Jeremy Hylton | 3e0055f | 2005-10-20 19:59:25 +0000 | [diff] [blame^] | 1 | #include "Python.h" |
| 2 | #include "code.h" |
| 3 | #include "structmember.h" |
| 4 | |
| 5 | #define NAME_CHARS \ |
| 6 | "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ_abcdefghijklmnopqrstuvwxyz" |
| 7 | |
| 8 | /* all_name_chars(s): true iff all chars in s are valid NAME_CHARS */ |
| 9 | |
| 10 | static int |
| 11 | all_name_chars(unsigned char *s) |
| 12 | { |
| 13 | static char ok_name_char[256]; |
| 14 | static unsigned char *name_chars = (unsigned char *)NAME_CHARS; |
| 15 | |
| 16 | if (ok_name_char[*name_chars] == 0) { |
| 17 | unsigned char *p; |
| 18 | for (p = name_chars; *p; p++) |
| 19 | ok_name_char[*p] = 1; |
| 20 | } |
| 21 | while (*s) { |
| 22 | if (ok_name_char[*s++] == 0) |
| 23 | return 0; |
| 24 | } |
| 25 | return 1; |
| 26 | } |
| 27 | |
| 28 | static void |
| 29 | intern_strings(PyObject *tuple) |
| 30 | { |
| 31 | int i; |
| 32 | |
| 33 | for (i = PyTuple_GET_SIZE(tuple); --i >= 0; ) { |
| 34 | PyObject *v = PyTuple_GET_ITEM(tuple, i); |
| 35 | if (v == NULL || !PyString_CheckExact(v)) { |
| 36 | Py_FatalError("non-string found in code slot"); |
| 37 | } |
| 38 | PyString_InternInPlace(&PyTuple_GET_ITEM(tuple, i)); |
| 39 | } |
| 40 | } |
| 41 | |
| 42 | |
| 43 | PyCodeObject * |
| 44 | PyCode_New(int argcount, int nlocals, int stacksize, int flags, |
| 45 | PyObject *code, PyObject *consts, PyObject *names, |
| 46 | PyObject *varnames, PyObject *freevars, PyObject *cellvars, |
| 47 | PyObject *filename, PyObject *name, int firstlineno, |
| 48 | PyObject *lnotab) |
| 49 | { |
| 50 | PyCodeObject *co; |
| 51 | int i; |
| 52 | /* Check argument types */ |
| 53 | if (argcount < 0 || nlocals < 0 || |
| 54 | code == NULL || |
| 55 | consts == NULL || !PyTuple_Check(consts) || |
| 56 | names == NULL || !PyTuple_Check(names) || |
| 57 | varnames == NULL || !PyTuple_Check(varnames) || |
| 58 | freevars == NULL || !PyTuple_Check(freevars) || |
| 59 | cellvars == NULL || !PyTuple_Check(cellvars) || |
| 60 | name == NULL || !PyString_Check(name) || |
| 61 | filename == NULL || !PyString_Check(filename) || |
| 62 | lnotab == NULL || !PyString_Check(lnotab) || |
| 63 | !PyObject_CheckReadBuffer(code)) { |
| 64 | PyErr_BadInternalCall(); |
| 65 | return NULL; |
| 66 | } |
| 67 | intern_strings(names); |
| 68 | intern_strings(varnames); |
| 69 | intern_strings(freevars); |
| 70 | intern_strings(cellvars); |
| 71 | /* Intern selected string constants */ |
| 72 | for (i = PyTuple_Size(consts); --i >= 0; ) { |
| 73 | PyObject *v = PyTuple_GetItem(consts, i); |
| 74 | if (!PyString_Check(v)) |
| 75 | continue; |
| 76 | if (!all_name_chars((unsigned char *)PyString_AS_STRING(v))) |
| 77 | continue; |
| 78 | PyString_InternInPlace(&PyTuple_GET_ITEM(consts, i)); |
| 79 | } |
| 80 | co = PyObject_NEW(PyCodeObject, &PyCode_Type); |
| 81 | if (co != NULL) { |
| 82 | co->co_argcount = argcount; |
| 83 | co->co_nlocals = nlocals; |
| 84 | co->co_stacksize = stacksize; |
| 85 | co->co_flags = flags; |
| 86 | Py_INCREF(code); |
| 87 | co->co_code = code; |
| 88 | Py_INCREF(consts); |
| 89 | co->co_consts = consts; |
| 90 | Py_INCREF(names); |
| 91 | co->co_names = names; |
