summaryrefslogtreecommitdiffstats
path: root/Source/cmPolicies.cxx
blob: 3eb19bb96579bbc224662741886b3d22a4774bb1 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
#include "cmPolicies.h"
#include "cmake.h"
#include "cmMakefile.h"
#include "cmVersion.h"
#include "cmVersionMacros.h"
#include "cmAlgorithms.h"
#include <map>
#include <set>
#include <queue>
#include <assert.h>

static bool stringToId(const char* input, cmPolicies::PolicyID& pid)
{
  assert(input);
  if (strlen(input) != 7)
    {
    return false;
    }
  if (!cmHasLiteralPrefix(input, "CMP"))
    {
    return false;
    }
  if (cmHasLiteralSuffix(input, "0000"))
    {
    pid = cmPolicies::CMP0000;
    return true;
    }
  for (int i = 3; i < 7; ++i)
    {
    if (!isdigit(*(input + i)))
      {
      return false;
      }
    }
  long id;
  if (!cmSystemTools::StringToLong(input + 3, &id))
    {
    return false;
    }
  if (id >= cmPolicies::CMPCOUNT)
    {
    return false;
    }
  pid = cmPolicies::PolicyID(id);
  return true;
}

#define CM_SELECT_ID_VERSION(F, A1, A2, A3, A4, A5, A6) F(A1, A3, A4, A5)
#define CM_FOR_EACH_POLICY_ID_VERSION(POLICY) \
  CM_FOR_EACH_POLICY_TABLE(POLICY, CM_SELECT_ID_VERSION)

#define CM_SELECT_ID_DOC(F, A1, A2, A3, A4, A5, A6) F(A1, A2)
#define CM_FOR_EACH_POLICY_ID_DOC(POLICY) \
  CM_FOR_EACH_POLICY_TABLE(POLICY, CM_SELECT_ID_DOC)

static const char* idToString(cmPolicies::PolicyID id)
{
  switch(id)
    {
#define POLICY_CASE(ID) \
    case cmPolicies::ID: \
      return #ID;
  CM_FOR_EACH_POLICY_ID(POLICY_CASE)
#undef POLICY_CASE
    case cmPolicies::CMPCOUNT:
      return 0;
    }
  return 0;
}

static const char* idToVersion(cmPolicies::PolicyID id)
{
  switch(id)
    {
#define POLICY_CASE(ID, V_MAJOR, V_MINOR, V_PATCH) \
    case cmPolicies::ID: \
      return #V_MAJOR "." #V_MINOR "." #V_PATCH;
  CM_FOR_EACH_POLICY_ID_VERSION(POLICY_CASE)
#undef POLICY_CASE
    case cmPolicies::CMPCOUNT:
      return 0;
    }
  return 0;
}

static bool isPolicyNewerThan(cmPolicies::PolicyID id,
                       unsigned int majorV,
                       unsigned int minorV,
                       unsigned int patchV)
{
  switch(id)
    {
#define POLICY_CASE(ID, V_MAJOR, V_MINOR, V_PATCH) \
    case cmPolicies::ID: \
      return (majorV < V_MAJOR || \
             (majorV == V_MAJOR && \
              minorV + 1 < V_MINOR + 1) || \
             (majorV == V_MAJOR && \
              minorV == V_MINOR && \
              patchV + 1 < V_PATCH + 1));
  CM_FOR_EACH_POLICY_ID_VERSION(POLICY_CASE)
#undef POLICY_CASE
    case cmPolicies::CMPCOUNT:
      return false;
    }
  return false;
}

const char* idToShortDescription(cmPolicies::PolicyID id)
{
  switch(id)
    {
#define POLICY_CASE(ID, SHORT_DESCRIPTION) \
    case cmPolicies::ID: \
      return SHORT_DESCRIPTION;
  CM_FOR_EACH_POLICY_ID_DOC(POLICY_CASE)
#undef POLICY_CASE
    case cmPolicies::CMPCOUNT:
      return 0;
    }
  return 0;
}

//----------------------------------------------------------------------------
static void DiagnoseAncientPolicies(
    std::vector<cmPolicies::PolicyID> const& ancient,
    unsigned int majorVer,
    unsigned int minorVer,
    unsigned int patchVer,
    cmMakefile* mf)
{
  std::ostringstream e;
  e << "The project requests behavior compatible with CMake version \""
    << majorVer << "." << minorVer << "." << patchVer
    << "\", which requires the OLD behavior for some policies:\n";
  for(std::vector<cmPolicies::PolicyID>::const_iterator
        i = ancient.begin(); i != ancient.end(); ++i)
    {
    e << "  " << idToString(*i) << ": " << idToShortDescription(*i) << "\n";
    }
  e << "However, this version of CMake no longer supports the OLD "
    << "behavior for these policies.  "
    << "Please either update your CMakeLists.txt files to conform to "
    << "the new behavior or use an older version of CMake that still "
    << "supports the old behavior.";
  mf->IssueMessage(cmake::FATAL_ERROR, e.str());
}

//----------------------------------------------------------------------------
static bool GetPolicyDefault(cmMakefile* mf, std::string const& policy,
                             cmPolicies::PolicyStatus* defaultSetting)
{
  std::string defaultVar = "CMAKE_POLICY_DEFAULT_" + policy;
  std::string defaultValue = mf->GetSafeDefinition(defaultVar);
  if(defaultValue == "NEW")
    {
    *defaultSetting = cmPolicies::NEW;
    }
  else if(defaultValue == "OLD")
    {
    *defaultSetting = cmPolicies::OLD;
    }
  else if(defaultValue == "")
    {
    *defaultSetting = cmPolicies::WARN;
    }
  else
    {
    std::ostringstream e;
    e << defaultVar << " has value \"" << defaultValue
      << "\" but must be \"OLD\", \"NEW\", or \"\" (empty).";
    mf->IssueMessage(cmake::FATAL_ERROR, e.str());
    return false;
    }

  return true;
}

//----------------------------------------------------------------------------
bool cmPolicies::ApplyPolicyVersion(cmMakefile *mf,
                                    const char *version)
{
  std::string ver = "2.4.0";

  if (version && strlen(version) > 0)
    {
    ver = version;
    }

  unsigned int majorVer = 2;
  unsigned int minorVer = 0;
  unsigned int patchVer = 0;
  unsigned int tweakVer = 0;

  // parse the string
  if(sscanf(ver.c_str(), "%u.%u.%u.%u",
            &majorVer, &minorVer, &patchVer, &tweakVer) < 2)
    {
    std::ostringstream e;
    e << "Invalid policy version value \"" << ver << "\".  "
      << "A numeric major.minor[.patch[.tweak]] must be given.";
    mf->IssueMessage(cmake::FATAL_ERROR, e.str());
    return false;
    }

  // it is an error if the policy version is less than 2.4
  if (majorVer < 2 || (majorVer == 2 && minorVer < 4))
    {
    mf->IssueMessage(cmake::FATAL_ERROR,
      "Compatibility with CMake < 2.4 is not supported by CMake >= 3.0.  "
      "For compatibility with older versions please use any CMake 2.8.x "
      "release or lower.");
    return false;
    }

  // It is an error if the policy version is greater than the running
  // CMake.
  if (majorVer > cmVersion::GetMajorVersion() ||
      (majorVer == cmVersion::GetMajorVersion() &&
       minorVer > cmVersion::GetMinorVersion()) ||
      (majorVer == cmVersion::GetMajorVersion() &&
       minorVer == cmVersion::GetMinorVersion() &&
       patchVer > cmVersion::GetPatchVersion()) ||
      (majorVer == cmVersion::GetMajorVersion() &&
       minorVer == cmVersion::GetMinorVersion() &&
       patchVer == cmVersion::GetPatchVersion() &&
       tweakVer > cmVersion::GetTweakVersion()))
    {
    std::ostringstream e;
    e << "An attempt was made to set the policy version of CMake to \""
      << version << "\" which is greater than this version of CMake.  "
      << "This is not allowed because the greater version may have new "
      << "policies not known to this CMake.  "
      << "You may need a newer CMake version to build this project.";
    mf->IssueMessage(cmake::FATAL_ERROR, e.str());
    return false;
    }

  // now loop over all the policies and set them as appropriate
  std::vector<cmPolicies::PolicyID> ancientPolicies;
  for(PolicyID pid = cmPolicies::CMP0000;
      pid != cmPolicies::CMPCOUNT; pid = PolicyID(pid+1))
    {
    if (isPolicyNewerThan(pid, majorVer, minorVer, patchVer))
      {
      if(cmPolicies::GetPolicyStatus(pid) == cmPolicies::REQUIRED_ALWAYS)
        {
        ancientPolicies.push_back(pid);
        }
      else
        {
        cmPolicies::PolicyStatus status = cmPolicies::WARN;
        if(!GetPolicyDefault(mf, idToString(pid), &status) ||
           !mf->SetPolicy(pid, status))
          {
          return false;
          }
        if(pid == cmPolicies::CMP0001 &&
           (status == cmPolicies::WARN || status == cmPolicies::OLD))
          {
          if(!(mf->GetState()
               ->GetInitializedCacheValue("CMAKE_BACKWARDS_COMPATIBILITY")))
            {
            // Set it to 2.4 because that is the last version where the
            // variable had meaning.
            mf->AddCacheDefinition
              ("CMAKE_BACKWARDS_COMPATIBILITY", "2.4",
               "For backwards compatibility, what version of CMake "
               "commands and "
               "syntax should this version of CMake try to support.",
               cmState::STRING);
            }
          }
        }
      }
    else
      {
      if (!mf->SetPolicy(pid, cmPolicies::NEW))
        {
        return false;
        }
      }
    }

  // Make sure the project does not use any ancient policies.
  if(!ancientPolicies.empty())
    {
    DiagnoseAncientPolicies(ancientPolicies,
                            majorVer, minorVer, patchVer, mf);
    cmSystemTools::SetFatalErrorOccured();
    return false;
    }

  return true;
}

bool cmPolicies::GetPolicyID(const char *id, cmPolicies::PolicyID &pid)
{
  return stringToId(id, pid);
}

///! return a warning string for a given policy
std::string cmPolicies::GetPolicyWarning(cmPolicies::PolicyID id)
{
  std::ostringstream msg;
  msg <<
    "Policy " << idToString(id) << " is not set: "
    "" << idToShortDescription(id) << "  "
    "Run \"cmake --help-policy " << idToString(id) << "\" for "
    "policy details.  "
    "Use the cmake_policy command to set the policy "
    "and suppress this warning.";
  return msg.str();
}


///! return an error string for when a required policy is unspecified
std::string cmPolicies::GetRequiredPolicyError(cmPolicies::PolicyID id)
{
  std::ostringstream error;
  error <<
    "Policy " << idToString(id) << " is not set to NEW: "
    "" << idToShortDescription(id) << "  "
    "Run \"cmake --help-policy " << idToString(id) << "\" for "
    "policy details.  "
    "CMake now requires this policy to be set to NEW by the project.  "
    "The policy may be set explicitly using the code\n"
    "  cmake_policy(SET " << idToString(id) << " NEW)\n"
    "or by upgrading all policies with the code\n"
    "  cmake_policy(VERSION " << idToVersion(id) <<
    ") # or later\n"
    "Run \"cmake --help-command cmake_policy\" for more information.";
  return error.str();
}

///! Get the default status for a policy
cmPolicies::PolicyStatus
cmPolicies::GetPolicyStatus(cmPolicies::PolicyID)
{
  return cmPolicies::WARN;
}

//----------------------------------------------------------------------------
std::string
cmPolicies::GetRequiredAlwaysPolicyError(cmPolicies::PolicyID id)
{
  std::string pid = idToString(id);
  std::ostringstream e;
  e << "Policy " << pid << " may not be set to OLD behavior because this "
    << "version of CMake no longer supports it.  "
    << "The policy was introduced in "
    << "CMake version " << idToVersion(id)
    << ", and use of NEW behavior is now required."
    << "\n"
    << "Please either update your CMakeLists.txt files to conform to "
    << "the new behavior or use an older version of CMake that still "
    << "supports the old behavior.  "
    << "Run cmake --help-policy " << pid << " for more information.";
  return e.str();
}

cmPolicies::PolicyStatus
cmPolicies::PolicyMap::Get(cmPolicies::PolicyID id) const
{
  PolicyStatus status = cmPolicies::WARN;

  if (this->Status[(POLICY_STATUS_COUNT * id) + OLD])
    {
    status = cmPolicies::OLD;
    }
  else if (this->Status[(POLICY_STATUS_COUNT * id) + NEW])
    {
    status = cmPolicies::NEW;
    }
  return status;
}

void cmPolicies::PolicyMap::Set(cmPolicies::PolicyID id,
                                cmPolicies::PolicyStatus status)
{
  this->Status[(POLICY_STATUS_COUNT * id) + OLD] = (status == OLD);
  this->Status[(POLICY_STATUS_COUNT * id) + WARN] = (status == WARN);
  this->Status[(POLICY_STATUS_COUNT * id) + NEW] = (status == NEW);
}

bool cmPolicies::PolicyMap::IsDefined(cmPolicies::PolicyID id) const
{
  return this->Status[(POLICY_STATUS_COUNT * id) + OLD]
      || this->Status[(POLICY_STATUS_COUNT * id) + WARN]
      || this->Status[(POLICY_STATUS_COUNT * id) + NEW];
}

bool cmPolicies::PolicyMap::IsEmpty() const
{
  return this->Status.none();
}
a id='n627' href='#n627'>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
/*
 * tclCompExpr.c --
 *
 *	This file contains the code to parse and compile Tcl expressions
 *	and implementations of the Tcl commands corresponding to expression
 *	operators, such as the command ::tcl::mathop::+ .
 *
 * Contributions from Don Porter, NIST, 2006-2007. (not subject to US copyright)
 *
 * See the file "license.terms" for information on usage and redistribution of
 * this file, and for a DISCLAIMER OF ALL WARRANTIES.
 *
 * RCS: @(#) $Id: tclCompExpr.c,v 1.89 2007/11/11 19:32:14 msofer Exp $
 */

