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
|
/*
* range_source.h
*/
#ifndef RANGE_SOURCE_H_
#define RANGE_SOURCE_H_
#include <stdint.h>
#include <queue>
#include <vector>
#include "ds.h"
#include "ebwt.h"
#include "range.h"
#include "pool.h"
#include "edit.h"
enum AdvanceUntil {
ADV_FOUND_RANGE = 1,
ADV_COST_CHANGES,
ADV_STEP
};
/**
* List of Edits that automatically expands as edits are added.
*/
struct EditList {
EditList() : sz_(0), moreEdits_(NULL), yetMoreEdits_(NULL) { }
/**
* Add an edit to the edit list.
*/
bool add(const Edit& e, AllocOnlyPool<Edit>& pool, size_t qlen) {
assert_lt(sz_, qlen + 10);
if(sz_ < numEdits) {
assert(moreEdits_ == NULL);
assert(yetMoreEdits_ == NULL);
edits_[sz_++] = e;
} else if(sz_ == numEdits) {
assert(moreEdits_ == NULL);
assert(yetMoreEdits_ == NULL);
moreEdits_ = pool.alloc(numMoreEdits);
if(moreEdits_ == NULL) {
return false;
}
assert(moreEdits_ != NULL);
moreEdits_[0] = e;
sz_++;
} else if(sz_ < (numEdits + numMoreEdits)) {
assert(moreEdits_ != NULL);
assert(yetMoreEdits_ == NULL);
moreEdits_[sz_ - numEdits] = e;
sz_++;
} else if(sz_ == (numEdits + numMoreEdits)) {
assert(moreEdits_ != NULL);
assert(yetMoreEdits_ == NULL);
yetMoreEdits_ = pool.alloc((uint32_t)qlen + 10 - numMoreEdits - numEdits);
if(yetMoreEdits_ == NULL) {
return false;
}
assert(yetMoreEdits_ != NULL);
yetMoreEdits_[0] = e;
sz_++;
} else {
assert(moreEdits_ != NULL);
assert(yetMoreEdits_ != NULL);
yetMoreEdits_[sz_ - numEdits - numMoreEdits] = e;
sz_++;
}
return true;
}
/**
* Return a const reference to the ith Edit in the list.
*/
const Edit& get(size_t i) const {
assert_lt(i, sz_);
if(i < numEdits) {
return edits_[i];
} else if(i < (numEdits + numMoreEdits)) {
assert(moreEdits_ != NULL);
return moreEdits_[i-numEdits];
} else {
assert(moreEdits_ != NULL);
assert(yetMoreEdits_ != NULL);
return yetMoreEdits_[i-numEdits-numMoreEdits];
}
}
/**
* Get most recently added Edit.
*/
const Edit& top() const {
assert_gt(size(), 0);
return get(size()-1);
}
/**
* Return true iff no Edits have been added.
*/
bool empty() const { return size() == 0; }
/**
* Set a particular element of the EditList.
*/
void set(size_t i, const Edit& e) {
assert_lt(i, sz_);
if(i < numEdits) {
edits_[i] = e;
} else if(i < (numEdits + numMoreEdits)) {
assert(moreEdits_ != NULL);
moreEdits_[i-numEdits] = e;
} else {
assert(moreEdits_ != NULL);
assert(yetMoreEdits_ != NULL);
yetMoreEdits_[i-numEdits-numMoreEdits] = e;
}
}
/**
* Remove all Edits from the list.
*/
void clear() {
sz_ = 0;
moreEdits_ = NULL;
yetMoreEdits_ = NULL;
}
/**
* Return number of Edits in the List.
*/
size_t size() const { return sz_; }
/**
* Free all the heap-allocated edit lists
*/
void free(AllocOnlyPool<Edit>& epool, size_t qlen) {
if(yetMoreEdits_ != NULL)
epool.free(yetMoreEdits_, (uint32_t)qlen + 10 - numMoreEdits - numEdits);
if(moreEdits_ != NULL)
epool.free(moreEdits_, numMoreEdits);
}
const static size_t numEdits = 6; // part of object allocation
const static size_t numMoreEdits = 16; // first extra allocation
size_t sz_; // number of Edits stored in the EditList
Edit edits_[numEdits]; // first 4 edits; typically, no more are needed
Edit *moreEdits_; // if used, size is dictated by numMoreEdits
Edit *yetMoreEdits_; // if used, size is dictated by length of read
};
/**
* Holds per-position information about what outgoing paths have been
* eliminated and what the quality value(s) is (are) at the position.
*/
union ElimsAndQual {
/**
* Assuming qual A/C/G/T are already set, set quallo and quallo2
* to the additional cost incurred by the least and second-least
* costly paths.
*/
void updateLo() {
flags.quallo = 127;
flags.quallo2 = 127;
if(!flags.mmA) {
// A mismatch to an A in the genome has not been ruled out
if(flags.qualA < flags.quallo) {
//flags.quallo2 = flags.quallo;
flags.quallo = flags.qualA;
}
//else if(flags.qualA == flags.quallo) {
// flags.quallo2 = flags.quallo;
//} else if(flags.qualA < flags.quallo2) {
// flags.quallo2 = flags.qualA;
//}
}
if(!flags.mmC) {
// A mismatch to a C in the genome has not been ruled out
if(flags.qualC < flags.quallo) {
flags.quallo2 = flags.quallo;
flags.quallo = flags.qualC;
} else if(flags.qualC == flags.quallo) {
flags.quallo2 = flags.quallo;
} else if(flags.qualC < flags.quallo2) {
flags.quallo2 = flags.qualC;
}
}
if(!flags.mmG) {
// A mismatch to a G in the genome has not been ruled out
if(flags.qualG < flags.quallo) {
flags.quallo2 = flags.quallo;
flags.quallo = flags.qualG;
} else if(flags.qualG == flags.quallo) {
flags.quallo2 = flags.quallo;
} else if(flags.qualG < flags.quallo2) {
flags.quallo2 = flags.qualG;
}
}
if(!flags.mmT) {
// A mismatch to a T in the genome has not been ruled out
if(flags.qualT < flags.quallo) {
flags.quallo2 = flags.quallo;
flags.quallo = flags.qualT;
} else if(flags.qualT == flags.quallo) {
flags.quallo2 = flags.quallo;
} else if(flags.qualT < flags.quallo2) {
flags.quallo2 = flags.qualT;
}
}
assert(repOk());
}
/**
* Set all 13 elimination bits of the flags field to 1, indicating
* that all outgoing paths are eliminated.
*/
inline void eliminate() {
join.elims = ((1 << 13) - 1);
}
/**
* Internal consistency check. Basically just checks that lo and
* lo2 are set correctly.
*/
bool repOk() const {
uint8_t lo = 127;
uint8_t lo2 = 127;
assert_lt(flags.qualA, 127);
assert_lt(flags.qualC, 127);
assert_lt(flags.qualG, 127);
assert_lt(flags.qualT, 127);
if(!flags.mmA) {
if(flags.qualA < lo) {
lo = flags.qualA;
}
//else if(flags.qualA == lo || flags.qualA < lo2) {
// lo2 = flags.qualA;
//}
}
if(!flags.mmC) {
if(flags.qualC < lo) {
lo2 = lo;
lo = flags.qualC;
} else if(flags.qualC == lo || flags.qualC < lo2) {
lo2 = flags.qualC;
}
}
if(!flags.mmG) {
if(flags.qualG < lo) {
lo2 = lo;
lo = flags.qualG;
} else if(flags.qualG == lo || flags.qualG < lo2) {
lo2 = flags.qualG;
}
}
if(!flags.mmT) {
if(flags.qualT < lo) {
lo2 = lo;
lo = flags.qualT;
} else if(flags.qualT == lo || flags.qualT < lo2) {
lo2 = flags.qualT;
}
}
assert_eq((int)lo, (int)flags.quallo);
assert_eq((int)lo2, (int)flags.quallo2);
return true;
}
struct {
uint64_t mmA : 1; // A in ref aligns to non-A char in read
uint64_t mmC : 1; // C in ref aligns to non-C char in read
uint64_t mmG : 1; // G in ref aligns to non-G char in read
uint64_t mmT : 1; // T in ref aligns to non-T char in read
uint64_t snpA : 1; // Same as mmA, but we think it's a SNP and not a miscall
uint64_t snpC : 1; // Same as mmC, but we think it's a SNP and not a miscall
uint64_t snpG : 1; // Same as mmG, but we think it's a SNP and not a miscall
uint64_t snpT : 1; // Same as mmT, but we think it's a SNP and not a miscall
uint64_t insA : 1; // A insertion in reference w/r/t read
uint64_t insC : 1; // C insertion in reference w/r/t read
uint64_t insG : 1; // G insertion in reference w/r/t read
uint64_t insT : 1; // T insertion in reference w/r/t read
uint64_t del : 1; // deletion of read character
uint64_t qualA : 7; // quality penalty for picking A at this position
uint64_t qualC : 7; // quality penalty for picking C at this position
uint64_t qualG : 7; // quality penalty for picking G at this position
uint64_t qualT : 7; // quality penalty for picking T at this position
uint64_t quallo : 7; // lowest quality penalty at this position
uint64_t quallo2 : 7; // 2nd-lowest quality penalty at this position
uint64_t reserved : 9;
} flags;
struct {
uint64_t elims : 13; // all of the edit-elim flags bundled together
uint64_t quals : 42; // quality of positions
uint64_t reserved : 9;
} join;
struct {
uint64_t mmElims : 4; // substitution flags bundled together
uint64_t snpElims : 4; // substitution flags bundled together
uint64_t insElims : 4; // inserts-in-reference flags bundled together
uint64_t delElims : 1; // deletion of read character
uint64_t quals : 42; // quality of positions
uint64_t reserved : 9;
} join2;
};
/**
* All per-position state, including the ranges calculated for each
* character, the quality value at the position, and a set of flags
* recording whether we've tried each way of proceeding from this
* position.
