The array of offsets from the first step is used
    //  to compute the new operand values.
    int32_t src;
    int32_t dst = 0;
    for (src=0; src<end; src++) {
        int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(src);
        int32_t opType = URX_TYPE(op);
        switch (opType) {
        case URX_NOP:
            break;

        case URX_STATE_SAVE:
        case URX_JMP:
        case URX_CTR_LOOP:
        case URX_CTR_LOOP_NG:
        case URX_RELOC_OPRND:
        case URX_JMPX:
        case URX_JMP_SAV:
        case URX_JMP_SAV_X:
            // These are instructions with operands that refer to code locations.
            {
                int32_t  operandAddress = URX_VAL(op);
                U_ASSERT(operandAddress>=0 && operandAddress<deltas.size());
                int32_t fixedOperandAddress = operandAddress - deltas.elementAti(operandAddress);
                op = buildOp(opType, fixedOperandAddress);
                fRXPat->fCompiledPat->setElementAt(op, dst);
                dst++;
                break;
            }

        case URX_BACKREF:
        case URX_BACKREF_I:
            {
                int32_t where = URX_VAL(op);
                if (where > fRXPat->fGroupMap->size()) {
                    error(U_REGEX_INVALID_BACK_REF);
                    break;
                }
                where = fRXPat->fGroupMap->elementAti(where-1);
                op    = buildOp(opType, where);
                fRXPat->fCompiledPat->setElementAt(op, dst);
                dst++;

                fRXPat->fNeedsAltInput = TRUE;
                break;
            }
        case URX_RESERVED_OP:
        case URX_RESERVED_OP_N:
        case URX_BACKTRACK:
        case URX_END:
        case URX_ONECHAR:
        case URX_STRING:
        case URX_STRING_LEN:
        case URX_START_CAPTURE:
        case URX_END_CAPTURE:
        case URX_STATIC_SETREF:
        case URX_STAT_SETREF_N:
        case URX_SETREF:
        case URX_DOTANY:
        case URX_FAIL:
        case URX_BACKSLASH_B:
        case URX_BACKSLASH_BU:
        case URX_BACKSLASH_G:
        case URX_BACKSLASH_X:
        case URX_BACKSLASH_Z:
        case URX_DOTANY_ALL:
        case URX_BACKSLASH_D:
        case URX_CARET:
        case URX_DOLLAR:
        case URX_CTR_INIT:
        case URX_CTR_INIT_NG:
        case URX_DOTANY_UNIX:
        case URX_STO_SP:
        case URX_LD_SP:
        case URX_STO_INP_LOC:
        case URX_LA_START:
        case URX_LA_END:
        case URX_ONECHAR_I:
        case URX_STRING_I:
        case URX_DOLLAR_M:
        case URX_CARET_M:
        case URX_CARET_M_UNIX:
        case URX_LB_START:
        case URX_LB_CONT:
        case URX_LB_END:
        case URX_LBN_CONT:
        case URX_LBN_END:
        case URX_LOOP_SR_I:
        case URX_LOOP_DOT_I:
        case URX_LOOP_C:
        case URX_DOLLAR_D:
        case URX_DOLLAR_MD:
        case URX_BACKSLASH_H:
        case URX_BACKSLASH_R:
        case URX_BACKSLASH_V:
            // These instructions are unaltered by the relocation.
            fRXPat->fCompiledPat->setElementAt(op, dst);
            dst++;
            break;

        default:
            // Some op is unaccounted for.
            U_ASSERT(FALSE);
            error(U_REGEX_INTERNAL_ERROR);
        }
    }