#include "tclInt.h"
#include "tclCompile.h"		/* CompileEnv */

/*
 * Expression parsing takes place in the routine ParseExpr().  It takes a
 * string as input, parses that string, and generates a representation of
 * the expression in the form of a tree of operators, a list of literals,
 * a list of function names, and an array of Tcl_Token's within a Tcl_Parse
 * struct.  The tree is composed of OpNodes.
 */

typedef struct OpNode {
    int left;			/* "Pointer" to the left operand. */
    int right;			/* "Pointer" to the right operand. */
    union {
	int parent;		/* "Pointer" to the parent operand. */
	int prev;		/* "Pointer" joining incomplete tree stack */
    } p;
    unsigned char lexeme;	/* Code that identifies the operator. */
    unsigned char precedence;	/* Precedence of the operator */
    unsigned char mark;		/* Mark used to control traversal. */
    unsigned char constant;	/* Flag marking constant subexpressions. */
} OpNode;

/*
 * The storage for the tree is dynamically allocated array of OpNodes.  The
 * array is grown as parsing needs dictate according to a scheme similar to
 * Tcl's string growth algorithm, so that the resizing costs are O(N) and so
 * that we use at least half the memory allocated as expressions get large.
 *
 * Each OpNode in the tree represents an operator in the expression, either
 * unary or binary.  When parsing is completed successfully, a binary operator
 * OpNode will have its left and right fields filled with "pointers" to its
 * left and right operands.  A unary operator OpNode will have its right field
 * filled with a pointer to its single operand.  When an operand is a
 * subexpression the "pointer" takes the form of the index -- a non-negative
 * integer -- into the OpNode storage array where the root of that
 * subexpression parse tree is found.  
 *
 * Non-operator elements of the expression do not get stored in the OpNode
 * tree.  They are stored in the other structures according to their type.
 * Literal values get appended to the literal list.  Elements that denote
 * forms of quoting or substitution known to the Tcl parser get stored as
 * Tcl_Tokens.  These non-operator elements of the expression are the
 * leaves of the completed parse tree.  When an operand of an OpNode is
 * one of these leaf elements, the following negative integer codes are used
 * to indicate which kind of elements it is.
 */

enum OperandTypes {
    OT_LITERAL = -3,	/* Operand is a literal in the literal list */
    OT_TOKENS = -2,	/* Operand is sequence of Tcl_Tokens */
    OT_EMPTY = -1	/* "Operand" is an empty string.  This is a
			 * special case used only to represent the
			 * EMPTY lexeme.  See below. */
};

/*
 * Readable macros to test whether a "pointer" value points to an operator.
 * They operate on the "non-negative integer -> operator; negative integer ->
 * a non-operator OperandType" distinction.
 */

#define IsOperator(l)	((l) >= 0)
#define NotOperator(l)	((l) < 0)

/*
 * Note that it is sufficient to store in the tree just the type of leaf
 * operand, without any explicit pointer to which leaf.  This is true because
 * the traversals of the completed tree we perform are known to visit
 * the leaves in the same order as the original parse.
 *
 * In a completed parse tree, those OpNodes that are themselves (roots of
 * subexpression trees that are) operands of some operator store in their
 * p.parent field a "pointer" to the OpNode of that operator.  The p.parent
 * field permits a traversal of the tree within a * non-recursive routine
 * (ConvertTreeToTokens() and CompileExprTree()).  This means that even
 * expression trees of great depth pose no risk of blowing the C stack.
 *
 * While the parse tree is being constructed, the same memory space is used
 * to hold the p.prev field which chains together a stack of incomplete
 * trees awaiting their right operands.
 *
 * The lexeme field is filled in with the lexeme of the operator that is
 * returned by the ParseLexeme() routine.  Only lexemes for unary and
 * binary operators get stored in an OpNode.  Other lexmes get different
 * treatement.
 *
 * The precedence field provides a place to store the precedence of the
 * operator, so it need not be looked up again and again.
 *
 * The mark field is use to control the traversal of the tree, so
 * that it can be done non-recursively.  The mark values are:
 */

enum Marks {
    MARK_LEFT,		/* Next step of traversal is to visit left subtree */
    MARK_RIGHT,		/* Next step of traversal is to visit right subtree */
    MARK_PARENT,	/* Next step of traversal is to return to parent */
};

/*
 * The constant field is a boolean flag marking which subexpressions are
 * completely known at compile time, and are eligible for computing then
 * rather than waiting until run time.
 */

/*
 * Each lexeme belongs to one of four categories, which determine
 * its place in the parse tree.  We use the two high bits of the
 * (unsigned char) value to store a NODE_TYPE code.
 */

#define NODE_TYPE	0xC0

/*
 * The four category values are LEAF, UNARY, and BINARY, explained below,
 * and "uncategorized", which is used either temporarily, until context
 * determines which of the other three categories is correct, or for
 * lexemes like INVALID, which aren't really lexemes at all, but indicators
 * of a parsing error.  Note that the codes must be distinct to distinguish
 * categories, but need not take the form of a bit array.
 */

#define BINARY		0x40	/* This lexeme is a binary operator.  An
				 * OpNode representing it should go into the
				 * parse tree, and two operands should be
				 * parsed for it in the expression.  */
#define UNARY		0x80	/* This lexeme is a unary operator.  An OpNode
				 * representing it should go into the parse
				 * tree, and one operand should be parsed for
				 * it in the expression. */
#define LEAF		0xC0	/* This lexeme is a leaf operand in the parse
				 * tree.  No OpNode will be placed in the tree
				 * for it.  Either a literal value will be
				 * appended to the list of literals in this
				 * expression, or appropriate Tcl_Tokens will
				 * be appended in a Tcl_Parse struct to 
				 * represent those leaves that require some
				 * form of substitution.
				 */

/* Uncategorized lexemes */

#define PLUS		1	/* Ambiguous.  Resolves to UNARY_PLUS or
				 * BINARY_PLUS according to context. */
#define MINUS		2	/* Ambiguous.  Resolves to UNARY_MINUS or
				 * BINARY_MINUS according to context. */
#define BAREWORD	3	/* Ambigous.  Resolves to BOOLEAN or to
				 * FUNCTION or a parse error according to
				 * context and value. */
#define INCOMPLETE	4	/* A parse error.  Used only when the single
				 * "=" is encountered.  */
#define INVALID		5	/* A parse error.  Used when any punctuation
				 * appears that's not a supported operator. */

/* Leaf lexemes */

#define NUMBER		( LEAF | 1)	/* For literal numbers */
#define SCRIPT		( LEAF | 2)	/* Script substitution; [foo] */
#define BOOLEAN		( LEAF | BAREWORD)	/* For literal booleans */
#define BRACED		( LEAF | 4)	/* Braced string; {foo bar} */
#define VARIABLE	( LEAF | 5)	/* Variable substitution; $x */
#define QUOTED		( LEAF | 6)	/* Quoted string; "foo $bar [soom]" */
#define EMPTY		( LEAF | 7)	/* Used only for an empty argument
					 * list to a function.  Represents
					 * the empty string within parens in
					 * the expression: rand() */

/* Unary operator lexemes */

#define UNARY_PLUS	( UNARY | PLUS)
#define UNARY_MINUS	( UNARY | MINUS)
#define FUNCTION	( UNARY | BAREWORD)	/* This is a bit of "creative
					 * interpretation" on the part of the
					 * parser.  A function call is parsed
					 * into the parse tree according to
					 * the perspective that the function
					 * name is a unary operator and its
					 * argument list, enclosed in parens,
					 * is its operand.  The additional
					 * requirements not implied generally
					 * by treatment as a unary operator --
					 * for example, the requirement that
					 * the operand be enclosed in parens --
					 * are hard coded in the relevant
					 * portions of ParseExpr().  We trade
					 * off the need to include such
					 * exceptional handling in the code
					 * against the need we would otherwise
					 * have for more lexeme categories. */
#define START		( UNARY | 4)	/* This lexeme isn't parsed from the
					 * expression text at all.  It
					 * represents the start of the
					 * expression and sits at the root of
					 * the parse tree where it serves as
					 * the start/end point of traversals. */
#define OPEN_PAREN	( UNARY | 5)	/* Another bit of creative
					 * interpretation, where we treat "("
					 * as a unary operator with the
					 * sub-expression between it and its
					 * matching ")" as its operand. See
					 * CLOSE_PAREN below. */
#define NOT		( UNARY | 6)
#define BIT_NOT		( UNARY | 7)

/* Binary operator lexemes */

#define BINARY_PLUS	( BINARY |  PLUS)
#define BINARY_MINUS	( BINARY |  MINUS)
#define COMMA		( BINARY |  3)	/* The "," operator is a low precedence
					 * binary operator that separates the
					 * arguments in a function call.  The
					 * additional constraint that this
					 * operator can only legally appear
					 * at the right places within a
					 * function call argument list are
					 * hard coded within ParseExpr().  */
#define MULT		( BINARY |  4)
#define DIVIDE		( BINARY |  5)
#define MOD		( BINARY |  6)
#define LESS		( BINARY |  7)
#define GREATER		( BINARY |  8)
#define BIT_AND		( BINARY |  9)
#define BIT_XOR		( BINARY | 10)
#define BIT_OR		( BINARY | 11)
#define QUESTION	( BINARY | 12)	/* These two lexemes make up the */
#define COLON		( BINARY | 13)	/* ternary conditional operator,
					 * $x ? $y : $z .  We treat them as
					 * two binary operators to avoid
					 * another lexeme category, and
					 * code the additional constraints
					 * directly in ParseExpr().  For
					 * instance, the right operand of
					 * a "?" operator must be a ":"
					 * operator. */
#define LEFT_SHIFT	( BINARY | 14)
#define RIGHT_SHIFT	( BINARY | 15)
#define LEQ		( BINARY | 16)
#define GEQ		( BINARY | 17)
#define EQUAL		( BINARY | 18)
#define NEQ		( BINARY | 19)
#define AND		( BINARY | 20)
#define OR		( BINARY | 21)
#define STREQ		( BINARY | 22)
#define STRNEQ		( BINARY | 23)
#define EXPON		( BINARY | 24)	/* Unlike the other binary operators,
					 * EXPON is right associative and this
					 * distinction is coded directly in
					 * ParseExpr(). */
#define IN_LIST		( BINARY | 25)
#define NOT_IN_LIST	( BINARY | 26)
#define CLOSE_PAREN	( BINARY | 27)	/* By categorizing the CLOSE_PAREN
					 * lexeme as a BINARY operator, the
					 * normal parsing rules for binary
					 * operators assure that a close paren
					 * will not directly follow another
					 * operator, and the machinery already
					 * in place to connect operands to
					 * operators according to precedence
					 * performs most of the work of
					 * matching open and close parens for
					 * us.  In the end though, a close
					 * paren is not really a binary
					 * operator, and some special coding
					 * in ParseExpr() make sure we never
					 * put an actual CLOSE_PAREN node
					 * in the parse tree.   The
					 * sub-expression between parens
					 * becomes the single argument of
					 * the matching OPEN_PAREN unary
					 * operator. */
#define END		( BINARY | 28)	/* This lexeme represents the end of
					 * the string being parsed.  Treating
					 * it as a binary operator follows the
					 * same logic as the CLOSE_PAREN lexeme
					 * and END pairs with START, in the
					 * same way that CLOSE_PAREN pairs with
					 * OPEN_PAREN. */
/*
 * When ParseExpr() builds the parse tree it must choose which operands to
 * connect to which operators.  This is done according to operator precedence.
 * The greater an operator's precedence the greater claim it has to link to
 * an available operand.  The Precedence enumeration lists the precedence
 * values used by Tcl expression operators, from lowest to highest claim.
 * Each precedence level is commented with the operators that hold that
 * precedence.
 */

enum Precedence {
    PREC_END = 1,	/* END */
    PREC_START,		/* START */
    PREC_CLOSE_PAREN,	/* ")" */
    PREC_OPEN_PAREN,	/* "(" */
    PREC_COMMA,		/* "," */
    PREC_CONDITIONAL,	/* "?", ":" */
    PREC_OR,		/* "||" */
    PREC_AND,		/* "&&" */
    PREC_BIT_OR,	/* "|" */
    PREC_BIT_XOR,	/* "^" */
    PREC_BIT_AND,	/* "&" */
    PREC_EQUAL,		/* "==", "!=", "eq", "ne", "in", "ni" */
    PREC_COMPARE,	/* "<", ">", "<=", ">=" */
    PREC_SHIFT,		/* "<<", ">>" */
    PREC_ADD,		/* "+", "-" */
    PREC_MULT,		/* "*", "/", "%" */
    PREC_EXPON,		/* "**" */
    PREC_UNARY		/* "+", "-", FUNCTION, "!", "~" */
};

