##// END OF EJS Templates
parsers: move index_get_parents's declaration higher...
Laurent Charignon -
r25311:d2e88f96 default
parent child Browse files
Show More
@@ -1,2648 +1,2649 b''
1 1 /*
2 2 parsers.c - efficient content parsing
3 3
4 4 Copyright 2008 Matt Mackall <mpm@selenic.com> and others
5 5
6 6 This software may be used and distributed according to the terms of
7 7 the GNU General Public License, incorporated herein by reference.
8 8 */
9 9
10 10 #include <Python.h>
11 11 #include <ctype.h>
12 12 #include <stddef.h>
13 13 #include <string.h>
14 14
15 15 #include "util.h"
16 16
17 17 static char *versionerrortext = "Python minor version mismatch";
18 18
19 19 static int8_t hextable[256] = {
20 20 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
21 21 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
22 22 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
23 23 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, -1, -1, -1, -1, /* 0-9 */
24 24 -1, 10, 11, 12, 13, 14, 15, -1, -1, -1, -1, -1, -1, -1, -1, -1, /* A-F */
25 25 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
26 26 -1, 10, 11, 12, 13, 14, 15, -1, -1, -1, -1, -1, -1, -1, -1, -1, /* a-f */
27 27 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
28 28 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
29 29 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
30 30 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
31 31 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
32 32 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
33 33 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
34 34 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
35 35 -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1
36 36 };
37 37
38 38 static char lowertable[128] = {
39 39 '\x00', '\x01', '\x02', '\x03', '\x04', '\x05', '\x06', '\x07',
40 40 '\x08', '\x09', '\x0a', '\x0b', '\x0c', '\x0d', '\x0e', '\x0f',
41 41 '\x10', '\x11', '\x12', '\x13', '\x14', '\x15', '\x16', '\x17',
42 42 '\x18', '\x19', '\x1a', '\x1b', '\x1c', '\x1d', '\x1e', '\x1f',
43 43 '\x20', '\x21', '\x22', '\x23', '\x24', '\x25', '\x26', '\x27',
44 44 '\x28', '\x29', '\x2a', '\x2b', '\x2c', '\x2d', '\x2e', '\x2f',
45 45 '\x30', '\x31', '\x32', '\x33', '\x34', '\x35', '\x36', '\x37',
46 46 '\x38', '\x39', '\x3a', '\x3b', '\x3c', '\x3d', '\x3e', '\x3f',
47 47 '\x40',
48 48 '\x61', '\x62', '\x63', '\x64', '\x65', '\x66', '\x67', /* A-G */
49 49 '\x68', '\x69', '\x6a', '\x6b', '\x6c', '\x6d', '\x6e', '\x6f', /* H-O */
50 50 '\x70', '\x71', '\x72', '\x73', '\x74', '\x75', '\x76', '\x77', /* P-W */
51 51 '\x78', '\x79', '\x7a', /* X-Z */
52 52 '\x5b', '\x5c', '\x5d', '\x5e', '\x5f',
53 53 '\x60', '\x61', '\x62', '\x63', '\x64', '\x65', '\x66', '\x67',
54 54 '\x68', '\x69', '\x6a', '\x6b', '\x6c', '\x6d', '\x6e', '\x6f',
55 55 '\x70', '\x71', '\x72', '\x73', '\x74', '\x75', '\x76', '\x77',
56 56 '\x78', '\x79', '\x7a', '\x7b', '\x7c', '\x7d', '\x7e', '\x7f'
57 57 };
58 58
59 59 static char uppertable[128] = {
60 60 '\x00', '\x01', '\x02', '\x03', '\x04', '\x05', '\x06', '\x07',
61 61 '\x08', '\x09', '\x0a', '\x0b', '\x0c', '\x0d', '\x0e', '\x0f',