    fRXPat->fCompiledPat->setSize(dst);
}




//------------------------------------------------------------------------------
//
//  Error         Report a rule parse error.
//                Only report it if no previous error has been recorded.
//
//------------------------------------------------------------------------------
void RegexCompile::error(UErrorCode e) {
    if (U_SUCCESS(*fStatus)) {
        *fStatus = e;
        // Hmm. fParseErr (UParseError) line & offset fields are int32_t in public
        // API (see common/unicode/parseerr.h), while fLineNum and fCharNum are
        // int64_t. If the values of the latter are out of range for the former,
        // set them to the appropriate "field not supported" values.
        if (fLineNum > 0x7FFFFFFF) {
            fParseErr->line   = 0;
            fParseErr->offset = -1;
        } else if (fCharNum > 0x7FFFFFFF) {
            fParseErr->line   = (int32_t)fLineNum;
            fParseErr->offset = -1;
        } else {
            fParseErr->line   = (int32_t)fLineNum;
            fParseErr->offset = (int32_t)fCharNum;
        }

        UErrorCode status = U_ZERO_ERROR; // throwaway status for extracting context

        // Fill in the context.
        //   Note: extractBetween() pins supplied indicies to the string bounds.
        uprv_memset(fParseErr->preContext,  0, sizeof(fParseErr->preContext));
        uprv_memset(fParseErr->postContext, 0, sizeof(fParseErr->postContext));
        utext_extract(fRXPat->fPattern, fScanIndex-U_PARSE_CONTEXT_LEN+1, fScanIndex, fParseErr->preContext, U_PARSE_CONTEXT_LEN, &status);
        utext_extract(fRXPat->fPattern, fScanIndex, fScanIndex+U_PARSE_CONTEXT_LEN-1, fParseErr->postContext, U_PARSE_CONTEXT_LEN, &status);
    }
}


//
//  Assorted Unicode character constants.
//     Numeric because there is no portable way to enter them as literals.
//     (Think EBCDIC).
//
static const UChar      chCR        = 0x0d;      // New lines, for terminating comments.
static const UChar      chLF        = 0x0a;      // Line Feed
static const UChar      chPound     = 0x23;      // '#', introduces a comment.
static const UChar      chDigit0    = 0x30;      // '0'
static const UChar      chDigit7    = 0x37;      // '9'
static const UChar      chColon     = 0x3A;      // ':'
static const UChar      chE         = 0x45;      // 'E'
static const UChar      chQ         = 0x51;      // 'Q'
//static const UChar      chN         = 0x4E;      // 'N'
static const UChar      chP         = 0x50;      // 'P'
static const UChar      chBackSlash = 0x5c;      // '\'  introduces a char escape
//static const UChar      chLBracket  = 0x5b;      // '['
static const UChar      chRBracket  = 0x5d;      // ']'
static const UChar      chUp        = 0x5e;      // '^'
static const UChar      chLowerP    = 0x70;
static const UChar      chLBrace    = 0x7b;      // '{'
static const UChar      chRBrace    = 0x7d;      // '}'
static const UChar      chNEL       = 0x85;      //    NEL newline variant
static const UChar      chLS        = 0x2028;    //    Unicode Line Separator


//------------------------------------------------------------------------------
//
//  nextCharLL    Low Level Next Char from the regex pattern.
//                Get a char from the string, keep track of input position
//                     for error reporting.
//
//------------------------------------------------------------------------------
UChar32  RegexCompile::nextCharLL() {
    UChar32       ch;

    if (fPeekChar != -1) {
        ch = fPeekChar;
        fPeekChar = -1;
        return ch;
    }

    // assume we're already in the right place
    ch = UTEXT_NEXT32(fRXPat->fPattern);
    if (ch == U_SENTINEL) {
        return ch;
    }

    if (ch == chCR ||
        ch == chNEL ||
        ch == chLS   ||
        (ch == chLF && fLastChar != chCR)) {
        // Character is starting a new line.  Bump up the line number, and
        //  reset the column to 0.
        fLineNum++;
        fCharNum=0;
    }
    else {
        // Character is not starting a new line.  Except in the case of a
        //   LF following a CR, increment the column position.
        if (ch != chLF) {
            fCharNum++;
        }
    }
    fLastChar = ch;
    return ch;
}