/*
 * Here the same information contained in the comments above is stored
 * in inverted form, so that given a lexeme, one can quickly look up 
 * its precedence value.
 */

static const unsigned char prec[] = {
    /* Non-operator lexemes */
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,
    /* Binary operator lexemes */
    PREC_ADD,		/* BINARY_PLUS */
    PREC_ADD,		/* BINARY_MINUS */
    PREC_COMMA,		/* COMMA */
    PREC_MULT,		/* MULT */
    PREC_MULT,		/* DIVIDE */
    PREC_MULT,		/* MOD */
    PREC_COMPARE,	/* LESS */
    PREC_COMPARE,	/* GREATER */
    PREC_BIT_AND,	/* BIT_AND */
    PREC_BIT_XOR,	/* BIT_XOR */
    PREC_BIT_OR,	/* BIT_OR */
    PREC_CONDITIONAL,	/* QUESTION */
    PREC_CONDITIONAL,	/* COLON */
    PREC_SHIFT,		/* LEFT_SHIFT */
    PREC_SHIFT,		/* RIGHT_SHIFT */
    PREC_COMPARE,	/* LEQ */
    PREC_COMPARE,	/* GEQ */
    PREC_EQUAL,		/* EQUAL */
    PREC_EQUAL,		/* NEQ */
    PREC_AND,		/* AND */
    PREC_OR,		/* OR */
    PREC_EQUAL,		/* STREQ */
    PREC_EQUAL,		/* STRNEQ */
    PREC_EXPON,		/* EXPON */
    PREC_EQUAL,		/* IN_LIST */
    PREC_EQUAL,		/* NOT_IN_LIST */
    PREC_CLOSE_PAREN,	/* CLOSE_PAREN */
    PREC_END,		/* END */
    /* Expansion room for more binary operators */
    0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  
    /* Unary operator lexemes */
    PREC_UNARY,		/* UNARY_PLUS */
    PREC_UNARY,		/* UNARY_MINUS */
    PREC_UNARY,		/* FUNCTION */
    PREC_START,		/* START */
    PREC_OPEN_PAREN,	/* OPEN_PAREN */
    PREC_UNARY,		/* NOT*/
    PREC_UNARY,		/* BIT_NOT*/
};

/*
 * A table mapping lexemes to bytecode instructions, used by CompileExprTree().
 */

static const unsigned char instruction[] = {
    /* Non-operator lexemes */
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,
    /* Binary operator lexemes */
    INST_ADD,		/* BINARY_PLUS */
    INST_SUB,		/* BINARY_MINUS */
    0,			/* COMMA */
    INST_MULT,		/* MULT */
    INST_DIV,		/* DIVIDE */
    INST_MOD,		/* MOD */
    INST_LT,		/* LESS */
    INST_GT,		/* GREATER */
    INST_BITAND,	/* BIT_AND */
    INST_BITXOR,	/* BIT_XOR */
    INST_BITOR,		/* BIT_OR */
    0,			/* QUESTION */
    0,			/* COLON */
    INST_LSHIFT,	/* LEFT_SHIFT */
    INST_RSHIFT,	/* RIGHT_SHIFT */
    INST_LE,		/* LEQ */
    INST_GE,		/* GEQ */
    INST_EQ,		/* EQUAL */
    INST_NEQ,		/* NEQ */
    0,			/* AND */
    0,			/* OR */
    INST_STR_EQ,	/* STREQ */
    INST_STR_NEQ,	/* STRNEQ */
    INST_EXPON,		/* EXPON */
    INST_LIST_IN,	/* IN_LIST */
    INST_LIST_NOT_IN,	/* NOT_IN_LIST */
    0,			/* CLOSE_PAREN */
    0,			/* END */
    /* Expansion room for more binary operators */
    0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  
    /* Unary operator lexemes */
    INST_UPLUS,		/* UNARY_PLUS */
    INST_UMINUS,	/* UNARY_MINUS */
    0,			/* FUNCTION */
    0,			/* START */
    0,			/* OPEN_PAREN */
    INST_LNOT,		/* NOT*/
    INST_BITNOT,	/* BIT_NOT*/
};

/*
 * A table mapping a byte value to the corresponding lexeme for use by
 * ParseLexeme().
 */

static unsigned char Lexeme[] = {
	INVALID		/* NUL */,	INVALID		/* SOH */,
	INVALID		/* STX */,	INVALID		/* ETX */,
	INVALID		/* EOT */,	INVALID		/* ENQ */,
	INVALID		/* ACK */,	INVALID		/* BEL */,
	INVALID		/* BS */,	INVALID		/* HT */,
	INVALID		/* LF */,	INVALID		/* VT */,
	INVALID		/* FF */,	INVALID		/* CR */,
	INVALID		/* SO */,	INVALID		/* SI */,
	INVALID		/* DLE */,	INVALID		/* DC1 */,
	INVALID		/* DC2 */,	INVALID		/* DC3 */,
	INVALID		/* DC4 */,	INVALID		/* NAK */,
	INVALID		/* SYN */,	INVALID		/* ETB */,
	INVALID		/* CAN */,	INVALID		/* EM */,
	INVALID		/* SUB */,	INVALID		/* ESC */,
	INVALID		/* FS */,	INVALID		/* GS */,
	INVALID		/* RS */,	INVALID		/* US */,
	INVALID		/* SPACE */,	0 		/* ! or != */,
	QUOTED		/* " */,	INVALID		/* # */,
	VARIABLE	/* $ */,	MOD		/* % */,
	0		/* & or && */,	INVALID		/* ' */,
	OPEN_PAREN	/* ( */,	CLOSE_PAREN	/* ) */,
	0		/* * or ** */,	PLUS		/* + */,
	COMMA		/* , */,	MINUS		/* - */,
	0		/* . */,	DIVIDE		/* / */,
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0,			/* 0-9 */
	COLON		/* : */,	INVALID		/* ; */,
	0		/* < or << or <= */,
	0		/* == or INVALID */,
	0		/* > or >> or >= */,
	QUESTION	/* ? */,	INVALID		/* @ */,
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,		/* A-M */
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,		/* N-Z */
	SCRIPT		/* [ */,	INVALID		/* \ */,
	INVALID		/* ] */,	BIT_XOR		/* ^ */,
	INVALID		/* _ */,	INVALID		/* ` */,
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,		/* a-m */
	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,		/* n-z */
	BRACED		/* { */,	0		/* | or || */,
	INVALID		/* } */,	BIT_NOT		/* ~ */,
	INVALID		/* DEL */
};

/*
 * The JumpList struct is used to create a stack of data needed for the
 * TclEmitForwardJump() and TclFixupForwardJump() calls that are performed
 * when compiling the short-circuiting operators QUESTION/COLON, AND, and OR.
 * Keeping a stack permits the CompileExprTree() routine to be non-recursive.
 */

typedef struct JumpList {
    JumpFixup jump;		/* Pass this argument to matching calls of
				 * TclEmitForwardJump() and 
				 * TclFixupForwardJump(). */
    int depth;			/* Remember the currStackDepth of the
				 * CompileEnv here. */
    int offset;			/* Data used to compute jump lengths to pass
				 * to TclFixupForwardJump() */
    int convert;		/* Temporary storage used to compute whether
				 * numeric conversion will be needed following
				 * the operator we're compiling. */
    struct JumpList *next;	/* Point to next item on the stack */
} JumpList;

/*
 * Declarations for local functions to this file:
 */

static void		CompileExprTree(Tcl_Interp *interp, OpNode *nodes,
			    int index, Tcl_Obj *const **litObjvPtr,
			    Tcl_Obj *const *funcObjv, Tcl_Token *tokenPtr,
			    CompileEnv *envPtr, int optimize);
static void		ConvertTreeToTokens(const char *start, int numBytes,
			    OpNode *nodes, Tcl_Token *tokenPtr,
			    Tcl_Parse *parsePtr);
static int		ExecConstantExprTree(Tcl_Interp *interp, OpNode *nodes,
			    int index, Tcl_Obj * const **litObjvPtr);
static int		ParseExpr(Tcl_Interp *interp, const char *start,
			    int numBytes, OpNode **opTreePtr,
			    Tcl_Obj *litList, Tcl_Obj *funcList,
			    Tcl_Parse *parsePtr, int parseOnly);
static int		ParseLexeme(const char *start, int numBytes,
			    unsigned char *lexemePtr, Tcl_Obj **literalPtr);


/*
 *----------------------------------------------------------------------
 *
 * ParseExpr --
 *
 *	Given a string, the numBytes bytes starting at start, this function
 *	parses it as a Tcl expression and constructs a tree representing
 *	the structure of the expression.  The caller must pass in empty
 * 	lists as the funcList and litList arguments.  The elements of the
 *	parsed expression are returned to the caller as that tree, a list of
 *	literal values, a list of function names, and in Tcl_Tokens
 *	added to a Tcl_Parse struct passed in by the caller.
 *
 * Results:
 *	If the string is successfully parsed as a valid Tcl expression, TCL_OK
 *	is returned, and data about the expression structure is written to
 *	the last four arguments.  If the string cannot be parsed as a valid
 *	Tcl expression, TCL_ERROR is returned, and if interp is non-NULL, an
 *	error message is written to interp.
 *
 * Side effects:
 *	Memory will be allocated.  If TCL_OK is returned, the caller must
 *	clean up the returned data structures.  The (OpNode *) value written
 *	to opTreePtr should be passed to ckfree() and the parsePtr argument
 *	should be passed to Tcl_FreeParse().  The elements appended to the
 *	litList and funcList will automatically be freed whenever the
 *	refcount on those lists indicates they can be freed.
 *
 *----------------------------------------------------------------------
 */

static int
ParseExpr(
    Tcl_Interp *interp,		/* Used for error reporting. */
    const char *start,		/* Start of source string to parse. */
    int numBytes,		/* Number of bytes in string. */
    OpNode **opTreePtr,		/* Points to space where a pointer to the
				 * allocated OpNode tree should go. */
    Tcl_Obj *litList,		/* List to append literals to. */
    Tcl_Obj *funcList,		/* List to append function names to. */
    Tcl_Parse *parsePtr,	/* Structure to fill with tokens representing
				 * those operands that require run time
				 * substitutions. */
    int parseOnly)		/* A boolean indicating whether the caller's
				 * aim is just a parse, or whether it will go
				 * on to compile the expression.  Different
				 * optimizations are appropriate for the
				 * two scenarios. */
{
    OpNode *nodes = NULL;	/* Pointer to the OpNode storage array where
				 * we build the parse tree. */
    int nodesAvailable = 64;	/* Initial size of the storage array.  This
				 * value establishes a minimum tree memory cost
				 * of only about 1 kibyte, and is large enough
				 * for most expressions to parse with no need
				 * for array growth and reallocation. */
    int nodesUsed = 0;		/* Number of OpNodes filled. */
    int scanned = 0;		/* Capture number of byte scanned by 
				 * parsing routines. */
    int lastParsed;		/* Stores info about what the lexeme parsed
				 * the previous pass through the parsing loop
				 * was.  If it was an operator, lastParsed is
				 * the index of the OpNode for that operator.
				 * If it was not an operator, lastParsed holds
				 * an OperandTypes value encoding what we
				 * need to know about it. */
    int incomplete;		/* Index of the most recent incomplete tree
				 * in the OpNode array.  Heads a stack of
				 * incomplete trees linked by p.prev. */
    int complete = OT_EMPTY;	/* "Index" of the complete tree (that is, a
				 * complete subexpression) determined at the
				 * moment.   OT_EMPTY is a nonsense value
				 * used only to silence compiler warnings.
				 * During a parse, complete will always hold
				 * an index or an OperandTypes value pointing
				 * to an actual leaf at the time the complete
				 * tree is needed. */

    /* These variables control generation of the error message. */
    Tcl_Obj *msg = NULL;	/* The error message. */
    Tcl_Obj *post = NULL;	/* In a few cases, an additional postscript
				 * for the error message, supplying more
				 * information after the error msg and
				 * location have been reported. */
    const char *mark = "_@_";	/* In the portion of the complete error message
				 * where the error location is reported, this
				 * "mark" substring is inserted into the
				 * string being parsed to aid in pinpointing
				 * the location of the syntax error in the
				 * expression. */
    int insertMark = 0;		/* A boolean controlling whether the "mark"
				 * should be inserted. */
    const int limit = 25;	/* Portions of the error message are
				 * constructed out of substrings of the
				 * original expression.  In order to keep the
				 * error message readable, we impose this limit
				 * on the substring size we extract. */

    TclParseInit(interp, start, numBytes, parsePtr);

    nodes = (OpNode *) attemptckalloc(nodesAvailable * sizeof(OpNode));
    if (nodes == NULL) {
	TclNewLiteralStringObj(msg, "not enough memory to parse expression");
	goto error;
    }

    /* Initialize the parse tree with the special "START" node. */
    nodes->lexeme = START;
    nodes->precedence = prec[START];
    nodes->mark = MARK_RIGHT;
    nodes->constant = 1;
    incomplete = lastParsed = nodesUsed;
    nodesUsed++;