62 62 '\x10', '\x11', '\x12', '\x13', '\x14', '\x15', '\x16', '\x17',
63 63 '\x18', '\x19', '\x1a', '\x1b', '\x1c', '\x1d', '\x1e', '\x1f',
64 64 '\x20', '\x21', '\x22', '\x23', '\x24', '\x25', '\x26', '\x27',
65 65 '\x28', '\x29', '\x2a', '\x2b', '\x2c', '\x2d', '\x2e', '\x2f',
66 66 '\x30', '\x31', '\x32', '\x33', '\x34', '\x35', '\x36', '\x37',
67 67 '\x38', '\x39', '\x3a', '\x3b', '\x3c', '\x3d', '\x3e', '\x3f',
68 68 '\x40', '\x41', '\x42', '\x43', '\x44', '\x45', '\x46', '\x47',
69 69 '\x48', '\x49', '\x4a', '\x4b', '\x4c', '\x4d', '\x4e', '\x4f',
70 70 '\x50', '\x51', '\x52', '\x53', '\x54', '\x55', '\x56', '\x57',
71 71 '\x58', '\x59', '\x5a', '\x5b', '\x5c', '\x5d', '\x5e', '\x5f',
72 72 '\x60',
73 73 '\x41', '\x42', '\x43', '\x44', '\x45', '\x46', '\x47', /* a-g */
74 74 '\x48', '\x49', '\x4a', '\x4b', '\x4c', '\x4d', '\x4e', '\x4f', /* h-o */
75 75 '\x50', '\x51', '\x52', '\x53', '\x54', '\x55', '\x56', '\x57', /* p-w */
76 76 '\x58', '\x59', '\x5a', /* x-z */
77 77 '\x7b', '\x7c', '\x7d', '\x7e', '\x7f'
78 78 };
79 79
80 80 static inline int hexdigit(const char *p, Py_ssize_t off)
81 81 {
82 82 int8_t val = hextable[(unsigned char)p[off]];
83 83
84 84 if (val >= 0) {
85 85 return val;
86 86 }
87 87
88 88 PyErr_SetString(PyExc_ValueError, "input contains non-hex character");
89 89 return 0;
90 90 }
91 91
92 92 /*
93 93 * Turn a hex-encoded string into binary.
94 94 */
95 95 PyObject *unhexlify(const char *str, int len)
96 96 {
97 97 PyObject *ret;
98 98 char *d;
99 99 int i;
100 100
101 101 ret = PyBytes_FromStringAndSize(NULL, len / 2);
102 102
103 103 if (!ret)
104 104 return NULL;
105 105
106 106 d = PyBytes_AsString(ret);
107 107
108 108 for (i = 0; i < len;) {
109 109 int hi = hexdigit(str, i++);
110 110 int lo = hexdigit(str, i++);
111 111 *d++ = (hi << 4) | lo;
112 112 }
113 113
114 114 return ret;
115 115 }
116 116
117 117 static inline PyObject *_asciitransform(PyObject *str_obj,
118 118 const char table[128],
119 119 PyObject *fallback_fn)
120 120 {
121 121 char *str, *newstr;
122 122 Py_ssize_t i, len;
123 123 PyObject *newobj = NULL;
124 124 PyObject *ret = NULL;
125 125
126 126 str = PyBytes_AS_STRING(str_obj);
127 127 len = PyBytes_GET_SIZE(str_obj);
128 128
129 129 newobj = PyBytes_FromStringAndSize(NULL, len);
130 130 if (!newobj)
131 131 goto quit;
132 132
133 133 newstr = PyBytes_AS_STRING(newobj);
134 134
135 135 for (i = 0; i < len; i++) {
136 136 char c = str[i];
137 137 if (c & 0x80) {
138 138 if (fallback_fn != NULL) {
139 139 ret = PyObject_CallFunctionObjArgs(fallback_fn,
140 140 str_obj, NULL);
141 141 } else {
142 142 PyObject *err = PyUnicodeDecodeError_Create(
143 143 "ascii", str, len, i, (i + 1),
144 144 "unexpected code byte");
145 145 PyErr_SetObject(PyExc_UnicodeDecodeError, err);
146 146 Py_XDECREF(err);
147 147 }
148 148 goto quit;
149 149 }
150 150 newstr[i] = table[(unsigned char)c];
151 151 }
152 152
153 153 ret = newobj;
154 154 Py_INCREF(ret);
155 155 quit:
156 156 Py_XDECREF(newobj);
157 157 return ret;
158 158 }
159 159
160 160 static PyObject *asciilower(PyObject *self, PyObject *args)
161 161 {
162 162 PyObject *str_obj;