//------------------------------------------------------------------------------
//
//   peekCharLL    Low Level Character Scanning, sneak a peek at the next
//                 character without actually getting it.
//
//------------------------------------------------------------------------------
UChar32  RegexCompile::peekCharLL() {
    if (fPeekChar == -1) {
        fPeekChar = nextCharLL();
    }
    return fPeekChar;
}


//------------------------------------------------------------------------------
//
//   nextChar     for pattern scanning.  At this level, we handle stripping
//                out comments and processing some backslash character escapes.
//                The rest of the pattern grammar is handled at the next level up.
//
//------------------------------------------------------------------------------
void RegexCompile::nextChar(RegexPatternChar &c) {

    fScanIndex = UTEXT_GETNATIVEINDEX(fRXPat->fPattern);
    c.fChar    = nextCharLL();
    c.fQuoted  = FALSE;

    if (fQuoteMode) {
        c.fQuoted = TRUE;
        if ((c.fChar==chBackSlash && peekCharLL()==chE && ((fModeFlags & UREGEX_LITERAL) == 0)) ||
            c.fChar == (UChar32)-1) {
            fQuoteMode = FALSE;  //  Exit quote mode,
            nextCharLL();        // discard the E
            nextChar(c);         // recurse to get the real next char
        }
    }
    else if (fInBackslashQuote) {
        // The current character immediately follows a '\'
        // Don't check for any further escapes, just return it as-is.
        // Don't set c.fQuoted, because that would prevent the state machine from
        //    dispatching on the character.
        fInBackslashQuote = FALSE;
    }
    else
    {
        // We are not in a \Q quoted region \E of the source.
        //
        if (fModeFlags & UREGEX_COMMENTS) {
            //
            // We are in free-spacing and comments mode.
            //  Scan through any white space and comments, until we
            //  reach a significant character or the end of inut.
            for (;;) {
                if (c.fChar == (UChar32)-1) {
                    break;     // End of Input
                }
                if  (c.fChar == chPound && fEOLComments == TRUE) {
                    // Start of a comment.  Consume the rest of it, until EOF or a new line
                    for (;;) {
                        c.fChar = nextCharLL();
                        if (c.fChar == (UChar32)-1 ||  // EOF
                            c.fChar == chCR        ||
                            c.fChar == chLF        ||
                            c.fChar == chNEL       ||
                            c.fChar == chLS)       {
                            break;
                        }
                    }
                }
                // TODO:  check what Java & Perl do with non-ASCII white spaces.  Ticket 6061.
                if (PatternProps::isWhiteSpace(c.fChar) == FALSE) {
                    break;
                }
                c.fChar = nextCharLL();
            }
        }

        //
        //  check for backslash escaped characters.
        //
        if (c.fChar == chBackSlash) {
            int64_t pos = UTEXT_GETNATIVEINDEX(fRXPat->fPattern);
            if (RegexStaticSets::gStaticSets->fUnescapeCharSet.contains(peekCharLL())) {
                //
                // A '\' sequence that is handled by ICU's standard unescapeAt function.
                //   Includes \uxxxx, \n, \r, many others.
                //   Return the single equivalent character.
                //
                nextCharLL();                 // get & discard the peeked char.
                c.fQuoted = TRUE;

                if (UTEXT_FULL_TEXT_IN_CHUNK(fRXPat->fPattern, fPatternLength)) {
                    int32_t endIndex = (int32_t)pos;
                    c.fChar = u_unescapeAt(uregex_ucstr_unescape_charAt, &endIndex, (int32_t)fPatternLength, (void *)fRXPat->fPattern->chunkContents);

                    if (endIndex == pos) {
                        error(U_REGEX_BAD_ESCAPE_SEQUENCE);
                    }
                    fCharNum += endIndex - pos;
                    UTEXT_SETNATIVEINDEX(fRXPat->fPattern, endIndex);
                } else {
                    int32_t offset = 0;
                    struct URegexUTextUnescapeCharContext context = U_REGEX_UTEXT_UNESCAPE_CONTEXT(fRXPat->fPattern);