    /*
     * Main parsing loop parses one lexeme per iteration.  We exit the
     * loop only when there's a syntax error with a "goto error" which
     * takes us to the error handling code following the loop, or when
     * we've successfully completed the parse and we return to the caller.
     */

    while (1) {
	OpNode *nodePtr;	/* Points to the OpNode we may fill this
				 * pass through the loop. */
	unsigned char lexeme;	/* The lexeme we parse this iteration. */
	Tcl_Obj *literal;	/* Filled by the ParseLexeme() call when
				 * a literal is parsed that has a Tcl_Obj
				 * rep worth preserving. */
	const char *lastStart = start - scanned;
				/* Compute where the lexeme parsed the
				 * previous pass through the loop began.
				 * This is helpful for detecting invalid
				 * octals and providing more complete error
				 * messages. */

	/*
	 * Each pass through this loop adds up to one more OpNode. Allocate
	 * space for one if required.
	 */

	if (nodesUsed >= nodesAvailable) {
	    int size = nodesUsed * 2;
	    OpNode *newPtr;

	    do {
		newPtr = (OpNode *) attemptckrealloc((char *) nodes,
			(unsigned int) size * sizeof(OpNode));
	    } while ((newPtr == NULL)
		    && ((size -= (size - nodesUsed) / 2) > nodesUsed));
	    if (newPtr == NULL) {
		TclNewLiteralStringObj(msg,
			"not enough memory to parse expression");
		goto error;
	    }
	    nodesAvailable = size;
	    nodes = newPtr;
	}
	nodePtr = nodes + nodesUsed;

	/* Skip white space between lexemes. */
	scanned = TclParseAllWhiteSpace(start, numBytes);
	start += scanned;
	numBytes -= scanned;

	scanned = ParseLexeme(start, numBytes, &lexeme, &literal);

	/* Use context to categorize the lexemes that are ambiguous. */
	if ((NODE_TYPE & lexeme) == 0) {
	    switch (lexeme) {
	    case INVALID:
		msg = Tcl_ObjPrintf(
			"invalid character \"%.*s\"", scanned, start);
		goto error;
	    case INCOMPLETE:
		msg = Tcl_ObjPrintf(
			"incomplete operator \"%.*s\"", scanned, start);
		goto error;
	    case BAREWORD:

		/*
		 * Most barewords in an expression are a syntax error.
		 * The exceptions are that when a bareword is followed by
		 * an open paren, it might be a function call, and when the
		 * bareword is a legal literal boolean value, we accept that 
		 * as well.
		 */

		if (start[scanned+TclParseAllWhiteSpace(
			start+scanned, numBytes-scanned)] == '(') {
		    lexeme = FUNCTION;

		    /*
		     * When we compile the expression we'll need the function
		     * name, and there's no place in the parse tree to store
		     * it, so we keep a separate list of all the function
		     * names we've parsed in the order we found them.
		     */

		    Tcl_ListObjAppendElement(NULL, funcList, literal);
		} else {
		    int b;
		    if (Tcl_GetBooleanFromObj(NULL, literal, &b) == TCL_OK) {
			lexeme = BOOLEAN;
		    } else {
			Tcl_DecrRefCount(literal);
			msg = Tcl_ObjPrintf(
				"invalid bareword \"%.*s%s\"",
				(scanned < limit) ? scanned : limit - 3, start,
				(scanned < limit) ? "" : "...");
			post = Tcl_ObjPrintf(
				"should be \"$%.*s%s\" or \"{%.*s%s}\"",
				(scanned < limit) ? scanned : limit - 3,
				start, (scanned < limit) ? "" : "...",
				(scanned < limit) ? scanned : limit - 3,
				start, (scanned < limit) ? "" : "...");
			Tcl_AppendPrintfToObj(post,
				" or \"%.*s%s(...)\" or ...",
				(scanned < limit) ? scanned : limit - 3,
				start, (scanned < limit) ? "" : "...");
			if (NotOperator(lastParsed)) {
			    if ((lastStart[0] == '0')
				    && ((lastStart[1] == 'o')
				    || (lastStart[1] == 'O'))
				    && (lastStart[2] >= '0')
				    && (lastStart[2] <= '9')) {
				const char *end = lastStart + 2;
				Tcl_Obj* copy;
				while (isdigit(*end)) {
				    end++;
				}
				copy = Tcl_NewStringObj(lastStart,
					end - lastStart);
				if (TclCheckBadOctal(NULL,
					Tcl_GetString(copy))) {
					TclNewLiteralStringObj(post,
						"(invalid octal number?)");
				}
				Tcl_DecrRefCount(copy);
			    }
			    scanned = 0;
			    insertMark = 1;
			    parsePtr->errorType = TCL_PARSE_BAD_NUMBER;
			}
			goto error;
		    }
		}
		break;
	    case PLUS:
	    case MINUS:
		if (IsOperator(lastParsed)) {

		    /*
		     * A "+" or "-" coming just after another operator
		     * must be interpreted as a unary operator.
		     */

		    lexeme |= UNARY;
		} else {
		    lexeme |= BINARY;
		}
	    }
	}	/* Uncategorized lexemes */

	/* Handle lexeme based on its category. */
	switch (NODE_TYPE & lexeme) {

	/*
	 * Each LEAF results in either a literal getting appended to the
	 * litList, or a sequence of Tcl_Tokens representing a Tcl word
	 * getting appended to the parsePtr->tokens.  No OpNode is filled
	 * for this lexeme.
	 */

	case LEAF: {
	    Tcl_Token *tokenPtr;
	    const char *end = start;
	    int wordIndex;
	    int code = TCL_OK;

	    /*
	     * A leaf operand appearing just after something that's not an
	     * operator is a syntax error.
	     */

	    if (NotOperator(lastParsed)) {
		msg = Tcl_ObjPrintf("missing operator at %s", mark);
		if (lastStart[0] == '0') {
		    Tcl_Obj *copy = Tcl_NewStringObj(lastStart,
			    start + scanned - lastStart);
		    if (TclCheckBadOctal(NULL, Tcl_GetString(copy))) {
			TclNewLiteralStringObj(post,
				"looks like invalid octal number");
		    }
		    Tcl_DecrRefCount(copy);
		}
		scanned = 0;
		insertMark = 1;
		parsePtr->errorType = TCL_PARSE_BAD_NUMBER;

		/* Free any literal to avoid a memleak. */
		if ((lexeme == NUMBER) || (lexeme == BOOLEAN)) {
		    Tcl_DecrRefCount(literal);
		}
		goto error;
	    }

	    switch (lexeme) {
	    case NUMBER:
	    case BOOLEAN: {
		if (interp) {
		    int new;
		    /* LiteralEntry *lePtr; */
		    Tcl_Obj *objPtr = TclCreateLiteral((Interp *)interp,
			    (char *)start, scanned,
			    /* hash */ (unsigned int) -1, &new,
			    /* nsPtr */ NULL, /* flags */ 0,
			    NULL /* &lePtr */);
		    if (objPtr->typePtr != literal->typePtr) {
			/*
			 * What we would like to do is this:
			 *
			 * lePtr->objPtr = literal;
			 * Tcl_IncrRefCount(literal);
			 * Tcl_DecrRefCount(objPtr);
			 *
			 * However, the design of the "global" and "local"
			 * LiteralTable does not permit the value of
			 * lePtr->objPtr to be changed.  So rather than
			 * replace lePtr->objPtr, we do surgery to transfer
			 * the intrep of literal into it.  Ugly stuff here
			 * that's generally unsafe, but ok here since we know
			 * the Tcl_ObjTypes literal might possibly have.
			 */
			Tcl_Obj *toFree = literal;
			literal = objPtr;
			TclFreeIntRep(literal);
			literal->typePtr = toFree->typePtr;
			literal->internalRep = toFree->internalRep;
			toFree->typePtr = NULL;
			Tcl_DecrRefCount(toFree);
		    }
		}

		Tcl_ListObjAppendElement(NULL, litList, literal);
		complete = lastParsed = OT_LITERAL;
		start += scanned;
		numBytes -= scanned;
		continue;
	    }
	    default:
		break;
	    }

	    /*
	     * Remaining LEAF cases may involve filling Tcl_Tokens, so
	     * make room for at least 2 more tokens.
	     */

	    if (parsePtr->numTokens+1 >= parsePtr->tokensAvailable) {
		TclExpandTokenArray(parsePtr);
	    }
	    wordIndex = parsePtr->numTokens;
	    tokenPtr = parsePtr->tokenPtr + wordIndex;
	    tokenPtr->type = TCL_TOKEN_WORD;
	    tokenPtr->start = start;
	    parsePtr->numTokens++;

	    switch (lexeme) {
	    case QUOTED:
		code = Tcl_ParseQuotedString(NULL, start, numBytes,
			parsePtr, 1, &end);
		scanned = end - start;
		break;

	    case BRACED:
		code = Tcl_ParseBraces(NULL, start, numBytes,
			    parsePtr, 1, &end);
		scanned = end - start;
		break;

	    case VARIABLE:
		code = Tcl_ParseVarName(NULL, start, numBytes, parsePtr, 1);

		/*
		 * Handle the quirk that Tcl_ParseVarName reports a successful
		 * parse even when it gets only a "$" with no variable name.
		 */

		tokenPtr = parsePtr->tokenPtr + wordIndex + 1;
		if (code == TCL_OK && tokenPtr->type != TCL_TOKEN_VARIABLE) {
		    TclNewLiteralStringObj(msg, "invalid character \"$\"");
		    goto error;
		}
		scanned = tokenPtr->size;
		break;

	    case SCRIPT: {
		Tcl_Parse *nestedPtr =
			(Tcl_Parse *) TclStackAlloc(interp, sizeof(Tcl_Parse));

		tokenPtr = parsePtr->tokenPtr + parsePtr->numTokens;
		tokenPtr->type = TCL_TOKEN_COMMAND;
		tokenPtr->start = start;
		tokenPtr->numComponents = 0;

		end = start + numBytes;
		start++;
		while (1) {
		    code = Tcl_ParseCommand(interp, start, (end - start), 1,
			    nestedPtr);
		    if (code != TCL_OK) {
			parsePtr->term = nestedPtr->term;
			parsePtr->errorType = nestedPtr->errorType;
			parsePtr->incomplete = nestedPtr->incomplete;
			break;
		    }
		    start = (nestedPtr->commandStart + nestedPtr->commandSize);
		    Tcl_FreeParse(nestedPtr);
		    if ((nestedPtr->term < end) && (*(nestedPtr->term) == ']')
			    && !(nestedPtr->incomplete)) {
			break;
		    }

		    if (start == end) {
			TclNewLiteralStringObj(msg, "missing close-bracket");
			parsePtr->term = tokenPtr->start;
			parsePtr->errorType = TCL_PARSE_MISSING_BRACKET;
			parsePtr->incomplete = 1;
			code = TCL_ERROR;
			break;
		    }
		}
		TclStackFree(interp, nestedPtr);
		end = start;
		start = tokenPtr->start;
		scanned = end - start;
		tokenPtr->size = scanned;
		parsePtr->numTokens++;
		break;
	    }
	    }
	    if (code != TCL_OK) {

		/*
		 * Here we handle all the syntax errors generated by
		 * the Tcl_Token generating parsing routines called in the
		 * switch just above.  If the value of parsePtr->incomplete
		 * is 1, then the error was an unbalanced '[', '(', '{',
		 * or '"' and parsePtr->term is pointing to that unbalanced
		 * character.  If the value of parsePtr->incomplete is 0,
		 * then the error is one of lacking whitespace following a
		 * quoted word, for example: expr {[an error {foo}bar]},
		 * and parsePtr->term points to where the whitespace is
		 * missing.  We reset our values of start and scanned so that
		 * when our error message is constructed, the location of
		 * the syntax error is sure to appear in it, even if the
		 * quoted expression is truncated.
		 */

		start = parsePtr->term;
		scanned = parsePtr->incomplete;
		goto error;
	    }

	    tokenPtr = parsePtr->tokenPtr + wordIndex;
	    tokenPtr->size = scanned;
	    tokenPtr->numComponents = parsePtr->numTokens - wordIndex - 1;
	    if (!parseOnly && ((lexeme == QUOTED) || (lexeme == BRACED))) {

		/*
		 * When this expression is destined to be compiled, and a
		 * braced or quoted word within an expression is known at
		 * compile time (no runtime substitutions in it), we can
		 * store it as a literal rather than in its tokenized form.
		 * This is an advantage since the compiled bytecode is going
		 * to need the argument in Tcl_Obj form eventually, so it's
		 * just as well to get there now.  Another advantage is that
		 * with this conversion, larger constant expressions might
		 * be grown and optimized.
		 *
		 * On the contrary, if the end goal of this parse is to
		 * fill a Tcl_Parse for a caller of Tcl_ParseExpr(), then it's
		 * wasteful to convert to a literal only to convert back again
		 * later.
		 */

		literal = Tcl_NewObj();
		if (TclWordKnownAtCompileTime(tokenPtr, literal)) {
		    Tcl_ListObjAppendElement(NULL, litList, literal);
		    complete = lastParsed = OT_LITERAL;
		    parsePtr->numTokens = wordIndex;
		    break;
		}
		Tcl_DecrRefCount(literal);
	    }
	    complete = lastParsed = OT_TOKENS;
	    break;
	} /* case LEAF */

	case UNARY:

	    /*
	     * A unary operator appearing just after something that's not an
	     * operator is a syntax error -- something trying to be the left
	     * operand of an operator that doesn't take one.
	     */

	    if (NotOperator(lastParsed)) {
		msg = Tcl_ObjPrintf("missing operator at %s", mark);
		scanned = 0;
		insertMark = 1;
		goto error;
	    }