163 163 if (!PyArg_ParseTuple(args, "O!:asciilower", &PyBytes_Type, &str_obj))
164 164 return NULL;
165 165 return _asciitransform(str_obj, lowertable, NULL);
166 166 }
167 167
168 168 static PyObject *asciiupper(PyObject *self, PyObject *args)
169 169 {
170 170 PyObject *str_obj;
171 171 if (!PyArg_ParseTuple(args, "O!:asciiupper", &PyBytes_Type, &str_obj))
172 172 return NULL;
173 173 return _asciitransform(str_obj, uppertable, NULL);
174 174 }
175 175
176 176 static PyObject *make_file_foldmap(PyObject *self, PyObject *args)
177 177 {
178 178 PyObject *dmap, *spec_obj, *normcase_fallback;
179 179 PyObject *file_foldmap = NULL;
180 180 enum normcase_spec spec;
181 181 PyObject *k, *v;
182 182 dirstateTupleObject *tuple;
183 183 Py_ssize_t pos = 0;
184 184 const char *table;
185 185
186 186 if (!PyArg_ParseTuple(args, "O!O!O!:make_file_foldmap",
187 187 &PyDict_Type, &dmap,
188 188 &PyInt_Type, &spec_obj,
189 189 &PyFunction_Type, &normcase_fallback))
190 190 goto quit;
191 191
192 192 spec = (int)PyInt_AS_LONG(spec_obj);
193 193 switch (spec) {
194 194 case NORMCASE_LOWER:
195 195 table = lowertable;
196 196 break;
197 197 case NORMCASE_UPPER:
198 198 table = uppertable;
199 199 break;
200 200 case NORMCASE_OTHER:
201 201 table = NULL;
202 202 break;
203 203 default:
204 204 PyErr_SetString(PyExc_TypeError, "invalid normcasespec");
205 205 goto quit;
206 206 }
207 207
208 208 #if PY_VERSION_HEX >= 0x02060000
209 209 /* _PyDict_NewPresized expects a minused parameter, but it actually
210 210 creates a dictionary that's the nearest power of two bigger than the
211 211 parameter. For example, with the initial minused = 1000, the
212 212 dictionary created has size 1024. Of course in a lot of cases that
213 213 can be greater than the maximum load factor Python's dict object
214 214 expects (= 2/3), so as soon as we cross the threshold we'll resize
215 215 anyway. So create a dictionary that's 3/2 the size. Also add some
216 216 more to deal with additions outside this function. */
217 217 file_foldmap = _PyDict_NewPresized((PyDict_Size(dmap) / 5) * 8);
218 218 #else
219 219 file_foldmap = PyDict_New();
220 220 #endif
221 221
222 222 if (file_foldmap == NULL)
223 223 goto quit;
224 224
225 225 while (PyDict_Next(dmap, &pos, &k, &v)) {
226 226 if (!dirstate_tuple_check(v)) {
227 227 PyErr_SetString(PyExc_TypeError,
228 228 "expected a dirstate tuple");
229 229 goto quit;
230 230 }
231 231
232 232 tuple = (dirstateTupleObject *)v;
233 233 if (tuple->state != 'r') {
234 234 PyObject *normed;
235 235 if (table != NULL) {
236 236 normed = _asciitransform(k, table,
237 237 normcase_fallback);
238 238 } else {
239 239 normed = PyObject_CallFunctionObjArgs(
240 240 normcase_fallback, k, NULL);
241 241 }
242 242
243 243 if (normed == NULL)
244 244 goto quit;
245 245 if (PyDict_SetItem(file_foldmap, normed, k) == -1)
246 246 goto quit;
247 247 }
248 248 }
249 249 return file_foldmap;
250 250 quit:
251 251 Py_XDECREF(file_foldmap);
252 252 return NULL;
253 253 }
254 254
255 255 /*
256 256 * This code assumes that a manifest is stitched together with newline
257 257 * ('\n') characters.
