DXR is a code search and navigation tool aimed at making sense of large projects. It supports full-text and regex searches as well as structural queries.

Implementation

Mercurial (a5710687f9b4)

VCS Links

Line Code
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
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*-
 * vim: set ts=8 sts=2 et sw=2 tw=80:
 * This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

/*
 * Streaming access to the raw tokens of JavaScript source.
 *
 * Because JS tokenization is context-sensitive -- a '/' could be either a
 * regular expression *or* a division operator depending on context -- the
 * various token stream classes are mostly not useful outside of the Parser
 * where they reside.  We should probably eventually merge the two concepts.
 */
#ifndef frontend_TokenStream_h
#define frontend_TokenStream_h

/*
 * [SMDOC] Parser Token Stream
 *
 * A token stream exposes the raw tokens -- operators, names, numbers,
 * keywords, and so on -- of JavaScript source code.
 *
 * These are the components of the overall token stream concept:
 * TokenStreamShared, TokenStreamAnyChars, TokenStreamCharsBase<Unit>,
 * TokenStreamChars<Unit>, and TokenStreamSpecific<Unit, AnyCharsAccess>.
 *
 * == TokenStreamShared → ∅ ==
 *
 * Certain aspects of tokenizing are used everywhere:
 *
 *   * modifiers (used to select which context-sensitive interpretation of a
 *     character should be used to decide what token it is) and modifier
 *     assertion handling;
 *   * flags on the overall stream (have we encountered any characters on this
 *     line?  have we hit a syntax error?  and so on);
 *   * and certain token-count constants.
 *
 * These are all defined in TokenStreamShared.  (They could be namespace-
 * scoped, but it seems tentatively better not to clutter the namespace.)
 *
 * == TokenStreamAnyChars → TokenStreamShared ==
 *
 * Certain aspects of tokenizing have meaning independent of the character type
 * of the source text being tokenized: line/column number information, tokens
 * in lookahead from determining the meaning of a prior token, compilation
 * options, the filename, flags, source map URL, access to details of the
 * current and next tokens (is the token of the given type?  what name or
 * number is contained in the token?  and other queries), and others.
 *
 * All this data/functionality *could* be duplicated for both single-byte and
 * double-byte tokenizing, but there are two problems.  First, it's potentially
 * wasteful if the compiler doesnt recognize it can unify the concepts.  (And
 * if any-character concepts are intermixed with character-specific concepts,
 * potentially the compiler *can't* unify them because offsets into the
 * hypothetical TokenStream<Unit>s would differ.)  Second, some of this stuff
 * needs to be accessible in ParserBase, the aspects of JS language parsing
 * that have meaning independent of the character type of the source text being
 * parsed.  So we need a separate data structure that ParserBase can hold on to
 * for it.  (ParserBase isn't the only instance of this, but it's certainly the
 * biggest case of it.)  Ergo, TokenStreamAnyChars.
 *
 * == TokenStreamCharsShared → ∅ ==
 *
 * Some functionality has meaning independent of character type, yet has no use
 * *unless* you know the character type in actual use.  It *could* live in
 * TokenStreamAnyChars, but it makes more sense to live in a separate class
 * that character-aware token information can simply inherit.
 *
 * This class currently exists only to contain a char16_t buffer, transiently
 * used to accumulate strings in tricky cases that can't just be read directly
 * from source text.  It's not used outside character-aware tokenizing, so it
 * doesn't make sense in TokenStreamAnyChars.
 *
 * == TokenStreamCharsBase<Unit> → TokenStreamCharsShared ==
 *
 * Certain data structures in tokenizing are character-type-specific: namely,
 * the various pointers identifying the source text (including current offset
 * and end).
 *
 * Additionally, some functions operating on this data are defined the same way
 * no matter what character type you have (e.g. current offset in code units
 * into the source text) or share a common interface regardless of character
 * type (e.g. consume the next code unit if it has a given value).
 *
 * All such functionality lives in TokenStreamCharsBase<Unit>.
 *
 * == SpecializedTokenStreamCharsBase<Unit> → TokenStreamCharsBase<Unit> ==
 *
 * Certain tokenizing functionality is specific to a single character type.
 * For example, JS's UTF-16 encoding recognizes no coding errors, because lone
 * surrogates are not an error; but a UTF-8 encoding must recognize a variety
 * of validation errors.  Such functionality is defined only in the appropriate
 * SpecializedTokenStreamCharsBase specialization.
 *
 * == GeneralTokenStreamChars<Unit, AnyCharsAccess> →
 *    SpecializedTokenStreamCharsBase<Unit> ==
 *
 * Some functionality operates differently on different character types, just
 * as for TokenStreamCharsBase, but additionally requires access to character-
 * type-agnostic information in TokenStreamAnyChars.  For example, getting the
 * next character performs different steps for different character types and
 * must access TokenStreamAnyChars to update line break information.
 *
 * Such functionality, if it can be defined using the same algorithm for all
 * character types, lives in GeneralTokenStreamChars<Unit, AnyCharsAccess>.
 * The AnyCharsAccess parameter provides a way for a GeneralTokenStreamChars
 * instance to access its corresponding TokenStreamAnyChars, without inheriting
 * from it.
 *
 * GeneralTokenStreamChars<Unit, AnyCharsAccess> is just functionality, no
 * actual member data.
 *
 * Such functionality all lives in TokenStreamChars<Unit, AnyCharsAccess>, a
 * declared-but-not-defined template class whose specializations have a common
 * public interface (plus whatever private helper functions are desirable).
 *
 * == TokenStreamChars<Unit, AnyCharsAccess> →
 *    GeneralTokenStreamChars<Unit, AnyCharsAccess> ==
 *
 * Some functionality is like that in GeneralTokenStreamChars, *but* it's
 * defined entirely differently for different character types.
 *
 * For example, consider "match a multi-code unit code point" (hypothetically:
 * we've only implemented two-byte tokenizing right now):
 *
 *   * For two-byte text, there must be two code units to get, the leading code
 *     unit must be a UTF-16 lead surrogate, and the trailing code unit must be
 *     a UTF-16 trailing surrogate.  (If any of these fail to hold, a next code
 *     unit encodes that code point and is not multi-code unit.)
 *   * For single-byte Latin-1 text, there are no multi-code unit code points.
 *   * For single-byte UTF-8 text, the first code unit must have N > 1 of its
 *     highest bits set (and the next unset), and |N - 1| successive code units
 *     must have their high bit set and next-highest bit unset, *and*
 *     concatenating all unconstrained bits together must not produce a code
 *     point value that could have been encoded in fewer code units.
 *
 * This functionality can't be implemented as member functions in
 * GeneralTokenStreamChars because we'd need to *partially specialize* those
 * functions -- hold Unit constant while letting AnyCharsAccess vary.  But
 * C++ forbids function template partial specialization like this: either you
 * fix *all* parameters or you fix none of them.
 *
 * Fortunately, C++ *does* allow *class* template partial specialization.  So
 * TokenStreamChars is a template class with one specialization per Unit.
 * Functions can be defined differently in the different specializations,
 * because AnyCharsAccess as the only template parameter on member functions
 * *can* vary.
 *
 * All TokenStreamChars<Unit, AnyCharsAccess> specializations, one per Unit,
 * are just functionality, no actual member data.
 *
 * == TokenStreamSpecific<Unit, AnyCharsAccess> →
 *    TokenStreamChars<Unit, AnyCharsAccess>, TokenStreamShared,
 *    ErrorReporter ==
 *
 * TokenStreamSpecific is operations that are parametrized on character type
 * but implement the *general* idea of tokenizing, without being intrinsically
 * tied to character type.  Notably, this includes all operations that can
 * report warnings or errors at particular offsets, because we include a line
 * of context with such errors -- and that necessarily accesses the raw
 * characters of their specific type.
 *
 * Much TokenStreamSpecific operation depends on functionality in
 * TokenStreamAnyChars.  The obvious solution is to inherit it -- but this
 * doesn't work in Parser: its ParserBase base class needs some
 * TokenStreamAnyChars functionality without knowing character type.
 *
 * The AnyCharsAccess type parameter is a class that statically converts from a
 * TokenStreamSpecific* to its corresponding TokenStreamAnyChars.  The
 * TokenStreamSpecific in Parser<ParseHandler, Unit> can then specify a class
 * that properly converts from TokenStreamSpecific Parser::tokenStream to
 * TokenStreamAnyChars ParserBase::anyChars.
 *
 * Could we hardcode one set of offset calculations for this and eliminate
 * AnyCharsAccess?  No.  Offset calculations possibly could be hardcoded if
 * TokenStreamSpecific were present in Parser before Parser::handler, assuring
 * the same offsets in all Parser-related cases.  But there's still a separate
 * TokenStream class, that requires different offset calculations.  So even if
 * we wanted to hardcode this (it's not clear we would, because forcing the
 * TokenStreamSpecific declarer to specify this is more explicit), we couldn't.
 */

#include "mozilla/ArrayUtils.h"
#include "mozilla/Assertions.h"
#include "mozilla/Attributes.h"
#include "mozilla/Casting.h"
#include "mozilla/DebugOnly.h"
#include "mozilla/Maybe.h"
#include "mozilla/MemoryChecking.h"
#include "mozilla/PodOperations.h"
#include "mozilla/Span.h"
#include "mozilla/TextUtils.h"
#include "mozilla/TypeTraits.h"
#include "mozilla/Unused.h"
#include "mozilla/Utf8.h"

#include <algorithm>
#include <stdarg.h>
#include <stddef.h>
#include <stdint.h>
#include <stdio.h>

#include "jspubtd.h"

#include "frontend/ErrorReporter.h"
#include "frontend/Token.h"
#include "frontend/TokenKind.h"
#include "js/CompileOptions.h"
#include "js/HashTable.h"    // js::HashMap
#include "js/RegExpFlags.h"  // JS::RegExpFlags
#include "js/UniquePtr.h"
#include "js/Vector.h"
#include "util/Text.h"
#include "util/Unicode.h"
#include "vm/ErrorReporting.h"
#include "vm/JSAtom.h"
#include "vm/StringType.h"

struct JSContext;
struct KeywordInfo;

namespace js {

class AutoKeepAtoms;

namespace frontend {

extern TokenKind ReservedWordTokenKind(PropertyName* str);

extern const char* ReservedWordToCharZ(PropertyName* str);

extern const char* ReservedWordToCharZ(TokenKind tt);

struct TokenStreamFlags {
  bool isEOF : 1;           // Hit end of file.
  bool isDirtyLine : 1;     // Non-whitespace since start of line.
  bool sawOctalEscape : 1;  // Saw an octal character escape.
  bool hadError : 1;        // Hit a syntax error, at start or during a
                            // token.

  TokenStreamFlags() : isEOF(), isDirtyLine(), sawOctalEscape(), hadError() {}
};

template <typename Unit>
class TokenStreamPosition;

/**
 * TokenStream types and constants that are used in both TokenStreamAnyChars
 * and TokenStreamSpecific.  Do not add any non-static data members to this
 * class!
 */
class TokenStreamShared {
 protected:
  static constexpr size_t ntokens = 4;  // 1 current + 2 lookahead, rounded
                                        // to power of 2 to avoid divmod by 3

  static constexpr unsigned ntokensMask = ntokens - 1;

  template <typename Unit>
  friend class TokenStreamPosition;

 public:
  static constexpr unsigned maxLookahead = 2;

  using Modifier = Token::Modifier;
  static constexpr Modifier SlashIsDiv = Token::SlashIsDiv;
  static constexpr Modifier SlashIsRegExp = Token::SlashIsRegExp;
  static constexpr Modifier SlashIsInvalid = Token::SlashIsInvalid;

  static void verifyConsistentModifier(Modifier modifier,
                                       const Token& nextToken) {
    MOZ_ASSERT(
        modifier == nextToken.modifier || modifier == SlashIsInvalid,
        "This token was scanned with both SlashIsRegExp and SlashIsDiv, "
        "indicating the parser is confused about how to handle a slash here. "
        "See comment at Token::Modifier.");
  }
};

static_assert(mozilla::IsEmpty<TokenStreamShared>::value,
              "TokenStreamShared shouldn't bloat classes that inherit from it");

template <typename Unit, class AnyCharsAccess>
class TokenStreamSpecific;

template <typename Unit>
class MOZ_STACK_CLASS TokenStreamPosition final {
 public:
  // The JS_HAZ_ROOTED is permissible below because: 1) the only field in
  // TokenStreamPosition that can keep GC things alive is Token, 2) the only
  // GC things Token can keep alive are atoms, and 3) the AutoKeepAtoms&
  // passed to the constructor here represents that collection of atoms
  // is disabled while atoms in Tokens in this Position are alive.  DON'T
  // ADD NON-ATOM GC THING POINTERS HERE!  They would create a rooting
  // hazard that JS_HAZ_ROOTED will cause to be ignored.
  template <class AnyCharsAccess>
  inline TokenStreamPosition(
      AutoKeepAtoms& keepAtoms,
      TokenStreamSpecific<Unit, AnyCharsAccess>& tokenStream);

 private:
  TokenStreamPosition(const TokenStreamPosition&) = delete;

  // Technically only TokenStreamSpecific<Unit, AnyCharsAccess>::seek with
  // Unit constant and AnyCharsAccess varying must be friended, but 1) it's
  // hard to friend one function in template classes, and 2) C++ doesn't
  // allow partial friend specialization to target just that single class.
  template <typename Char, class AnyCharsAccess>
  friend class TokenStreamSpecific;

  const Unit* buf;
  TokenStreamFlags flags;
  unsigned lineno;
  size_t linebase;
  size_t prevLinebase;
  Token currentToken;
  unsigned lookahead;
  Token lookaheadTokens[TokenStreamShared::maxLookahead];
} JS_HAZ_ROOTED;

template <typename Unit>
class SourceUnits;