	    /* Create an OpNode for the unary operator */
	    nodePtr->lexeme = lexeme;
	    nodePtr->precedence = prec[lexeme];
	    nodePtr->mark = MARK_RIGHT;

	    /*
	     * A FUNCTION cannot be a constant expression, because Tcl allows
	     * functions to return variable results with the same arguments;
	     * for example, rand().  Other unary operators can root a constant
	     * expression, so long as the argument is a constant expression.
	     */

	    nodePtr->constant = (lexeme != FUNCTION);

	    /*
	     * This unary operator is a new incomplete tree, so push it
	     * onto our stack of incomplete trees.  Also remember it as
	     * the last lexeme we parsed.
	     */

	    nodePtr->p.prev = incomplete;
	    incomplete = lastParsed = nodesUsed;
	    nodesUsed++;
	    break;

	case BINARY: {
	    OpNode *incompletePtr;
	    unsigned char precedence = prec[lexeme];

	    /*
	     * A binary operator appearing just after another operator is a
	     * syntax error -- one of the two operators is missing an operand.
	     */

	    if (IsOperator(lastParsed)) {
		if ((lexeme == CLOSE_PAREN)
			&& (nodePtr[-1].lexeme == OPEN_PAREN)) {
		    if (nodePtr[-2].lexeme == FUNCTION) {

			/*
			 * Normally, "()" is a syntax error, but as a special
			 * case accept it as an argument list for a function.
			 * Treat this as a special LEAF lexeme, and restart
			 * the parsing loop with zero characters scanned.
			 * We'll parse the ")" again the next time through,
			 * but with the OT_EMPTY leaf as the subexpression
			 * between the parens.
			 */

			scanned = 0;
			complete = lastParsed = OT_EMPTY;
			break;
		    }
		    msg = Tcl_ObjPrintf("empty subexpression at %s", mark);
		    scanned = 0;
		    insertMark = 1;
		    goto error;
		}

		if (nodePtr[-1].precedence > precedence) {
		    if (nodePtr[-1].lexeme == OPEN_PAREN) {
			TclNewLiteralStringObj(msg, "unbalanced open paren");
			parsePtr->errorType = TCL_PARSE_MISSING_PAREN;
		    } else if (nodePtr[-1].lexeme == COMMA) {
			msg = Tcl_ObjPrintf(
				"missing function argument at %s", mark);
			scanned = 0;
			insertMark = 1;
		    } else if (nodePtr[-1].lexeme == START) {
			TclNewLiteralStringObj(msg, "empty expression");
		    }
		} else {
		    if (lexeme == CLOSE_PAREN) {
			TclNewLiteralStringObj(msg, "unbalanced close paren");
		    } else if ((lexeme == COMMA)
			    && (nodePtr[-1].lexeme == OPEN_PAREN)
			    && (nodePtr[-2].lexeme == FUNCTION)) {
			msg = Tcl_ObjPrintf(
				"missing function argument at %s", mark);
			scanned = 0;
			insertMark = 1;
		    }
		}
		if (msg == NULL) {
		    msg = Tcl_ObjPrintf("missing operand at %s", mark);
		    scanned = 0;
		    insertMark = 1;
		}
		goto error;
	    }

	    /*
	     * Here is where the tree comes together.  At this point, we
	     * have a stack of incomplete trees corresponding to 
	     * substrings that are incomplete expressions, followed by
	     * a complete tree corresponding to a substring that is itself
	     * a complete expression, followed by the binary operator we have
	     * just parsed.  The incomplete trees can each be completed by
	     * adding a right operand.
	     *
	     * To illustrate with an example, when we parse the expression
	     * "1+2*3-4" and we reach this point having just parsed the "-"
	     * operator, we have these incomplete trees: START, "1+", and
	     * "2*".  Next we have the complete subexpression "3".  Last is
	     * the "-" we've just parsed.
	     *
	     * The next step is to join our complete tree to an operator.
	     * The choice is governed by the precedence and associativity
	     * of the competing operators.  If we connect it as the right
	     * operand of our most recent incomplete tree, we get a new
	     * complete tree, and we can repeat the process.  The while
	     * loop following repeats this until precedence indicates it
	     * is time to join the complete tree as the left operand of
	     * the just parsed binary operator.
	     *
	     * Continuing the example, the first pass through the loop
	     * will join "3" to "2*"; the next pass will join "2*3" to
	     * "1+".  Then we'll exit the loop and join "1+2*3" to "-".
	     * When we return to parse another lexeme, our stack of
	     * incomplete trees is START and "1+2*3-".
	     */

	    while (1) {
		incompletePtr = nodes + incomplete;

		if (incompletePtr->precedence < precedence) {
		    break;
		}

		if (incompletePtr->precedence == precedence) {

		    /* Right association rules for exponentiation. */
		    if (lexeme == EXPON) {
			break;
		    }

		    /*
		     * Special association rules for the conditional operators.
		     * The "?" and ":" operators have equal precedence, but
		     * must be linked up in sensible pairs.
		     */

		    if ((incompletePtr->lexeme == QUESTION)
			    && (NotOperator(complete)
			    || (nodes[complete].lexeme != COLON))) {
			break;
		    }
		    if ((incompletePtr->lexeme == COLON)
			    && (lexeme == QUESTION)) {
			break;
		    }
		}

		/* Some special syntax checks... */

		/* Parens must balance */
		if ((incompletePtr->lexeme == OPEN_PAREN)
			&& (lexeme != CLOSE_PAREN)) {
		    TclNewLiteralStringObj(msg, "unbalanced open paren");
		    parsePtr->errorType = TCL_PARSE_MISSING_PAREN;
		    goto error;
		}

		/* Right operand of "?" must be ":" */
		if ((incompletePtr->lexeme == QUESTION)
			&& (NotOperator(complete)
			|| (nodes[complete].lexeme != COLON))) {
		    msg = Tcl_ObjPrintf(
			    "missing operator \":\" at %s", mark);
		    scanned = 0;
		    insertMark = 1;
		    goto error;
		}

		/* Operator ":" may only be right operand of "?" */
		if (IsOperator(complete)
			&& (nodes[complete].lexeme == COLON)
			&& (incompletePtr->lexeme != QUESTION)) {
		    TclNewLiteralStringObj(msg,
			    "unexpected operator \":\" "
			    "without preceding \"?\"");
		    goto error;
		}

		/*
		 * Attach complete tree as right operand of most recent
		 * incomplete tree.
		 */

		incompletePtr->right = complete;
		if (IsOperator(complete)) {
		    nodes[complete].p.parent = incomplete;
		    incompletePtr->constant = incompletePtr->constant
			    && nodes[complete].constant;
		} else {
		    incompletePtr->constant = incompletePtr->constant
			    && (complete == OT_LITERAL);
		}

		/*
		 * The QUESTION/COLON and FUNCTION/OPEN_PAREN combinations each
		 * make up a single operator.  Force them to agree whether they
		 * have a constant expression.
		 */

		if ((incompletePtr->lexeme == QUESTION)
			|| (incompletePtr->lexeme == FUNCTION)) {
		    nodes[complete].constant = incompletePtr->constant;
		}

		if (incompletePtr->lexeme == START) {

		    /*
		     * Completing the START tree indicates we're done.
		     * Transfer the parse tree to the caller and return.
		     */

		    *opTreePtr = nodes;
		    return TCL_OK;
		}

		/*
		 * With a right operand attached, last incomplete tree has
		 * become the complete tree.  Pop it from the incomplete
		 * tree stack.
		 */

		complete = incomplete;
		incomplete = incompletePtr->p.prev;

		/* CLOSE_PAREN can only close one OPEN_PAREN. */
		if (incompletePtr->lexeme == OPEN_PAREN) {
		    break;
		}
	    }

	    /* More syntax checks... */

	    /* Parens must balance. */
	    if (lexeme == CLOSE_PAREN) {
		if (incompletePtr->lexeme != OPEN_PAREN) {
		    TclNewLiteralStringObj(msg, "unbalanced close paren");
		    goto error;
		}
	    }

	    /* Commas must appear only in function argument lists. */
	    if (lexeme == COMMA) {
		if  ((incompletePtr->lexeme != OPEN_PAREN)
			|| (incompletePtr[-1].lexeme != FUNCTION)) {
		    TclNewLiteralStringObj(msg,
			    "unexpected \",\" outside function argument list");
		    goto error;
		}
	    }

	    /* Operator ":" may only be right operand of "?" */
	    if (IsOperator(complete) && (nodes[complete].lexeme == COLON)) {
		TclNewLiteralStringObj(msg,
			"unexpected operator \":\" without preceding \"?\"");
		goto error;
	    }

	    /* Create no node for a CLOSE_PAREN lexeme. */
	    if (lexeme == CLOSE_PAREN) {
		break;
	    }

	    /* Link complete tree as left operand of new node. */
	    nodePtr->lexeme = lexeme;
	    nodePtr->precedence = precedence;
	    nodePtr->mark = MARK_LEFT;
	    nodePtr->left = complete;

	    /* 
	     * The COMMA operator cannot be optimized, since the function
	     * needs all of its arguments, and optimization would reduce
	     * the number.  Other binary operators root constant expressions
	     * when both arguments are constant expressions.
	     */

	    nodePtr->constant = (lexeme != COMMA);

	    if (IsOperator(complete)) {
		nodes[complete].p.parent = nodesUsed;
		nodePtr->constant = nodePtr->constant
			&& nodes[complete].constant;
	    } else {
		nodePtr->constant = nodePtr->constant
			&& (complete == OT_LITERAL);
	    }

	    /*
	     * With a left operand attached and a right operand missing,
	     * the just-parsed binary operator is root of a new incomplete
	     * tree.  Push it onto the stack of incomplete trees.
	     */

	    nodePtr->p.prev = incomplete;
	    incomplete = lastParsed = nodesUsed;
	    nodesUsed++;
	    break;
	}	/* case BINARY */
	}	/* lexeme handler */

	/* Advance past the just-parsed lexeme */
	start += scanned;
	numBytes -= scanned;
    }	/* main parsing loop */

  error:

    /*
     * We only get here if there's been an error.
     * Any errors that didn't get a suitable parsePtr->errorType,
     * get recorded as syntax errors.
     */

    if (parsePtr->errorType == TCL_PARSE_SUCCESS) {
	parsePtr->errorType = TCL_PARSE_SYNTAX;
    }

    /* Free any partial parse tree we've built. */
    if (nodes != NULL) {
	ckfree((char*) nodes);
    }

    if (interp == NULL) {

	/* Nowhere to report an error message, so just free it */
	if (msg) {
	    Tcl_DecrRefCount(msg);
	}
    } else {

	/*
	 * Construct the complete error message.  Start with the simple
	 * error message, pulled from the interp result if necessary...
	 */

	if (msg == NULL) {
	    msg = Tcl_GetObjResult(interp);
	}

	/*
	 * Add a detailed quote from the bad expression, displaying and
	 * sometimes marking the precise location of the syntax error.
	 */

	Tcl_AppendPrintfToObj(msg, "\nin expression \"%s%.*s%.*s%s%s%.*s%s\"",
		((start - limit) < parsePtr->string) ? "" : "...",
		((start - limit) < parsePtr->string)
			? (start - parsePtr->string) : limit - 3,
		((start - limit) < parsePtr->string)
			? parsePtr->string : start - limit + 3,
		(scanned < limit) ? scanned : limit - 3, start,
		(scanned < limit) ? "" : "...", insertMark ? mark : "",
		(start + scanned + limit > parsePtr->end)
			? parsePtr->end - (start + scanned) : limit-3,
		start + scanned,
		(start + scanned + limit > parsePtr->end) ? "" : "...");

	/* Next, append any postscript message. */
	if (post != NULL) {
	    Tcl_AppendToObj(msg, ";\n", -1);
	    Tcl_AppendObjToObj(msg, post);
	    Tcl_DecrRefCount(post);
	}
	Tcl_SetObjResult(interp, msg);

	/* Finally, place context information in the errorInfo. */
	numBytes = parsePtr->end - parsePtr->string;
	Tcl_AppendObjToErrorInfo(interp, Tcl_ObjPrintf(
		"\n    (parsing expression \"%.*s%s\")",
		(numBytes < limit) ? numBytes : limit - 3,
		parsePtr->string, (numBytes < limit) ? "" : "..."));
    }

    return TCL_ERROR;
}

/*
 *----------------------------------------------------------------------
 *
 * ConvertTreeToTokens --
 *
 *	Given a string, the numBytes bytes starting at start, and an OpNode
 *	tree and Tcl_Token array created by passing that same string to
 *	ParseExpr(), this function writes into *parsePtr the sequence of
 * 	Tcl_Tokens needed so to satisfy the historical interface provided
 * 	by Tcl_ParseExpr().  Note that this routine exists only for the sake
 *	of the public Tcl_ParseExpr() routine.  It is not used by Tcl itself
 * 	at all.
 *
 * Results:
 *	None.
 *
 * Side effects:
 *	The Tcl_Parse *parsePtr is filled with Tcl_Tokens representing the
 *	parsed expression.
 *
 *----------------------------------------------------------------------
 */

static void
ConvertTreeToTokens(
    const char *start,
    int numBytes,
    OpNode *nodes,
    Tcl_Token *tokenPtr,
    Tcl_Parse *parsePtr)
{
    int subExprTokenIdx = 0;
    OpNode *nodePtr = nodes;
    int next = nodePtr->right;

    while (1) {
	Tcl_Token *subExprTokenPtr;
	int scanned, parentIdx;
	unsigned char lexeme;