| 92 | Py_INCREF(varnames); |
| 93 | co->co_varnames = varnames; |
| 94 | Py_INCREF(freevars); |
| 95 | co->co_freevars = freevars; |
| 96 | Py_INCREF(cellvars); |
| 97 | co->co_cellvars = cellvars; |
| 98 | Py_INCREF(filename); |
| 99 | co->co_filename = filename; |
| 100 | Py_INCREF(name); |
| 101 | co->co_name = name; |
| 102 | co->co_firstlineno = firstlineno; |
| 103 | Py_INCREF(lnotab); |
| 104 | co->co_lnotab = lnotab; |
| 105 | } |
| 106 | return co; |
| 107 | } |
| 108 | |
| 109 | |
| 110 | #define OFF(x) offsetof(PyCodeObject, x) |
| 111 | |
| 112 | static PyMemberDef code_memberlist[] = { |
| 113 | {"co_argcount", T_INT, OFF(co_argcount), READONLY}, |
| 114 | {"co_nlocals", T_INT, OFF(co_nlocals), READONLY}, |
| 115 | {"co_stacksize",T_INT, OFF(co_stacksize), READONLY}, |
| 116 | {"co_flags", T_INT, OFF(co_flags), READONLY}, |
| 117 | {"co_code", T_OBJECT, OFF(co_code), READONLY}, |
| 118 | {"co_consts", T_OBJECT, OFF(co_consts), READONLY}, |
| 119 | {"co_names", T_OBJECT, OFF(co_names), READONLY}, |
| 120 | {"co_varnames", T_OBJECT, OFF(co_varnames), READONLY}, |
| 121 | {"co_freevars", T_OBJECT, OFF(co_freevars), READONLY}, |
| 122 | {"co_cellvars", T_OBJECT, OFF(co_cellvars), READONLY}, |
| 123 | {"co_filename", T_OBJECT, OFF(co_filename), READONLY}, |
| 124 | {"co_name", T_OBJECT, OFF(co_name), READONLY}, |
| 125 | {"co_firstlineno", T_INT, OFF(co_firstlineno), READONLY}, |
| 126 | {"co_lnotab", T_OBJECT, OFF(co_lnotab), READONLY}, |
| 127 | {NULL} /* Sentinel */ |
| 128 | }; |
| 129 | |
| 130 | /* Helper for code_new: return a shallow copy of a tuple that is |
| 131 | guaranteed to contain exact strings, by converting string subclasses |
| 132 | to exact strings and complaining if a non-string is found. */ |
| 133 | static PyObject* |
| 134 | validate_and_copy_tuple(PyObject *tup) |
| 135 | { |
| 136 | PyObject *newtuple; |
| 137 | PyObject *item; |
| 138 | int i, len; |
| 139 | |
| 140 | len = PyTuple_GET_SIZE(tup); |
| 141 | newtuple = PyTuple_New(len); |
| 142 | if (newtuple == NULL) |
| 143 | return NULL; |
| 144 | |
| 145 | for (i = 0; i < len; i++) { |
| 146 | item = PyTuple_GET_ITEM(tup, i); |
| 147 | if (PyString_CheckExact(item)) { |
| 148 | Py_INCREF(item); |
| 149 | } |
| 150 | else if (!PyString_Check(item)) { |
| 151 | PyErr_Format( |
| 152 | PyExc_TypeError, |
| 153 | "name tuples must contain only " |
| 154 | "strings, not '%.500s'", |
| 155 | item->ob_type->tp_name); |
| 156 | Py_DECREF(newtuple); |
| 157 | return NULL; |
| 158 | } |
| 159 | else { |
| 160 | item = PyString_FromStringAndSize( |
| 161 | PyString_AS_STRING(item), |
| 162 | PyString_GET_SIZE(item)); |
| 163 | if (item == NULL) { |
| 164 | Py_DECREF(newtuple); |
| 165 | return NULL; |
| 166 | } |
| 167 | } |
| 168 | PyTuple_SET_ITEM(newtuple, i, item); |
| 169 | } |
| 170 | |
| 171 | return newtuple; |
| 172 | } |
| 173 | |
| 174 | PyDoc_STRVAR(code_doc, |
| 175 | "code(argcount, nlocals, stacksize, flags, codestring, constants, names,\n\ |
| 176 | varnames, filename, name, firstlineno, lnotab[, freevars[, cellvars]])\n\ |
| 177 | \n\ |
| 178 | Create a code object. Not for the faint of heart."); |
| 179 | |
| 180 | static PyObject * |
| 181 | code_new(PyTypeObject *type, PyObject *args, PyObject *kw) |