// Column numbers *ought* be in terms of counts of code points, but in the past
// we counted code units.  Set this to 0 to keep returning counts of code units
// (even for UTF-8, which is clearly wrong, but we don't ship UTF-8 yet so this
// is fine until we can fix users that depend on code-unit counting).
#define JS_COLUMN_DIMENSION_IS_CODE_POINTS() 1

class TokenStreamAnyChars : public TokenStreamShared {
 public:
  TokenStreamAnyChars(JSContext* cx, const JS::ReadOnlyCompileOptions& options,
                      StrictModeGetter* smg);

  template <typename Unit, class AnyCharsAccess>
  friend class GeneralTokenStreamChars;
  template <typename Unit, class AnyCharsAccess>
  friend class TokenStreamChars;
  template <typename Unit, class AnyCharsAccess>
  friend class TokenStreamSpecific;

  template <typename Unit>
  friend class TokenStreamPosition;

  // Accessors.
  unsigned cursor() const { return cursor_; }
  unsigned nextCursor() const { return (cursor_ + 1) & ntokensMask; }
  unsigned aheadCursor(unsigned steps) const {
    return (cursor_ + steps) & ntokensMask;
  }

  const Token& currentToken() const { return tokens[cursor()]; }
  bool isCurrentTokenType(TokenKind type) const {
    return currentToken().type == type;
  }

  MOZ_MUST_USE bool checkOptions();

 private:
  PropertyName* reservedWordToPropertyName(TokenKind tt) const;

 public:
  PropertyName* currentName() const {
    if (isCurrentTokenType(TokenKind::Name) ||
        isCurrentTokenType(TokenKind::PrivateName)) {
      return currentToken().name();
    }

    MOZ_ASSERT(TokenKindIsPossibleIdentifierName(currentToken().type));
    return reservedWordToPropertyName(currentToken().type);
  }

  bool currentNameHasEscapes() const {
    if (isCurrentTokenType(TokenKind::Name) ||
        isCurrentTokenType(TokenKind::PrivateName)) {
      TokenPos pos = currentToken().pos;
      return (pos.end - pos.begin) != currentToken().name()->length();
    }

    MOZ_ASSERT(TokenKindIsPossibleIdentifierName(currentToken().type));
    return false;
  }

  bool isCurrentTokenAssignment() const {
    return TokenKindIsAssignment(currentToken().type);
  }

  // Flag methods.
  bool isEOF() const { return flags.isEOF; }
  bool sawOctalEscape() const { return flags.sawOctalEscape; }
  bool hadError() const { return flags.hadError; }
  void clearSawOctalEscape() { flags.sawOctalEscape = false; }

  bool hasInvalidTemplateEscape() const {
    return invalidTemplateEscapeType != InvalidEscapeType::None;
  }
  void clearInvalidTemplateEscape() {
    invalidTemplateEscapeType = InvalidEscapeType::None;
  }

 private:
  // This is private because it should only be called by the tokenizer while
  // tokenizing not by, for example, BytecodeEmitter.
  bool strictMode() const {
    return strictModeGetter && strictModeGetter->strictMode();
  }

  void setInvalidTemplateEscape(uint32_t offset, InvalidEscapeType type) {
    MOZ_ASSERT(type != InvalidEscapeType::None);
    if (invalidTemplateEscapeType != InvalidEscapeType::None) {
      return;
    }
    invalidTemplateEscapeOffset = offset;
    invalidTemplateEscapeType = type;
  }

  uint32_t invalidTemplateEscapeOffset = 0;
  InvalidEscapeType invalidTemplateEscapeType = InvalidEscapeType::None;

 public:
  // Call this immediately after parsing an OrExpression to allow scanning the
  // next token with SlashIsRegExp without asserting (even though we just
  // peeked at it in SlashIsDiv mode).
  //
  // It's OK to disable the assertion because the places where this is called
  // have peeked at the next token in SlashIsDiv mode, and checked that it is
  // *not* a Div token.
  //
  // To see why it is necessary to disable the assertion, consider these two
  // programs:
  //
  //     x = arg => q       // per spec, this is all one statement, and the
  //     /a/g;              // slashes are division operators
  //
  //     x = arg => {}      // per spec, ASI at the end of this line
  //     /a/g;              // and that's a regexp literal
  //
  // The first program shows why orExpr() has use SlashIsDiv mode when peeking
  // ahead for the next operator after parsing `q`. The second program shows
  // why matchOrInsertSemicolon() must use SlashIsRegExp mode when scanning
  // ahead for a semicolon.
  void allowGettingNextTokenWithSlashIsRegExp() {
#ifdef DEBUG
    // Check the precondition: Caller already peeked ahead at the next token,
    // in SlashIsDiv mode, and it is *not* a Div token.
    MOZ_ASSERT(hasLookahead());
    const Token& next = nextToken();
    MOZ_ASSERT(next.modifier == SlashIsDiv);
    MOZ_ASSERT(next.type != TokenKind::Div);
    tokens[nextCursor()].modifier = SlashIsRegExp;
#endif
  }

#ifdef DEBUG
  inline bool debugHasNoLookahead() const { return lookahead == 0; }
#endif

  bool hasDisplayURL() const { return displayURL_ != nullptr; }

  char16_t* displayURL() { return displayURL_.get(); }

  bool hasSourceMapURL() const { return sourceMapURL_ != nullptr; }

  char16_t* sourceMapURL() { return sourceMapURL_.get(); }

  // This class maps a sourceUnits offset (which is 0-indexed) to a line
  // number (which is 1-indexed) and an offset (which is 0-indexed) in
  // code *units* (not code points, not bytes) into the line.
  //
  // If you need a column number (i.e. an offset in code *points*),
  // GeneralTokenStreamChars<Unit> contains functions that consult this and
  // use Unit-specific source text to determine the correct column.
  class SourceCoords {
    // For a given buffer holding source code, |lineStartOffsets_| has one
    // element per line of source code, plus one sentinel element.  Each
    // non-sentinel element holds the buffer offset for the start of the
    // corresponding line of source code.  For this example script,
    // assuming an initialLineOffset of 0:
    //
    // 1  // xyz            [line starts at offset 0]
    // 2  var x;            [line starts at offset 7]
    // 3                    [line starts at offset 14]
    // 4  var y;            [line starts at offset 15]
    //
    // |lineStartOffsets_| is:
    //
    //   [0, 7, 14, 15, MAX_PTR]
    //
    // To convert a "line number" to an "index" into |lineStartOffsets_|,
    // subtract |initialLineNum_|.  E.g. line 3's index is
    // (3 - initialLineNum_), which is 2.  Therefore lineStartOffsets_[2]
    // holds the buffer offset for the start of line 3, which is 14.  (Note
    // that |initialLineNum_| is often 1, but not always.
    //
    // The first element is always initialLineOffset, passed to the
    // constructor, and the last element is always the MAX_PTR sentinel.
    //
    // Offset-to-{line,offset-into-line} lookups are O(log n) in the worst
    // case (binary search), but in practice they're heavily clustered and
    // we do better than that by using the previous lookup's result
    // (lastIndex_) as a starting point.
    //
    // Checking if an offset lies within a particular line number
    // (isOnThisLine()) is O(1).
    //
    Vector<uint32_t, 128> lineStartOffsets_;

    /** The line number on which the source text begins. */
    uint32_t initialLineNum_;

    /**
     * The index corresponding to the last offset lookup -- used so that if
     * offset lookups proceed in increasing order, and and the offset appears
     * in the next couple lines from the last offset, we can avoid a full
     * binary-search.
     *
     * This is mutable because it's modified on every search, but that fact
     * isn't visible outside this class.
     */
    mutable uint32_t lastIndex_;

    uint32_t indexFromOffset(uint32_t offset) const;

    static const uint32_t MAX_PTR = UINT32_MAX;

    uint32_t lineNumberFromIndex(uint32_t index) const {
      return index + initialLineNum_;
    }

    uint32_t indexFromLineNumber(uint32_t lineNum) const {
      return lineNum - initialLineNum_;
    }

   public:
    SourceCoords(JSContext* cx, uint32_t initialLineNumber,
                 uint32_t initialOffset);

    MOZ_MUST_USE bool add(uint32_t lineNum, uint32_t lineStartOffset);
    MOZ_MUST_USE bool fill(const SourceCoords& other);

    bool isOnThisLine(uint32_t offset, uint32_t lineNum,
                      bool* onThisLine) const {
      uint32_t index = indexFromLineNumber(lineNum);
      if (index + 1 >= lineStartOffsets_.length()) {  // +1 due to sentinel
        return false;
      }
      *onThisLine = lineStartOffsets_[index] <= offset &&
                    offset < lineStartOffsets_[index + 1];
      return true;
    }

    /**
     * A token, computed for an offset in source text, that can be used to
     * access line number and line-offset information for that offset.
     *
     * LineToken *alone* exposes whether the corresponding offset is in the
     * the first line of source (which may not be 1, depending on
     * |initialLineNumber|), and whether it's in the same line as
     * another LineToken.
     */
    class LineToken {
      uint32_t index;
#ifdef DEBUG
      uint32_t offset_;  // stored for consistency-of-use assertions
#endif

      friend class SourceCoords;

     public:
      explicit LineToken(uint32_t index, uint32_t offset)
          : index(index)
#ifdef DEBUG
            ,
            offset_(offset)
#endif
      {
      }

      bool isFirstLine() const { return index == 0; }

      bool isSameLine(LineToken other) const { return index == other.index; }

      void assertConsistentOffset(uint32_t offset) const {
        MOZ_ASSERT(offset_ == offset);
      }
    };

    /**
     * Compute a token usable to access information about the line at the
     * given offset.
     *
     * The only information directly accessible in a token is whether it
     * corresponds to the first line of source text (which may not be line
     * 1, depending on the |initialLineNumber| value used to construct
     * this).  Use |lineNumber(LineToken)| to compute the actual line
     * number (incorporating the contribution of |initialLineNumber|).
     */
    LineToken lineToken(uint32_t offset) const;

    /** Compute the line number for the given token. */
    uint32_t lineNumber(LineToken lineToken) const {
      return lineNumberFromIndex(lineToken.index);
    }

    /** Return the offset of the start of the line for |lineToken|. */
    uint32_t lineStart(LineToken lineToken) const {
      MOZ_ASSERT(lineToken.index + 1 < lineStartOffsets_.length(),
                 "recorded line-start information must be available");
      return lineStartOffsets_[lineToken.index];
    }
  };

  SourceCoords srcCoords;

  JSContext* context() const { return cx; }

  using LineToken = SourceCoords::LineToken;

  LineToken lineToken(uint32_t offset) const {
    return srcCoords.lineToken(offset);
  }

  uint32_t lineNumber(LineToken lineToken) const {
    return srcCoords.lineNumber(lineToken);
  }

  uint32_t lineStart(LineToken lineToken) const {
    return srcCoords.lineStart(lineToken);
  }

  /**
   * Fill in |err|.
   *
   * If the token stream doesn't have location info for this error, use the
   * caller's location (including line/column number) and return false.  (No
   * line of context is set.)
   *
   * Otherwise fill in everything in |err| except 1) line/column numbers and
   * 2) line-of-context-related fields and return true.  The caller *must*
   * fill in the line/column number; filling the line of context is optional.
   */
  bool fillExceptingContext(ErrorMetadata* err, uint32_t offset);

  MOZ_ALWAYS_INLINE void updateFlagsForEOL() { flags.isDirtyLine = false; }

 private:
#if JS_COLUMN_DIMENSION_IS_CODE_POINTS()
  /**
   * Compute the "partial" column number in Unicode code points of the absolute
   * |offset| within source text on the line of |lineToken| (which must have
   * been computed from |offset|).
   *
   * A partial column number on a line that isn't the first line is just the
   * actual column number.  But a partial column number on the first line is the
   * column number *ignoring the initial line/column of the script*.  For
   * example, consider this HTML with line/column number keys:
   *
   *                 1         2            3
   *       0123456789012345678901234   567890
   *     ------------------------------------
   *   1 | <html>
   *   2 | <head>
   *   3 |   <script>var x = 3;  x &lt; 4;
   *   4 | const y = 7;</script>
   *   5 | </head>
   *   6 | <body></body>
   *   7 | </html>
   *
   * The script would be compiled specifying initial (line, column) of (3, 10)
   * using |JS::ReadOnlyCompileOptions::{lineno,column}|.  And the column
   * reported by |computeColumn| for the "v" of |var| would be 10.  But the
   * partial column number of the "v" in |var|, that this function returns,
   * would be 0.  On the other hand, the column reported by |computeColumn| and
   * the partial column number returned by this function for the "c" in |const|
   * would both be 0, because it's not in the first line of source text.
   *
   * The partial column is with respect *only* to the JavaScript source text as
   * SpiderMonkey sees it.  In the example, the "&lt;" is converted to "<" by
   * the browser before SpiderMonkey would see it.  So the partial column of the
   * "4" in the inequality would be 16, not 19.
   *
   * Code points are not all equal length, so counting requires *some* kind of
   * linear-time counting from the start of the line.  This function attempts
   * various tricks to reduce this cost.  If these optimizations succeed,
   * repeated calls to this function on a line will pay a one-time cost linear
   * in the length of the line, then each call pays a separate constant-time
   * cost.  If the optimizations do not succeed, this function works in time
   * linear in the length of the line.
   *
   * It's unusual for a function in *this* class to be |Unit|-templated, but
   * while this operation manages |Unit|-agnostic fields in this class and in
   * |srcCoords|, it must *perform* |Unit|-sensitive computations to fill them.
   * And this is the best place to do that.
   */
  template <typename Unit>
  uint32_t computePartialColumn(const LineToken lineToken,
                                const uint32_t offset,
                                const SourceUnits<Unit>& sourceUnits) const;
#endif  // JS_COLUMN_DIMENSION_IS_CODE_POINTS()