258 258 */
259 259 static PyObject *parse_manifest(PyObject *self, PyObject *args)
260 260 {
261 261 PyObject *mfdict, *fdict;
262 262 char *str, *start, *end;
263 263 int len;
264 264
265 265 if (!PyArg_ParseTuple(args, "O!O!s#:parse_manifest",
266 266 &PyDict_Type, &mfdict,
267 267 &PyDict_Type, &fdict,
268 268 &str, &len))
269 269 goto quit;
270 270
271 271 start = str;
272 272 end = str + len;
273 273 while (start < end) {
274 274 PyObject *file = NULL, *node = NULL;
275 275 PyObject *flags = NULL;
276 276 char *zero = NULL, *newline = NULL;
277 277 ptrdiff_t nlen;
278 278
279 279 zero = memchr(start, '\0', end - start);
280 280 if (!zero) {
281 281 PyErr_SetString(PyExc_ValueError,
282 282 "manifest entry has no separator");
283 283 goto quit;
284 284 }
285 285
286 286 newline = memchr(zero + 1, '\n', end - (zero + 1));
287 287 if (!newline) {
288 288 PyErr_SetString(PyExc_ValueError,
289 289 "manifest contains trailing garbage");
290 290 goto quit;
291 291 }
292 292
293 293 file = PyBytes_FromStringAndSize(start, zero - start);
294 294
295 295 if (!file)
296 296 goto bail;
297 297
298 298 nlen = newline - zero - 1;
299 299
300 300 node = unhexlify(zero + 1, nlen > 40 ? 40 : (int)nlen);
301 301 if (!node)
302 302 goto bail;
303 303
304 304 if (nlen > 40) {
305 305 flags = PyBytes_FromStringAndSize(zero + 41,
306 306 nlen - 40);
307 307 if (!flags)
308 308 goto bail;
309 309
310 310 if (PyDict_SetItem(fdict, file, flags) == -1)
311 311 goto bail;
312 312 }
313 313
314 314 if (PyDict_SetItem(mfdict, file, node) == -1)
315 315 goto bail;
316 316
317 317 start = newline + 1;
318 318
319 319 Py_XDECREF(flags);
320 320 Py_XDECREF(node);
321 321 Py_XDECREF(file);
322 322 continue;
323 323 bail:
324 324 Py_XDECREF(flags);
325 325 Py_XDECREF(node);
326 326 Py_XDECREF(file);
327 327 goto quit;
328 328 }
329 329
330 330 Py_INCREF(Py_None);
331 331 return Py_None;
332 332 quit:
333 333 return NULL;
334 334 }
335 335
336 336 static inline dirstateTupleObject *make_dirstate_tuple(char state, int mode,
337 337 int size, int mtime)
338 338 {
339 339 dirstateTupleObject *t = PyObject_New(dirstateTupleObject,
340 340 &dirstateTupleType);
341 341 if (!t)
342 342 return NULL;
343 343 t->state = state;
344 344 t->mode = mode;
345 345 t->size = size;
346 346 t->mtime = mtime;
347 347 return t;
348 348 }
349 349
350 350 static PyObject *dirstate_tuple_new(PyTypeObject *subtype, PyObject *args,
351 351 PyObject *kwds)
352 352 {
353 353 /* We do all the initialization here and not a tp_init function because
354 354 * dirstate_tuple is immutable. */
355 355 dirstateTupleObject *t;
356 356 char state;
357 357 int size, mode, mtime;
358 358 if (!PyArg_ParseTuple(args, "ciii", &state, &mode, &size, &mtime))
359 359 return NULL;
360 360
361 361 t = (dirstateTupleObject *)subtype->tp_alloc(subtype, 1);
362 362 if (!t)
363 363 return NULL;
364 364 t->state = state;
365 365 t->mode = mode;
366 366 t->size = size;
367 367 t->mtime = mtime;
368 368
369 369 return (PyObject *)t;
370 370 }
371 371
372 372 static void dirstate_tuple_dealloc(PyObject *o)