*/
struct RangeState {
/**
* Using randomness when picking from among multiple choices, pick
* an edit to make. TODO: Only knows how to pick mismatches for
* now.
*/
Edit pickEdit(int pos, RandomSource& rand,
TIndexOffU& top, TIndexOffU& bot, bool indels,
bool& last)
{
Edit ret;
ret.type = EDIT_TYPE_MM;
ret.pos = pos;
ret.chr = 0;
ret.qchr = 0;
ret.reserved = 0;
assert(!eliminated_);
assert(!eq.flags.mmA || !eq.flags.mmC || !eq.flags.mmG || !eq.flags.mmT);
int num = !eq.flags.mmA + !eq.flags.mmC + !eq.flags.mmG + !eq.flags.mmT;
assert_leq(num, 4);
assert_gt(num, 0);
if(num == 2) eq.flags.quallo2 = 127;
// Only need to pick randomly if there's a quality tie
if(num > 1) {
last = false; // not the last at this pos
// Sum up range sizes and do a random weighted pick
TIndexOffU tot = 0;
bool candA = !eq.flags.mmA; bool candC = !eq.flags.mmC;
bool candG = !eq.flags.mmG; bool candT = !eq.flags.mmT;
bool candInsA = false, candInsC = false;
bool candInsG = false, candInsT = false;
bool candDel = false;
if(indels) {
// Insertions and deletions can only be candidates
// if the user asked for indels
candInsA = !eq.flags.insA;
candInsC = !eq.flags.insC;
candInsG = !eq.flags.insG;
candInsT = !eq.flags.insT;
candDel = !eq.flags.del;
}
ASSERT_ONLY(int origNum = num);
if(candA) {
assert_gt(bots[0], tops[0]);
tot += (bots[0] - tops[0]);
assert_gt(num, 0);
ASSERT_ONLY(num--);
}
if(candC) {
assert_gt(bots[1], tops[1]);
tot += (bots[1] - tops[1]);
assert_gt(num, 0);
ASSERT_ONLY(num--);
}
if(candG) {
assert_gt(bots[2], tops[2]);
tot += (bots[2] - tops[2]);
assert_gt(num, 0);
ASSERT_ONLY(num--);
}
if(candT) {
assert_gt(bots[3], tops[3]);
tot += (bots[3] - tops[3]);
assert_gt(num, 0);
ASSERT_ONLY(num--);
}
if(indels) {
if(candInsA) {
assert_gt(bots[0], tops[0]);
tot += (bots[0] - tops[0]);
assert_gt(num, 0);
ASSERT_ONLY(num--);
}
if(candInsC) {
assert_gt(bots[1], tops[1]);
tot += (bots[1] - tops[1]);
assert_gt(num, 0);
ASSERT_ONLY(num--);
}
if(candInsG) {
assert_gt(bots[2], tops[2]);
tot += (bots[2] - tops[2]);
assert_gt(num, 0);
ASSERT_ONLY(num--);
}
if(candInsT) {
assert_gt(bots[3], tops[3]);
tot += (bots[3] - tops[3]);
assert_gt(num, 0);
ASSERT_ONLY(num--);
}
if(candDel) {
// Always a candidate?
// Always a candidate just within the window?
// We can eliminate some indels based on the content downstream, but not many
}
}
assert_geq(num, 0);
assert_lt(num, origNum);
// Throw a dart randomly that hits one of the possible
// substitutions, with likelihoods weighted by range size
uint32_t dart = rand.nextU32() % tot;
if(candA) {
if(dart < (bots[0] - tops[0])) {
// Eliminate A mismatch
top = tops[0];
bot = bots[0];
eq.flags.mmA = 1;
assert_lt(eq.join2.mmElims, 15);
ret.chr = 'A';
return ret;
}
dart -= (bots[0] - tops[0]);
}
if(candC) {
if(dart < (bots[1] - tops[1])) {
// Eliminate C mismatch
top = tops[1];
bot = bots[1];
eq.flags.mmC = 1;
assert_lt(eq.join2.mmElims, 15);
ret.chr = 'C';
return ret;
}
dart -= (bots[1] - tops[1]);
}
if(candG) {
if(dart < (bots[2] - tops[2])) {
// Eliminate G mismatch
top = tops[2];
bot = bots[2];
eq.flags.mmG = 1;
assert_lt(eq.join2.mmElims, 15);
ret.chr = 'G';
return ret;
}
dart -= (bots[2] - tops[2]);
}
if(candT) {
assert_lt(dart, (bots[3] - tops[3]));
// Eliminate T mismatch
top = tops[3];
bot = bots[3];
eq.flags.mmT = 1;
assert_lt(eq.join2.mmElims, 15);
ret.chr = 'T';
}
} else {
last = true; // last at this pos
// There's only one; pick it!
int chr = -1;
if(!eq.flags.mmA) {
chr = 0;
} else if(!eq.flags.mmC) {
chr = 1;
} else if(!eq.flags.mmG) {
chr = 2;
} else {
assert(!eq.flags.mmT);
chr = 3;
}
ret.chr = "ACGT"[chr];
top = tops[chr];
bot = bots[chr];
//assert_eq(15, eq.join2.mmElims);
// Mark entire position as eliminated
eliminated_ = true;
}
return ret;
}
/**
* Return true (without assertion) iff this RangeState is
* internally consistent.
*/
bool repOk() {
if(eliminated_) return true;
// Uneliminated chars must have non-empty ranges
if(!eq.flags.mmA || !eq.flags.insA) assert_gt(bots[0], tops[0]);
if(!eq.flags.mmC || !eq.flags.insC) assert_gt(bots[1], tops[1]);
if(!eq.flags.mmG || !eq.flags.insG) assert_gt(bots[2], tops[2]);
if(!eq.flags.mmT || !eq.flags.insT) assert_gt(bots[3], tops[3]);
return true;
}
// Outgoing ranges; if the position being described is not a
// legitimate jumping-off point for a branch, tops[] and bots[]
// will be filled with 0s and all possibilities in eq will be
// eliminated
TIndexOffU tops[4]; // A, C, G, T top offsets
TIndexOffU bots[4]; // A, C, G, T bot offsets
ElimsAndQual eq; // Which outgoing paths have been tried already
bool eliminated_; // Whether all outgoing paths have been eliminated
};
/**
* Encapsulates a "branch" of the search space; i.e. all of the
* information deduced by walking along a path with only matches, along
* with information about the decisions that lead to the root of that
* path.
*/
class Branch {
typedef std::pair<TIndexOffU, TIndexOffU> UPair;
public:
Branch() :
delayedCost_(0), curtailed_(false), exhausted_(false),
prepped_(false), delayedIncrease_(false) { }
/**
* Initialize a new branch object with an empty path.
*/
bool init(AllocOnlyPool<RangeState>& rsPool,
AllocOnlyPool<Edit>& epool,
uint32_t id,
uint32_t qlen,
uint16_t depth0,
uint16_t depth1,
uint16_t depth2,
uint16_t depth3,
uint16_t rdepth,
uint16_t len,
uint16_t cost,
uint16_t ham,
TIndexOffU itop,
TIndexOffU ibot,
const EbwtParams& ep,
const uint8_t* ebwt,
const EditList* edits = NULL)
{
id_ = id;
delayedCost_ = 0;
depth0_ = depth0;
depth1_ = depth1;
depth2_ = depth2;
depth3_ = depth3;
assert_gt(depth3_, 0);
rdepth_ = rdepth;
len_ = len;
cost_ = cost;
ham_ = ham;
top_ = itop;
bot_ = ibot;
if(ibot > itop+1) {
// Care about both top and bot
SideLocus::initFromTopBot(itop, ibot, ep, ebwt, ltop_, lbot_);
} else if(ibot > itop) {
// Only care about top
ltop_.initFromRow(itop, ep, ebwt);
lbot_.invalidate();
}
if(qlen - rdepth_ > 0) {
ranges_ = rsPool.allocC(qlen - rdepth_); // allocated from the RangeStatePool
if(ranges_ == NULL) {
return false; // RangeStatePool exhausted
}
rangesSz_ = qlen - rdepth_;
} else {
ranges_ = NULL;
rangesSz_ = 0;
}
#ifndef NDEBUG
for(size_t i = 0; i < (qlen - rdepth_); i++) {
for(int j = 0; j < 4; j++) {
assert_eq(0, ranges_[i].tops[j]);
assert_eq(0, ranges_[i].bots[j]);
}
}
#endif
curtailed_ = false;
exhausted_ = false;
prepped_ = true;
delayedIncrease_ = false;
edits_.clear();
if(edits != NULL) {
const size_t numEdits = edits->size();
for(size_t i = 0; i < numEdits; i++) {
edits_.add(edits->get(i), epool, qlen);
}
}
// If we're starting with a non-zero length, that means we're
// jumping over a bunch of unrevisitable positions.
for(size_t i = 0; i < len_; i++) {
ranges_[i].eliminated_ = true;
assert(eliminated((int)i));
}
assert(repOk(qlen));
return true;
}
/**
* Depth of the deepest tip of the branch.