  /**
   * Update line/column information for the start of a new line at
   * |lineStartOffset|.
   */
  MOZ_MUST_USE MOZ_ALWAYS_INLINE bool internalUpdateLineInfoForEOL(
      uint32_t lineStartOffset);

 public:
  const Token& nextToken() const {
    MOZ_ASSERT(hasLookahead());
    return tokens[nextCursor()];
  }

  bool hasLookahead() const { return lookahead > 0; }

  void advanceCursor() { cursor_ = (cursor_ + 1) & ntokensMask; }

  void retractCursor() { cursor_ = (cursor_ - 1) & ntokensMask; }

  Token* allocateToken() {
    advanceCursor();

    Token* tp = &tokens[cursor()];
    MOZ_MAKE_MEM_UNDEFINED(tp, sizeof(*tp));

    return tp;
  }

  // Push the last scanned token back into the stream.
  void ungetToken() {
    MOZ_ASSERT(lookahead < maxLookahead);
    lookahead++;
    retractCursor();
  }

 public:
  // Compute error metadata for an error at no offset.
  void computeErrorMetadataNoOffset(ErrorMetadata* err);

  // ErrorReporter API Helpers

  // Provide minimal set of error reporting API given we cannot use
  // ErrorReportMixin here. "report" prefix is added to avoid conflict with
  // ErrorReportMixin methods in TokenStream class.
  void reportErrorNoOffset(unsigned errorNumber, ...);
  void reportErrorNoOffsetVA(unsigned errorNumber, va_list* args);

  const JS::ReadOnlyCompileOptions& options() const { return options_; }

  const char* getFilename() const { return filename_; }

 private:
#if JS_COLUMN_DIMENSION_IS_CODE_POINTS()
  static constexpr uint32_t ColumnChunkLength = 128;

  enum class UnitsType : unsigned char{
      PossiblyMultiUnit = 0,
      GuaranteedSingleUnit = 1,
  };

  class ChunkInfo {
   public:
    ChunkInfo(uint32_t col, UnitsType type)
        : unitsType_(static_cast<unsigned char>(type)) {
      memcpy(column_, &col, sizeof(col));
    }

    uint32_t column() const {
      uint32_t col;
      memcpy(&col, column_, sizeof(uint32_t));
      return col;
    }

    UnitsType unitsType() const {
      MOZ_ASSERT(unitsType_ <= 1, "unitsType_ must be 0 or 1");
      return static_cast<UnitsType>(unitsType_);
    }

    void guaranteeSingleUnits() {
      MOZ_ASSERT(unitsType() == UnitsType::PossiblyMultiUnit,
                 "should only be setting to possibly optimize from the "
                 "pessimistic case");
      unitsType_ = static_cast<unsigned char>(UnitsType::GuaranteedSingleUnit);
    }

   private:
    // Store everything in |unsigned char|s so everything packs.
    unsigned char column_[sizeof(uint32_t)];
    unsigned char unitsType_;
  };

  /**
   * Line number (of lines at least |ColumnChunkLength| code units long) to
   * a sequence of the column numbers at |ColumnChunkLength| boundaries rewound
   * (if needed) to the nearest code point boundary.
   *
   * Entries appear in this map only when a column computation of sufficient
   * distance is performed on a line, and the vectors are lazily filled as
   * greater offsets within lines require column computations.
   */
  mutable HashMap<uint32_t, Vector<ChunkInfo>> longLineColumnInfo_;
#endif  // JS_COLUMN_DIMENSION_IS_CODE_POINTS()

 protected:
  // Options used for parsing/tokenizing.
  const JS::ReadOnlyCompileOptions& options_;

  Token tokens[ntokens];  // circular token buffer
 private:
  unsigned cursor_;  // index of last parsed token
 protected:
  unsigned lookahead;      // count of lookahead tokens
  unsigned lineno;         // current line number
  TokenStreamFlags flags;  // flags -- see above
  size_t linebase;         // start of current line
  size_t
      prevLinebase;  // start of previous line;  size_t(-1) if on the first line
  const char* filename_;             // input filename or null
  UniqueTwoByteChars displayURL_;    // the user's requested source URL or null
  UniqueTwoByteChars sourceMapURL_;  // source map's filename or null

  /**
   * An array storing whether a TokenKind observed while attempting to extend
   * a valid AssignmentExpression into an even longer AssignmentExpression
   * (e.g., extending '3' to '3 + 5') will terminate it without error.
   *
   * For example, ';' always ends an AssignmentExpression because it ends a
   * Statement or declaration.  '}' always ends an AssignmentExpression
   * because it terminates BlockStatement, FunctionBody, and embedded
   * expressions in TemplateLiterals.  Therefore both entries are set to true
   * in TokenStreamAnyChars construction.
   *
   * But e.g. '+' *could* extend an AssignmentExpression, so its entry here
   * is false.  Meanwhile 'this' can't extend an AssignmentExpression, but
   * it's only valid after a line break, so its entry here must be false.
   *
   * NOTE: This array could be static, but without C99's designated
   *       initializers it's easier zeroing here and setting the true entries
   *       in the constructor body.  (Having this per-instance might also aid
   *       locality.)  Don't worry!  Initialization time for each TokenStream
   *       is trivial.  See bug 639420.
   */
  bool isExprEnding[size_t(TokenKind::Limit)] = {};  // all-false initially

  JSContext* const cx;
  bool mutedErrors;
  StrictModeGetter* strictModeGetter;  // used to test for strict mode

#if JS_COLUMN_DIMENSION_IS_CODE_POINTS()
  // Computing accurate column numbers requires at *some* point linearly
  // iterating through prior source units in the line to properly account for
  // multi-unit code points.  This is quadratic if counting happens repeatedly.
  //
  // But usually we need columns for advancing offsets through scripts.  By
  // caching the last ((line number, offset) => relative column) mapping (in
  // similar manner to how |SourceUnits::lastIndex_| is used to cache
  // (offset => line number) mappings) we can usually avoid re-iterating through
  // the common line prefix.
  //
  // Additionally, we avoid hash table lookup costs by caching the
  // |Vector<ChunkInfo>*| for the line of the last lookup.  (|nullptr| means we
  // have to look it up -- or it hasn't been created yet.)  This pointer is
  // invalidated when a lookup on a new line occurs, but as it's not a pointer
  // at literal element data, it's *not* invalidated when new entries are added
  // to such a vector.
  mutable uint32_t lineOfLastColumnComputation_ = UINT32_MAX;
  mutable Vector<ChunkInfo>* lastChunkVectorForLine_ = nullptr;
  mutable uint32_t lastOffsetOfComputedColumn_ = UINT32_MAX;
  mutable uint32_t lastComputedColumn_ = 0;
#endif  // JS_COLUMN_DIMENSION_IS_CODE_POINTS()
};

constexpr char16_t CodeUnitValue(char16_t unit) { return unit; }

constexpr uint8_t CodeUnitValue(mozilla::Utf8Unit unit) {
  return unit.toUint8();
}

template <typename Unit>
class TokenStreamCharsBase;

template <typename T>
inline bool IsLineTerminator(T) = delete;

inline bool IsLineTerminator(char32_t codePoint) {
  return codePoint == '\n' || codePoint == '\r' ||
         codePoint == unicode::LINE_SEPARATOR ||
         codePoint == unicode::PARA_SEPARATOR;
}

inline bool IsLineTerminator(char16_t unit) {
  // Every LineTerminator fits in char16_t, so this is exact.
  return IsLineTerminator(static_cast<char32_t>(unit));
}

template <typename Unit>
struct SourceUnitTraits;

template <>
struct SourceUnitTraits<char16_t> {
 public:
  static constexpr uint8_t maxUnitsLength = 2;

  static constexpr size_t lengthInUnits(char32_t codePoint) {
    return codePoint < unicode::NonBMPMin ? 1 : 2;
  }
};

template <>
struct SourceUnitTraits<mozilla::Utf8Unit> {
 public:
  static constexpr uint8_t maxUnitsLength = 4;

  static constexpr size_t lengthInUnits(char32_t codePoint) {
    return codePoint < 0x80
               ? 1
               : codePoint < 0x800 ? 2 : codePoint < 0x10000 ? 3 : 4;
  }
};

/**
 * PeekedCodePoint represents the result of peeking ahead in some source text
 * to determine the next validly-encoded code point.
 *
 * If there isn't a valid code point, then |isNone()|.
 *
 * But if there *is* a valid code point, then |!isNone()|, the code point has
 * value |codePoint()| and its length in code units is |lengthInUnits()|.
 *
 * Conceptually, this class is |Maybe<struct { char32_t v; uint8_t len; }>|.
 */
template <typename Unit>
class PeekedCodePoint final {
  char32_t codePoint_ = 0;
  uint8_t lengthInUnits_ = 0;

 private:
  using SourceUnitTraits = frontend::SourceUnitTraits<Unit>;

  PeekedCodePoint() = default;

 public:
  /**
   * Create a peeked code point with the given value and length in code
   * units.
   *
   * While the latter value is computable from the former for both UTF-8 and
   * JS's version of UTF-16, the caller likely computed a length in units in
   * the course of determining the peeked value.  Passing both here avoids
   * recomputation and lets us do a consistency-checking assertion.
   */
  PeekedCodePoint(char32_t codePoint, uint8_t lengthInUnits)
      : codePoint_(codePoint), lengthInUnits_(lengthInUnits) {
    MOZ_ASSERT(codePoint <= unicode::NonBMPMax);
    MOZ_ASSERT(lengthInUnits != 0, "bad code point length");
    MOZ_ASSERT(lengthInUnits == SourceUnitTraits::lengthInUnits(codePoint));
  }

  /** Create a PeekedCodeUnit that represents no valid code point. */
  static PeekedCodePoint none() { return PeekedCodePoint(); }

  /** True if no code point was found, false otherwise. */
  bool isNone() const { return lengthInUnits_ == 0; }

  /** If a code point was found, its value. */
  char32_t codePoint() const {
    MOZ_ASSERT(!isNone());
    return codePoint_;
  }

  /** If a code point was found, its length in code units. */
  uint8_t lengthInUnits() const {
    MOZ_ASSERT(!isNone());
    return lengthInUnits_;
  }
};

inline PeekedCodePoint<char16_t> PeekCodePoint(const char16_t* const ptr,
                                               const char16_t* const end) {
  if (MOZ_UNLIKELY(ptr >= end)) {
    return PeekedCodePoint<char16_t>::none();
  }

  char16_t lead = ptr[0];

  char32_t c;
  uint8_t len;
  if (MOZ_LIKELY(!unicode::IsLeadSurrogate(lead)) ||
      MOZ_UNLIKELY(ptr + 1 >= end || !unicode::IsTrailSurrogate(ptr[1]))) {
    c = lead;
    len = 1;
  } else {
    c = unicode::UTF16Decode(lead, ptr[1]);
    len = 2;
  }

  return PeekedCodePoint<char16_t>(c, len);
}

inline PeekedCodePoint<mozilla::Utf8Unit> PeekCodePoint(
    const mozilla::Utf8Unit* const ptr, const mozilla::Utf8Unit* const end) {
  if (MOZ_UNLIKELY(ptr >= end)) {
    return PeekedCodePoint<mozilla::Utf8Unit>::none();
  }

  const mozilla::Utf8Unit lead = ptr[0];
  if (mozilla::IsAscii(lead)) {
    return PeekedCodePoint<mozilla::Utf8Unit>(lead.toUint8(), 1);
  }

  const mozilla::Utf8Unit* afterLead = ptr + 1;
  mozilla::Maybe<char32_t> codePoint =
      mozilla::DecodeOneUtf8CodePoint(lead, &afterLead, end);
  if (codePoint.isNothing()) {
    return PeekedCodePoint<mozilla::Utf8Unit>::none();
  }

  auto len =
      mozilla::AssertedCast<uint8_t>(mozilla::PointerRangeSize(ptr, afterLead));
  MOZ_ASSERT(len <= 4);

  return PeekedCodePoint<mozilla::Utf8Unit>(codePoint.value(), len);
}

inline bool IsSingleUnitLineTerminator(mozilla::Utf8Unit unit) {
  // BEWARE: The Unicode line/paragraph separators don't fit in a single
  //         UTF-8 code unit, so this test is exact for Utf8Unit but inexact
  //         for UTF-8 as a whole.  Users must handle |unit| as start of a
  //         Unicode LineTerminator themselves!
  return unit == mozilla::Utf8Unit('\n') || unit == mozilla::Utf8Unit('\r');
}