                    UTEXT_SETNATIVEINDEX(fRXPat->fPattern, pos);
                    c.fChar = u_unescapeAt(uregex_utext_unescape_charAt, &offset, INT32_MAX, &context);

                    if (offset == 0) {
                        error(U_REGEX_BAD_ESCAPE_SEQUENCE);
                    } else if (context.lastOffset == offset) {
                        UTEXT_PREVIOUS32(fRXPat->fPattern);
                    } else if (context.lastOffset != offset-1) {
                        utext_moveIndex32(fRXPat->fPattern, offset - context.lastOffset - 1);
                    }
                    fCharNum += offset;
                }
            }
            else if (peekCharLL() == chDigit0) {
                //  Octal Escape, using Java Regexp Conventions
                //    which are \0 followed by 1-3 octal digits.
                //    Different from ICU Unescape handling of Octal, which does not
                //    require the leading 0.
                //  Java also has the convention of only consuming 2 octal digits if
                //    the three digit number would be > 0xff
                //
                c.fChar = 0;
                nextCharLL();    // Consume the initial 0.
                int index;
                for (index=0; index<3; index++) {
                    int32_t ch = peekCharLL();
                    if (ch<chDigit0 || ch>chDigit7) {
                        if (index==0) {
                           // \0 is not followed by any octal digits.
                           error(U_REGEX_BAD_ESCAPE_SEQUENCE);
                        }
                        break;
                    }
                    c.fChar <<= 3;
                    c.fChar += ch&7;
                    if (c.fChar <= 255) {
                        nextCharLL();
                    } else {
                        // The last digit made the number too big.  Forget we saw it.
                        c.fChar >>= 3;
                    }
                }
                c.fQuoted = TRUE;
            }
            else if (peekCharLL() == chQ) {
                //  "\Q"  enter quote mode, which will continue until "\E"
                fQuoteMode = TRUE;
                nextCharLL();       // discard the 'Q'.
                nextChar(c);        // recurse to get the real next char.
            }
            else
            {
                // We are in a '\' escape that will be handled by the state table scanner.
                // Just return the backslash, but remember that the following char is to
                //  be taken literally.
                fInBackslashQuote = TRUE;
            }
        }
    }

    // re-enable # to end-of-line comments, in case they were disabled.
    // They are disabled by the parser upon seeing '(?', but this lasts for
    //  the fetching of the next character only.
    fEOLComments = TRUE;

    // putc(c.fChar, stdout);
}



//------------------------------------------------------------------------------
//
//  scanNamedChar
//            Get a UChar32 from a \N{UNICODE CHARACTER NAME} in the pattern.
//
//             The scan position will be at the 'N'.  On return
//             the scan position should be just after the '}'
//
//             Return the UChar32
//
//------------------------------------------------------------------------------
UChar32  RegexCompile::scanNamedChar() {
    if (U_FAILURE(*fStatus)) {
        return 0;
    }

    nextChar(fC);
    if (fC.fChar != chLBrace) {
        error(U_REGEX_PROPERTY_SYNTAX);
        return 0;
    }

    UnicodeString  charName;
    for (;;) {
        nextChar(fC);
        if (fC.fChar == chRBrace) {
            break;
        }
        if (fC.fChar == -1) {
            error(U_REGEX_PROPERTY_SYNTAX);
            return 0;
        }
        charName.append(fC.fChar);
    }

    char name[100];
    if (!uprv_isInvariantUString(charName.getBuffer(), charName.length()) ||
         (uint32_t)charName.length()>=sizeof(name)) {
        // All Unicode character names have only invariant characters.
        // The API to get a character, given a name, accepts only char *, forcing us to convert,
        //   which requires this error check
        error(U_REGEX_PROPERTY_SYNTAX);
        return 0;
    }
    charName.extract(0, charName.length(), name, sizeof(name), US_INV);

    UChar32  theChar = u_charFromName(U_UNICODE_CHAR_NAME, name, fStatus);
    if (U_FAILURE(*fStatus)) {
        error(U_REGEX_PROPERTY_SYNTAX);
    }

    nextChar(fC);      // Continue overall regex pattern processing with char after the '}'
    return theChar;
}