*/
uint16_t tipDepth() const {
return rdepth_ + len_;
}
/**
* Return true iff all outgoing edges from position i have been
* eliminated.
*/
inline bool eliminated(int i) const {
assert(!exhausted_);
if(i <= len_ && i < rangesSz_) {
assert(ranges_ != NULL);
#ifndef NDEBUG
if(!ranges_[i].eliminated_) {
// Someone must be as-yet-uneliminated
assert(!ranges_[i].eq.flags.mmA ||
!ranges_[i].eq.flags.mmC ||
!ranges_[i].eq.flags.mmG ||
!ranges_[i].eq.flags.mmT);
assert_lt(ranges_[i].eq.flags.quallo, 127);
}
#endif
return ranges_[i].eliminated_;
}
return true;
}
/**
* Split off a new branch by selecting a good outgoing path and
* creating a new Branch object for it and inserting that branch
* into the priority queue. Mark that outgoing path from the
* parent branch as eliminated. If the second-best outgoing path
* cost more, add the difference to the cost of this branch (since
* that's the best we can do starting from here from now on).
*/
Branch* splitBranch(AllocOnlyPool<RangeState>& rpool,
AllocOnlyPool<Edit>& epool,
AllocOnlyPool<Branch>& bpool,
uint32_t pmSz,
RandomSource& rand, uint32_t qlen,
uint32_t qualLim, int seedLen,
bool qualOrder, const EbwtParams& ep,
const uint8_t* ebwt, bool ebwtFw,
bool verbose,
bool quiet)
{
assert(!exhausted_);
assert(ranges_ != NULL);
assert(curtailed_);
assert_gt(pmSz, 0);
Branch *newBranch = bpool.alloc();
if(newBranch == NULL) {
return NULL;
}
assert(newBranch != NULL);
uint32_t id = bpool.lastId();
int tiedPositions[3];
int numTiedPositions = 0;
// Lowest marginal cost incurred by any of the positions with
// non-eliminated outgoing edges
uint16_t bestCost = 0xffff;
// Next-lowest
uint16_t nextCost = 0xffff;
int numNotEliminated = 0;
int i = (int)depth0_;
i = max(0, i - rdepth_);
// Iterate over revisitable positions in the path
for(; i <= len_; i++) {
// If there are still valid options for leaving out of this
// position
if(!eliminated(i)) {
numNotEliminated++;
uint16_t stratum = (rdepth_ + i < seedLen) ? (1 << 14) : 0;
uint16_t cost = stratum;
cost |= (qualOrder ? ranges_[i].eq.flags.quallo : 0);
if(cost < bestCost) {
// Demote the old best to the next-best
nextCost = bestCost;
// Update the new best
bestCost = cost;
numTiedPositions = 1;
tiedPositions[0] = i;
} else if(cost == bestCost) {
// As good as the best so far
assert_gt(numTiedPositions, 0);
if(numTiedPositions < 3) {
tiedPositions[numTiedPositions++] = i;
} else {
tiedPositions[0] = tiedPositions[1];
tiedPositions[1] = tiedPositions[2];
tiedPositions[2] = i;
}
} else if(cost < nextCost) {
// 'cost' isn't beter than the best, but it is
// better than the next-best
nextCost = cost;
}
}
}
assert_gt(numNotEliminated, 0);
assert_gt(numTiedPositions, 0);
//if(nextCost != 0xffff) assert_gt(nextCost, bestCost);
int r = 0;
if(numTiedPositions > 1) {
r = rand.nextU32() % numTiedPositions;
}
int pos = tiedPositions[r];
bool last = false;
// Pick an edit from among the edits tied for lowest cost
// (using randomness to break ties). If the selected edit is
// the last remaining one at this position, 'last' is set to
// true.
TIndexOffU top = 0, bot = 0;
Edit e = ranges_[pos].pickEdit(pos + rdepth_, rand, top, bot, false, last);
assert_gt(bot, top);
// Create and initialize a new Branch
uint16_t newRdepth = rdepth_ + pos + 1;
assert_lt((bestCost >> 14), 4);
uint32_t hamadd = (bestCost & ~0xc000);
uint16_t depth = pos + rdepth_;
assert_geq(depth, depth0_);
uint16_t newDepth0 = depth0_;
uint16_t newDepth1 = depth1_;
uint16_t newDepth2 = depth2_;
uint16_t newDepth3 = depth3_;
if(depth < depth1_) newDepth0 = depth1_;
if(depth < depth2_) newDepth1 = depth2_;
if(depth < depth3_) newDepth2 = depth3_;
assert_eq((uint32_t)(cost_ & ~0xc000), (uint32_t)(ham_ + hamadd));
if(!newBranch->init(
rpool, epool, id, qlen,
newDepth0, newDepth1, newDepth2, newDepth3,
newRdepth, 0, cost_, ham_ + hamadd,
top, bot, ep, ebwt, &edits_))
{
return NULL;
}
// Add the new edit
newBranch->edits_.add(e, epool, qlen);
if(numNotEliminated == 1 && last) {
// This branch is totally exhausted; there are no more
// valid outgoing paths from any positions within it.
// Remove it from the PathManager and mark it as exhausted.
// The caller should delete it.
exhausted_ = true;
if(ranges_ != NULL) {
assert_gt(rangesSz_, 0);
if(rpool.free(ranges_, rangesSz_)) {
ranges_ = NULL;
rangesSz_ = 0;
}
}
}
else if(numTiedPositions == 1 && last) {
// We exhausted the last outgoing edge at the current best
// cost; update the best cost to be the next-best
assert_neq(0xffff, nextCost);
if(bestCost != nextCost) {
assert_gt(nextCost, bestCost);
delayedCost_ = (cost_ - bestCost + nextCost);
delayedIncrease_ = true;
}
}
return newBranch;
}
/**
* Free a branch and all its contents.
*/
void free(uint32_t qlen,
AllocOnlyPool<RangeState>& rpool,
AllocOnlyPool<Edit>& epool,
AllocOnlyPool<Branch>& bpool)
{
edits_.free(epool, qlen);
if(ranges_ != NULL) {
assert_gt(rangesSz_, 0);
rpool.free(ranges_, rangesSz_);
ranges_ = NULL;
rangesSz_ = 0;
}
bpool.free(this);
}
/**
* Pretty-print the state of this branch.
*/
void print(const BTDnaString& qry,
const BTString& quals,
uint16_t minCost,
std::ostream& out,
bool halfAndHalf,
bool seeded,
bool fw,
bool ebwtFw)
{
size_t editidx = 0;
size_t printed = 0;
const size_t qlen = qry.length();
if(exhausted_) out << "E ";
else if(curtailed_) out << "C ";
else out << " ";
if(ebwtFw) out << "<";
else out << ">";
if(fw) out << "F ";
else out << "R ";
std::stringstream ss;
ss << cost_;
string s = ss.str();
if(s.length() < 6) {
for(size_t i = 0; i < 6 - s.length(); i++) {
out << "0";
}
}
out << s << " ";
std::stringstream ss2;
ss2 << minCost;
s = ss2.str();
if(s.length() < 6) {
for(size_t i = 0; i < 6 - s.length(); i++) {
out << "0";
}
}
out << s;
if(halfAndHalf) out << " h ";
else if(seeded) out << " s ";
else out << " ";
std::stringstream ss3;
const size_t numEdits = edits_.size();
if(rdepth_ > 0) {
for(size_t i = 0; i < rdepth_; i++) {
if(editidx < numEdits && edits_.get(editidx).pos == i) {
ss3 << " " << (char)tolower(edits_.get(editidx).chr);
editidx++;
} else {
ss3 << " " << (char)qry.toChar(qlen - i - 1);
}
printed++;
}
ss3 << "|";
} else {
ss3 << " ";
}
for(size_t i = 0; i < len_; i++) {
if(editidx < numEdits && edits_.get(editidx).pos == printed) {
ss3 << (char)tolower(edits_.get(editidx).chr) << " ";
editidx++;
} else {
ss3 << (char)qry.toChar(qlen - printed - 1) << " ";
}
printed++;
}
assert_eq(editidx, edits_.size());
for(size_t i = printed; i < qlen; i++) {
ss3 << "= ";
}
s = ss3.str();
if(ebwtFw) {
std::reverse(s.begin(), s.end());
}
out << s << endl;
}
/**
* Called when the most recent branch extension resulted in an
* empty range or some other constraint violation (e.g., a
* half-and-half constraint).