373 373 {
374 374 PyObject_Del(o);
375 375 }
376 376
377 377 static Py_ssize_t dirstate_tuple_length(PyObject *o)
378 378 {
379 379 return 4;
380 380 }
381 381
382 382 static PyObject *dirstate_tuple_item(PyObject *o, Py_ssize_t i)
383 383 {
384 384 dirstateTupleObject *t = (dirstateTupleObject *)o;
385 385 switch (i) {
386 386 case 0:
387 387 return PyBytes_FromStringAndSize(&t->state, 1);
388 388 case 1:
389 389 return PyInt_FromLong(t->mode);
390 390 case 2:
391 391 return PyInt_FromLong(t->size);
392 392 case 3:
393 393 return PyInt_FromLong(t->mtime);
394 394 default:
395 395 PyErr_SetString(PyExc_IndexError, "index out of range");
396 396 return NULL;
397 397 }
398 398 }
399 399
400 400 static PySequenceMethods dirstate_tuple_sq = {
401 401 dirstate_tuple_length, /* sq_length */
402 402 0, /* sq_concat */
403 403 0, /* sq_repeat */
404 404 dirstate_tuple_item, /* sq_item */
405 405 0, /* sq_ass_item */
406 406 0, /* sq_contains */
407 407 0, /* sq_inplace_concat */
408 408 0 /* sq_inplace_repeat */
409 409 };
410 410
411 411 PyTypeObject dirstateTupleType = {
412 412 PyVarObject_HEAD_INIT(NULL, 0)
413 413 "dirstate_tuple", /* tp_name */
414 414 sizeof(dirstateTupleObject),/* tp_basicsize */
415 415 0, /* tp_itemsize */
416 416 (destructor)dirstate_tuple_dealloc, /* tp_dealloc */
417 417 0, /* tp_print */
418 418 0, /* tp_getattr */
419 419 0, /* tp_setattr */
420 420 0, /* tp_compare */
421 421 0, /* tp_repr */
422 422 0, /* tp_as_number */
423 423 &dirstate_tuple_sq, /* tp_as_sequence */
424 424 0, /* tp_as_mapping */
425 425 0, /* tp_hash */
426 426 0, /* tp_call */
427 427 0, /* tp_str */
428 428 0, /* tp_getattro */
429 429 0, /* tp_setattro */
430 430 0, /* tp_as_buffer */
431 431 Py_TPFLAGS_DEFAULT, /* tp_flags */
432 432 "dirstate tuple", /* tp_doc */
433 433 0, /* tp_traverse */
434 434 0, /* tp_clear */
435 435 0, /* tp_richcompare */
436 436 0, /* tp_weaklistoffset */
437 437 0, /* tp_iter */
438 438 0, /* tp_iternext */
439 439 0, /* tp_methods */
440 440 0, /* tp_members */
441 441 0, /* tp_getset */
442 442 0, /* tp_base */
443 443 0, /* tp_dict */
444 444 0, /* tp_descr_get */
445 445 0, /* tp_descr_set */
446 446 0, /* tp_dictoffset */
447 447 0, /* tp_init */
448 448 0, /* tp_alloc */
449 449 dirstate_tuple_new, /* tp_new */
450 450 };
451 451
452 452 static PyObject *parse_dirstate(PyObject *self, PyObject *args)
453 453 {
454 454 PyObject *dmap, *cmap, *parents = NULL, *ret = NULL;
455 455 PyObject *fname = NULL, *cname = NULL, *entry = NULL;
456 456 char state, *cur, *str, *cpos;
457 457 int mode, size, mtime;
458 458 unsigned int flen, len, pos = 40;
459 459 int readlen;
460 460
461 461 if (!PyArg_ParseTuple(args, "O!O!s#:parse_dirstate",
462 462 &PyDict_Type, &dmap,
463 463 &PyDict_Type, &cmap,
464 464 &str, &readlen))
465 465 goto quit;
466 466
467 467 if (readlen < 0)
468 468 goto quit;
469 469
470 470 len = readlen;
471 471
472 472 /* read parents */
473 473 if (len < 40)
474 474 goto quit;
475 475
476 476 parents = Py_BuildValue("s#s#", str, 20, str + 20, 20);