// This is the low-level interface to the JS source code buffer.  It just gets
// raw Unicode code units -- 16-bit char16_t units of source text that are not
// (always) full code points, and 8-bit units of UTF-8 source text soon.
// TokenStreams functions are layered on top and do some extra stuff like
// converting all EOL sequences to '\n', tracking the line number, and setting
// |flags.isEOF|.  (The "raw" in "raw Unicode code units" refers to the lack of
// EOL sequence normalization.)
//
// buf[0..length-1] often represents a substring of some larger source,
// where we have only the substring in memory. The |startOffset| argument
// indicates the offset within this larger string at which our string
// begins, the offset of |buf[0]|.
template <typename Unit>
class SourceUnits {
 public:
  SourceUnits(const Unit* units, size_t length, size_t startOffset)
      : base_(units),
        startOffset_(startOffset),
        limit_(units + length),
        ptr(units) {}

  bool atStart() const {
    MOZ_ASSERT(!isPoisoned(), "shouldn't be using if poisoned");
    return ptr == base_;
  }

  bool atEnd() const {
    MOZ_ASSERT(!isPoisoned(), "shouldn't be using if poisoned");
    MOZ_ASSERT(ptr <= limit_, "shouldn't have overrun");
    return ptr >= limit_;
  }

  size_t remaining() const {
    MOZ_ASSERT(!isPoisoned(),
               "can't get a count of remaining code units if poisoned");
    return mozilla::PointerRangeSize(ptr, limit_);
  }

  size_t startOffset() const { return startOffset_; }

  size_t offset() const {
    return startOffset_ + mozilla::PointerRangeSize(base_, ptr);
  }

  const Unit* codeUnitPtrAt(size_t offset) const {
    MOZ_ASSERT(!isPoisoned(), "shouldn't be using if poisoned");
    MOZ_ASSERT(startOffset_ <= offset);
    MOZ_ASSERT(offset - startOffset_ <=
               mozilla::PointerRangeSize(base_, limit_));
    return base_ + (offset - startOffset_);
  }

  const Unit* current() const { return ptr; }

  const Unit* limit() const { return limit_; }

  Unit previousCodeUnit() {
    MOZ_ASSERT(!isPoisoned(), "can't get previous code unit if poisoned");
    MOZ_ASSERT(!atStart(), "must have a previous code unit to get");
    return *(ptr - 1);
  }

  Unit getCodeUnit() {
    return *ptr++;  // this will nullptr-crash if poisoned
  }

  Unit peekCodeUnit() const {
    return *ptr;  // this will nullptr-crash if poisoned
  }

  /**
   * Determine the next code point in source text.  The code point is not
   * normalized: '\r', '\n', '\u2028', and '\u2029' are returned literally.
   * If there is no next code point because |atEnd()|, or if an encoding
   * error is encountered, return a |PeekedCodePoint| that |isNone()|.
   *
   * This function does not report errors: code that attempts to get the next
   * code point must report any error.
   *
   * If a next code point is found, it may be consumed by passing it to
   * |consumeKnownCodePoint|.
   */
  PeekedCodePoint<Unit> peekCodePoint() const {
    return PeekCodePoint(ptr, limit_);
  }

 private:
#ifdef DEBUG
  void assertNextCodePoint(const PeekedCodePoint<Unit>& peeked);
#endif

 public:
  /**
   * Consume a peeked code point that |!isNone()|.
   *
   * This call DOES NOT UPDATE LINE-STATUS.  You may need to call
   * |updateLineInfoForEOL()| and |updateFlagsForEOL()| if this consumes a
   * LineTerminator.  Note that if this consumes '\r', you also must consume
   * an optional '\n' (i.e. a full LineTerminatorSequence) before doing so.
   */
  void consumeKnownCodePoint(const PeekedCodePoint<Unit>& peeked) {
    MOZ_ASSERT(!peeked.isNone());
    MOZ_ASSERT(peeked.lengthInUnits() <= remaining());

#ifdef DEBUG
    assertNextCodePoint(peeked);
#endif

    ptr += peeked.lengthInUnits();
  }

  /** Match |n| hexadecimal digits and store their value in |*out|. */
  bool matchHexDigits(uint8_t n, char16_t* out) {
    MOZ_ASSERT(!isPoisoned(), "shouldn't peek into poisoned SourceUnits");
    MOZ_ASSERT(n <= 4, "hexdigit value can't overflow char16_t");
    if (n > remaining()) {
      return false;
    }

    char16_t v = 0;
    for (uint8_t i = 0; i < n; i++) {
      auto unit = CodeUnitValue(ptr[i]);
      if (!mozilla::IsAsciiHexDigit(unit)) {
        return false;
      }

      v = (v << 4) | mozilla::AsciiAlphanumericToNumber(unit);
    }

    *out = v;
    ptr += n;
    return true;
  }

  bool matchCodeUnits(const char* chars, uint8_t length) {
    MOZ_ASSERT(!isPoisoned(), "shouldn't match into poisoned SourceUnits");
    if (length > remaining()) {
      return false;
    }

    const Unit* start = ptr;
    const Unit* end = ptr + length;
    while (ptr < end) {
      if (*ptr++ != Unit(*chars++)) {
        ptr = start;
        return false;
      }
    }

    return true;
  }

  void skipCodeUnits(uint32_t n) {
    MOZ_ASSERT(!isPoisoned(), "shouldn't use poisoned SourceUnits");
    MOZ_ASSERT(n <= remaining(), "shouldn't skip beyond end of SourceUnits");
    ptr += n;
  }

  void unskipCodeUnits(uint32_t n) {
    MOZ_ASSERT(!isPoisoned(), "shouldn't use poisoned SourceUnits");
    MOZ_ASSERT(n <= mozilla::PointerRangeSize(base_, ptr),
               "shouldn't unskip beyond start of SourceUnits");
    ptr -= n;
  }

 private:
  friend class TokenStreamCharsBase<Unit>;

  bool internalMatchCodeUnit(Unit c) {
    MOZ_ASSERT(!isPoisoned(), "shouldn't use poisoned SourceUnits");
    if (MOZ_LIKELY(!atEnd()) && *ptr == c) {
      ptr++;
      return true;
    }
    return false;
  }

 public:
  void consumeKnownCodeUnit(Unit c) {
    MOZ_ASSERT(!isPoisoned(), "shouldn't use poisoned SourceUnits");
    MOZ_ASSERT(*ptr == c, "consuming the wrong code unit");
    ptr++;
  }

  /**
   * Unget the '\n' (CR) that precedes a '\n' (LF), when ungetting a line
   * terminator that's a full "\r\n" sequence.  If the prior code unit isn't
   * '\r', do nothing.
   */
  void ungetOptionalCRBeforeLF() {
    MOZ_ASSERT(!isPoisoned(),
               "shouldn't unget a '\\r' from poisoned SourceUnits");
    MOZ_ASSERT(*ptr == Unit('\n'),
               "function should only be called when a '\\n' was just "
               "ungotten, and any '\\r' preceding it must also be "
               "ungotten");
    if (*(ptr - 1) == Unit('\r')) {
      ptr--;
    }
  }

  /** Unget U+2028 LINE SEPARATOR or U+2029 PARAGRAPH SEPARATOR. */
  inline void ungetLineOrParagraphSeparator();

  void ungetCodeUnit() {
    MOZ_ASSERT(!isPoisoned(), "can't unget from poisoned units");
    MOZ_ASSERT(!atStart(), "can't unget if currently at start");
    ptr--;
  }

  const Unit* addressOfNextCodeUnit(bool allowPoisoned = false) const {
    MOZ_ASSERT_IF(!allowPoisoned, !isPoisoned());
    return ptr;
  }

  // Use this with caution!
  void setAddressOfNextCodeUnit(const Unit* a, bool allowPoisoned = false) {
    MOZ_ASSERT_IF(!allowPoisoned, a);
    ptr = a;
  }

  // Poison the SourceUnits so they can't be accessed again.
  void poisonInDebug() {
#ifdef DEBUG
    ptr = nullptr;
#endif
  }

 private:
  bool isPoisoned() const {
#ifdef DEBUG
    // |ptr| can be null for unpoisoned SourceUnits if this was initialized with
    // |units == nullptr| and |length == 0|.  In that case, for lack of any
    // better options, consider this to not be poisoned.
    return ptr == nullptr && ptr != limit_;
#else
    return false;
#endif
  }

 public:
  /**
   * Consume the rest of a single-line comment (but not the EOL/EOF that
   * terminates it).
   *
   * If an encoding error is encountered -- possible only for UTF-8 because
   * JavaScript's conception of UTF-16 encompasses any sequence of 16-bit
   * code units -- valid code points prior to the encoding error are consumed
   * and subsequent invalid code units are not consumed.  For example, given
   * these UTF-8 code units:
   *
   *   'B'   'A'  'D'  ':'   <bad code unit sequence>
   *   0x42  0x41 0x44 0x3A  0xD0 0x00 ...
   *
   * the first four code units are consumed, but 0xD0 and 0x00 are not
   * consumed because 0xD0 encodes a two-byte lead unit but 0x00 is not a
   * valid trailing code unit.
   *
   * It is expected that the caller will report such an encoding error when
   * it attempts to consume the next code point.
   */
  void consumeRestOfSingleLineComment();

  /**
   * The maximum radius of code around the location of an error that should
   * be included in a syntax error message -- this many code units to either
   * side.  The resulting window of data is then accordinngly trimmed so that
   * the window contains only validly-encoded data.
   *
   * Because this number is the same for both UTF-8 and UTF-16, windows in
   * UTF-8 may contain fewer code points than windows in UTF-16.  As we only
   * use this for error messages, we don't particularly care.
   */
  static constexpr size_t WindowRadius = ErrorMetadata::lineOfContextRadius;

  /**
   * From absolute offset |offset|, search backward to find an absolute
   * offset within source text, no further than |WindowRadius| code units
   * away from |offset|, such that all code points from that offset to
   * |offset| are valid, non-LineTerminator code points.
   */
  size_t findWindowStart(size_t offset) const;

  /**
   * From absolute offset |offset|, find an absolute offset within source
   * text, no further than |WindowRadius| code units away from |offset|, such
   * that all code units from |offset| to that offset are valid,
   * non-LineTerminator code points.
   */
  size_t findWindowEnd(size_t offset) const;

  /**
   * Given a |window| of |encodingSpecificWindowLength| units encoding valid
   * Unicode text, with index |encodingSpecificTokenOffset| indicating a
   * particular code point boundary in |window|, compute the corresponding
   * token offset and length if |window| were encoded in UTF-16.  For
   * example:
   *
   *   // U+03C0 GREEK SMALL LETTER PI is encoded as 0xCF 0x80.
   *   const Utf8Unit* encodedWindow =
   *     reinterpret_cast<const Utf8Unit*>(u8"ππππ = @ FAIL");
   *   size_t encodedTokenOffset = 11; // 2 * 4 + ' = '.length
   *   size_t encodedWindowLength = 17; // 2 * 4 + ' = @ FAIL'.length
   *   size_t utf16Offset, utf16Length;
   *   computeWindowOffsetAndLength(encodedWindow,
   *                                encodedTokenOffset, &utf16Offset,
   *                                encodedWindowLength, &utf16Length);
   *   MOZ_ASSERT(utf16Offset == 7);
   *   MOZ_ASSERT(utf16Length = 13);
   *
   * This function asserts if called for UTF-16: the sole caller can avoid
   * computing UTF-16 offsets when they're definitely the same as the encoded
   * offsets.
   */
  inline void computeWindowOffsetAndLength(const Unit* encodeWindow,
                                           size_t encodingSpecificTokenOffset,
                                           size_t* utf16TokenOffset,
                                           size_t encodingSpecificWindowLength,
                                           size_t* utf16WindowLength);

 private:
  /** Base of buffer. */
  const Unit* base_;

  /** Offset of base_[0]. */
  uint32_t startOffset_;

  /** Limit for quick bounds check. */
  const Unit* limit_;

  /** Next char to get. */
  const Unit* ptr;
};

template <>
inline void SourceUnits<char16_t>::ungetLineOrParagraphSeparator() {
#ifdef DEBUG
  char16_t prev = previousCodeUnit();
#endif
  MOZ_ASSERT(prev == unicode::LINE_SEPARATOR ||
             prev == unicode::PARA_SEPARATOR);

  ungetCodeUnit();
}

template <>
inline void SourceUnits<mozilla::Utf8Unit>::ungetLineOrParagraphSeparator() {
  unskipCodeUnits(3);

  MOZ_ASSERT(ptr[0].toUint8() == 0xE2);
  MOZ_ASSERT(ptr[1].toUint8() == 0x80);

#ifdef DEBUG
  uint8_t last = ptr[2].toUint8();
#endif
  MOZ_ASSERT(last == 0xA8 || last == 0xA9);
}

class TokenStreamCharsShared {
  // Using char16_t (not Unit) is a simplifying decision that hopefully
  // eliminates the need for a UTF-8 regular expression parser and makes
  // |copyCharBufferTo| markedly simpler.
  using CharBuffer = Vector<char16_t, 32>;

 protected:
  /**
   * Buffer transiently used to store sequences of identifier or string code
   * points when such can't be directly processed from the original source
   * text (e.g. because it contains escapes).
   */
  CharBuffer charBuffer;

 protected:
  explicit TokenStreamCharsShared(JSContext* cx) : charBuffer(cx) {}

  MOZ_MUST_USE bool appendCodePointToCharBuffer(uint32_t codePoint);

  MOZ_MUST_USE bool copyCharBufferTo(
      JSContext* cx, UniquePtr<char16_t[], JS::FreePolicy>* destination);

  /**
   * Determine whether a code unit constitutes a complete ASCII code point.
   * (The code point's exact value might not be used, however, if subsequent
   * code observes that |unit| is part of a LineTerminatorSequence.)
   */
  static constexpr MOZ_ALWAYS_INLINE MOZ_MUST_USE bool isAsciiCodePoint(
      int32_t unit) {
    return mozilla::IsAscii(unit);
  }