*/
void curtail(AllocOnlyPool<RangeState>& rpool, int seedLen, bool qualOrder) {
assert(!curtailed_);
assert(!exhausted_);
if(ranges_ == NULL) {
exhausted_ = true;
curtailed_ = true;
return;
}
uint16_t lowestCost = 0xffff;
// Iterate over positions in the path looking for the cost of
// the lowest-cost non-eliminated position
uint32_t eliminatedStretch = 0;
int i = (int)depth0_;
i = max(0, i - rdepth_);
// TODO: It matters whether an insertion/deletion at given
// position would be a gap open or a gap extension
for(; i <= len_; i++) {
if(!eliminated(i)) {
eliminatedStretch = 0;
uint16_t stratum = (rdepth_ + i < seedLen) ? (1 << 14) : 0;
uint16_t cost = (qualOrder ? /*TODO*/ ranges_[i].eq.flags.quallo : 0) | stratum;
if(cost < lowestCost) lowestCost = cost;
} else if(i < rangesSz_) {
eliminatedStretch++;
}
}
if(lowestCost > 0 && lowestCost != 0xffff) {
// This branch's cost will change when curtailed; the
// caller should re-insert it into the priority queue so
// that the new cost takes effect.
cost_ += lowestCost;
} else if(lowestCost == 0xffff) {
// This branch is totally exhausted; there are no more
// valid outgoing paths from any positions within it.
// Remove it from the PathManager and mark it as exhausted.
// The caller should delete it.
exhausted_ = true;
if(ranges_ != NULL) {
assert_gt(rangesSz_, 0);
if(rpool.free(ranges_, rangesSz_)) {
ranges_ = NULL;
rangesSz_ = 0;
}
}
} else {
// Just mark it as curtailed and keep the same cost
}
if(ranges_ != NULL) {
// Try to trim off no-longer-relevant elements of the
// ranges_ array
assert(!exhausted_);
assert_gt(rangesSz_, 0);
uint32_t trim = (rangesSz_ - len_ - 1) + eliminatedStretch;
assert_leq(trim, rangesSz_);
if(rpool.free(ranges_ + rangesSz_ - trim, trim)) {
rangesSz_ -= trim;
if(rangesSz_ == 0) {
ranges_ = NULL;
}
}
}
curtailed_ = true;
}
/**
* Prep this branch for the next extension by calculating the
* SideLocus information and prefetching cache lines from the
* appropriate loci.
*/
void prep(const EbwtParams& ep, const uint8_t* ebwt) {
if(bot_ > top_+1) {
SideLocus::initFromTopBot(top_, bot_, ep, ebwt, ltop_, lbot_);
} else if(bot_ > top_) {
ltop_.initFromRow(top_, ep, ebwt);
lbot_.invalidate();
}
prepped_ = true;
}
/**
* Get the furthest-out RangeState.
*/
RangeState* rangeState() {
assert(!exhausted_);
assert(ranges_ != NULL);
assert_lt(len_, rangesSz_);
return &ranges_[len_];
}
/**
* Set the elims to match the ranges in ranges_[len_], already
* calculated by the caller. Only does mismatches for now.
*/
int installRanges(int c, int nextc, uint32_t qAllow, const uint8_t* qs) {
assert(!exhausted_);
assert(ranges_ != NULL);
RangeState& r = ranges_[len_];
int ret = 0;
r.eliminated_ = true; // start with everything eliminated
r.eq.eliminate(); // set all elim flags to 1
assert_lt(qs[0], 127);
assert_lt(qs[1], 127);
assert_lt(qs[2], 127);
assert_lt(qs[3], 127);
assert_eq(qs[0], qs[1]);
assert_eq(qs[0], qs[2]);
assert_eq(qs[0], qs[3]);
r.eq.flags.quallo = qs[0];
if(qs[0] > qAllow) return 0;
// Set one/both of these to true to do the accounting for
// insertions and deletions as well as mismatches
bool doInserts = false;
bool doDeletes = false;
// We can proceed on an A
if(c != 0 && r.bots[0] > r.tops[0] && qs[0] <= qAllow) {
r.eliminated_ = false;
r.eq.flags.mmA = 0; // A substitution is an option
if(doInserts) r.eq.flags.insA = 0;
if(doDeletes && nextc == 0) r.eq.flags.del = 0;
ret++;
}
// We can proceed on a C
if(c != 1 && r.bots[1] > r.tops[1] && qs[1] <= qAllow) {
r.eliminated_ = false;
r.eq.flags.mmC = 0; // C substitution is an option
if(doInserts) r.eq.flags.insC = 0;
if(doDeletes && nextc == 1) r.eq.flags.del = 0;
ret++;
}
// We can proceed on a G
if(c != 2 && r.bots[2] > r.tops[2] && qs[2] <= qAllow) {
r.eliminated_ = false;
r.eq.flags.mmG = 0; // G substitution is an option
if(doInserts) r.eq.flags.insG = 0;
if(doDeletes && nextc == 2) r.eq.flags.del = 0;
ret++;
}
// We can proceed on a T
if(c != 3 && r.bots[3] > r.tops[3] && qs[3] <= qAllow) {
r.eliminated_ = false;
r.eq.flags.mmT = 0; // T substitution is an option
if(doInserts) r.eq.flags.insT = 0;
if(doDeletes && nextc == 3) r.eq.flags.del = 0;
ret++;
}
return ret;
}
/**
* Extend this branch by one position.
*/
void extend() {
assert(!exhausted_);
assert(!curtailed_);
assert(ranges_ != NULL);
assert(repOk());
prepped_ = false;
len_++;
}
/**
* Do an internal consistency check
*/
bool repOk(uint32_t qlen = 0) const{
assert_leq(depth0_, depth1_);
assert_leq(depth1_, depth2_);
assert_leq(depth2_, depth3_);
assert_gt(depth3_, 0);
if(qlen > 0) {
assert_leq(edits_.size(), qlen); // might have to relax this with inserts
assert_leq(rdepth_, qlen);
}
for(int i = 0; i < len_; i++) {
if(!eliminated(i)) {
assert_lt(i, (int)(len_));
assert(ranges_[i].repOk());
}
}
const size_t numEdits = edits_.size();
for(size_t i = 0; i < numEdits; i++) {
for(size_t j = i+1; j < numEdits; j++) {
// No two edits should be at the same position (might
// have to relax this with inserts)
assert_neq(edits_.get(i).pos, edits_.get(j).pos);
}
}
assert_lt((cost_ >> 14), 4);
return true;
}
uint32_t id_; // branch id; needed to make the ordering of
// branches that are tied in the priority queue
// totally unambiguous. Otherwise, things start
// getting non-deterministic.
uint16_t depth0_; // no edits at depths < depth0
uint16_t depth1_; // at most 1 edit at depths < depth1
uint16_t depth2_; // at most 2 edits at depths < depth2
uint16_t depth3_; // at most 3 edits at depths < depth3
uint16_t rdepth_; // offset in read space from root of search space
uint16_t len_; // length of the branch
uint16_t cost_; // top 2 bits = stratum, bottom 14 = qual ham
// it's up to Branch to keep this updated with the
// cumulative cost of the best path leaving the
// branch; if the branch hasn't been fully
// extended yet, then that path will always be the
// one that extends it by one more
uint16_t ham_; // quality-weighted hamming distance so far
RangeState *ranges_; // Allocated from the RangeStatePool
uint16_t rangesSz_;
TIndexOffU top_; // top offset leading to the root of this subtree
TIndexOffU bot_; // bot offset leading to the root of this subtree
SideLocus ltop_;
SideLocus lbot_;
EditList edits_; // edits leading to the root of the branch
uint16_t delayedCost_;
bool curtailed_; // can't be extended anymore without using edits
bool exhausted_; // all outgoing edges exhausted, including all edits
bool prepped_; // whether SideLocus's are inited
bool delayedIncrease_;
};
/**
* Order two Branches based on cost.
*/
class CostCompare {
public:
/**
* true -> b before a
* false -> a before b
*/
bool operator()(const Branch* a, const Branch* b) const {
bool aUnextendable = a->curtailed_ || a->exhausted_;
bool bUnextendable = b->curtailed_ || b->exhausted_;
// Branch with the best cost
if(a->cost_ == b->cost_) {
// If one or the other is curtailed, take the one that's
// still getting extended
if(bUnextendable && !aUnextendable) {
// a still being extended, return false
return false;
}
if(aUnextendable && !bUnextendable) {
// b still being extended, return true
return true;
}
// Either both are curtailed or both are still being
// extended, pick based on which one is deeper
if(a->tipDepth() != b->tipDepth()) {
// Expression is true if b is deeper
return a->tipDepth() < b->tipDepth();
}
// Keep things deterministic by providing an unambiguous
// order using the id_ field
assert_neq(b->id_, a->id_);
return b->id_ < a->id_;
} else {
return b->cost_ < a->cost_;
}
}
static bool equal(const Branch* a, const Branch* b) {
return a->cost_ == b->cost_ && a->curtailed_ == b->curtailed_ && a->tipDepth() == b->tipDepth();
}
};
/**
* A priority queue for Branch objects; makes it easy to process
* branches in a best-first manner by prioritizing branches with lower
* cumulative costs over branches with higher cumulative costs.
*/
class BranchQueue {
typedef std::pair<int, int> TIntPair;
typedef std::priority_queue<Branch*, std::vector<Branch*>, CostCompare> TBranchQueue;
public:
BranchQueue(bool verbose, bool quiet) :
sz_(0), branchQ_(), patid_(0), verbose_(verbose), quiet_(quiet)
{ }
/**
* Return the front (highest-priority) element of the queue.