477 477 if (!parents)
478 478 goto quit;
479 479
480 480 /* read filenames */
481 481 while (pos >= 40 && pos < len) {
482 482 cur = str + pos;
483 483 /* unpack header */
484 484 state = *cur;
485 485 mode = getbe32(cur + 1);
486 486 size = getbe32(cur + 5);
487 487 mtime = getbe32(cur + 9);
488 488 flen = getbe32(cur + 13);
489 489 pos += 17;
490 490 cur += 17;
491 491 if (flen > len - pos) {
492 492 PyErr_SetString(PyExc_ValueError, "overflow in dirstate");
493 493 goto quit;
494 494 }
495 495
496 496 entry = (PyObject *)make_dirstate_tuple(state, mode, size,
497 497 mtime);
498 498 cpos = memchr(cur, 0, flen);
499 499 if (cpos) {
500 500 fname = PyBytes_FromStringAndSize(cur, cpos - cur);
501 501 cname = PyBytes_FromStringAndSize(cpos + 1,
502 502 flen - (cpos - cur) - 1);
503 503 if (!fname || !cname ||
504 504 PyDict_SetItem(cmap, fname, cname) == -1 ||
505 505 PyDict_SetItem(dmap, fname, entry) == -1)
506 506 goto quit;
507 507 Py_DECREF(cname);
508 508 } else {
509 509 fname = PyBytes_FromStringAndSize(cur, flen);
510 510 if (!fname ||
511 511 PyDict_SetItem(dmap, fname, entry) == -1)
512 512 goto quit;
513 513 }
514 514 Py_DECREF(fname);
515 515 Py_DECREF(entry);
516 516 fname = cname = entry = NULL;
517 517 pos += flen;
518 518 }
519 519
520 520 ret = parents;
521 521 Py_INCREF(ret);
522 522 quit:
523 523 Py_XDECREF(fname);
524 524 Py_XDECREF(cname);
525 525 Py_XDECREF(entry);
526 526 Py_XDECREF(parents);
527 527 return ret;
528 528 }
529 529
530 530 /*
531 531 * Efficiently pack a dirstate object into its on-disk format.
532 532 */
533 533 static PyObject *pack_dirstate(PyObject *self, PyObject *args)
534 534 {
535 535 PyObject *packobj = NULL;
536 536 PyObject *map, *copymap, *pl, *mtime_unset = NULL;
537 537 Py_ssize_t nbytes, pos, l;
538 538 PyObject *k, *v = NULL, *pn;
539 539 char *p, *s;
540 540 double now;
541 541
542 542 if (!PyArg_ParseTuple(args, "O!O!Od:pack_dirstate",
543 543 &PyDict_Type, &map, &PyDict_Type, &copymap,
544 544 &pl, &now))
545 545 return NULL;
546 546
547 547 if (!PySequence_Check(pl) || PySequence_Size(pl) != 2) {
548 548 PyErr_SetString(PyExc_TypeError, "expected 2-element sequence");
549 549 return NULL;
550 550 }
551 551
552 552 /* Figure out how much we need to allocate. */
553 553 for (nbytes = 40, pos = 0; PyDict_Next(map, &pos, &k, &v);) {
554 554 PyObject *c;
555 555 if (!PyString_Check(k)) {
556 556 PyErr_SetString(PyExc_TypeError, "expected string key");
557 557 goto bail;
558 558 }
559 559 nbytes += PyString_GET_SIZE(k) + 17;
560 560 c = PyDict_GetItem(copymap, k);
561 561 if (c) {
562 562 if (!PyString_Check(c)) {
563 563 PyErr_SetString(PyExc_TypeError,
564 564 "expected string key");
565 565 goto bail;
566 566 }
567 567 nbytes += PyString_GET_SIZE(c) + 1;
568 568 }
569 569 }
570 570
571 571 packobj = PyString_FromStringAndSize(NULL, nbytes);
572 572 if (packobj == NULL)
573 573 goto bail;
574 574
575 575 p = PyString_AS_STRING(packobj);
576 576
577 577 pn = PySequence_ITEM(pl, 0);
578 578 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