| 182 | { |
| 183 | int argcount; |
| 184 | int nlocals; |
| 185 | int stacksize; |
| 186 | int flags; |
| 187 | PyObject *co = NULL; |
| 188 | PyObject *code; |
| 189 | PyObject *consts; |
| 190 | PyObject *names, *ournames = NULL; |
| 191 | PyObject *varnames, *ourvarnames = NULL; |
| 192 | PyObject *freevars = NULL, *ourfreevars = NULL; |
| 193 | PyObject *cellvars = NULL, *ourcellvars = NULL; |
| 194 | PyObject *filename; |
| 195 | PyObject *name; |
| 196 | int firstlineno; |
| 197 | PyObject *lnotab; |
| 198 | |
| 199 | if (!PyArg_ParseTuple(args, "iiiiSO!O!O!SSiS|O!O!:code", |
| 200 | &argcount, &nlocals, &stacksize, &flags, |
| 201 | &code, |
| 202 | &PyTuple_Type, &consts, |
| 203 | &PyTuple_Type, &names, |
| 204 | &PyTuple_Type, &varnames, |
| 205 | &filename, &name, |
| 206 | &firstlineno, &lnotab, |
| 207 | &PyTuple_Type, &freevars, |
| 208 | &PyTuple_Type, &cellvars)) |
| 209 | return NULL; |
| 210 | |
| 211 | if (argcount < 0) { |
| 212 | PyErr_SetString( |
| 213 | PyExc_ValueError, |
| 214 | "code: argcount must not be negative"); |
| 215 | goto cleanup; |
| 216 | } |
| 217 | |
| 218 | if (nlocals < 0) { |
| 219 | PyErr_SetString( |
| 220 | PyExc_ValueError, |
| 221 | "code: nlocals must not be negative"); |
| 222 | goto cleanup; |
| 223 | } |
| 224 | |
| 225 | ournames = validate_and_copy_tuple(names); |
| 226 | if (ournames == NULL) |
| 227 | goto cleanup; |
| 228 | ourvarnames = validate_and_copy_tuple(varnames); |
| 229 | if (ourvarnames == NULL) |
| 230 | goto cleanup; |
| 231 | if (freevars) |
| 232 | ourfreevars = validate_and_copy_tuple(freevars); |
| 233 | else |
| 234 | ourfreevars = PyTuple_New(0); |
| 235 | if (ourfreevars == NULL) |
| 236 | goto cleanup; |
| 237 | if (cellvars) |
| 238 | ourcellvars = validate_and_copy_tuple(cellvars); |
| 239 | else |
| 240 | ourcellvars = PyTuple_New(0); |
| 241 | if (ourcellvars == NULL) |
| 242 | goto cleanup; |
| 243 | |
| 244 | co = (PyObject *)PyCode_New(argcount, nlocals, stacksize, flags, |
| 245 | code, consts, ournames, ourvarnames, |
| 246 | ourfreevars, ourcellvars, filename, |
| 247 | name, firstlineno, lnotab); |
| 248 | cleanup: |
| 249 | Py_XDECREF(ournames); |
| 250 | Py_XDECREF(ourvarnames); |
| 251 | Py_XDECREF(ourfreevars); |
| 252 | Py_XDECREF(ourcellvars); |
| 253 | return co; |
| 254 | } |
| 255 | |
| 256 | static void |
| 257 | code_dealloc(PyCodeObject *co) |
| 258 | { |
| 259 | Py_XDECREF(co->co_code); |
| 260 | Py_XDECREF(co->co_consts); |
| 261 | Py_XDECREF(co->co_names); |
| 262 | Py_XDECREF(co->co_varnames); |
| 263 | Py_XDECREF(co->co_freevars); |
| 264 | Py_XDECREF(co->co_cellvars); |
| 265 | Py_XDECREF(co->co_filename); |
| 266 | Py_XDECREF(co->co_name); |
| 267 | Py_XDECREF(co->co_lnotab); |
| 268 | PyObject_DEL(co); |
| 269 | } |
| 270 | |
| 271 | static PyObject * |
| 272 | code_repr(PyCodeObject *co) |
| 273 | { |
| 274 | char buf[500]; |
| 275 | int lineno = -1; |
| 276 | char *filename = "???"; |
| 277 | char *name = "???"; |
| 278 | |
| 279 | if (co->co_firstlineno != 0) |
| 280 | lineno = co->co_firstlineno; |
| 281 | if (co->co_filename && PyString_Check(co->co_filename)) |
| 282 | filename = PyString_AS_STRING(co->co_filename); |