//------------------------------------------------------------------------------
//
//  scanProp   Construct a UnicodeSet from the text at the current scan
//             position, which will be of the form \p{whaterver}
//
//             The scan position will be at the 'p' or 'P'.  On return
//             the scan position should be just after the '}'
//
//             Return a UnicodeSet, constructed from the \P pattern,
//             or NULL if the pattern is invalid.
//
//------------------------------------------------------------------------------
UnicodeSet *RegexCompile::scanProp() {
    UnicodeSet    *uset = NULL;

    if (U_FAILURE(*fStatus)) {
        return NULL;
    }
    (void)chLowerP;   // Suppress compiler unused variable warning.
    U_ASSERT(fC.fChar == chLowerP || fC.fChar == chP);
    UBool negated = (fC.fChar == chP);

    UnicodeString propertyName;
    nextChar(fC);
    if (fC.fChar != chLBrace) {
        error(U_REGEX_PROPERTY_SYNTAX);
        return NULL;
    }
    for (;;) {
        nextChar(fC);
        if (fC.fChar == chRBrace) {
            break;
        }
        if (fC.fChar == -1) {
            // Hit the end of the input string without finding the closing '}'
            error(U_REGEX_PROPERTY_SYNTAX);
            return NULL;
        }
        propertyName.append(fC.fChar);
    }
    uset = createSetForProperty(propertyName, negated);
    nextChar(fC);    // Move input scan to position following the closing '}'
    return uset;
}

//------------------------------------------------------------------------------
//
//  scanPosixProp   Construct a UnicodeSet from the text at the current scan
//             position, which is expected be of the form [:property expression:]
//
//             The scan position will be at the opening ':'.  On return
//             the scan position must be on the closing ']'
//
//             Return a UnicodeSet constructed from the pattern,
//             or NULL if this is not a valid POSIX-style set expression.
//             If not a property expression, restore the initial scan position
//                (to the opening ':')
//
//               Note:  the opening '[:' is not sufficient to guarantee that
//                      this is a [:property:] expression.
//                      [:'+=,] is a perfectly good ordinary set expression that
//                              happens to include ':' as one of its characters.
//
//------------------------------------------------------------------------------
UnicodeSet *RegexCompile::scanPosixProp() {
    UnicodeSet    *uset = NULL;

    if (U_FAILURE(*fStatus)) {
        return NULL;
    }

    U_ASSERT(fC.fChar == chColon);

    // Save the scanner state.
    // TODO:  move this into the scanner, with the state encapsulated in some way.  Ticket 6062
    int64_t     savedScanIndex        = fScanIndex;
    int64_t     savedNextIndex        = UTEXT_GETNATIVEINDEX(fRXPat->fPattern);
    UBool       savedQuoteMode        = fQuoteMode;
    UBool       savedInBackslashQuote = fInBackslashQuote;
    UBool       savedEOLComments      = fEOLComments;
    int64_t     savedLineNum          = fLineNum;
    int64_t     savedCharNum          = fCharNum;
    UChar32     savedLastChar         = fLastChar;
    UChar32     savedPeekChar         = fPeekChar;
    RegexPatternChar savedfC          = fC;

    // Scan for a closing ].   A little tricky because there are some perverse
    //   edge cases possible.  "[:abc\Qdef:] \E]"  is a valid non-property expression,
    //   ending on the second closing ].