579 579 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
580 580 goto bail;
581 581 }
582 582 memcpy(p, s, l);
583 583 p += 20;
584 584 pn = PySequence_ITEM(pl, 1);
585 585 if (PyString_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
586 586 PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
587 587 goto bail;
588 588 }
589 589 memcpy(p, s, l);
590 590 p += 20;
591 591
592 592 for (pos = 0; PyDict_Next(map, &pos, &k, &v); ) {
593 593 dirstateTupleObject *tuple;
594 594 char state;
595 595 uint32_t mode, size, mtime;
596 596 Py_ssize_t len, l;
597 597 PyObject *o;
598 598 char *t;
599 599
600 600 if (!dirstate_tuple_check(v)) {
601 601 PyErr_SetString(PyExc_TypeError,
602 602 "expected a dirstate tuple");
603 603 goto bail;
604 604 }
605 605 tuple = (dirstateTupleObject *)v;
606 606
607 607 state = tuple->state;
608 608 mode = tuple->mode;
609 609 size = tuple->size;
610 610 mtime = tuple->mtime;
611 611 if (state == 'n' && mtime == (uint32_t)now) {
612 612 /* See pure/parsers.py:pack_dirstate for why we do
613 613 * this. */
614 614 mtime = -1;
615 615 mtime_unset = (PyObject *)make_dirstate_tuple(
616 616 state, mode, size, mtime);
617 617 if (!mtime_unset)
618 618 goto bail;
619 619 if (PyDict_SetItem(map, k, mtime_unset) == -1)
620 620 goto bail;
621 621 Py_DECREF(mtime_unset);
622 622 mtime_unset = NULL;
623 623 }
624 624 *p++ = state;
625 625 putbe32(mode, p);
626 626 putbe32(size, p + 4);
627 627 putbe32(mtime, p + 8);
628 628 t = p + 12;
629 629 p += 16;
630 630 len = PyString_GET_SIZE(k);
631 631 memcpy(p, PyString_AS_STRING(k), len);
632 632 p += len;
633 633 o = PyDict_GetItem(copymap, k);
634 634 if (o) {
635 635 *p++ = '\0';
636 636 l = PyString_GET_SIZE(o);
637 637 memcpy(p, PyString_AS_STRING(o), l);
638 638 p += l;
639 639 len += l + 1;
640 640 }
641 641 putbe32((uint32_t)len, t);
642 642 }
643 643
644 644 pos = p - PyString_AS_STRING(packobj);
645 645 if (pos != nbytes) {
646 646 PyErr_Format(PyExc_SystemError, "bad dirstate size: %ld != %ld",
647 647 (long)pos, (long)nbytes);
648 648 goto bail;
649 649 }
650 650
651 651 return packobj;
652 652 bail:
653 653 Py_XDECREF(mtime_unset);
654 654 Py_XDECREF(packobj);
655 655 Py_XDECREF(v);
656 656 return NULL;
657 657 }
658 658
659 659 /*
660 660 * A base-16 trie for fast node->rev mapping.
661 661 *
662 662 * Positive value is index of the next node in the trie
663 663 * Negative value is a leaf: -(rev + 1)
664 664 * Zero is empty
665 665 */
666 666 typedef struct {
667 667 int children[16];
668 668 } nodetree;
669 669
670 670 /*
671 671 * This class has two behaviours.
672 672 *
673 673 * When used in a list-like way (with integer keys), we decode an
674 674 * entry in a RevlogNG index file on demand. Our last entry is a
675 675 * sentinel, always a nullid. We have limited support for
676 676 * integer-keyed insert and delete, only at elements right before the
677 677 * sentinel.
678 678 *
679 679 * With string keys, we lazily perform a reverse mapping from node to
680 680 * rev, using a base-16 trie.