| 283 | if (co->co_name && PyString_Check(co->co_name)) |
| 284 | name = PyString_AS_STRING(co->co_name); |
| 285 | PyOS_snprintf(buf, sizeof(buf), |
| 286 | "<code object %.100s at %p, file \"%.300s\", line %d>", |
| 287 | name, co, filename, lineno); |
| 288 | return PyString_FromString(buf); |
| 289 | } |
| 290 | |
| 291 | static int |
| 292 | code_compare(PyCodeObject *co, PyCodeObject *cp) |
| 293 | { |
| 294 | int cmp; |
| 295 | cmp = PyObject_Compare(co->co_name, cp->co_name); |
| 296 | if (cmp) return cmp; |
| 297 | cmp = co->co_argcount - cp->co_argcount; |
| 298 | if (cmp) goto normalize; |
| 299 | cmp = co->co_nlocals - cp->co_nlocals; |
| 300 | if (cmp) goto normalize; |
| 301 | cmp = co->co_flags - cp->co_flags; |
| 302 | if (cmp) goto normalize; |
| 303 | cmp = co->co_firstlineno - cp->co_firstlineno; |
| 304 | if (cmp) goto normalize; |
| 305 | cmp = PyObject_Compare(co->co_code, cp->co_code); |
| 306 | if (cmp) return cmp; |
| 307 | cmp = PyObject_Compare(co->co_consts, cp->co_consts); |
| 308 | if (cmp) return cmp; |
| 309 | cmp = PyObject_Compare(co->co_names, cp->co_names); |
| 310 | if (cmp) return cmp; |
| 311 | cmp = PyObject_Compare(co->co_varnames, cp->co_varnames); |
| 312 | if (cmp) return cmp; |
| 313 | cmp = PyObject_Compare(co->co_freevars, cp->co_freevars); |
| 314 | if (cmp) return cmp; |
| 315 | cmp = PyObject_Compare(co->co_cellvars, cp->co_cellvars); |
| 316 | return cmp; |
| 317 | |
| 318 | normalize: |
| 319 | if (cmp > 0) |
| 320 | return 1; |
| 321 | else if (cmp < 0) |
| 322 | return -1; |
| 323 | else |
| 324 | return 0; |
| 325 | } |
| 326 | |
| 327 | static long |
| 328 | code_hash(PyCodeObject *co) |
| 329 | { |
| 330 | long h, h0, h1, h2, h3, h4, h5, h6; |
| 331 | h0 = PyObject_Hash(co->co_name); |
| 332 | if (h0 == -1) return -1; |
| 333 | h1 = PyObject_Hash(co->co_code); |
| 334 | if (h1 == -1) return -1; |
| 335 | h2 = PyObject_Hash(co->co_consts); |
| 336 | if (h2 == -1) return -1; |
| 337 | h3 = PyObject_Hash(co->co_names); |
| 338 | if (h3 == -1) return -1; |
| 339 | h4 = PyObject_Hash(co->co_varnames); |
| 340 | if (h4 == -1) return -1; |
| 341 | h5 = PyObject_Hash(co->co_freevars); |
| 342 | if (h5 == -1) return -1; |
| 343 | h6 = PyObject_Hash(co->co_cellvars); |
| 344 | if (h6 == -1) return -1; |
| 345 | h = h0 ^ h1 ^ h2 ^ h3 ^ h4 ^ h5 ^ h6 ^ |
| 346 | co->co_argcount ^ co->co_nlocals ^ co->co_flags; |
| 347 | if (h == -1) h = -2; |
| 348 | return h; |
| 349 | } |
| 350 | |
| 351 | /* XXX code objects need to participate in GC? */ |
| 352 | |
| 353 | PyTypeObject PyCode_Type = { |
| 354 | PyObject_HEAD_INIT(&PyType_Type) |
| 355 | 0, |
| 356 | "code", |
| 357 | sizeof(PyCodeObject), |
| 358 | 0, |
| 359 | (destructor)code_dealloc, /* tp_dealloc */ |
| 360 | 0, /* tp_print */ |
| 361 | 0, /* tp_getattr */ |
| 362 | 0, /* tp_setattr */ |
| 363 | (cmpfunc)code_compare, /* tp_compare */ |
| 364 | (reprfunc)code_repr, /* tp_repr */ |
| 365 | 0, /* tp_as_number */ |
| 366 | 0, /* tp_as_sequence */ |
| 367 | 0, /* tp_as_mapping */ |
| 368 | (hashfunc)code_hash, /* tp_hash */ |
| 369 | 0, /* tp_call */ |
| 370 | 0, /* tp_str */ |
| 371 | PyObject_GenericGetAttr, /* tp_getattro */ |
| 372 | 0, /* tp_setattro */ |
| 373 | 0, /* tp_as_buffer */ |