  JSAtom* drainCharBufferIntoAtom(JSContext* cx) {
    JSAtom* atom = AtomizeChars(cx, charBuffer.begin(), charBuffer.length());
    charBuffer.clear();
    return atom;
  }

 public:
  CharBuffer& getCharBuffer() { return charBuffer; }
};

inline mozilla::Span<const char> ToCharSpan(
    mozilla::Span<const mozilla::Utf8Unit> codeUnits) {
  static_assert(alignof(char) == alignof(mozilla::Utf8Unit),
                "must have equal alignment to reinterpret_cast<>");
  static_assert(sizeof(char) == sizeof(mozilla::Utf8Unit),
                "must have equal size to reinterpret_cast<>");

  // This cast is safe for two reasons.
  //
  // First, per C++11 [basic.lval]p10 it is permitted to access any object's
  // memory through |char|.
  //
  // Second, Utf8Unit *contains* a |char|.  Examining that memory as |char|
  // is simply, per C++11 [basic.lval]p10, to access the memory according to
  // the dynamic type of the object: essentially trivially safe.
  return mozilla::MakeSpan(reinterpret_cast<const char*>(codeUnits.data()),
                           codeUnits.size());
}

template <typename Unit>
class TokenStreamCharsBase : public TokenStreamCharsShared {
 protected:
  TokenStreamCharsBase(JSContext* cx, const Unit* units, size_t length,
                       size_t startOffset);

  /**
   * Convert a non-EOF code unit returned by |getCodeUnit()| or
   * |peekCodeUnit()| to a Unit code unit.
   */
  inline Unit toUnit(int32_t codeUnitValue);

  void ungetCodeUnit(int32_t c) {
    if (c == EOF) {
      return;
    }

    sourceUnits.ungetCodeUnit();
  }

  static MOZ_ALWAYS_INLINE JSAtom* atomizeSourceChars(
      JSContext* cx, mozilla::Span<const Unit> units);

  using SourceUnits = frontend::SourceUnits<Unit>;

  /**
   * Try to match a non-LineTerminator ASCII code point.  Return true iff it
   * was matched.
   */
  bool matchCodeUnit(char expect) {
    MOZ_ASSERT(mozilla::IsAscii(expect));
    MOZ_ASSERT(expect != '\r');
    MOZ_ASSERT(expect != '\n');
    return this->sourceUnits.internalMatchCodeUnit(Unit(expect));
  }

  /**
   * Try to match an ASCII LineTerminator code point.  Return true iff it was
   * matched.
   */
  bool matchLineTerminator(char expect) {
    MOZ_ASSERT(expect == '\r' || expect == '\n');
    return this->sourceUnits.internalMatchCodeUnit(Unit(expect));
  }

  template <typename T>
  bool matchCodeUnit(T) = delete;
  template <typename T>
  bool matchLineTerminator(T) = delete;

  int32_t peekCodeUnit() {
    return MOZ_LIKELY(!sourceUnits.atEnd())
               ? CodeUnitValue(sourceUnits.peekCodeUnit())
               : EOF;
  }

  /** Consume a known, non-EOF code unit. */
  inline void consumeKnownCodeUnit(int32_t unit);

  // Forbid accidental calls to consumeKnownCodeUnit *not* with the single
  // unit-or-EOF type.  Unit should use SourceUnits::consumeKnownCodeUnit;
  // CodeUnitValue() results should go through toUnit(), or better yet just
  // use the original Unit.
  template <typename T>
  inline void consumeKnownCodeUnit(T) = delete;

  /**
   * Accumulate the provided range of already-validated text (valid UTF-8, or
   * anything if Unit is char16_t because JS allows lone surrogates) into
   * |charBuffer|.  Normalize '\r', '\n', and "\r\n" into '\n'.
   */
  MOZ_MUST_USE bool fillCharBufferFromSourceNormalizingAsciiLineBreaks(
      const Unit* cur, const Unit* end);

  /**
   * Add a null-terminated line of context to error information, for the line
   * in |sourceUnits| that contains |offset|.  Also record the window's
   * length and the offset of the error in the window.  (Don't bother adding
   * a line of context if it would be empty.)
   *
   * The window will contain no LineTerminators of any kind, and it will not
   * extend more than |SourceUnits::WindowRadius| to either side of |offset|,
   * nor into the previous or next lines.
   *
   * This function is quite internal, and you probably should be calling one
   * of its existing callers instead.
   */
  MOZ_MUST_USE bool addLineOfContext(ErrorMetadata* err, uint32_t offset);

 protected:
  /** Code units in the source code being tokenized. */
  SourceUnits sourceUnits;
};

template <>
inline char16_t TokenStreamCharsBase<char16_t>::toUnit(int32_t codeUnitValue) {
  MOZ_ASSERT(codeUnitValue != EOF, "EOF is not a Unit");
  return mozilla::AssertedCast<char16_t>(codeUnitValue);
}

template <>
inline mozilla::Utf8Unit TokenStreamCharsBase<mozilla::Utf8Unit>::toUnit(
    int32_t value) {
  MOZ_ASSERT(value != EOF, "EOF is not a Unit");
  return mozilla::Utf8Unit(mozilla::AssertedCast<unsigned char>(value));
}

template <typename Unit>
inline void TokenStreamCharsBase<Unit>::consumeKnownCodeUnit(int32_t unit) {
  sourceUnits.consumeKnownCodeUnit(toUnit(unit));
}

template <>
/* static */ MOZ_ALWAYS_INLINE JSAtom*
TokenStreamCharsBase<char16_t>::atomizeSourceChars(
    JSContext* cx, mozilla::Span<const char16_t> units) {
  return AtomizeChars(cx, units.data(), units.size());
}

template <>
/* static */ MOZ_ALWAYS_INLINE JSAtom*
TokenStreamCharsBase<mozilla::Utf8Unit>::atomizeSourceChars(
    JSContext* cx, mozilla::Span<const mozilla::Utf8Unit> units) {
  auto chars = ToCharSpan(units);
  return AtomizeUTF8Chars(cx, chars.data(), chars.size());
}

template <typename Unit>
class SpecializedTokenStreamCharsBase;

template <>
class SpecializedTokenStreamCharsBase<char16_t>
    : public TokenStreamCharsBase<char16_t> {
  using CharsBase = TokenStreamCharsBase<char16_t>;

 protected:
  using TokenStreamCharsShared::isAsciiCodePoint;
  // Deliberately don't |using| |sourceUnits| because of bug 1472569.  :-(

  using typename CharsBase::SourceUnits;

 protected:
  // These APIs are only usable by UTF-16-specific code.

  /**
   * Given |lead| already consumed, consume and return the code point encoded
   * starting from it.  Infallible because lone surrogates in JS encode a
   * "code point" of the same value.
   */
  char32_t infallibleGetNonAsciiCodePointDontNormalize(char16_t lead) {
    MOZ_ASSERT(!isAsciiCodePoint(lead));
    MOZ_ASSERT(this->sourceUnits.previousCodeUnit() == lead);

    // Handle single-unit code points and lone trailing surrogates.
    if (MOZ_LIKELY(!unicode::IsLeadSurrogate(lead)) ||
        // Or handle lead surrogates not paired with trailing surrogates.
        MOZ_UNLIKELY(
            this->sourceUnits.atEnd() ||
            !unicode::IsTrailSurrogate(this->sourceUnits.peekCodeUnit()))) {
      return lead;
    }

    // Otherwise it's a multi-unit code point.
    return unicode::UTF16Decode(lead, this->sourceUnits.getCodeUnit());
  }

 protected:
  // These APIs are in both SpecializedTokenStreamCharsBase specializations
  // and so are usable in subclasses no matter what Unit is.

  using CharsBase::CharsBase;
};

template <>
class SpecializedTokenStreamCharsBase<mozilla::Utf8Unit>
    : public TokenStreamCharsBase<mozilla::Utf8Unit> {
  using CharsBase = TokenStreamCharsBase<mozilla::Utf8Unit>;

 protected:
  // Deliberately don't |using| |sourceUnits| because of bug 1472569.  :-(

 protected:
  // These APIs are only usable by UTF-8-specific code.

  using typename CharsBase::SourceUnits;

  /**
   * A mutable iterator-wrapper around |SourceUnits| that translates
   * operators to calls to |SourceUnits::getCodeUnit()| and similar.
   *
   * This class is expected to be used in concert with |SourceUnitsEnd|.
   */
  class SourceUnitsIterator {
    SourceUnits& sourceUnits_;
#ifdef DEBUG
    // In iterator copies created by the post-increment operator, a pointer
    // at the next source text code unit when the post-increment operator
    // was called, cleared when the iterator is dereferenced.
    mutable mozilla::Maybe<const mozilla::Utf8Unit*>
        currentBeforePostIncrement_;
#endif

   public:
    explicit SourceUnitsIterator(SourceUnits& sourceUnits)
        : sourceUnits_(sourceUnits) {}

    mozilla::Utf8Unit operator*() const {
      // operator* is expected to get the *next* value from an iterator
      // not pointing at the end of the underlying range.  However, the
      // sole use of this is in the context of an expression of the form
      // |*iter++|, that performed the |sourceUnits_.getCodeUnit()| in
      // the |operator++(int)| below -- so dereferencing acts on a
      // |sourceUnits_| already advanced.  Therefore the correct unit to
      // return is the previous one.
      MOZ_ASSERT(currentBeforePostIncrement_.value() + 1 ==
                 sourceUnits_.current());
#ifdef DEBUG
      currentBeforePostIncrement_.reset();
#endif
      return sourceUnits_.previousCodeUnit();
    }

    SourceUnitsIterator operator++(int) {
      MOZ_ASSERT(currentBeforePostIncrement_.isNothing(),
                 "the only valid operation on a post-incremented "
                 "iterator is dereferencing a single time");

      SourceUnitsIterator copy = *this;
#ifdef DEBUG
      copy.currentBeforePostIncrement_.emplace(sourceUnits_.current());
#endif

      sourceUnits_.getCodeUnit();
      return copy;
    }

    void operator-=(size_t n) {
      MOZ_ASSERT(currentBeforePostIncrement_.isNothing(),
                 "the only valid operation on a post-incremented "
                 "iterator is dereferencing a single time");
      sourceUnits_.unskipCodeUnits(n);
    }

    mozilla::Utf8Unit operator[](ptrdiff_t index) {
      MOZ_ASSERT(currentBeforePostIncrement_.isNothing(),
                 "the only valid operation on a post-incremented "
                 "iterator is dereferencing a single time");
      MOZ_ASSERT(index == -1,
                 "must only be called to verify the value of the "
                 "previous code unit");
      return sourceUnits_.previousCodeUnit();
    }

    size_t remaining() const {
      MOZ_ASSERT(currentBeforePostIncrement_.isNothing(),
                 "the only valid operation on a post-incremented "
                 "iterator is dereferencing a single time");
      return sourceUnits_.remaining();
    }
  };

  /** A sentinel representing the end of |SourceUnits| data. */
  class SourceUnitsEnd {};

  friend inline size_t operator-(const SourceUnitsEnd& aEnd,
                                 const SourceUnitsIterator& aIter);

 protected:
  // These APIs are in both SpecializedTokenStreamCharsBase specializations
  // and so are usable in subclasses no matter what Unit is.

  using CharsBase::CharsBase;
};

inline size_t operator-(const SpecializedTokenStreamCharsBase<
                            mozilla::Utf8Unit>::SourceUnitsEnd& aEnd,
                        const SpecializedTokenStreamCharsBase<
                            mozilla::Utf8Unit>::SourceUnitsIterator& aIter) {
  return aIter.remaining();
}

/** A small class encapsulating computation of the start-offset of a Token. */
class TokenStart {
  uint32_t startOffset_;

 public:
  /**
   * Compute a starting offset that is the current offset of |sourceUnits|,
   * offset by |adjust|.  (For example, |adjust| of -1 indicates the code
   * unit one backwards from |sourceUnits|'s current offset.)
   */
  template <class SourceUnits>
  TokenStart(const SourceUnits& sourceUnits, ptrdiff_t adjust)
      : startOffset_(sourceUnits.offset() + adjust) {}

  TokenStart(const TokenStart&) = default;

  uint32_t offset() const { return startOffset_; }
};

template <typename Unit, class AnyCharsAccess>
class GeneralTokenStreamChars : public SpecializedTokenStreamCharsBase<Unit> {
  using CharsBase = TokenStreamCharsBase<Unit>;
  using SpecializedCharsBase = SpecializedTokenStreamCharsBase<Unit>;

  using LineToken = TokenStreamAnyChars::LineToken;

 private:
  Token* newTokenInternal(TokenKind kind, TokenStart start, TokenKind* out);

  /**
   * Allocates a new Token from the given offset to the current offset,
   * ascribes it the given kind, and sets |*out| to that kind.
   */
  Token* newToken(TokenKind kind, TokenStart start,
                  TokenStreamShared::Modifier modifier, TokenKind* out) {
    Token* token = newTokenInternal(kind, start, out);