	/*
	 * Advance the mark so the next exit from this node won't retrace
	 * steps over ground already covered.
	 */

	nodePtr->mark++;

	/* Handle next child node or leaf */
	switch (next) {
	case OT_EMPTY:

	    /* No tokens and no characters for the OT_EMPTY leaf. */
	    break;

	case OT_LITERAL:

	    /* Skip any white space that comes before the literal */
	    scanned = TclParseAllWhiteSpace(start, numBytes);
	    start +=scanned;
	    numBytes -= scanned;

	    /* Reparse the literal to get pointers into source string */
	    scanned = ParseLexeme(start, numBytes, &lexeme, NULL);

	    if (parsePtr->numTokens + 1 >= parsePtr->tokensAvailable) {
		TclExpandTokenArray(parsePtr);
	    }
	    subExprTokenPtr = parsePtr->tokenPtr + parsePtr->numTokens;
	    subExprTokenPtr->type = TCL_TOKEN_SUB_EXPR;
	    subExprTokenPtr->start = start;
	    subExprTokenPtr->size = scanned;
	    subExprTokenPtr->numComponents = 1;
	    subExprTokenPtr[1].type = TCL_TOKEN_TEXT;
	    subExprTokenPtr[1].start = start;
	    subExprTokenPtr[1].size = scanned;
	    subExprTokenPtr[1].numComponents = 0;

	    parsePtr->numTokens += 2;
	    start +=scanned;
	    numBytes -= scanned;
	    break;

	case OT_TOKENS: {

	    /*
	     * tokenPtr points to a token sequence that came from parsing
	     * a Tcl word.  A Tcl word is made up of a sequence of one or
	     * more elements.  When the word is only a single element, it's
	     * been the historical practice to replace the TCL_TOKEN_WORD
	     * token directly with a TCL_TOKEN_SUB_EXPR token.  However,
	     * when the word has multiple elements, a TCL_TOKEN_WORD token
	     * is kept as a grouping device so that TCL_TOKEN_SUB_EXPR
	     * always has only one element.  Wise or not, these are the
	     * rules the Tcl expr parser has followed, and for the sake
	     * of those few callers of Tcl_ParseExpr() we do not change
	     * them now.  Internally, we can do better.
	     */
	
	    int toCopy = tokenPtr->numComponents + 1;

	    if (tokenPtr->numComponents == tokenPtr[1].numComponents + 1) {

		/*
		 * Single element word.  Copy tokens and convert the leading
		 * token to TCL_TOKEN_SUB_EXPR.
		 */

		while (parsePtr->numTokens + toCopy - 1
			>= parsePtr->tokensAvailable) {
		    TclExpandTokenArray(parsePtr);
		}
		subExprTokenPtr = parsePtr->tokenPtr + parsePtr->numTokens;
		memcpy(subExprTokenPtr, tokenPtr,
			(size_t) toCopy * sizeof(Tcl_Token));
		subExprTokenPtr->type = TCL_TOKEN_SUB_EXPR;
		parsePtr->numTokens += toCopy;
	    } else {

		/* 
		 * Multiple element word.  Create a TCL_TOKEN_SUB_EXPR
		 * token to lead, with fields initialized from the leading
		 * token, then copy entire set of word tokens.
		 */

		while (parsePtr->numTokens + toCopy
			>= parsePtr->tokensAvailable) {
		    TclExpandTokenArray(parsePtr);
		}
		subExprTokenPtr = parsePtr->tokenPtr + parsePtr->numTokens;
		*subExprTokenPtr = *tokenPtr;
		subExprTokenPtr->type = TCL_TOKEN_SUB_EXPR;
		subExprTokenPtr->numComponents++;
		subExprTokenPtr++;
		memcpy(subExprTokenPtr, tokenPtr,
			(size_t) toCopy * sizeof(Tcl_Token));
		parsePtr->numTokens += toCopy + 1;
	    }

	    scanned = tokenPtr->start + tokenPtr->size - start;
	    start +=scanned;
	    numBytes -= scanned;
	    tokenPtr += toCopy;
	    break;
	}

	default:

	    /* Advance to the child node, which is an operator. */
	    nodePtr = nodes + next;

	    /* Skip any white space that comes before the subexpression */
	    scanned = TclParseAllWhiteSpace(start, numBytes);
	    start +=scanned;
	    numBytes -= scanned;

	    /* Generate tokens for the operator / subexpression... */
	    switch (nodePtr->lexeme) {
	    case OPEN_PAREN:
	    case COMMA:
	    case COLON:

		/* 
		 * Historical practice has been to have no Tcl_Tokens for
		 * these operators.
		 */

		break;

	    default: {

		/*
		 * Remember the index of the last subexpression we were
		 * working on -- that of our parent.  We'll stack it later.
		 */

		parentIdx = subExprTokenIdx;

		/*
		 * Verify space for the two leading Tcl_Tokens representing
		 * the subexpression rooted by this operator.  The first
		 * Tcl_Token will be of type TCL_TOKEN_SUB_EXPR; the second
		 * of type TCL_TOKEN_OPERATOR.
		 */

		if (parsePtr->numTokens + 1 >= parsePtr->tokensAvailable) {
		    TclExpandTokenArray(parsePtr);
		}
		subExprTokenIdx = parsePtr->numTokens;
		subExprTokenPtr = parsePtr->tokenPtr + subExprTokenIdx;
		parsePtr->numTokens += 2;
		subExprTokenPtr->type = TCL_TOKEN_SUB_EXPR;
		subExprTokenPtr[1].type = TCL_TOKEN_OPERATOR;

		/*
		 * Our current position scanning the string is the starting
		 * point for this subexpression.
		 */

		subExprTokenPtr->start = start;

		/*
		 * Eventually, we know that the numComponents field of the
		 * Tcl_Token of type TCL_TOKEN_OPERATOR will be 0.  This means
		 * we can make other use of this field for now to track the
		 * stack of subexpressions we have pending.
		 */

		subExprTokenPtr[1].numComponents = parentIdx;
		break;
	    }
	    }
	    break;
	}

	/* Determine which way to exit the node on this pass. */
    router:
	switch (nodePtr->mark) {
	case MARK_LEFT:
	    next = nodePtr->left;
	    break;

	case MARK_RIGHT:
	    next = nodePtr->right;

	    /* Skip any white space that comes before the operator */
	    scanned = TclParseAllWhiteSpace(start, numBytes);
	    start +=scanned;
	    numBytes -= scanned;

	    /*
	     * Here we scan from the string the operator corresponding to
	     * nodePtr->lexeme.
	     */

	    scanned = ParseLexeme(start, numBytes, &lexeme, NULL);

	    switch(nodePtr->lexeme) {
	    case OPEN_PAREN:
	    case COMMA:
	    case COLON:

		/* No tokens for these lexemes -> nothing to do. */
		break;

	    default:

		/*
		 * Record in the TCL_TOKEN_OPERATOR token the pointers into
		 * the string marking where the operator is.
		 */

		subExprTokenPtr = parsePtr->tokenPtr + subExprTokenIdx;
		subExprTokenPtr[1].start = start;
		subExprTokenPtr[1].size = scanned;
		break;
	    }

	    start +=scanned;
	    numBytes -= scanned;
	    break;

	case MARK_PARENT:
	    switch (nodePtr->lexeme) {
	    case START:

		/* When we get back to the START node, we're done. */
		return;

	    case COMMA:
	    case COLON:

		/* No tokens for these lexemes -> nothing to do. */
		break;

	    case OPEN_PAREN:

		/* Skip past matching close paren. */
		scanned = TclParseAllWhiteSpace(start, numBytes);
		start +=scanned;
		numBytes -= scanned;
		scanned = ParseLexeme(start, numBytes, &lexeme, NULL);
		start +=scanned;
		numBytes -= scanned;
		break;

	    default: {

		/*
		 * Before we leave this node/operator/subexpression for the
		 * last time, finish up its tokens....
		 * 
		 * Our current position scanning the string is where the
		 * substring for the subexpression ends.
		 */

		subExprTokenPtr = parsePtr->tokenPtr + subExprTokenIdx;
		subExprTokenPtr->size = start - subExprTokenPtr->start;

		/*
		 * All the Tcl_Tokens allocated and filled belong to
		 * this subexpresion.  The first token is the leading
		 * TCL_TOKEN_SUB_EXPR token, and all the rest (one fewer)
		 * are its components.
		 */

		subExprTokenPtr->numComponents =
			(parsePtr->numTokens - subExprTokenIdx) - 1;

		/*
		 * Finally, as we return up the tree to our parent, pop the
		 * parent subexpression off our subexpression stack, and
		 * fill in the zero numComponents for the operator Tcl_Token.
		 */

		parentIdx = subExprTokenPtr[1].numComponents;
		subExprTokenPtr[1].numComponents = 0;
		subExprTokenIdx = parentIdx;
		break;
	    }
	    }

	    /* Since we're returning to parent, skip child handling code. */
	    nodePtr = nodes + nodePtr->p.parent;
	    goto router;
	}
    }
}

/*
 *----------------------------------------------------------------------
 *
 * Tcl_ParseExpr --
 *
 *	Given a string, the numBytes bytes starting at start, this function
 *	parses it as a Tcl expression and stores information about the
 *	structure of the expression in the Tcl_Parse struct indicated by the
 *	caller.
 *
 * Results:
 *	If the string is successfully parsed as a valid Tcl expression, TCL_OK
 *	is returned, and data about the expression structure is written to
 *	*parsePtr. If the string cannot be parsed as a valid Tcl expression,
 *	TCL_ERROR is returned, and if interp is non-NULL, an error message is
 *	written to interp.
 *
 * Side effects:
 *	If there is insufficient space in parsePtr to hold all the information
 *	about the expression, then additional space is malloc-ed. If the
 *	function returns TCL_OK then the caller must eventually invoke
 *	Tcl_FreeParse to release any additional space that was allocated.
 *
 *----------------------------------------------------------------------
 */

int
Tcl_ParseExpr(
    Tcl_Interp *interp,		/* Used for error reporting. */
    const char *start,		/* Start of source string to parse. */
    int numBytes,		/* Number of bytes in string. If < 0, the
				 * string consists of all bytes up to the
				 * first null character. */
    Tcl_Parse *parsePtr)	/* Structure to fill with information about
				 * the parsed expression; any previous
				 * information in the structure is ignored. */
{
    int code;
    OpNode *opTree = NULL;	/* Will point to the tree of operators */
    Tcl_Obj *litList = Tcl_NewObj();	/* List to hold the literals */
    Tcl_Obj *funcList = Tcl_NewObj();	/* List to hold the functon names*/
    Tcl_Parse *exprParsePtr =
	    (Tcl_Parse *) TclStackAlloc(interp, sizeof(Tcl_Parse));
				/* Holds the Tcl_Tokens of substitutions */

    if (numBytes < 0) {
	numBytes = (start ? strlen(start) : 0);
    }

    code = ParseExpr(interp, start, numBytes, &opTree, litList,
	    funcList, exprParsePtr, 1 /* parseOnly */);
    Tcl_DecrRefCount(funcList);
    Tcl_DecrRefCount(litList);

    TclParseInit(interp, start, numBytes, parsePtr);
    if (code == TCL_OK) {
	ConvertTreeToTokens(start, numBytes,
		opTree, exprParsePtr->tokenPtr, parsePtr);
    } else {
	parsePtr->term = exprParsePtr->term;
	parsePtr->errorType = exprParsePtr->errorType;
    }

    Tcl_FreeParse(exprParsePtr);
    TclStackFree(interp, exprParsePtr);
    ckfree((char *) opTree);
    return code;
}

/*
 *----------------------------------------------------------------------
 *
 * ParseLexeme --
 *
 *	Parse a single lexeme from the start of a string, scanning no more
 *	than numBytes bytes.
 *
 * Results:
 *	Returns the number of bytes scanned to produce the lexeme.
 *
 * Side effects:
 *	Code identifying lexeme parsed is writen to *lexemePtr.
 *
 *----------------------------------------------------------------------
 */

static int
ParseLexeme(
    const char *start,		/* Start of lexeme to parse. */
    int numBytes,		/* Number of bytes in string. */
    unsigned char *lexemePtr,	/* Write code of parsed lexeme to this
				 * storage. */
    Tcl_Obj **literalPtr)	/* Write corresponding literal value to this
				   storage, if non-NULL. */
{
    const char *end;
    int scanned;
    Tcl_UniChar ch;
    Tcl_Obj *literal = NULL;
    unsigned char byte;

    if (numBytes == 0) {
	*lexemePtr = END;
	return 0;
    }
    byte = (unsigned char)(*start);
    if (byte < sizeof(Lexeme) && Lexeme[byte] != 0) {
	*lexemePtr = Lexeme[byte];
	return 1;
    }
    switch (byte) {
    case '*':
	if ((numBytes > 1) && (start[1] == '*')) {
	    *lexemePtr = EXPON;
	    return 2;
	}
	*lexemePtr = MULT;
	return 1;

    case '=':
	if ((numBytes > 1) && (start[1] == '=')) {
	    *lexemePtr = EQUAL;
	    return 2;
	}
	*lexemePtr = INCOMPLETE;
	return 1;

    case '!':
	if ((numBytes > 1) && (start[1] == '=')) {
	    *lexemePtr = NEQ;
	    return 2;
	}
	*lexemePtr = NOT;
	return 1;

    case '&':
	if ((numBytes > 1) && (start[1] == '&')) {
	    *lexemePtr = AND;
	    return 2;
	}
	*lexemePtr = BIT_AND;
	return 1;

    case '|':
	if ((numBytes > 1) && (start[1] == '|')) {
	    *lexemePtr = OR;
	    return 2;
	}
	*lexemePtr = BIT_OR;
	return 1;

    case '<':
	if (numBytes > 1) {
	    switch (start[1]) {
	    case '<':
		*lexemePtr = LEFT_SHIFT;
		return 2;
	    case '=':
		*lexemePtr = LEQ;
		return 2;
	    }
	}
	*lexemePtr = LESS;
	return 1;

    case '>':
	if (numBytes > 1) {
	    switch (start[1]) {
	    case '>':
		*lexemePtr = RIGHT_SHIFT;
		return 2;
	    case '=':
		*lexemePtr = GEQ;
		return 2;
	    }
	}
	*lexemePtr = GREATER;
	return 1;

    case 'i':
	if ((numBytes > 1) && (start[1] == 'n')
		&& ((numBytes == 2) || !isalpha(UCHAR(start[2])))) {