    UnicodeString propName;
    UBool         negated  = FALSE;

    // Check for and consume the '^' in a negated POSIX property, e.g.  [:^Letter:]
    nextChar(fC);
    if (fC.fChar == chUp) {
       negated = TRUE;
       nextChar(fC);
    }

    // Scan for the closing ":]", collecting the property name along the way.
    UBool  sawPropSetTerminator = FALSE;
    for (;;) {
        propName.append(fC.fChar);
        nextChar(fC);
        if (fC.fQuoted || fC.fChar == -1) {
            // Escaped characters or end of input - either says this isn't a [:Property:]
            break;
        }
        if (fC.fChar == chColon) {
            nextChar(fC);
            if (fC.fChar == chRBracket) {
                sawPropSetTerminator = TRUE;
            }
            break;
        }
    }

    if (sawPropSetTerminator) {
        uset = createSetForProperty(propName, negated);
    }
    else
    {
        // No closing ":]".
        //  Restore the original scan position.
        //  The main scanner will retry the input as a normal set expression,
        //    not a [:Property:] expression.
        fScanIndex        = savedScanIndex;
        fQuoteMode        = savedQuoteMode;
        fInBackslashQuote = savedInBackslashQuote;
        fEOLComments      = savedEOLComments;
        fLineNum          = savedLineNum;
        fCharNum          = savedCharNum;
        fLastChar         = savedLastChar;
        fPeekChar         = savedPeekChar;
        fC                = savedfC;
        UTEXT_SETNATIVEINDEX(fRXPat->fPattern, savedNextIndex);
    }
    return uset;
}

static inline void addIdentifierIgnorable(UnicodeSet *set, UErrorCode& ec) {
    set->add(0, 8).add(0x0e, 0x1b).add(0x7f, 0x9f);
    addCategory(set, U_GC_CF_MASK, ec);
}

//
//  Create a Unicode Set from a Unicode Property expression.
//     This is common code underlying both \p{...} ane [:...:] expressions.
//     Includes trying the Java "properties" that aren't supported as
//     normal ICU UnicodeSet properties
//
static const UChar posSetPrefix[] = {0x5b, 0x5c, 0x70, 0x7b, 0}; // "[\p{"
static const UChar negSetPrefix[] = {0x5b, 0x5c, 0x50, 0x7b, 0}; // "[\P{"
UnicodeSet *RegexCompile::createSetForProperty(const UnicodeString &propName, UBool negated) {
    UnicodeString   setExpr;
    UnicodeSet      *set;
    uint32_t        usetFlags = 0;

    if (U_FAILURE(*fStatus)) {
        return NULL;
    }

    //
    //  First try the property as we received it
    //
    if (negated) {
        setExpr.append(negSetPrefix, -1);
    } else {
        setExpr.append(posSetPrefix, -1);
    }
    setExpr.append(propName);
    setExpr.append(chRBrace);
    setExpr.append(chRBracket);
    if (fModeFlags & UREGEX_CASE_INSENSITIVE) {
        usetFlags |= USET_CASE_INSENSITIVE;
    }
    set = new UnicodeSet(setExpr, usetFlags, NULL, *fStatus);
    if (U_SUCCESS(*fStatus)) {
       return set;
    }
    delete set;
    set = NULL;

    //
    //  The property as it was didn't work.

    //  Do [:word:]. It is not recognized as a property by UnicodeSet.  "word" not standard POSIX
    //     or standard Java, but many other regular expression packages do recognize it.

    if (propName.caseCompare(UNICODE_STRING_SIMPLE("word"), 0) == 0) {
        *fStatus = U_ZERO_ERROR;
        set = new UnicodeSet(*(fRXPat->fStaticSets[URX_ISWORD_SET]));
        if (set == NULL) {
            *fStatus = U_MEMORY_ALLOCATION_ERROR;
            return set;
        }
        if (negated) {
            set->complement();
        }
        return set;
    }