*/
Branch *front() {
Branch *b = branchQ_.top();
if(verbose_) {
stringstream ss;
ss << patid_ << ": Fronting " << b->id_ << ", " << b << ", " << b->cost_ << ", " << b->exhausted_ << ", " << b->curtailed_ << ", " << sz_ << "->" << (sz_-1);
glog.msg(ss.str());
}
return b;
}
/**
* Remove and return the front (highest-priority) element of the
* queue.
*/
Branch *pop() {
Branch *b = branchQ_.top(); // get it
branchQ_.pop(); // remove it
if(verbose_) {
stringstream ss;
ss << patid_ << ": Popping " << b->id_ << ", " << b << ", " << b->cost_ << ", " << b->exhausted_ << ", " << b->curtailed_ << ", " << sz_ << "->" << (sz_-1);
glog.msg(ss.str());
}
sz_--;
return b;
}
/**
* Insert a new Branch into the sorted priority queue.
*/
void push(Branch *b) {
#ifndef NDEBUG
bool bIsBetter = empty() || !CostCompare()(b, branchQ_.top());
#endif
if(verbose_) {
stringstream ss;
ss << patid_ << ": Pushing " << b->id_ << ", " << b << ", " << b->cost_ << ", " << b->exhausted_ << ", " << b->curtailed_ << ", " << sz_ << "->" << (sz_+1);
glog.msg(ss.str());
}
branchQ_.push(b);
#ifndef NDEBUG
assert(bIsBetter || branchQ_.top() != b || CostCompare::equal(branchQ_.top(), b));
assert(!bIsBetter || branchQ_.top() == b || CostCompare::equal(branchQ_.top(), b));
#endif
sz_++;
}
/**
* Empty the priority queue and reset the count.
*/
void reset(uint32_t patid) {
patid_ = patid;
branchQ_ = TBranchQueue();
sz_ = 0;
}
/**
* Return true iff the priority queue of branches is empty.
*/
bool empty() const {
bool ret = branchQ_.empty();
assert(ret || sz_ > 0);
assert(!ret || sz_ == 0);
return ret;
}
/**
* Return the number of Branches in the queue.
*/
uint32_t size() const {
return sz_;
}
#ifndef NDEBUG
/**
* Consistency check.
*/
bool repOk(std::set<Branch*>& bset) {
TIntPair pair = bestStratumAndHam(bset);
Branch *b = branchQ_.top();
assert_eq(pair.first, (b->cost_ >> 14));
assert_eq(pair.second, (b->cost_ & ~0xc000));
std::set<Branch*>::iterator it;
for(it = bset.begin(); it != bset.end(); it++) {
assert_gt((*it)->depth3_, 0);
}
return true;
}
#endif
protected:
#ifndef NDEBUG
/**
* Return the stratum and quality-weight (sum of qualities of all
* edited positions) of the lowest-cost branch.
*/
TIntPair bestStratumAndHam(std::set<Branch*>& bset) const {
TIntPair ret = make_pair(0xffff, 0xffff);
std::set<Branch*>::iterator it;
for(it = bset.begin(); it != bset.end(); it++) {
Branch *b = *it;
int stratum = b->cost_ >> 14;
assert_lt(stratum, 4);
int qual = b->cost_ & ~0xc000;
if(stratum < ret.first ||
(stratum == ret.first && qual < ret.second))
{
ret.first = stratum;
ret.second = qual;
}
}
return ret;
}
#endif
uint32_t sz_;
TBranchQueue branchQ_; // priority queue of branches
uint32_t patid_;
bool verbose_;
bool quiet_;
};
/**
* A class that both contains Branches and determines how those
* branches are extended to form longer paths. The overall goal is to
* find the best full alignment(s) as quickly as possible so that a
* successful search can finish quickly. A second goal is to ensure
* that the most "promising" paths are tried first so that, if there is
* a limit on the amount of effort spent searching before we give up,
* we can be as sensitive as possible given that limit.
*
* The quality (or cost) of a given alignment path will ultimately be
* configurable. The default cost model is:
*
* 1. Mismatches incur one "edit" penalty and a "quality" penalty with
* magnitude equal to the Phred quality score of the read position
* involved in the edit (note that insertions into the read are a
* little trickier).
* 2. Edit penalties are all more costly than any quality penalty; i.e.
* the policy sorts alignments first by edit penalty, then by
* quality penalty.
* 3. For the Maq-like alignment policy, edit penalties saturate (don't
* get any greater) after leaving the seed region of the alignment.
*/
class PathManager {
public:
PathManager(ChunkPool* cpool_, int *btCnt, bool verbose, bool quiet) :
branchQ_(verbose, quiet),
cpool(cpool_),
bpool(cpool, "branch"),
rpool(cpool, "range state"),
epool(cpool, "edit"),
minCost(0), btCnt_(btCnt),
verbose_(verbose),
quiet_(quiet)
{ }
~PathManager() { }
/**
* Return the "front" (highest-priority) branch in the collection.
*/
Branch* front() {
assert(!empty());
assert_gt(branchQ_.front()->depth3_, 0);
return branchQ_.front();
}
/**
* Pop the highest-priority (lowest cost) element from the
* priority queue.
*/
Branch* pop() {
Branch* b = branchQ_.pop();
assert_gt(b->depth3_, 0);
#ifndef NDEBUG
// Also remove it from the set
assert(branchSet_.find(b) != branchSet_.end());
ASSERT_ONLY(size_t setSz = branchSet_.size());
branchSet_.erase(branchSet_.find(b));
assert_eq(setSz-1, branchSet_.size());
if(!branchQ_.empty()) {
// Top shouldn't be b any more
Branch *newtop = branchQ_.front();
assert(b != newtop);
}
#endif
// Update this PathManager's cost
minCost = branchQ_.front()->cost_;
assert(repOk());
return b;
}
/**
* Push a new element onto the priority queue.
*/
void push(Branch *b) {
assert(!b->exhausted_);
assert_gt(b->depth3_, 0);
branchQ_.push(b);
#ifndef NDEBUG
// Also insert it into the set
assert(branchSet_.find(b) == branchSet_.end());
branchSet_.insert(b);
#endif
// Update this PathManager's cost
minCost = branchQ_.front()->cost_;
}
/**
* Return the number of active branches in the best-first
* BranchQueue.
*/
uint32_t size() {
return branchQ_.size();
}
/**
* Reset the PathManager, clearing out the priority queue and
* resetting the RangeStatePool.
*/
void reset(uint32_t patid) {
branchQ_.reset(patid); // reset the priority queue
assert(branchQ_.empty());
bpool.reset(); // reset the Branch pool
epool.reset(); // reset the Edit pool
rpool.reset(); // reset the RangeState pool
assert(bpool.empty());
assert(epool.empty());
assert(rpool.empty());
ASSERT_ONLY(branchSet_.clear());
assert_eq(0, branchSet_.size());
assert_eq(0, branchQ_.size());
minCost = 0;
}
#ifndef NDEBUG
/**
* Return true iff Branch b is in the priority queue;
*/
bool contains(Branch *b) const {
bool ret = branchSet_.find(b) != branchSet_.end();
assert(!ret || !b->exhausted_);
return ret;
}
/**
* Do a consistenty-check on the collection of branches contained
* in this PathManager.
*/
bool repOk() {
if(empty()) return true;
assert(branchQ_.repOk(branchSet_));
return true;
}
#endif
/**
* Return true iff the priority queue of branches is empty.
*/
bool empty() const {
bool ret = branchQ_.empty();
assert_eq(ret, branchSet_.empty());
return ret;
}
/**
* Curtail the given branch, and possibly remove it from or
* re-insert it into the priority queue.
*/
void curtail(Branch *br, uint32_t qlen, int seedLen, bool qualOrder) {
assert(!br->exhausted_);
assert(!br->curtailed_);
uint16_t origCost = br->cost_;
br->curtail(rpool, seedLen, qualOrder);
assert(br->curtailed_);
assert_geq(br->cost_, origCost);
if(br->exhausted_) {
assert(br == front());
ASSERT_ONLY(Branch *popped =) pop();
assert(popped == br);
br->free(qlen, rpool, epool, bpool);
} else if(br->cost_ != origCost) {
// Re-insert the newly-curtailed branch
assert(br == front());
Branch *popped = pop();
assert(popped == br);
push(popped);
}
}
/**
* If the frontmost branch is a curtailed branch, split off an
* extendable branch and add it to the queue.
*/
bool splitAndPrep(RandomSource& rand, uint32_t qlen,
uint32_t qualLim, int seedLen,
bool qualOrder,
const EbwtParams& ep, const uint8_t* ebwt,
bool ebwtFw)
{
if(empty()) return true;
// This counts as a backtrack
if(btCnt_ != NULL && (*btCnt_ == 0)) {
// Abruptly end search
return false;
}
Branch *f = front();
assert(!f->exhausted_);
while(f->delayedIncrease_) {
assert(!f->exhausted_);
if(f->delayedIncrease_) {
assert_neq(0, f->delayedCost_);
ASSERT_ONLY(Branch *popped =) pop();
assert(popped == f);
f->cost_ = f->delayedCost_;
f->delayedIncrease_ = false;
f->delayedCost_ = 0;
push(f); // re-insert it
assert(!empty());
}
f = front();
assert(!f->exhausted_);
}
if(f->curtailed_) {
ASSERT_ONLY(uint16_t origCost = f->cost_);
// This counts as a backtrack
if(btCnt_ != NULL) {
if(--(*btCnt_) == 0) {
// Abruptly end search
return false;
}
}
Branch* newbr = splitBranch(
f, rand, qlen, qualLim, seedLen, qualOrder, ep, ebwt,
ebwtFw);
if(newbr == NULL) {
return false;
}
// If f is exhausted, get rid of it immediately
if(f->exhausted_) {
assert(!f->delayedIncrease_);
ASSERT_ONLY(Branch *popped =) pop();
assert(popped == f);
f->free(qlen, rpool, epool, bpool);
}
assert_eq(origCost, f->cost_);
assert(newbr != NULL);
push(newbr);
assert(newbr == front());
}
prep(ep, ebwt);
return true;
}
/**
* Return true iff the front element of the queue is prepped.