#ifdef DEBUG
    // Save the modifier used to get this token, so that if an ungetToken()
    // occurs and then the token is re-gotten (or peeked, etc.), we can
    // assert both gets used compatible modifiers.
    token->modifier = modifier;
#endif

    return token;
  }

  uint32_t matchUnicodeEscape(uint32_t* codePoint);
  uint32_t matchExtendedUnicodeEscape(uint32_t* codePoint);

 protected:
  using CharsBase::addLineOfContext;
  using CharsBase::fillCharBufferFromSourceNormalizingAsciiLineBreaks;
  using CharsBase::matchCodeUnit;
  using CharsBase::matchLineTerminator;
  using TokenStreamCharsShared::drainCharBufferIntoAtom;
  using TokenStreamCharsShared::isAsciiCodePoint;
  // Deliberately don't |using CharsBase::sourceUnits| because of bug 1472569.
  // :-(
  using CharsBase::toUnit;

  using typename CharsBase::SourceUnits;

 protected:
  using SpecializedCharsBase::SpecializedCharsBase;

  TokenStreamAnyChars& anyCharsAccess() {
    return AnyCharsAccess::anyChars(this);
  }

  const TokenStreamAnyChars& anyCharsAccess() const {
    return AnyCharsAccess::anyChars(this);
  }

  using TokenStreamSpecific =
      frontend::TokenStreamSpecific<Unit, AnyCharsAccess>;

  TokenStreamSpecific* asSpecific() {
    static_assert(
        mozilla::IsBaseOf<GeneralTokenStreamChars, TokenStreamSpecific>::value,
        "static_cast below presumes an inheritance relationship");

    return static_cast<TokenStreamSpecific*>(this);
  }

 protected:
  /**
   * Compute the column number in Unicode code points of the absolute |offset|
   * within source text on the line corresponding to |lineToken|.
   *
   * |offset| must be a code point boundary, preceded only by validly-encoded
   * source units.  (It doesn't have to be *followed* by valid source units.)
   */
  uint32_t computeColumn(LineToken lineToken, uint32_t offset) const;
  void computeLineAndColumn(uint32_t offset, uint32_t* line,
                            uint32_t* column) const;

  /**
   * Fill in |err| completely, except for line-of-context information.
   *
   * Return true if the caller can compute a line of context from the token
   * stream.  Otherwise return false.
   */
  MOZ_MUST_USE bool fillExceptingContext(ErrorMetadata* err, uint32_t offset) {
    if (anyCharsAccess().fillExceptingContext(err, offset)) {
      computeLineAndColumn(offset, &err->lineNumber, &err->columnNumber);
      return true;
    }
    return false;
  }

  void newSimpleToken(TokenKind kind, TokenStart start,
                      TokenStreamShared::Modifier modifier, TokenKind* out) {
    newToken(kind, start, modifier, out);
  }

  void newNumberToken(double dval, DecimalPoint decimalPoint, TokenStart start,
                      TokenStreamShared::Modifier modifier, TokenKind* out) {
    Token* token = newToken(TokenKind::Number, start, modifier, out);
    token->setNumber(dval, decimalPoint);
  }

  void newBigIntToken(TokenStart start, TokenStreamShared::Modifier modifier,
                      TokenKind* out) {
    newToken(TokenKind::BigInt, start, modifier, out);
  }

  void newAtomToken(TokenKind kind, JSAtom* atom, TokenStart start,
                    TokenStreamShared::Modifier modifier, TokenKind* out) {
    MOZ_ASSERT(kind == TokenKind::String || kind == TokenKind::TemplateHead ||
               kind == TokenKind::NoSubsTemplate);

    Token* token = newToken(kind, start, modifier, out);
    token->setAtom(atom);
  }

  void newNameToken(PropertyName* name, TokenStart start,
                    TokenStreamShared::Modifier modifier, TokenKind* out) {
    Token* token = newToken(TokenKind::Name, start, modifier, out);
    token->setName(name);
  }

  void newPrivateNameToken(PropertyName* name, TokenStart start,
                           TokenStreamShared::Modifier modifier,
                           TokenKind* out) {
    Token* token = newToken(TokenKind::PrivateName, start, modifier, out);
    token->setName(name);
  }

  void newRegExpToken(JS::RegExpFlags reflags, TokenStart start,
                      TokenKind* out) {
    Token* token = newToken(TokenKind::RegExp, start,
                            TokenStreamShared::SlashIsRegExp, out);
    token->setRegExpFlags(reflags);
  }

  MOZ_COLD bool badToken();

  /**
   * Get the next code unit -- the next numeric sub-unit of source text,
   * possibly smaller than a full code point -- without updating line/column
   * counters or consuming LineTerminatorSequences.
   *
   * Because of these limitations, only use this if (a) the resulting code
   * unit is guaranteed to be ungotten (by ungetCodeUnit()) if it's an EOL,
   * and (b) the line-related state (lineno, linebase) is not used before
   * it's ungotten.
   */
  int32_t getCodeUnit() {
    if (MOZ_LIKELY(!this->sourceUnits.atEnd())) {
      return CodeUnitValue(this->sourceUnits.getCodeUnit());
    }

    anyCharsAccess().flags.isEOF = true;
    return EOF;
  }

  void ungetCodeUnit(int32_t c) {
    MOZ_ASSERT_IF(c == EOF, anyCharsAccess().flags.isEOF);

    CharsBase::ungetCodeUnit(c);
  }

  /**
   * Given a just-consumed ASCII code unit/point |lead|, consume a full code
   * point or LineTerminatorSequence (normalizing it to '\n') and store it in
   * |*codePoint|.  Return true on success, otherwise return false and leave
   * |*codePoint| undefined on failure.
   *
   * If a LineTerminatorSequence was consumed, also update line/column info.
   *
   * This may change the current |sourceUnits| offset.
   */
  MOZ_MUST_USE bool getFullAsciiCodePoint(int32_t lead, int32_t* codePoint) {
    MOZ_ASSERT(isAsciiCodePoint(lead),
               "non-ASCII code units must be handled separately");
    MOZ_ASSERT(toUnit(lead) == this->sourceUnits.previousCodeUnit(),
               "getFullAsciiCodePoint called incorrectly");

    if (MOZ_UNLIKELY(lead == '\r')) {
      matchLineTerminator('\n');
    } else if (MOZ_LIKELY(lead != '\n')) {
      *codePoint = lead;
      return true;
    }

    *codePoint = '\n';
    bool ok = updateLineInfoForEOL();
    if (!ok) {
#ifdef DEBUG
      *codePoint = EOF;  // sentinel value to hopefully cause errors
#endif
      MOZ_MAKE_MEM_UNDEFINED(codePoint, sizeof(*codePoint));
    }
    return ok;
  }

  MOZ_MUST_USE MOZ_ALWAYS_INLINE bool updateLineInfoForEOL() {
    return anyCharsAccess().internalUpdateLineInfoForEOL(
        this->sourceUnits.offset());
  }

  uint32_t matchUnicodeEscapeIdStart(uint32_t* codePoint);
  bool matchUnicodeEscapeIdent(uint32_t* codePoint);
  bool matchIdentifierStart();

  /**
   * If possible, compute a line of context for an otherwise-filled-in |err|
   * at the given offset in this token stream.
   *
   * This function is very-internal: almost certainly you should use one of
   * its callers instead.  It basically exists only to make those callers
   * more readable.
   */
  MOZ_MUST_USE bool internalComputeLineOfContext(ErrorMetadata* err,
                                                 uint32_t offset) {
    // We only have line-start information for the current line.  If the error
    // is on a different line, we can't easily provide context.  (This means
    // any error in a multi-line token, e.g. an unterminated multiline string
    // literal, won't have context.)
    if (err->lineNumber != anyCharsAccess().lineno) {
      return true;
    }

    return addLineOfContext(err, offset);
  }

 public:
  /**
   * Consume any hashbang comment at the start of a Script or Module, if one is
   * present.  Stops consuming just before any terminating LineTerminator or
   * before an encoding error is encountered.
   */
  void consumeOptionalHashbangComment();

  JSAtom* getRawTemplateStringAtom() {
    TokenStreamAnyChars& anyChars = anyCharsAccess();

    MOZ_ASSERT(anyChars.currentToken().type == TokenKind::TemplateHead ||
               anyChars.currentToken().type == TokenKind::NoSubsTemplate);
    const Unit* cur =
        this->sourceUnits.codeUnitPtrAt(anyChars.currentToken().pos.begin + 1);
    const Unit* end;
    if (anyChars.currentToken().type == TokenKind::TemplateHead) {
      // Of the form    |`...${|   or   |}...${|
      end =
          this->sourceUnits.codeUnitPtrAt(anyChars.currentToken().pos.end - 2);
    } else {
      // NO_SUBS_TEMPLATE is of the form   |`...`|   or   |}...`|
      end =
          this->sourceUnits.codeUnitPtrAt(anyChars.currentToken().pos.end - 1);
    }

    // Template literals normalize only '\r' and "\r\n" to '\n'; Unicode
    // separators don't need special handling.
    // https://tc39.github.io/ecma262/#sec-static-semantics-tv-and-trv
    if (!fillCharBufferFromSourceNormalizingAsciiLineBreaks(cur, end)) {
      return nullptr;
    }

    return drainCharBufferIntoAtom(anyChars.cx);
  }
};

template <typename Unit, class AnyCharsAccess>
class TokenStreamChars;

template <class AnyCharsAccess>
class TokenStreamChars<char16_t, AnyCharsAccess>
    : public GeneralTokenStreamChars<char16_t, AnyCharsAccess> {
  using CharsBase = TokenStreamCharsBase<char16_t>;
  using SpecializedCharsBase = SpecializedTokenStreamCharsBase<char16_t>;
  using GeneralCharsBase = GeneralTokenStreamChars<char16_t, AnyCharsAccess>;
  using Self = TokenStreamChars<char16_t, AnyCharsAccess>;

  using GeneralCharsBase::asSpecific;

  using typename GeneralCharsBase::TokenStreamSpecific;

 protected:
  using CharsBase::matchLineTerminator;
  using GeneralCharsBase::anyCharsAccess;
  using GeneralCharsBase::getCodeUnit;
  using SpecializedCharsBase::infallibleGetNonAsciiCodePointDontNormalize;
  using TokenStreamCharsShared::isAsciiCodePoint;
  // Deliberately don't |using| |sourceUnits| because of bug 1472569.  :-(
  using GeneralCharsBase::ungetCodeUnit;
  using GeneralCharsBase::updateLineInfoForEOL;

 protected:
  using GeneralCharsBase::GeneralCharsBase;

  /**
   * Given the non-ASCII |lead| code unit just consumed, consume and return a
   * complete non-ASCII code point.  Line/column updates are not performed,
   * and line breaks are returned as-is without normalization.
   */
  MOZ_MUST_USE bool getNonAsciiCodePointDontNormalize(char16_t lead,
                                                      char32_t* codePoint) {
    // There are no encoding errors in 16-bit JS, so implement this so that
    // the compiler knows it, too.
    *codePoint = infallibleGetNonAsciiCodePointDontNormalize(lead);
    return true;
  }

  /**
   * Given a just-consumed non-ASCII code unit |lead| (which may also be a
   * full code point, for UTF-16), consume a full code point or
   * LineTerminatorSequence (normalizing it to '\n') and store it in
   * |*codePoint|.  Return true on success, otherwise return false and leave
   * |*codePoint| undefined on failure.
   *
   * If a LineTerminatorSequence was consumed, also update line/column info.
   *
   * This may change the current |sourceUnits| offset.
   */
  MOZ_MUST_USE bool getNonAsciiCodePoint(int32_t lead, int32_t* codePoint);
};

template <class AnyCharsAccess>
class TokenStreamChars<mozilla::Utf8Unit, AnyCharsAccess>
    : public GeneralTokenStreamChars<mozilla::Utf8Unit, AnyCharsAccess> {
  using CharsBase = TokenStreamCharsBase<mozilla::Utf8Unit>;
  using SpecializedCharsBase =
      SpecializedTokenStreamCharsBase<mozilla::Utf8Unit>;
  using GeneralCharsBase =
      GeneralTokenStreamChars<mozilla::Utf8Unit, AnyCharsAccess>;
  using Self = TokenStreamChars<mozilla::Utf8Unit, AnyCharsAccess>;

  using typename SpecializedCharsBase::SourceUnitsEnd;
  using typename SpecializedCharsBase::SourceUnitsIterator;

 protected:
  using GeneralCharsBase::anyCharsAccess;
  using GeneralCharsBase::computeLineAndColumn;
  using GeneralCharsBase::fillExceptingContext;
  using GeneralCharsBase::internalComputeLineOfContext;
  using TokenStreamCharsShared::isAsciiCodePoint;
  // Deliberately don't |using| |sourceUnits| because of bug 1472569.  :-(
  using GeneralCharsBase::updateLineInfoForEOL;

 private:
  static char toHexChar(uint8_t nibble) {
    MOZ_ASSERT(nibble < 16);
    return "0123456789ABCDEF"[nibble];
  }

  static void byteToString(uint8_t n, char* str) {
    str[0] = '0';
    str[1] = 'x';
    str[2] = toHexChar(n >> 4);
    str[3] = toHexChar(n & 0xF);
  }

  static void byteToTerminatedString(uint8_t n, char* str) {
    byteToString(n, str);
    str[4] = '\0';
  }

  /**
   * Report a UTF-8 encoding-related error for a code point starting AT THE
   * CURRENT OFFSET.
   *
   * |relevantUnits| indicates how many code units from the current offset
   * are potentially relevant to the reported error, such that they may be
   * included in the error message.  For example, if at the current offset we
   * have
   *
   *   0b1111'1111 ...
   *
   * a code unit never allowed in UTF-8, then |relevantUnits| might be 1
   * because only that unit is relevant.  Or if we have
   *
   *   0b1111'0111 0b1011'0101 0b0000'0000 ...
   *
   * where the first two code units are a valid prefix to a four-unit code
   * point but the third unit *isn't* a valid trailing code unit, then
   * |relevantUnits| might be 3.
   */
  MOZ_COLD void internalEncodingError(uint8_t relevantUnits,
                                      unsigned errorNumber, ...);

  // Don't use |internalEncodingError|!  Use one of the elaborated functions
  // that calls it, below -- all of which should be used to indicate an error
  // in a code point starting AT THE CURRENT OFFSET as with
  // |internalEncodingError|.