	    /*
	     * Must make this check so we can tell the difference between
	     * the "in" operator and the "int" function name and the
	     * "infinity" numeric value.
	     */

	    *lexemePtr = IN_LIST;
	    return 2;
	}
	break;

    case 'e':
	if ((numBytes > 1) && (start[1] == 'q')
		&& ((numBytes == 2) || !isalpha(UCHAR(start[2])))) {
	    *lexemePtr = STREQ;
	    return 2;
	}
	break;

    case 'n':
	if ((numBytes > 1) && ((numBytes == 2) || !isalpha(UCHAR(start[2])))) {
	    switch (start[1]) {
	    case 'e':
		*lexemePtr = STRNEQ;
		return 2;
	    case 'i':
		*lexemePtr = NOT_IN_LIST;
		return 2;
	    }
	}
    }

    literal = Tcl_NewObj();
    if (TclParseNumber(NULL, literal, NULL, start, numBytes, &end,
	    TCL_PARSE_NO_WHITESPACE) == TCL_OK) {
	TclInitStringRep(literal, start, end-start);
	*lexemePtr = NUMBER;
	if (literalPtr) {
	    *literalPtr = literal;
	} else {
	    Tcl_DecrRefCount(literal);
	}
	return (end-start);
    }

    if (Tcl_UtfCharComplete(start, numBytes)) {
	scanned = Tcl_UtfToUniChar(start, &ch);
    } else {
	char utfBytes[TCL_UTF_MAX];
	memcpy(utfBytes, start, (size_t) numBytes);
	utfBytes[numBytes] = '\0';
	scanned = Tcl_UtfToUniChar(utfBytes, &ch);
    }
    if (!isalpha(UCHAR(ch))) {
	*lexemePtr = INVALID;
	Tcl_DecrRefCount(literal);
	return scanned;
    }
    end = start;
    while (isalnum(UCHAR(ch)) || (UCHAR(ch) == '_')) {
	end += scanned;
	numBytes -= scanned;
	if (Tcl_UtfCharComplete(end, numBytes)) {
	    scanned = Tcl_UtfToUniChar(end, &ch);
	} else {
	    char utfBytes[TCL_UTF_MAX];
	    memcpy(utfBytes, end, (size_t) numBytes);
	    utfBytes[numBytes] = '\0';
	    scanned = Tcl_UtfToUniChar(utfBytes, &ch);
	}
    }
    *lexemePtr = BAREWORD;
    if (literalPtr) {
	Tcl_SetStringObj(literal, start, (int) (end-start));
	*literalPtr = literal;
    } else {
	Tcl_DecrRefCount(literal);
    }
    return (end-start);
}

/*
 *----------------------------------------------------------------------
 *
 * TclCompileExpr --
 *
 *	This procedure compiles a string containing a Tcl expression into Tcl
 *	bytecodes. 
 *
 * Results:
 *	None.
 *
 * Side effects:
 *	Adds instructions to envPtr to evaluate the expression at runtime.
 *
 *----------------------------------------------------------------------
 */

void
TclCompileExpr(
    Tcl_Interp *interp,		/* Used for error reporting. */
    const char *script,		/* The source script to compile. */
    int numBytes,		/* Number of bytes in script. */
    CompileEnv *envPtr)		/* Holds resulting instructions. */
{
    OpNode *opTree = NULL;	/* Will point to the tree of operators */
    Tcl_Obj *litList = Tcl_NewObj();	/* List to hold the literals */
    Tcl_Obj *funcList = Tcl_NewObj();	/* List to hold the functon names*/
    Tcl_Parse *parsePtr =
	    (Tcl_Parse *) TclStackAlloc(interp, sizeof(Tcl_Parse));
				/* Holds the Tcl_Tokens of substitutions */

    int code = ParseExpr(interp, script, numBytes, &opTree, litList,
	    funcList, parsePtr, 0 /* parseOnly */);

    if (code == TCL_OK) {

	/* Valid parse; compile the tree. */
	int objc;
	Tcl_Obj *const *litObjv;
	Tcl_Obj **funcObjv;

	/* TIP #280 : Track Lines within the expression */
	TclAdvanceLines(&envPtr->line, script,
		script + TclParseAllWhiteSpace(script, numBytes));

	TclListObjGetElements(NULL, litList, &objc, (Tcl_Obj ***)&litObjv);
	TclListObjGetElements(NULL, funcList, &objc, &funcObjv);
	CompileExprTree(interp, opTree, 0, &litObjv, funcObjv,
		parsePtr->tokenPtr, envPtr, 1 /* optimize */);
    } else {
	TclCompileSyntaxError(interp, envPtr);
    }

    Tcl_FreeParse(parsePtr);
    TclStackFree(interp, parsePtr);
    Tcl_DecrRefCount(funcList);
    Tcl_DecrRefCount(litList);
    ckfree((char *) opTree);
}

/*
 *----------------------------------------------------------------------
 *
 * ExecConstantExprTree --
 *	Compiles and executes bytecode for the subexpression tree at index
 *	in the nodes array.  This subexpression must be constant, made up
 *	of only constant operators (not functions) and literals.
 *
 * Results:
 *	A standard Tcl return code and result left in interp.
 *
 * Side effects:
 *	Consumes subtree of nodes rooted at index.  Advances the pointer
 *	*litObjvPtr.
 *
 *----------------------------------------------------------------------
 */

static int
ExecConstantExprTree(
    Tcl_Interp *interp,
    OpNode *nodes,
    int index,
    Tcl_Obj *const **litObjvPtr)
{
    CompileEnv *envPtr;
    ByteCode *byteCodePtr;
    int code;
    Tcl_Obj *byteCodeObj = Tcl_NewObj();

    /*
     * Note we are compiling an expression with literal arguments. This means
     * there can be no [info frame] calls when we execute the resulting
     * bytecode, so there's no need to tend to TIP 280 issues.
     */

    envPtr = (CompileEnv *) TclStackAlloc(interp, sizeof(CompileEnv));
    TclInitCompileEnv(interp, envPtr, NULL, 0, NULL, 0);
    CompileExprTree(interp, nodes, index, litObjvPtr, NULL, NULL, envPtr,
	    0 /* optimize */);
    TclEmitOpcode(INST_DONE, envPtr);
    Tcl_IncrRefCount(byteCodeObj);
    TclInitByteCodeObj(byteCodeObj, envPtr);
    TclFreeCompileEnv(envPtr);
    TclStackFree(interp, envPtr);
    byteCodePtr = (ByteCode *) byteCodeObj->internalRep.otherValuePtr;
    code = TclExecuteByteCode(interp, byteCodePtr);
    Tcl_DecrRefCount(byteCodeObj);
    return code;
}

/*
 *----------------------------------------------------------------------
 *
 * CompileExprTree --
 *	Compiles and writes to envPtr instructions for the subexpression
 *	tree at index in the nodes array.  (*litObjvPtr) must point to the
 *	proper location in a corresponding literals list.  Likewise, when
 *	non-NULL, funcObjv and tokenPtr must point into matching arrays of
 * 	function names and Tcl_Token's derived from earlier call to
 *	ParseExpr().  When optimize is true, any constant subexpressions
 *	will be precomputed.
 *
 * Results:
 *	None.
 *
 * Side effects:
 *	Adds instructions to envPtr to evaluate the expression at runtime.
 *	Consumes subtree of nodes rooted at index.  Advances the pointer
 *	*litObjvPtr.
 *
 *----------------------------------------------------------------------
 */

static void
CompileExprTree(
    Tcl_Interp *interp,
    OpNode *nodes,
    int index,
    Tcl_Obj *const **litObjvPtr,
    Tcl_Obj *const *funcObjv,
    Tcl_Token *tokenPtr,
    CompileEnv *envPtr,
    int optimize)
{
    OpNode *nodePtr = nodes + index;
    OpNode *rootPtr = nodePtr;
    int numWords = 0;
    JumpList *jumpPtr = NULL;
    int convert = 1;

    while (1) {
	int next;
	JumpList *freePtr, *newJump;

	if (nodePtr->mark == MARK_LEFT) {
	    next = nodePtr->left;

	    switch (nodePtr->lexeme) {
	    case QUESTION:
		newJump = (JumpList *) TclStackAlloc(interp, sizeof(JumpList));
		newJump->next = jumpPtr;
		jumpPtr = newJump;
		newJump = (JumpList *) TclStackAlloc(interp, sizeof(JumpList));
		newJump->next = jumpPtr;
		jumpPtr = newJump;
		jumpPtr->depth = envPtr->currStackDepth;
		convert = 1;
		break;
	    case AND:
	    case OR:
		newJump = (JumpList *) TclStackAlloc(interp, sizeof(JumpList));
		newJump->next = jumpPtr;
		jumpPtr = newJump;
		newJump = (JumpList *) TclStackAlloc(interp, sizeof(JumpList));
		newJump->next = jumpPtr;
		jumpPtr = newJump;
		newJump = (JumpList *) TclStackAlloc(interp, sizeof(JumpList));
		newJump->next = jumpPtr;
		jumpPtr = newJump;
		jumpPtr->depth = envPtr->currStackDepth;
		break;
	    }
	} else if (nodePtr->mark == MARK_RIGHT) {
	    next = nodePtr->right;

	    switch (nodePtr->lexeme) {
	    case FUNCTION: {
		Tcl_DString cmdName;
		const char *p;
		int length;

		Tcl_DStringInit(&cmdName);
		Tcl_DStringAppend(&cmdName, "tcl::mathfunc::", -1);
		p = TclGetStringFromObj(*funcObjv, &length);
		funcObjv++;
		Tcl_DStringAppend(&cmdName, p, length);
		TclEmitPush(TclRegisterNewNSLiteral(envPtr,
			Tcl_DStringValue(&cmdName),
			Tcl_DStringLength(&cmdName)), envPtr);
		Tcl_DStringFree(&cmdName);

		/*
		 * Start a count of the number of words in this function
		 * command invocation.  In case there's already a count
		 * in progress (nested functions), save it in our unused
		 * "left" field for restoring later.
		 */

		nodePtr->left = numWords;
		numWords = 2;	/* Command plus one argument */
		break;
	    }
	    case QUESTION:
		TclEmitForwardJump(envPtr, TCL_FALSE_JUMP, &(jumpPtr->jump));
		break;
	    case COLON:
		TclEmitForwardJump(envPtr, TCL_UNCONDITIONAL_JUMP,
			&(jumpPtr->next->jump));
		envPtr->currStackDepth = jumpPtr->depth;
		jumpPtr->offset = (envPtr->codeNext - envPtr->codeStart);
		jumpPtr->convert = convert;
		convert = 1;
		break;
	    case AND:
		TclEmitForwardJump(envPtr, TCL_FALSE_JUMP, &(jumpPtr->jump));
		break;
	    case OR:
		TclEmitForwardJump(envPtr, TCL_TRUE_JUMP, &(jumpPtr->jump));
		break;
	    }
	} else {
	    switch (nodePtr->lexeme) {
	    case START:
	    case QUESTION:
		if (convert && (nodePtr == rootPtr)) {
		    TclEmitOpcode(INST_TRY_CVT_TO_NUMERIC, envPtr);
		}
		break;
	    case OPEN_PAREN:

		/* do nothing */
		break;
	    case FUNCTION:

		/*
		 * Use the numWords count we've kept to invoke the
		 * function command with the correct number of arguments.
		 */
		
		if (numWords < 255) {
		    TclEmitInstInt1(INST_INVOKE_STK1, numWords, envPtr);
		} else {
		    TclEmitInstInt4(INST_INVOKE_STK4, numWords, envPtr);
		}

		/* Restore any saved numWords value. */
		numWords = nodePtr->left;
		convert = 1;
		break;
	    case COMMA:

		/* Each comma implies another function argument. */
		numWords++;
		break;
	    case COLON:
		if (TclFixupForwardJump(envPtr, &(jumpPtr->next->jump),
			(envPtr->codeNext - envPtr->codeStart)
			- jumpPtr->next->jump.codeOffset, 127)) {
		    jumpPtr->offset += 3;
		}
		TclFixupForwardJump(envPtr, &(jumpPtr->jump),
			jumpPtr->offset - jumpPtr->jump.codeOffset, 127);
		convert |= jumpPtr->convert;
		envPtr->currStackDepth = jumpPtr->depth + 1;
		freePtr = jumpPtr;
		jumpPtr = jumpPtr->next;
		TclStackFree(interp, freePtr);
		freePtr = jumpPtr;
		jumpPtr = jumpPtr->next;
		TclStackFree(interp, freePtr);
		break;
	    case AND:
	    case OR:
		TclEmitForwardJump(envPtr, (nodePtr->lexeme == AND)
			?  TCL_FALSE_JUMP : TCL_TRUE_JUMP,
			&(jumpPtr->next->jump));
		TclEmitPush(TclRegisterNewLiteral(envPtr,
			(nodePtr->lexeme == AND) ? "1" : "0", 1), envPtr);
		TclEmitForwardJump(envPtr, TCL_UNCONDITIONAL_JUMP,
			&(jumpPtr->next->next->jump));
		TclFixupForwardJumpToHere(envPtr, &(jumpPtr->next->jump), 127);
		if (TclFixupForwardJumpToHere(envPtr, &(jumpPtr->jump), 127)) {
		    jumpPtr->next->next->jump.codeOffset += 3;
		}
		TclEmitPush(TclRegisterNewLiteral(envPtr,
			(nodePtr->lexeme == AND) ? "0" : "1", 1), envPtr);
		TclFixupForwardJumpToHere(envPtr, &(jumpPtr->next->next->jump),
			127);
		convert = 0;
		envPtr->currStackDepth = jumpPtr->depth + 1;
		freePtr = jumpPtr;
		jumpPtr = jumpPtr->next;
		TclStackFree(interp, freePtr);
		freePtr = jumpPtr;
		jumpPtr = jumpPtr->next;
		TclStackFree(interp, freePtr);
		freePtr = jumpPtr;
		jumpPtr = jumpPtr->next;
		TclStackFree(interp, freePtr);
		break;
	    default:
		TclEmitOpcode(instruction[nodePtr->lexeme], envPtr);
		convert = 0;
		break;
	    }
	    if (nodePtr == rootPtr) {

		/* We're done */
		return;
	    }
	    nodePtr = nodes + nodePtr->p.parent;
	    continue;
	}

	nodePtr->mark++;
	switch (next) {
	case OT_EMPTY:
	    numWords = 1;	/* No arguments, so just the command */
	    break;
	case OT_LITERAL: {
	    Tcl_Obj *const *litObjv = *litObjvPtr;
	    Tcl_Obj *literal = *litObjv;
	    int length;
	    const char *bytes = TclGetStringFromObj(literal, &length);

	    TclEmitPush(TclRegisterNewLiteral(envPtr, bytes, length), envPtr);
	    (*litObjvPtr)++;
	    break;
	}
	case OT_TOKENS:
	    TclCompileTokens(interp, tokenPtr+1, tokenPtr->numComponents,
		    envPtr);
	    tokenPtr += tokenPtr->numComponents + 1;
	    break;
	default:
	    if (optimize && nodes[next].constant) {
		Tcl_InterpState save = Tcl_SaveInterpState(interp, TCL_OK);
		if (ExecConstantExprTree(interp, nodes, next, litObjvPtr)
			== TCL_OK) {
		    TclEmitPush(TclAddLiteralObj(envPtr,
			    Tcl_GetObjResult(interp), NULL), envPtr);
		} else {
		    TclCompileSyntaxError(interp, envPtr);
		}
		Tcl_RestoreInterpState(interp, save);
		convert = 0;
	    } else {
		nodePtr = nodes + next;
	    }
	}
    }
}

/*
 *----------------------------------------------------------------------
 *
 * TclSingleOpCmd --
 *	Implements the commands: ~, !, <<, >>, %, !=, ne, in, ni
 *	in the ::tcl::mathop namespace.  These commands have no
 *	extension to arbitrary arguments; they accept only exactly one
 *	or exactly two arguments as suitable for the operator.
 *
 * Results:
 *	A standard Tcl return code and result left in interp.
 *
 * Side effects:
 * 	None.
 *
 *----------------------------------------------------------------------
 */

int
TclSingleOpCmd(
    ClientData clientData,
    Tcl_Interp *interp,
    int objc,
    Tcl_Obj *const objv[])
{
    TclOpCmdClientData *occdPtr = (TclOpCmdClientData *)clientData;
    unsigned char lexeme;
    OpNode nodes[2];
    Tcl_Obj *const *litObjv = objv + 1;

    if (objc != 1+occdPtr->i.numArgs) {
	Tcl_WrongNumArgs(interp, 1, objv, occdPtr->expected);
	return TCL_ERROR;
    }

    ParseLexeme(occdPtr->operator, strlen(occdPtr->operator), &lexeme, NULL);
    nodes[0].lexeme = START;
    nodes[0].mark = MARK_RIGHT;
    nodes[0].right = 1;
    nodes[1].lexeme = lexeme;
    if (objc == 2) {
	nodes[1].mark = MARK_RIGHT;
    } else {
	nodes[1].mark = MARK_LEFT;
	nodes[1].left = OT_LITERAL;
    }
    nodes[1].right = OT_LITERAL;
    nodes[1].p.parent = 0;

    return ExecConstantExprTree(interp, nodes, 0, &litObjv);
}

/*
 *----------------------------------------------------------------------
 *
 * TclSortingOpCmd --
 *	Implements the commands: <, <=, >, >=, ==, eq 
 *	in the ::tcl::mathop namespace.  These commands are defined for
 *	arbitrary number of arguments by computing the AND of the base
 * 	operator applied to all neighbor argument pairs.
 *
 * Results:
 *	A standard Tcl return code and result left in interp.
 *
 * Side effects:
 * 	None.
 *
 *----------------------------------------------------------------------
 */

int
TclSortingOpCmd(
    ClientData clientData,
    Tcl_Interp *interp,
    int objc,
    Tcl_Obj *const objv[])
{
    int code = TCL_OK;

    if (objc < 3) {
	Tcl_SetObjResult(interp, Tcl_NewBooleanObj(1));
    } else {
	TclOpCmdClientData *occdPtr = (TclOpCmdClientData *)clientData;
	Tcl_Obj **litObjv = (Tcl_Obj **) TclStackAlloc(interp,
		2*(objc-2)*sizeof(Tcl_Obj *));
	OpNode *nodes = (OpNode *) TclStackAlloc(interp,
		2*(objc-2)*sizeof(OpNode));
	unsigned char lexeme;
	int i, lastAnd = 1;
	Tcl_Obj *const *litObjPtrPtr = litObjv;

	ParseLexeme(occdPtr->operator, strlen(occdPtr->operator),
		&lexeme, NULL);

	litObjv[0] = objv[1];
	nodes[0].lexeme = START;
	nodes[0].mark = MARK_RIGHT;
	for (i=2; i<objc-1; i++) {
	    litObjv[2*(i-1)-1] = objv[i];
	    nodes[2*(i-1)-1].lexeme = lexeme;
	    nodes[2*(i-1)-1].mark = MARK_LEFT;
	    nodes[2*(i-1)-1].left = OT_LITERAL;
	    nodes[2*(i-1)-1].right = OT_LITERAL;

	    litObjv[2*(i-1)] = objv[i];
	    nodes[2*(i-1)].lexeme = AND;
	    nodes[2*(i-1)].mark = MARK_LEFT;
	    nodes[2*(i-1)].left = lastAnd;
	    nodes[lastAnd].p.parent = 2*(i-1);

	    nodes[2*(i-1)].right = 2*(i-1)+1;
	    nodes[2*(i-1)+1].p.parent= 2*(i-1);

	    lastAnd = 2*(i-1);
	}
	litObjv[2*(objc-2)-1] = objv[objc-1];

	nodes[2*(objc-2)-1].lexeme = lexeme;
	nodes[2*(objc-2)-1].mark = MARK_LEFT;
	nodes[2*(objc-2)-1].left = OT_LITERAL;
	nodes[2*(objc-2)-1].right = OT_LITERAL;

	nodes[0].right = lastAnd;
	nodes[lastAnd].p.parent = 0;

	code = ExecConstantExprTree(interp, nodes, 0, &litObjPtrPtr);

	TclStackFree(interp, nodes);
	TclStackFree(interp, litObjv);
    }
    return code;
}

/*
 *----------------------------------------------------------------------
 *
 * TclVariadicOpCmd --
 *	Implements the commands: +, *, &, |, ^, **
 *	in the ::tcl::mathop namespace.  These commands are defined for
 *	arbitrary number of arguments by repeatedly applying the base
 *	operator with suitable associative rules.  When fewer than two
 *	arguments are provided, suitable identity values are returned.
 *
 * Results:
 *	A standard Tcl return code and result left in interp.
 *
 * Side effects:
 * 	None.
 *
 *----------------------------------------------------------------------
 */

int
TclVariadicOpCmd(
    ClientData clientData,
    Tcl_Interp *interp,
    int objc,
    Tcl_Obj *const objv[])
{
    TclOpCmdClientData *occdPtr = (TclOpCmdClientData *)clientData;
    unsigned char lexeme;
    int code;

    if (objc < 2) {
	Tcl_SetObjResult(interp, Tcl_NewIntObj(occdPtr->i.identity));
	return TCL_OK;
    }

    ParseLexeme(occdPtr->operator, strlen(occdPtr->operator), &lexeme, NULL);
    lexeme |= BINARY;

    if (objc == 2) {
	Tcl_Obj *litObjv[2];
	OpNode nodes[2];
	int decrMe = 0;
	Tcl_Obj *const *litObjPtrPtr = litObjv;

	if (lexeme == EXPON) {
	    litObjv[1] = Tcl_NewIntObj(occdPtr->i.identity);
	    Tcl_IncrRefCount(litObjv[1]);
	    decrMe = 1;
	    litObjv[0] = objv[1];
	    nodes[0].lexeme = START;
	    nodes[0].mark = MARK_RIGHT;
	    nodes[0].right = 1;
	    nodes[1].lexeme = lexeme;
	    nodes[1].mark = MARK_LEFT;
	    nodes[1].left = OT_LITERAL;
	    nodes[1].right = OT_LITERAL;
	    nodes[1].p.parent = 0;
	} else {
	    if (lexeme == DIVIDE) {
		litObjv[0] = Tcl_NewDoubleObj(1.0);
	    } else {
		litObjv[0] = Tcl_NewIntObj(occdPtr->i.identity);
	    }
	    Tcl_IncrRefCount(litObjv[0]);
	    litObjv[1] = objv[1];
	    nodes[0].lexeme = START;
	    nodes[0].mark = MARK_RIGHT;
	    nodes[0].right = 1;
	    nodes[1].lexeme = lexeme;
	    nodes[1].mark = MARK_LEFT;
	    nodes[1].left = OT_LITERAL;
	    nodes[1].right = OT_LITERAL;
	    nodes[1].p.parent = 0;
	}

	code = ExecConstantExprTree(interp, nodes, 0, &litObjPtrPtr);

	Tcl_DecrRefCount(litObjv[decrMe]);
	return code;
    } else {
	Tcl_Obj *const *litObjv = objv + 1;
	OpNode *nodes = (OpNode *) TclStackAlloc(interp,
		(objc-1)*sizeof(OpNode));
	int i, lastOp = OT_LITERAL;

	nodes[0].lexeme = START;
	nodes[0].mark = MARK_RIGHT;
	if (lexeme == EXPON) {
	    for (i=objc-2; i>0; i-- ) {
		nodes[i].lexeme = lexeme;
		nodes[i].mark = MARK_LEFT;
		nodes[i].left = OT_LITERAL;
		nodes[i].right = lastOp;
		if (lastOp >= 0) {
		    nodes[lastOp].p.parent = i;
		}
		lastOp = i;
	    }
	} else {
	    for (i=1; i<objc-1; i++ ) {
		nodes[i].lexeme = lexeme;
		nodes[i].mark = MARK_LEFT;
		nodes[i].left = lastOp;
		if (lastOp >= 0) {
		    nodes[lastOp].p.parent = i;
		}
		nodes[i].right = OT_LITERAL;
		lastOp = i;
	    }
	}
	nodes[0].right = lastOp;
	nodes[lastOp].p.parent = 0;

	code = ExecConstantExprTree(interp, nodes, 0, &litObjv);

	TclStackFree(interp, nodes);

	return code;
    }
}

/*
 *----------------------------------------------------------------------
 *
 * TclNoIdentOpCmd --
 *	Implements the commands: -, /
 *	in the ::tcl::mathop namespace.  These commands are defined for
 *	arbitrary non-zero number of arguments by repeatedly applying
 *	the base operator with suitable associative rules.  When no
 *	arguments are provided, an error is raised.
 *
 * Results:
 *	A standard Tcl return code and result left in interp.
 *
 * Side effects:
 * 	None.
 *
 *----------------------------------------------------------------------
 */

int
TclNoIdentOpCmd(
    ClientData clientData,
    Tcl_Interp *interp,
    int objc,
    Tcl_Obj *const objv[])
{
    TclOpCmdClientData *occdPtr = (TclOpCmdClientData *)clientData;
    if (objc < 2) {
	Tcl_WrongNumArgs(interp, 1, objv, occdPtr->expected);
	return TCL_ERROR;
    }
    return TclVariadicOpCmd(clientData, interp, objc, objv);
}
/*
 * Local Variables:
 * mode: c
 * c-basic-offset: 4
 * fill-column: 78
 * End:
 */