*/
bool prepped() {
return front()->prepped_;
}
protected:
/**
* Split off an extendable branch from a curtailed branch.
*/
Branch* splitBranch(Branch* src, RandomSource& rand, uint32_t qlen,
uint32_t qualLim, int seedLen, bool qualOrder,
const EbwtParams& ep, const uint8_t* ebwt,
bool ebwtFw)
{
Branch* dst = src->splitBranch(
rpool, epool, bpool, size(), rand,
qlen, qualLim, seedLen, qualOrder, ep, ebwt, ebwtFw,
verbose_, quiet_);
assert(dst->repOk());
return dst;
}
/**
* Prep the next branch to be extended in advanceBranch().
*/
void prep(const EbwtParams& ep, const uint8_t* ebwt) {
if(!branchQ_.empty()) {
branchQ_.front()->prep(ep, ebwt);
}
}
BranchQueue branchQ_; // priority queue for selecting lowest-cost Branch
// set of branches in priority queue, for sanity checks
ASSERT_ONLY(std::set<Branch*> branchSet_);
public:
ChunkPool *cpool; // pool for generic chunks of memory
AllocOnlyPool<Branch> bpool; // pool for allocating Branches
AllocOnlyPool<RangeState> rpool; // pool for allocating RangeStates
AllocOnlyPool<Edit> epool; // pool for allocating Edits
/// The minimum possible cost for any alignments obtained by
/// advancing further
uint16_t minCost;
protected:
/// Pointer to the aligner's per-read backtrack counter. We
/// increment it in splitBranch.
int *btCnt_;
bool verbose_;
bool quiet_;
};
/**
* Encapsulates an algorithm that navigates the Bowtie index to produce
* candidate ranges of alignments in the Burrows-Wheeler matrix. A
* higher authority is responsible for reporting hits out of those
* ranges, and stopping when the consumer is satisfied.
*/
class RangeSource {
public:
RangeSource() :
done(false), foundRange(false), curEbwt_(NULL) { }
virtual ~RangeSource() { }
/// Set query to find ranges for
virtual void setQuery(Read& r, Range *partial) = 0;
/// Set up the range search.
virtual void initBranch(PathManager& pm) = 0;
/// Advance the range search by one memory op
virtual void advanceBranch(int until, uint16_t minCost, PathManager& pm) = 0;
/// Return the last valid range found
virtual Range& range() = 0;
/// Return ptr to index this RangeSource is currently getting ranges from
const Ebwt *curEbwt() const { return curEbwt_; }
/// All searching w/r/t the current query is finished
bool done;
/// Set to true iff the last call to advance yielded a range
bool foundRange;
protected:
/// ptr to index this RangeSource is currently getting ranges from
const Ebwt *curEbwt_;
};
/**
* Abstract parent of RangeSourceDrivers.
*/
template<typename TRangeSource>
class RangeSourceDriver {
public:
RangeSourceDriver(bool _done, uint32_t minCostAdjustment = 0) :
foundRange(false), done(_done), minCostAdjustment_(minCostAdjustment)
{
minCost = minCostAdjustment_;
}
virtual ~RangeSourceDriver() { }
/**
* Prepare this aligner for the next read.
*/
virtual void setQuery(PatternSourcePerThread* patsrc, Range *r) {
// Clear our buffer of previously-dished-out top offsets
ASSERT_ONLY(allTops_.clear());
setQueryImpl(patsrc, r);
}
/**
* Prepare this aligner for the next read.
*/
virtual void setQueryImpl(PatternSourcePerThread* patsrc, Range *r) = 0;
/**
* Advance the aligner by one memory op. Return true iff we're
* done with this read.
*/
virtual void advance(int until) {
advanceImpl(until);
#ifndef NDEBUG
if(this->foundRange) {
// Assert that we have not yet dished out a range with this
// top offset
assert_gt(range().bot, range().top);
assert(range().ebwt != NULL);
TIndexOff top = (TIndexOff)range().top;
top++; // ensure it's not 0
if(!range().ebwt->fw()) top = -top;
assert(allTops_.find(top) == allTops_.end());
allTops_.insert(top);
}
#endif
}
/**
* Advance the aligner by one memory op. Return true iff we're
* done with this read.
*/
virtual void advanceImpl(int until) = 0;
/**
* Return the range found.
*/
virtual Range& range() = 0;
/**
* Return whether we're generating ranges for the first or the
* second mate.
*/
virtual bool mate1() const = 0;
/**
* Return true iff current pattern is forward-oriented.
*/
virtual bool fw() const = 0;
virtual void removeMate(int m) { }
/// Set to true iff we just found a range.
bool foundRange;
/**
* Set to true if all searching w/r/t the current query is
* finished or if there is no current query.
*/
bool done;
/**
* The minimum "combined" stratum/qual cost that could possibly
* result from subsequent calls to advance() for this driver.
*/
uint16_t minCost;
/**
* Adjustment to the minCost given by the caller that constructed
* the object. This is useful if we know the lowest-cost branch is
* likely to cost less than the any of the alignments that could
* possibly result from advancing (e.g. when we're going to force a
* mismatch somewhere down the line).
*/
uint16_t minCostAdjustment_;
protected:
#ifndef NDEBUG
std::set<TIndexOff> allTops_;
#endif
};
/**
* A concrete driver wrapper for a single RangeSource.
*/
template<typename TRangeSource>
class SingleRangeSourceDriver : public RangeSourceDriver<TRangeSource> {
public:
SingleRangeSourceDriver(
EbwtSearchParams& params,
TRangeSource* rs,
bool fw,
HitSink& sink,
HitSinkPerThread* sinkPt,
EList<BTRefString >& os,
bool verbose,
bool quiet,
bool mate1,
uint32_t minCostAdjustment,
ChunkPool* pool,
int *btCnt) :
RangeSourceDriver<TRangeSource>(true, minCostAdjustment),
len_(0), mate1_(mate1),
sinkPt_(sinkPt),
params_(params),
fw_(fw), rs_(rs),
ebwtFw_(rs_->curEbwt()->fw()),
pm_(pool, btCnt, verbose, quiet)
{
assert(rs_ != NULL);
}
virtual ~SingleRangeSourceDriver() {
delete rs_; rs_ = NULL;
}
/**
* Prepare this aligner for the next read.
*/
virtual void setQueryImpl(PatternSourcePerThread* patsrc, Range *r) {
this->done = false;
pm_.reset((uint32_t)patsrc->rdid());
Read* buf = mate1_ ? &patsrc->bufa() : &patsrc->bufb();
len_ = buf->length();
rs_->setQuery(*buf, r);
initRangeSource((fw_ == ebwtFw_) ? buf->qual : buf->qualRev);
assert_gt(len_, 0);
if(this->done) return;
ASSERT_ONLY(allTops_.clear());
if(!rs_->done) {
rs_->initBranch(pm_); // set up initial branch
}
uint16_t icost = (r != NULL) ? r->cost : 0;
this->minCost = max<uint16_t>(icost, this->minCostAdjustment_);
this->done = rs_->done;
this->foundRange = rs_->foundRange;
if(!pm_.empty()) {
assert(!pm_.front()->curtailed_);
assert(!pm_.front()->exhausted_);
}
}
/**
* Advance the aligner by one memory op. Return true iff we're
* done with this read.
*/
virtual void advanceImpl(int until) {
if(this->done || pm_.empty()) {
this->done = true;
return;
}
assert(!pm_.empty());
assert(!pm_.front()->curtailed_);
assert(!pm_.front()->exhausted_);
params_.setFw(fw_);
// Advance the RangeSource for the forward-oriented read
ASSERT_ONLY(uint16_t oldMinCost = this->minCost);
ASSERT_ONLY(uint16_t oldPmMinCost = pm_.minCost);
rs_->advanceBranch(until, this->minCost, pm_);
this->done = pm_.empty();
if(pm_.minCost != 0) {
this->minCost = max(pm_.minCost, this->minCostAdjustment_);
} else {
// pm_.minCost is 0 because we reset it due to exceptional
// circumstances
}
#ifndef NDEBUG
{
bool error = false;
if(pm_.minCost != 0 && pm_.minCost < oldPmMinCost) {
cerr << "PathManager's cost went down" << endl;
error = true;
}
if(this->minCost < oldMinCost) {
cerr << "this->minCost cost went down" << endl;
error = true;
}
if(error) {
cerr << "pm.minCost went from " << oldPmMinCost
<< " to " << pm_.minCost << endl;
cerr << "this->minCost went from " << oldMinCost
<< " to " << this->minCost << endl;
cerr << "this->minCostAdjustment_ == "
<< this->minCostAdjustment_ << endl;
}
assert(!error);
}
#endif
this->foundRange = rs_->foundRange;
#ifndef NDEBUG
if(this->foundRange) {
if(until >= ADV_COST_CHANGES) assert_eq(oldMinCost, range().cost);
// Assert that we have not yet dished out a range with this
// top offset
assert_gt(range().bot, range().top);
assert(range().ebwt != NULL);
TIndexOff top = (TIndexOff)range().top;
top++; // ensure it's not 0
if(!range().ebwt->fw()) top = -top;
assert(allTops_.find(top) == allTops_.end());
allTops_.insert(top);
}
if(!pm_.empty()) {
assert(!pm_.front()->curtailed_);
assert(!pm_.front()->exhausted_);
}
#endif
}
/**
* Return the range found.