| 374 | Py_TPFLAGS_DEFAULT, /* tp_flags */ |
| 375 | code_doc, /* tp_doc */ |
| 376 | 0, /* tp_traverse */ |
| 377 | 0, /* tp_clear */ |
| 378 | 0, /* tp_richcompare */ |
| 379 | 0, /* tp_weaklistoffset */ |
| 380 | 0, /* tp_iter */ |
| 381 | 0, /* tp_iternext */ |
| 382 | 0, /* tp_methods */ |
| 383 | code_memberlist, /* tp_members */ |
| 384 | 0, /* tp_getset */ |
| 385 | 0, /* tp_base */ |
| 386 | 0, /* tp_dict */ |
| 387 | 0, /* tp_descr_get */ |
| 388 | 0, /* tp_descr_set */ |
| 389 | 0, /* tp_dictoffset */ |
| 390 | 0, /* tp_init */ |
| 391 | 0, /* tp_alloc */ |
| 392 | code_new, /* tp_new */ |
| 393 | }; |
| 394 | |
| 395 | /* All about c_lnotab. |
| 396 | |
| 397 | c_lnotab is an array of unsigned bytes disguised as a Python string. In -O |
| 398 | mode, SET_LINENO opcodes aren't generated, and bytecode offsets are mapped |
| 399 | to source code line #s (when needed for tracebacks) via c_lnotab instead. |
| 400 | The array is conceptually a list of |
| 401 | (bytecode offset increment, line number increment) |
| 402 | pairs. The details are important and delicate, best illustrated by example: |
| 403 | |
| 404 | byte code offset source code line number |
| 405 | 0 1 |
| 406 | 6 2 |
| 407 | 50 7 |
| 408 | 350 307 |
| 409 | 361 308 |
| 410 | |
| 411 | The first trick is that these numbers aren't stored, only the increments |
| 412 | from one row to the next (this doesn't really work, but it's a start): |
| 413 | |
| 414 | 0, 1, 6, 1, 44, 5, 300, 300, 11, 1 |
| 415 | |
| 416 | The second trick is that an unsigned byte can't hold negative values, or |
| 417 | values larger than 255, so (a) there's a deep assumption that byte code |
| 418 | offsets and their corresponding line #s both increase monotonically, and (b) |
| 419 | if at least one column jumps by more than 255 from one row to the next, more |
| 420 | than one pair is written to the table. In case #b, there's no way to know |
| 421 | from looking at the table later how many were written. That's the delicate |
| 422 | part. A user of c_lnotab desiring to find the source line number |
| 423 | corresponding to a bytecode address A should do something like this |
| 424 | |
| 425 | lineno = addr = 0 |
| 426 | for addr_incr, line_incr in c_lnotab: |
| 427 | addr += addr_incr |
| 428 | if addr > A: |
| 429 | return lineno |
| 430 | lineno += line_incr |
| 431 | |
| 432 | In order for this to work, when the addr field increments by more than 255, |
| 433 | the line # increment in each pair generated must be 0 until the remaining addr |
| 434 | increment is < 256. So, in the example above, com_set_lineno should not (as |
| 435 | was actually done until 2.2) expand 300, 300 to 255, 255, 45, 45, but to |
| 436 | 255, 0, 45, 255, 0, 45. |
| 437 | */ |
| 438 | |
| 439 | int |
| 440 | PyCode_Addr2Line(PyCodeObject *co, int addrq) |
| 441 | { |
| 442 | int size = PyString_Size(co->co_lnotab) / 2; |
| 443 | unsigned char *p = (unsigned char*)PyString_AsString(co->co_lnotab); |
| 444 | int line = co->co_firstlineno; |
| 445 | int addr = 0; |
| 446 | while (--size >= 0) { |
| 447 | addr += *p++; |
| 448 | if (addr > addrq) |
| 449 | break; |
| 450 | line += *p++; |
| 451 | } |
| 452 | return line; |
| 453 | } |