    //    Do Java fixes -
    //       InGreek -> InGreek or Coptic, that being the official Unicode name for that block.
    //       InCombiningMarksforSymbols -> InCombiningDiacriticalMarksforSymbols.
    //
    //       Note on Spaces:  either "InCombiningMarksForSymbols" or "InCombining Marks for Symbols"
    //                        is accepted by Java.  The property part of the name is compared
    //                        case-insenstively.  The spaces must be exactly as shown, either
    //                        all there, or all omitted, with exactly one at each position
    //                        if they are present.  From checking against JDK 1.6
    //
    //       This code should be removed when ICU properties support the Java  compatibility names
    //          (ICU 4.0?)
    //
    UnicodeString mPropName = propName;
    if (mPropName.caseCompare(UNICODE_STRING_SIMPLE("InGreek"), 0) == 0) {
        mPropName = UNICODE_STRING_SIMPLE("InGreek and Coptic");
    }
    if (mPropName.caseCompare(UNICODE_STRING_SIMPLE("InCombining Marks for Symbols"), 0) == 0 ||
        mPropName.caseCompare(UNICODE_STRING_SIMPLE("InCombiningMarksforSymbols"), 0) == 0) {
        mPropName = UNICODE_STRING_SIMPLE("InCombining Diacritical Marks for Symbols");
    }
    else if (mPropName.compare(UNICODE_STRING_SIMPLE("all")) == 0) {
        mPropName = UNICODE_STRING_SIMPLE("javaValidCodePoint");
    }

    //    See if the property looks like a Java "InBlockName", which
    //    we will recast as "Block=BlockName"
    //
    static const UChar IN[] = {0x49, 0x6E, 0};  // "In"
    static const UChar BLOCK[] = {0x42, 0x6C, 0x6f, 0x63, 0x6b, 0x3d, 00};  // "Block="
    if (mPropName.startsWith(IN, 2) && propName.length()>=3) {
        setExpr.truncate(4);   // Leaves "[\p{", or "[\P{"
        setExpr.append(BLOCK, -1);
        setExpr.append(UnicodeString(mPropName, 2));  // Property with the leading "In" removed.
        setExpr.append(chRBrace);
        setExpr.append(chRBracket);
        *fStatus = U_ZERO_ERROR;
        set = new UnicodeSet(setExpr, usetFlags, NULL, *fStatus);
        if (U_SUCCESS(*fStatus)) {
            return set;
        }
        delete set;
        set = NULL;
    }

    if (propName.startsWith(UNICODE_STRING_SIMPLE("java")) ||
        propName.compare(UNICODE_STRING_SIMPLE("all")) == 0)
    {
        UErrorCode localStatus = U_ZERO_ERROR;
        //setExpr.remove();
        set = new UnicodeSet();
        //
        //  Try the various Java specific properties.
        //   These all begin with "java"
        //
        if (mPropName.compare(UNICODE_STRING_SIMPLE("javaDefined")) == 0) {
            addCategory(set, U_GC_CN_MASK, localStatus);
            set->complement();
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaDigit")) == 0) {
            addCategory(set, U_GC_ND_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaIdentifierIgnorable")) == 0) {
            addIdentifierIgnorable(set, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaISOControl")) == 0) {
            set->add(0, 0x1F).add(0x7F, 0x9F);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaJavaIdentifierPart")) == 0) {
            addCategory(set, U_GC_L_MASK, localStatus);
            addCategory(set, U_GC_SC_MASK, localStatus);
            addCategory(set, U_GC_PC_MASK, localStatus);
            addCategory(set, U_GC_ND_MASK, localStatus);
            addCategory(set, U_GC_NL_MASK, localStatus);
            addCategory(set, U_GC_MC_MASK, localStatus);
            addCategory(set, U_GC_MN_MASK, localStatus);
            addIdentifierIgnorable(set, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaJavaIdentifierStart")) == 0) {
            addCategory(set, U_GC_L_MASK, localStatus);
            addCategory(set, U_GC_NL_MASK, localStatus);
            addCategory(set, U_GC_SC_MASK, localStatus);
            addCategory(set, U_GC_PC_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLetter")) == 0) {
            addCategory(set, U_GC_L_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLetterOrDigit")) == 0) {
            addCategory(set, U_GC_L_MASK, localStatus);
            addCategory(set, U_GC_ND_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLowerCase")) == 0) {
            addCategory(set, U_GC_LL_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaMirrored")) == 0) {
            set->applyIntPropertyValue(UCHAR_BIDI_MIRRORED, 1, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaSpaceChar")) == 0) {
            addCategory(set, U_GC_Z_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaSupplementaryCodePoint")) == 0) {
            set->add(0x10000, UnicodeSet::MAX_VALUE);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaTitleCase")) == 0) {
            addCategory(set, U_GC_LT_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUnicodeIdentifierStart")) == 0) {
            addCategory(set, U_GC_L_MASK, localStatus);
            addCategory(set, U_GC_NL_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUnicodeIdentifierPart")) == 0) {
            addCategory(set, U_GC_L_MASK, localStatus);
            addCategory(set, U_GC_PC_MASK, localStatus);
            addCategory(set, U_GC_ND_MASK, localStatus);
            addCategory(set, U_GC_NL_MASK, localStatus);
            addCategory(set, U_GC_MC_MASK, localStatus);
            addCategory(set, U_GC_MN_MASK, localStatus);
            addIdentifierIgnorable(set, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUpperCase")) == 0) {
            addCategory(set, U_GC_LU_MASK, localStatus);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaValidCodePoint")) == 0) {
            set->add(0, UnicodeSet::MAX_VALUE);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaWhitespace")) == 0) {
            addCategory(set, U_GC_Z_MASK, localStatus);
            set->removeAll(UnicodeSet().add(0xa0).add(0x2007).add(0x202f));
            set->add(9, 0x0d).add(0x1c, 0x1f);
        }
        else if (mPropName.compare(UNICODE_STRING_SIMPLE("all")) == 0) {
            set->add(0, UnicodeSet::MAX_VALUE);
        }