*/
virtual Range& range() {
rs_->range().fw = fw_;
rs_->range().mate1 = mate1_;
return rs_->range();
}
/**
* Return whether we're generating ranges for the first or the
* second mate.
*/
bool mate1() const {
return mate1_;
}
/**
* Return true iff current pattern is forward-oriented.
*/
bool fw() const {
return fw_;
}
virtual void initRangeSource(const BTString& qual) = 0;
protected:
// Progress state
uint32_t len_;
bool mate1_;
// Temporary HitSink; to be deleted
HitSinkPerThread* sinkPt_;
// State for alignment
EbwtSearchParams& params_;
bool fw_;
TRangeSource* rs_; // delete this in destructor
bool ebwtFw_;
PathManager pm_;
ASSERT_ONLY(std::set<TIndexOff> allTops_);
};
/**
* Encapsulates an algorithm that navigates the Bowtie index to produce
* candidate ranges of alignments in the Burrows-Wheeler matrix. A
* higher authority is responsible for reporting hits out of those
* ranges, and stopping when the consumer is satisfied.
*/
template<typename TRangeSource>
class StubRangeSourceDriver : public RangeSourceDriver<TRangeSource> {
typedef EList<RangeSourceDriver<TRangeSource>*> TRangeSrcDrPtrVec;
public:
StubRangeSourceDriver() :
RangeSourceDriver<TRangeSource>(false)
{
this->done = true;
this->foundRange = false;
}
virtual ~StubRangeSourceDriver() { }
/// Set query to find ranges for
virtual void setQueryImpl(PatternSourcePerThread* patsrc, Range *r) { }
/// Advance the range search by one memory op
virtual void advanceImpl(int until) { }
/// Return the last valid range found
virtual Range& range() { throw 1; }
/**
* Return whether we're generating ranges for the first or the
* second mate.
*/
virtual bool mate1() const {
return true;
}
/**
* Return true iff current pattern is forward-oriented.
*/
virtual bool fw() const {
return true;
}
};
/**
* Encapsulates an algorithm that navigates the Bowtie index to produce
* candidate ranges of alignments in the Burrows-Wheeler matrix. A
* higher authority is responsible for reporting hits out of those
* ranges, and stopping when the consumer is satisfied.
*/
template<typename TRangeSource>
class ListRangeSourceDriver : public RangeSourceDriver<TRangeSource> {
typedef EList<RangeSourceDriver<TRangeSource>*> TRangeSrcDrPtrVec;
public:
ListRangeSourceDriver(const TRangeSrcDrPtrVec& rss) :
RangeSourceDriver<TRangeSource>(false),
cur_(0), ham_(0), rss_(rss) /* copy */,
patsrc_(NULL), seedRange_(NULL)
{
assert_gt(rss_.size(), 0);
assert(!this->done);
}
virtual ~ListRangeSourceDriver() {
for(size_t i = 0; i < rss_.size(); i++) {
delete rss_[i];
}
}
/// Set query to find ranges for
virtual void setQueryImpl(PatternSourcePerThread* patsrc, Range *r) {
cur_ = 0; // go back to first RangeSource in list
this->done = false;
rss_[cur_]->setQuery(patsrc, r);
patsrc_ = patsrc; // so that we can call setQuery on the other elements later
seedRange_ = r;
this->done = (cur_ == rss_.size()-1) && rss_[cur_]->done;
this->minCost = max(rss_[cur_]->minCost, this->minCostAdjustment_);
this->foundRange = rss_[cur_]->foundRange;
}
/// Advance the range search by one memory op
virtual void advanceImpl(int until) {
assert(!this->done);
assert_lt(cur_, rss_.size());
if(rss_[cur_]->done) {
// Move on to next RangeSourceDriver
if(cur_ < rss_.size()-1) {
rss_[++cur_]->setQuery(patsrc_, seedRange_);
this->minCost = max(rss_[cur_]->minCost, this->minCostAdjustment_);
this->foundRange = rss_[cur_]->foundRange;
} else {
// No RangeSources in list; done
cur_ = OFF_MASK;
this->done = true;
}
} else {
// Advance current RangeSource
rss_[cur_]->advance(until);
this->done = (cur_ == rss_.size()-1 && rss_[cur_]->done);
this->foundRange = rss_[cur_]->foundRange;
this->minCost = max(rss_[cur_]->minCost, this->minCostAdjustment_);
}
}
/// Return the last valid range found
virtual Range& range() { return rss_[cur_]->range(); }
/**
* Return whether we're generating ranges for the first or the
* second mate.
*/
virtual bool mate1() const {
return rss_[0]->mate1();
}
/**
* Return true iff current pattern is forward-oriented.
*/
virtual bool fw() const {
return rss_[0]->fw();
}
protected:
TIndexOffU cur_;
uint32_t ham_;
TRangeSrcDrPtrVec rss_;
PatternSourcePerThread* patsrc_;
Range *seedRange_;
};
/**
* A RangeSourceDriver that wraps a set of other RangeSourceDrivers and
* chooses which one to advance at any given moment by picking one with
* minimal "cumulative cost" so far.
*
* Note that costs have to be "adjusted" to account for the fact that
* the alignment policy for the underlying RangeSource might force
* mismatches.
*/
template<typename TRangeSource>
class CostAwareRangeSourceDriver : public RangeSourceDriver<TRangeSource> {
typedef RangeSourceDriver<TRangeSource>* TRangeSrcDrPtr;
typedef EList<TRangeSrcDrPtr> TRangeSrcDrPtrVec;
public:
CostAwareRangeSourceDriver(
bool strandFix,
const TRangeSrcDrPtrVec* rss,
bool verbose,
bool quiet,
bool mixesReads) :
RangeSourceDriver<TRangeSource>(false),
rss_(), active_(), strandFix_(strandFix),
lastRange_(NULL), delayedRange_(NULL), patsrc_(NULL),
verbose_(verbose), quiet_(quiet), mixesReads_(mixesReads)
{
if(rss != NULL) {
rss_ = (*rss);
}
paired_ = false;
this->foundRange = false;
this->done = false;
if(rss_.empty()) {
return;
}
calcPaired();
active_ = rss_;
this->minCost = 0;
}
/// Destroy all underlying RangeSourceDrivers
virtual ~CostAwareRangeSourceDriver() {
const size_t rssSz = rss_.size();
for(size_t i = 0; i < rssSz; i++) {
delete rss_[i];
}
rss_.clear();
active_.clear();
}
/// Set query to find ranges for
virtual void setQueryImpl(PatternSourcePerThread* patsrc, Range *r) {
this->done = false;
this->foundRange = false;
lastRange_ = NULL;
delayedRange_ = NULL;
ASSERT_ONLY(allTopsRc_.clear());
patsrc_ = patsrc;
rand_.init(patsrc->bufa().seed);
const size_t rssSz = rss_.size();
if(rssSz == 0) return;
for(size_t i = 0; i < rssSz; i++) {
// Assuming that all
rss_[i]->setQuery(patsrc, r);
}
active_ = rss_;
this->minCost = 0;
sortActives();
}
/**
* Add a new RangeSource to the list and re-sort it.
*/
void addSource(TRangeSrcDrPtr p, Range *r) {
assert(!this->foundRange);
this->lastRange_ = NULL;
this->delayedRange_ = NULL;
this->done = false;
if(patsrc_ != NULL) {
p->setQuery(patsrc_, r);
}
rss_.push_back(p);
active_.push_back(p);
calcPaired();
this->minCost = 0;
sortActives();
}
/**
* Free and remove all contained RangeSources.
*/
void clearSources() {
const size_t rssSz = rss_.size();
for(size_t i = 0; i < rssSz; i++) {
delete rss_[i];
}
rss_.clear();
active_.clear();
paired_ = false;
}
/**
* Return the number of RangeSources contained within.
*/
size_t size() const {
return rss_.size();
}
/**
* Return true iff no RangeSources are contained within.
*/
bool empty() const {
return rss_.empty();
}
/**
* Advance the aligner by one memory op. Return true iff we're
* done with this read.