  /** Report an error for an invalid lead code unit |lead|. */
  MOZ_COLD void badLeadUnit(mozilla::Utf8Unit lead);

  /**
   * Report an error when there aren't enough code units remaining to
   * constitute a full code point after |lead|: only |remaining| code units
   * were available for a code point starting with |lead|, when at least
   * |required| code units were required.
   */
  MOZ_COLD void notEnoughUnits(mozilla::Utf8Unit lead, uint8_t remaining,
                               uint8_t required);

  /**
   * Report an error for a bad trailing UTF-8 code unit, where the bad
   * trailing unit was the last of |unitsObserved| units examined from the
   * current offset.
   */
  MOZ_COLD void badTrailingUnit(uint8_t unitsObserved);

  // Helper used for both |badCodePoint| and |notShortestForm| for code units
  // that have all the requisite high bits set/unset in a manner that *could*
  // encode a valid code point, but the remaining bits encoding its actual
  // value do not define a permitted value.
  MOZ_COLD void badStructurallyValidCodePoint(uint32_t codePoint,
                                              uint8_t codePointLength,
                                              const char* reason);

  /**
   * Report an error for UTF-8 that encodes a UTF-16 surrogate or a number
   * outside the Unicode range.
   */
  MOZ_COLD void badCodePoint(uint32_t codePoint, uint8_t codePointLength) {
    MOZ_ASSERT(unicode::IsSurrogate(codePoint) ||
               codePoint > unicode::NonBMPMax);

    badStructurallyValidCodePoint(codePoint, codePointLength,
                                  unicode::IsSurrogate(codePoint)
                                      ? "it's a UTF-16 surrogate"
                                      : "the maximum code point is U+10FFFF");
  }

  /**
   * Report an error for UTF-8 that encodes a code point not in its shortest
   * form.
   */
  MOZ_COLD void notShortestForm(uint32_t codePoint, uint8_t codePointLength) {
    MOZ_ASSERT(!unicode::IsSurrogate(codePoint));
    MOZ_ASSERT(codePoint <= unicode::NonBMPMax);

    badStructurallyValidCodePoint(
        codePoint, codePointLength,
        "it wasn't encoded in shortest possible form");
  }

 protected:
  using GeneralCharsBase::GeneralCharsBase;

  /**
   * Given the non-ASCII |lead| code unit just consumed, consume the rest of
   * a non-ASCII code point.  The code point is not normalized: on success
   * |*codePoint| may be U+2028 LINE SEPARATOR or U+2029 PARAGRAPH SEPARATOR.
   *
   * Report an error if an invalid code point is encountered.
   */
  MOZ_MUST_USE bool getNonAsciiCodePointDontNormalize(mozilla::Utf8Unit lead,
                                                      char32_t* codePoint);

  /**
   * Given a just-consumed non-ASCII code unit |lead|, consume a full code
   * point or LineTerminatorSequence (normalizing it to '\n') and store it in
   * |*codePoint|.  Return true on success, otherwise return false and leave
   * |*codePoint| undefined on failure.
   *
   * If a LineTerminatorSequence was consumed, also update line/column info.
   *
   * This function will change the current |sourceUnits| offset.
   */
  MOZ_MUST_USE bool getNonAsciiCodePoint(int32_t lead, int32_t* codePoint);
};

// TokenStream is the lexical scanner for JavaScript source text.
//
// It takes a buffer of Unit code units (currently only char16_t encoding
// UTF-16, but we're adding either UTF-8 or Latin-1 single-byte text soon) and
// linearly scans it into |Token|s.
//
// Internally the class uses a four element circular buffer |tokens| of
// |Token|s. As an index for |tokens|, the member |cursor_| points to the
// current token. Calls to getToken() increase |cursor_| by one and return the
// new current token. If a TokenStream was just created, the current token is
// uninitialized. It's therefore important that one of the first four member
// functions listed below is called first. The circular buffer lets us go back
// up to two tokens from the last scanned token. Internally, the relative
// number of backward steps that were taken (via ungetToken()) after the last
// token was scanned is stored in |lookahead|.
//
// The following table lists in which situations it is safe to call each listed
// function. No checks are made by the functions in non-debug builds.
//
// Function Name     | Precondition; changes to |lookahead|
// ------------------+---------------------------------------------------------
// getToken          | none; if |lookahead > 0| then |lookahead--|
// peekToken         | none; if |lookahead == 0| then |lookahead == 1|
// peekTokenSameLine | none; if |lookahead == 0| then |lookahead == 1|
// matchToken        | none; if |lookahead > 0| and the match succeeds then
//                   |       |lookahead--|
// consumeKnownToken | none; if |lookahead > 0| then |lookahead--|
// ungetToken        | 0 <= |lookahead| <= |maxLookahead - 1|; |lookahead++|
//
// The behavior of the token scanning process (see getTokenInternal()) can be
// modified by calling one of the first four above listed member functions with
// an optional argument of type Modifier.  However, the modifier will be
// ignored unless |lookahead == 0| holds.  Due to constraints of the grammar,
// this turns out not to be a problem in practice. See the
// mozilla.dev.tech.js-engine.internals thread entitled 'Bug in the scanner?'
// for more details:
// https://groups.google.com/forum/?fromgroups=#!topic/mozilla.dev.tech.js-engine.internals/2JLH5jRcr7E).
//
// The method seek() allows rescanning from a previously visited location of
// the buffer, initially computed by constructing a Position local variable.
//
template <typename Unit, class AnyCharsAccess>
class MOZ_STACK_CLASS TokenStreamSpecific
    : public TokenStreamChars<Unit, AnyCharsAccess>,
      public TokenStreamShared,
      public ErrorReporter {
 public:
  using CharsBase = TokenStreamCharsBase<Unit>;
  using SpecializedCharsBase = SpecializedTokenStreamCharsBase<Unit>;
  using GeneralCharsBase = GeneralTokenStreamChars<Unit, AnyCharsAccess>;
  using SpecializedChars = TokenStreamChars<Unit, AnyCharsAccess>;

  using Position = TokenStreamPosition<Unit>;

  // Anything inherited through a base class whose type depends upon this
  // class's template parameters can only be accessed through a dependent
  // name: prefixed with |this|, by explicit qualification, and so on.  (This
  // is so that references to inherited fields are statically distinguishable
  // from references to names outside of the class.)  This is tedious and
  // onerous.
  //
  // As an alternative, we directly add every one of these functions to this
  // class, using explicit qualification to address the dependent-name
  // problem.  |this| or other qualification is no longer necessary -- at
  // cost of this ever-changing laundry list of |using|s.  So it goes.
 public:
  using GeneralCharsBase::anyCharsAccess;
  using GeneralCharsBase::computeLineAndColumn;

 private:
  using typename CharsBase::SourceUnits;

 private:
  using CharsBase::atomizeSourceChars;
  using GeneralCharsBase::badToken;
  using TokenStreamCharsShared::appendCodePointToCharBuffer;
  // Deliberately don't |using| |charBuffer| because of bug 1472569.  :-(
  using CharsBase::consumeKnownCodeUnit;
  using CharsBase::fillCharBufferFromSourceNormalizingAsciiLineBreaks;
  using CharsBase::matchCodeUnit;
  using CharsBase::matchLineTerminator;
  using CharsBase::peekCodeUnit;
  using GeneralCharsBase::computeColumn;
  using GeneralCharsBase::fillExceptingContext;
  using GeneralCharsBase::getCodeUnit;
  using GeneralCharsBase::getFullAsciiCodePoint;
  using GeneralCharsBase::internalComputeLineOfContext;
  using GeneralCharsBase::matchUnicodeEscapeIdent;
  using GeneralCharsBase::matchUnicodeEscapeIdStart;
  using GeneralCharsBase::newAtomToken;
  using GeneralCharsBase::newBigIntToken;
  using GeneralCharsBase::newNameToken;
  using GeneralCharsBase::newNumberToken;
  using GeneralCharsBase::newPrivateNameToken;
  using GeneralCharsBase::newRegExpToken;
  using GeneralCharsBase::newSimpleToken;
  using SpecializedChars::getNonAsciiCodePoint;
  using SpecializedChars::getNonAsciiCodePointDontNormalize;
  using TokenStreamCharsShared::copyCharBufferTo;
  using TokenStreamCharsShared::drainCharBufferIntoAtom;
  using TokenStreamCharsShared::isAsciiCodePoint;
  // Deliberately don't |using| |sourceUnits| because of bug 1472569.  :-(
  using CharsBase::toUnit;
  using GeneralCharsBase::ungetCodeUnit;
  using GeneralCharsBase::updateLineInfoForEOL;

  template <typename CharU>
  friend class TokenStreamPosition;

 public:
  TokenStreamSpecific(JSContext* cx, const JS::ReadOnlyCompileOptions& options,
                      const Unit* units, size_t length);

  /**
   * Get the next code point, converting LineTerminatorSequences to '\n' and
   * updating internal line-counter state if needed.  Return true on success
   * and store the code point in |*cp|.  Return false and leave |*cp|
   * undefined on failure.
   */
  MOZ_MUST_USE bool getCodePoint(int32_t* cp);

  // If there is an invalid escape in a template, report it and return false,
  // otherwise return true.
  bool checkForInvalidTemplateEscapeError() {
    if (anyCharsAccess().invalidTemplateEscapeType == InvalidEscapeType::None) {
      return true;
    }

    reportInvalidEscapeError(anyCharsAccess().invalidTemplateEscapeOffset,
                             anyCharsAccess().invalidTemplateEscapeType);
    return false;
  }

 public:
  // Implement ErrorReporter.

  void lineAndColumnAt(size_t offset, uint32_t* line,
                       uint32_t* column) const final {
    computeLineAndColumn(offset, line, column);
  }

  void currentLineAndColumn(uint32_t* line, uint32_t* column) const final {
    computeLineAndColumn(anyCharsAccess().currentToken().pos.begin, line,
                         column);
  }

  bool isOnThisLine(size_t offset, uint32_t lineNum,
                    bool* onThisLine) const final {
    return anyCharsAccess().srcCoords.isOnThisLine(offset, lineNum, onThisLine);
  }

  uint32_t lineAt(size_t offset) const final {
    const auto& anyChars = anyCharsAccess();
    auto lineToken = anyChars.lineToken(offset);
    return anyChars.lineNumber(lineToken);
  }

  uint32_t columnAt(size_t offset) const final {
    return computeColumn(anyCharsAccess().lineToken(offset), offset);
  }

  bool hasTokenizationStarted() const final;

  const char* getFilename() const final {
    return anyCharsAccess().getFilename();
  }

 private:
  // Implement ErrorReportMixin.

  JSContext* getContext() const override { return anyCharsAccess().cx; }

  MOZ_MUST_USE bool strictMode() const override {
    return anyCharsAccess().strictMode();
  }

 public:
  // Implement ErrorReportMixin.

  const JS::ReadOnlyCompileOptions& options() const final {
    return anyCharsAccess().options();
  }

  MOZ_MUST_USE bool computeErrorMetadata(
      ErrorMetadata* err, const ErrorOffset& errorOffset) override;

 private:
  void reportInvalidEscapeError(uint32_t offset, InvalidEscapeType type) {
    switch (type) {
      case InvalidEscapeType::None:
        MOZ_ASSERT_UNREACHABLE("unexpected InvalidEscapeType");
        return;
      case InvalidEscapeType::Hexadecimal:
        errorAt(offset, JSMSG_MALFORMED_ESCAPE, "hexadecimal");
        return;
      case InvalidEscapeType::Unicode:
        errorAt(offset, JSMSG_MALFORMED_ESCAPE, "Unicode");
        return;
      case InvalidEscapeType::UnicodeOverflow:
        errorAt(offset, JSMSG_UNICODE_OVERFLOW, "escape sequence");
        return;
      case InvalidEscapeType::Octal:
        errorAt(offset, JSMSG_DEPRECATED_OCTAL);
        return;
    }
  }

  MOZ_MUST_USE bool putIdentInCharBuffer(const Unit* identStart);

  using IsIntegerUnit = bool (*)(int32_t);
  MOZ_MUST_USE MOZ_ALWAYS_INLINE bool matchInteger(IsIntegerUnit isIntegerUnit,
                                                   int32_t* nextUnit);