        if (U_SUCCESS(localStatus) && !set->isEmpty()) {
            *fStatus = U_ZERO_ERROR;
            if (usetFlags & USET_CASE_INSENSITIVE) {
                set->closeOver(USET_CASE_INSENSITIVE);
            }
            if (negated) {
                set->complement();
            }
            return set;
        }
        delete set;
        set = NULL;
    }
    error(*fStatus);
    return NULL;
}



//
//  SetEval   Part of the evaluation of [set expressions].
//            Perform any pending (stacked) operations with precedence
//            equal or greater to that of the next operator encountered
//            in the expression.
//
void RegexCompile::setEval(int32_t nextOp) {
    UnicodeSet *rightOperand = NULL;
    UnicodeSet *leftOperand  = NULL;
    for (;;) {
        U_ASSERT(fSetOpStack.empty()==FALSE);
        int32_t pendingSetOperation = fSetOpStack.peeki();
        if ((pendingSetOperation&0xffff0000) < (nextOp&0xffff0000)) {
            break;
        }
        fSetOpStack.popi();
        U_ASSERT(fSetStack.empty() == FALSE);
        rightOperand = (UnicodeSet *)fSetStack.peek();
        switch (pendingSetOperation) {
            case setNegation:
                rightOperand->complement();
                break;
            case setCaseClose:
                // TODO: need a simple close function.  Ticket 6065
                rightOperand->closeOver(USET_CASE_INSENSITIVE);
                rightOperand->removeAllStrings();
                break;
            case setDifference1:
            case setDifference2:
                fSetStack.pop();
                leftOperand = (UnicodeSet *)fSetStack.peek();
                leftOperand->removeAll(*rightOperand);
                delete rightOperand;
                break;
            case setIntersection1:
            case setIntersection2:
                fSetStack.pop();
                leftOperand = (UnicodeSet *)fSetStack.peek();
                leftOperand->retainAll(*rightOperand);
                delete rightOperand;
                break;
            case setUnion:
                fSetStack.pop();
                leftOperand = (UnicodeSet *)fSetStack.peek();
                leftOperand->addAll(*rightOperand);
                delete rightOperand;
                break;
            default:
                U_ASSERT(FALSE);
                break;
            }
        }
    }

void RegexCompile::setPushOp(int32_t op) {
    setEval(op);
    fSetOpStack.push(op, *fStatus);
    fSetStack.push(new UnicodeSet(), *fStatus);
}

U_NAMESPACE_END
#endif  // !UCONFIG_NO_REGULAR_EXPRESSIONS