*/
virtual void advance(int until) {
ASSERT_ONLY(uint16_t precost = this->minCost);
assert(!this->done);
assert(!this->foundRange);
until = max<int>(until, ADV_COST_CHANGES);
advanceImpl(until);
assert(!this->foundRange || lastRange_ != NULL);
if(this->foundRange) {
assert_eq(range().cost, precost);
}
}
/// Advance the range search
virtual void advanceImpl(int until) {
lastRange_ = NULL;
ASSERT_ONLY(uint16_t iminCost = this->minCost);
const size_t actSz = active_.size();
assert(sortedActives());
if(delayedRange_ != NULL) {
assert_eq(iminCost, delayedRange_->cost);
lastRange_ = delayedRange_;
delayedRange_ = NULL;
this->foundRange = true;
assert_eq(range().cost, iminCost);
if(!active_.empty()) {
assert_geq(active_[0]->minCost, this->minCost);
this->minCost = max(active_[0]->minCost, this->minCost);
} else {
this->done = true;
}
return; // found a range
}
assert(delayedRange_ == NULL);
if(mateEliminated() || actSz == 0) {
// No more alternatoves; clear the active set and signal
// we're done
active_.clear();
this->done = true;
return;
}
// Advance lowest-cost RangeSourceDriver
TRangeSrcDrPtr p = active_[0];
uint16_t precost = p->minCost;
assert(!p->done || p->foundRange);
if(!p->foundRange) {
p->advance(until);
}
bool needsSort = false;
if(p->foundRange) {
Range *r = &p->range();
assert_eq(r->cost, iminCost);
needsSort = foundFirstRange(r); // may set delayedRange_; re-sorts active_
assert_eq(lastRange_->cost, iminCost);
if(delayedRange_ != NULL) assert_eq(delayedRange_->cost, iminCost);
p->foundRange = false;
}
if(p->done || (precost != p->minCost) || needsSort) {
sortActives();
if(mateEliminated() || active_.empty()) {
active_.clear();
this->done = (delayedRange_ == NULL);
}
}
assert(sortedActives());
assert(lastRange_ == NULL || lastRange_->cost == iminCost);
assert(delayedRange_ == NULL || delayedRange_->cost == iminCost);
}
/// Return the last valid range found
virtual Range& range() {
assert(lastRange_ != NULL);
return *lastRange_;
}
/**
* Return whether we're generating ranges for the first or the
* second mate.
*/
virtual bool mate1() const {
return rss_[0]->mate1();
}
/**
* Return true iff current pattern is forward-oriented.
*/
virtual bool fw() const {
return rss_[0]->fw();
}
virtual void removeMate(int m) {
bool qmate1 = (m == 1);
assert(paired_);
for(size_t i = 0; i < active_.size(); i++) {
if(active_[i]->mate1() == qmate1) {
active_[i]->done = true;
}
}
sortActives();
assert(mateEliminated());
}
protected:
/**
* Set paired_ to true iff there are mate1 and mate2 range sources
* in the rss_ vector.
*/
void calcPaired() {
const size_t rssSz = rss_.size();
bool saw1 = false;
bool saw2 = false;
for(size_t i = 0; i < rssSz; i++) {
if(rss_[i]->mate1()) saw1 = true;
else saw2 = true;
}
assert(saw1 || saw2);
paired_ = saw1 && saw2;
}
/**
* Return true iff one mate or the other has been eliminated.
*/
bool mateEliminated() {
if(!paired_) return false;
bool mate1sLeft = false;
bool mate2sLeft = false;
// If this RangeSourceDriver is done, shift everyone else
// up and delete it
const size_t rssSz = active_.size();
for(size_t i = 0; i < rssSz; i++) {
if(!active_[i]->done) {
if(active_[i]->mate1()) mate1sLeft = true;
else mate2sLeft = true;
}
}
return !mate1sLeft || !mate2sLeft;
}
#ifndef NDEBUG
/**
* Check that the given range has not yet been reported.
*/
bool checkRange(Range* r) {
// Assert that we have not yet dished out a range with this
// top offset
assert_gt(r->bot, r->top);
assert(r->ebwt != NULL);
TIndexOff top = (TIndexOff)r->top;
top++; // ensure it's not 0
if(!r->ebwt->fw()) top = -top;
if(r->fw) {
assert(this->allTops_.find(top) == this->allTops_.end());
if(!mixesReads_) this->allTops_.insert(top);
} else {
assert(this->allTopsRc_.find(top) == this->allTopsRc_.end());
if(!mixesReads_) this->allTopsRc_.insert(top);
}
return true;
}
#endif
/**
* We found a range; check whether we should attempt to find a
* range of equal quality from the opposite strand so that we can
* resolve the strand bias. Return true iff we modified the cost
* of one or more items after the first item.
*/
bool foundFirstRange(Range* r) {
assert(r != NULL);
assert(checkRange(r));
this->foundRange = true;
lastRange_ = r;
if(strandFix_) {
// We found a range but there may be an equally good range
// on the other strand; let's try to get it.
const size_t sz = active_.size();
for(size_t i = 1; i < sz; i++) {
// Same mate, different orientation?
if(rss_[i]->mate1() == r->mate1 && rss_[i]->fw() != r->fw) {
// Yes; see if it has the same cost
TRangeSrcDrPtr p = active_[i];
uint16_t minCost = max(this->minCost, p->minCost);
if(minCost > r->cost) break;
// Yes, it has the same cost
assert_eq(minCost, r->cost); // can't be better
// Advance it until it's done, we've found a range,
// or its cost increases
if(this->verbose_) cout << " Looking for opposite range to avoid strand bias:" << endl;
while(!p->done && !p->foundRange) {
p->advance(ADV_COST_CHANGES);
assert_geq(p->minCost, minCost);
if(p->minCost > minCost) break;
}
if(p->foundRange) {
// Found one! Now we have to choose which one
// to give out first; we choose randomly using
// the size of the ranges as weights.
delayedRange_ = &p->range();
assert(checkRange(delayedRange_));
size_t tot = (delayedRange_->bot - delayedRange_->top) +
(lastRange_->bot - lastRange_->top);
uint32_t rq = rand_.nextU32() % tot;
// We picked this range, not the first one
if(rq < (delayedRange_->bot - delayedRange_->top)) {
Range *tmp = lastRange_;
lastRange_ = delayedRange_;
delayedRange_ = tmp;
}
p->foundRange = false;
}
// Return true iff we need to force a re-sort
return true;
}
}
// OK, now we have a choice of two equally good ranges from
// each strand.
}
return false;
}
/**
* Sort all of the RangeSourceDriver ptrs in the rss_ array so that
* the one with the lowest cumulative cost is at the top. Break
* ties randomly. Just do selection sort for now; we don't expect
* the list to be long.
*/
void sortActives() {
TRangeSrcDrPtrVec& vec = active_;
size_t sz = vec.size();
// Selection sort / removal outer loop
for(size_t i = 0; i < sz;) {
// Remove elements that we're done with
if(vec[i]->done && !vec[i]->foundRange) {
vec.erase(i);
if(sz == 0) break;
else sz--;
continue;
}
uint16_t minCost = vec[i]->minCost;
size_t minOff = i;
// Selection sort inner loop
for(size_t j = i+1; j < sz; j++) {
if(vec[j]->done && !vec[j]->foundRange) {
// We'll get rid of this guy later
continue;
}
if(vec[j]->minCost < minCost) {
minCost = vec[j]->minCost;
minOff = j;
} else if(vec[j]->minCost == minCost) {
// Possibly randomly pick the other
if(rand_.nextU32() & 0x1000) {
minOff = j;
}
}
}
// Do the swap, if necessary
if(i != minOff) {
assert_leq(minCost, vec[i]->minCost);
TRangeSrcDrPtr tmp = vec[i];
vec[i] = vec[minOff];
vec[minOff] = tmp;
}
i++;
}
if(delayedRange_ == NULL && sz > 0) {
assert_geq(this->minCost, this->minCostAdjustment_);
assert_geq(vec[0]->minCost, this->minCost);
this->minCost = vec[0]->minCost;
}
assert(sortedActives());
}
#ifndef NDEBUG
/**
* Check that the rss_ array is sorted by minCost; assert if it's
* not.
*/
bool sortedActives() const {
// Selection sort outer loop
const TRangeSrcDrPtrVec& vec = active_;
const size_t sz = vec.size();
for(size_t i = 0; i < sz; i++) {
assert(!vec[i]->done || vec[i]->foundRange);
for(size_t j = i+1; j < sz; j++) {
assert(!vec[j]->done || vec[j]->foundRange);
assert_leq(vec[i]->minCost, vec[j]->minCost);
}
}
if(delayedRange_ == NULL && sz > 0) {
// Only assert this if there's no delayed range; if there's
// a delayed range, the minCost is its cost, not the 0th
// element's cost
assert_leq(vec[0]->minCost, this->minCost);
}
return true;
}
#endif
/// List of all the drivers
TRangeSrcDrPtrVec rss_;
/// List of all the as-yet-uneliminated drivers
TRangeSrcDrPtrVec active_;
/// Whether the list of drivers contains drivers for both mates 1 and 2
bool paired_;
/// If true, this driver will make an attempt to dish out ranges in
/// a way that approaches the right distribution based on the
/// number of hits on both strands.
bool strandFix_;
uint32_t randSeed_;
/// The random seed from the Aligner, which we use to randomly break ties
RandomSource rand_;
Range *lastRange_;
Range *delayedRange_;
PatternSourcePerThread* patsrc_;
bool verbose_;
bool quiet_;
bool mixesReads_;
ASSERT_ONLY(std::set<TIndexOff> allTopsRc_);
};
#endif /* RANGE_SOURCE_H_ */
|