  /**
   * Tokenize a decimal number that begins at |numStart| into the provided
   * token.
   *
   * |unit| must be one of these values:
   *
   *   1. The first decimal digit in the integral part of a decimal number
   *      not starting with '0' or '.', e.g. '1' for "17", '3' for "3.14", or
   *      '8' for "8.675309e6".
   *
   *   In this case, the next |getCodeUnit()| must return the code unit after
   *   |unit| in the overall number.
   *
   *   2. The '.' in a "."/"0."-prefixed decimal number or the 'e'/'E' in a
   *      "0e"/"0E"-prefixed decimal number, e.g. ".17", "0.42", or "0.1e3".
   *
   *   In this case, the next |getCodeUnit()| must return the code unit
   *   *after* the first decimal digit *after* the '.'.  So the next code
   *   unit would be '7' in ".17", '2' in "0.42", 'e' in "0.4e+8", or '/' in
   *   "0.5/2" (three separate tokens).
   *
   *   3. The code unit after the '0' where "0" is the entire number token.
   *
   *   In this case, the next |getCodeUnit()| would return the code unit
   *   after |unit|, but this function will never perform such call.
   *
   *   4. (Non-strict mode code only)  The first '8' or '9' in a "noctal"
   *      number that begins with a '0' but contains a non-octal digit in its
   *      integer part so is interpreted as decimal, e.g. '9' in "09.28" or
   *      '8' in "0386" or '9' in "09+7" (three separate tokens").
   *
   *   In this case, the next |getCodeUnit()| returns the code unit after
   *   |unit|: '.', '6', or '+' in the examples above.
   *
   * This interface is super-hairy and horribly stateful.  Unfortunately, its
   * hair merely reflects the intricacy of ECMAScript numeric literal syntax.
   * And incredibly, it *improves* on the goto-based horror that predated it.
   */
  MOZ_MUST_USE bool decimalNumber(int32_t unit, TokenStart start,
                                  const Unit* numStart, Modifier modifier,
                                  TokenKind* out);

  /** Tokenize a regular expression literal beginning at |start|. */
  MOZ_MUST_USE bool regexpLiteral(TokenStart start, TokenKind* out);

  /**
   * Slurp characters between |start| and sourceUnits.current() into
   * charBuffer, to later parse into a bigint.
   */
  MOZ_MUST_USE bool bigIntLiteral(TokenStart start, Modifier modifier,
                                  TokenKind* out);

 public:
  // Advance to the next token.  If the token stream encountered an error,
  // return false.  Otherwise return true and store the token kind in |*ttp|.
  MOZ_MUST_USE bool getToken(TokenKind* ttp, Modifier modifier = SlashIsDiv) {
    // Check for a pushed-back token resulting from mismatching lookahead.
    TokenStreamAnyChars& anyChars = anyCharsAccess();
    if (anyChars.lookahead != 0) {
      MOZ_ASSERT(!anyChars.flags.hadError);
      anyChars.lookahead--;
      anyChars.advanceCursor();
      TokenKind tt = anyChars.currentToken().type;
      MOZ_ASSERT(tt != TokenKind::Eol);
      verifyConsistentModifier(modifier, anyChars.currentToken());
      *ttp = tt;
      return true;
    }

    return getTokenInternal(ttp, modifier);
  }

  MOZ_MUST_USE bool peekToken(TokenKind* ttp, Modifier modifier = SlashIsDiv) {
    TokenStreamAnyChars& anyChars = anyCharsAccess();
    if (anyChars.lookahead > 0) {
      MOZ_ASSERT(!anyChars.flags.hadError);
      verifyConsistentModifier(modifier, anyChars.nextToken());
      *ttp = anyChars.nextToken().type;
      return true;
    }
    if (!getTokenInternal(ttp, modifier)) {
      return false;
    }
    anyChars.ungetToken();
    return true;
  }

  MOZ_MUST_USE bool peekTokenPos(TokenPos* posp,
                                 Modifier modifier = SlashIsDiv) {
    TokenStreamAnyChars& anyChars = anyCharsAccess();
    if (anyChars.lookahead == 0) {
      TokenKind tt;
      if (!getTokenInternal(&tt, modifier)) {
        return false;
      }
      anyChars.ungetToken();
      MOZ_ASSERT(anyChars.hasLookahead());
    } else {
      MOZ_ASSERT(!anyChars.flags.hadError);
      verifyConsistentModifier(modifier, anyChars.nextToken());
    }
    *posp = anyChars.nextToken().pos;
    return true;
  }

  MOZ_MUST_USE bool peekOffset(uint32_t* offset,
                               Modifier modifier = SlashIsDiv) {
    TokenPos pos;
    if (!peekTokenPos(&pos, modifier)) {
      return false;
    }
    *offset = pos.begin;
    return true;
  }

  // This is like peekToken(), with one exception:  if there is an EOL
  // between the end of the current token and the start of the next token, it
  // return true and store Eol in |*ttp|.  In that case, no token with
  // Eol is actually created, just a Eol TokenKind is returned, and
  // currentToken() shouldn't be consulted.  (This is the only place Eol
  // is produced.)
  MOZ_ALWAYS_INLINE MOZ_MUST_USE bool peekTokenSameLine(
      TokenKind* ttp, Modifier modifier = SlashIsDiv) {
    TokenStreamAnyChars& anyChars = anyCharsAccess();
    const Token& curr = anyChars.currentToken();

    // If lookahead != 0, we have scanned ahead at least one token, and
    // |lineno| is the line that the furthest-scanned token ends on.  If
    // it's the same as the line that the current token ends on, that's a
    // stronger condition than what we are looking for, and we don't need
    // to return Eol.
    if (anyChars.lookahead != 0) {
      bool onThisLine;
      if (!anyChars.srcCoords.isOnThisLine(curr.pos.end, anyChars.lineno,
                                           &onThisLine)) {
        error(JSMSG_OUT_OF_MEMORY);
        return false;
      }

      if (onThisLine) {
        MOZ_ASSERT(!anyChars.flags.hadError);
        verifyConsistentModifier(modifier, anyChars.nextToken());
        *ttp = anyChars.nextToken().type;
        return true;
      }
    }

    // The above check misses two cases where we don't have to return
    // Eol.
    // - The next token starts on the same line, but is a multi-line token.
    // - The next token starts on the same line, but lookahead==2 and there
    //   is a newline between the next token and the one after that.
    // The following test is somewhat expensive but gets these cases (and
    // all others) right.
    TokenKind tmp;
    if (!getToken(&tmp, modifier)) {
      return false;
    }

    const Token& next = anyChars.currentToken();
    anyChars.ungetToken();

    // Careful, |next| points to an initialized-but-not-allocated Token!
    // This is safe because we don't modify token data below.

    auto currentEndToken = anyChars.lineToken(curr.pos.end);
    auto nextBeginToken = anyChars.lineToken(next.pos.begin);

    *ttp =
        currentEndToken.isSameLine(nextBeginToken) ? next.type : TokenKind::Eol;
    return true;
  }

  // Get the next token from the stream if its kind is |tt|.
  MOZ_MUST_USE bool matchToken(bool* matchedp, TokenKind tt,
                               Modifier modifier = SlashIsDiv) {
    TokenKind token;
    if (!getToken(&token, modifier)) {
      return false;
    }
    if (token == tt) {
      *matchedp = true;
    } else {
      anyCharsAccess().ungetToken();
      *matchedp = false;
    }
    return true;
  }

  void consumeKnownToken(TokenKind tt, Modifier modifier = SlashIsDiv) {
    bool matched;
    MOZ_ASSERT(anyCharsAccess().hasLookahead());
    MOZ_ALWAYS_TRUE(matchToken(&matched, tt, modifier));
    MOZ_ALWAYS_TRUE(matched);
  }

  MOZ_MUST_USE bool nextTokenEndsExpr(bool* endsExpr) {
    TokenKind tt;
    if (!peekToken(&tt)) {
      return false;
    }

    *endsExpr = anyCharsAccess().isExprEnding[size_t(tt)];
    if (*endsExpr) {
      // If the next token ends an overall Expression, we'll parse this
      // Expression without ever invoking Parser::orExpr().  But we need that
      // function's DEBUG-only side effect of marking this token as safe to get
      // with SlashIsRegExp, so we have to do it manually here.
      anyCharsAccess().allowGettingNextTokenWithSlashIsRegExp();
    }
    return true;
  }

  MOZ_MUST_USE bool advance(size_t position);

  void seek(const Position& pos);
  MOZ_MUST_USE bool seek(const Position& pos, const TokenStreamAnyChars& other);

  const Unit* codeUnitPtrAt(size_t offset) const {
    return this->sourceUnits.codeUnitPtrAt(offset);
  }

  const Unit* rawLimit() const { return this->sourceUnits.limit(); }

  MOZ_MUST_USE bool identifierName(TokenStart start, const Unit* identStart,
                                   IdentifierEscapes escaping,
                                   Modifier modifier, NameVisibility visibility,
                                   TokenKind* out);

  MOZ_MUST_USE bool matchIdentifierStart(IdentifierEscapes* sawEscape);

  MOZ_MUST_USE bool getTokenInternal(TokenKind* const ttp,
                                     const Modifier modifier);

  MOZ_MUST_USE bool getStringOrTemplateToken(char untilChar, Modifier modifier,
                                             TokenKind* out);

  // Parse a TemplateMiddle or TemplateTail token (one of the string-like parts
  // of a template string) after already consuming the leading `RightCurly`.
  // (The spec says the `}` is the first character of the TemplateMiddle/
  // TemplateTail, but we treat it as a separate token because that's much
  // easier to implement in both TokenStream and the parser.)
  //
  // This consumes a token and sets the current token, like `getToken()`.  It
  // doesn't take a Modifier because there's no risk of encountering a division
  // operator or RegExp literal.
  //
  // On success, `*ttp` is either `TokenKind::TemplateHead` (if we got a
  // TemplateMiddle token) or `TokenKind::NoSubsTemplate` (if we got a
  // TemplateTail). That may seem strange; there are four different template
  // token types in the spec, but we only use two. We use `TemplateHead` for
  // TemplateMiddle because both end with `...${`, and `NoSubsTemplate` for
  // TemplateTail because both contain the end of the template, including the
  // closing quote mark. They're not treated differently, either in the parser
  // or in the tokenizer.
  MOZ_MUST_USE bool getTemplateToken(TokenKind* ttp) {
    MOZ_ASSERT(anyCharsAccess().currentToken().type == TokenKind::RightCurly);
    return getStringOrTemplateToken('`', SlashIsInvalid, ttp);
  }

  MOZ_MUST_USE bool getDirectives(bool isMultiline, bool shouldWarnDeprecated);
  MOZ_MUST_USE bool getDirective(
      bool isMultiline, bool shouldWarnDeprecated, const char* directive,
      uint8_t directiveLength, const char* errorMsgPragma,
      UniquePtr<char16_t[], JS::FreePolicy>* destination);
  MOZ_MUST_USE bool getDisplayURL(bool isMultiline, bool shouldWarnDeprecated);
  MOZ_MUST_USE bool getSourceMappingURL(bool isMultiline,
                                        bool shouldWarnDeprecated);
};

// It's preferable to define this in TokenStream.cpp, but its template-ness
// means we'd then have to *instantiate* this constructor for all possible
// (Unit, AnyCharsAccess) pairs -- and that gets super-messy as AnyCharsAccess
// *itself* is templated.  This symbol really isn't that huge compared to some
// defined inline in TokenStreamSpecific, so just rely on the linker commoning
// stuff up.
template <typename Unit>
template <class AnyCharsAccess>
inline TokenStreamPosition<Unit>::TokenStreamPosition(
    AutoKeepAtoms& keepAtoms,
    TokenStreamSpecific<Unit, AnyCharsAccess>& tokenStream)
    : currentToken(tokenStream.anyCharsAccess().currentToken()) {
  TokenStreamAnyChars& anyChars = tokenStream.anyCharsAccess();

  buf =
      tokenStream.sourceUnits.addressOfNextCodeUnit(/* allowPoisoned = */ true);
  flags = anyChars.flags;
  lineno = anyChars.lineno;
  linebase = anyChars.linebase;
  prevLinebase = anyChars.prevLinebase;
  lookahead = anyChars.lookahead;
  currentToken = anyChars.currentToken();
  for (unsigned i = 0; i < anyChars.lookahead; i++) {
    lookaheadTokens[i] = anyChars.tokens[anyChars.aheadCursor(1 + i)];
  }
}

class TokenStreamAnyCharsAccess {
 public:
  template <class TokenStreamSpecific>
  static inline TokenStreamAnyChars& anyChars(TokenStreamSpecific* tss);

  template <class TokenStreamSpecific>
  static inline const TokenStreamAnyChars& anyChars(
      const TokenStreamSpecific* tss);
};

class MOZ_STACK_CLASS TokenStream final
    : public TokenStreamAnyChars,
      public TokenStreamSpecific<char16_t, TokenStreamAnyCharsAccess> {
  using Unit = char16_t;

 public:
  TokenStream(JSContext* cx, const JS::ReadOnlyCompileOptions& options,
              const Unit* units, size_t length, StrictModeGetter* smg)
      : TokenStreamAnyChars(cx, options, smg),
        TokenStreamSpecific<Unit, TokenStreamAnyCharsAccess>(cx, options, units,
                                                             length) {}
};

template <class TokenStreamSpecific>
/* static */ inline TokenStreamAnyChars& TokenStreamAnyCharsAccess::anyChars(
    TokenStreamSpecific* tss) {
  auto* ts = static_cast<TokenStream*>(tss);
  return *static_cast<TokenStreamAnyChars*>(ts);
}

template <class TokenStreamSpecific>
/* static */ inline const TokenStreamAnyChars&
TokenStreamAnyCharsAccess::anyChars(const TokenStreamSpecific* tss) {
  const auto* ts = static_cast<const TokenStream*>(tss);
  return *static_cast<const TokenStreamAnyChars*>(ts);
}

extern const char* TokenKindToDesc(TokenKind tt);

}  // namespace frontend
}  // namespace js

extern JS_FRIEND_API int js_fgets(char* buf, int size, FILE* file);

#ifdef DEBUG
extern const char* TokenKindToString(js::frontend::TokenKind tt);
#endif

#endif /* frontend_TokenStream_h */