681 681 */
682 682 typedef struct {
683 683 PyObject_HEAD
684 684 /* Type-specific fields go here. */
685 685 PyObject *data; /* raw bytes of index */
686 686 PyObject **cache; /* cached tuples */
687 687 const char **offsets; /* populated on demand */
688 688 Py_ssize_t raw_length; /* original number of elements */
689 689 Py_ssize_t length; /* current number of elements */
690 690 PyObject *added; /* populated on demand */
691 691 PyObject *headrevs; /* cache, invalidated on changes */
692 692 PyObject *filteredrevs;/* filtered revs set */
693 693 nodetree *nt; /* base-16 trie */
694 694 int ntlength; /* # nodes in use */
695 695 int ntcapacity; /* # nodes allocated */
696 696 int ntdepth; /* maximum depth of tree */
697 697 int ntsplits; /* # splits performed */
698 698 int ntrev; /* last rev scanned */
699 699 int ntlookups; /* # lookups */
700 700 int ntmisses; /* # lookups that miss the cache */
701 701 int inlined;
702 702 } indexObject;
703 703
704 704 static Py_ssize_t index_length(const indexObject *self)
705 705 {
706 706 if (self->added == NULL)
707 707 return self->length;
708 708 return self->length + PyList_GET_SIZE(self->added);
709 709 }
710 710
711 711 static PyObject *nullentry;
712 712 static const char nullid[20];
713 713
714 714 static Py_ssize_t inline_scan(indexObject *self, const char **offsets);
715 715
716 716 #if LONG_MAX == 0x7fffffffL
717 717 static char *tuple_format = "Kiiiiiis#";
718 718 #else
719 719 static char *tuple_format = "kiiiiiis#";
720 720 #endif
721 721
722 722 /* A RevlogNG v1 index entry is 64 bytes long. */
723 723 static const long v1_hdrsize = 64;
724 724
725 725 /*
726 726 * Return a pointer to the beginning of a RevlogNG record.
727 727 */
728 728 static const char *index_deref(indexObject *self, Py_ssize_t pos)
729 729 {
730 730 if (self->inlined && pos > 0) {
731 731 if (self->offsets == NULL) {
732 732 self->offsets = malloc(self->raw_length *
733 733 sizeof(*self->offsets));
734 734 if (self->offsets == NULL)
735 735 return (const char *)PyErr_NoMemory();
736 736 inline_scan(self, self->offsets);
737 737 }
738 738 return self->offsets[pos];
739 739 }
740 740
741 741 return PyString_AS_STRING(self->data) + pos * v1_hdrsize;
742 742 }
743 743
744 static inline void index_get_parents(indexObject *self, Py_ssize_t rev,
745 int *ps)
746 {
747 if (rev >= self->length - 1) {
748 PyObject *tuple = PyList_GET_ITEM(self->added,
749 rev - self->length + 1);
750 ps[0] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 5));
751 ps[1] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 6));
752 } else {
753 const char *data = index_deref(self, rev);
754 ps[0] = getbe32(data + 24);
755 ps[1] = getbe32(data + 28);
756 }
757 }
758
759
744 760 /*
745 761 * RevlogNG format (all in big endian, data may be inlined):
746 762 * 6 bytes: offset
747 763 * 2 bytes: flags
748 764 * 4 bytes: compressed length
749 765 * 4 bytes: uncompressed length
750 766 * 4 bytes: base revision
751 767 * 4 bytes: link revision
752 768 * 4 bytes: parent 1 revision
753 769 * 4 bytes: parent 2 revision
754 770 * 32 bytes: nodeid (only 20 bytes used)
755 771 */
756 772 static PyObject *index_get(indexObject *self, Py_ssize_t pos)
757 773 {
758 774 uint64_t offset_flags;
759 775 int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2;
760 776 const char *c_node_id;
761 777 const char *data;
762 778 Py_ssize_t length = index_length(self);
763 779 PyObject *entry;
764 780
765 781 if (pos < 0)
766 782 pos += length;
767 783
768 784 if (pos < 0 || pos >= length) {
769 785 PyErr_SetString(PyExc_IndexError, "revlog index out of range");
770 786 return NULL;
771 787 }
772 788
773 789 if (pos == length - 1) {
774 790 Py_INCREF(nullentry);
775 791 return nullentry;
776 792 }
777 793
778 794 if (pos >= self->length - 1) {
779 795 PyObject *obj;
780 796 obj = PyList_GET_ITEM(self->added, pos - self->length + 1);
781 797 Py_INCREF(obj);
782 798 return obj;
783 799 }
784 800
785 801 if (self->cache) {
786 802 if (self->cache[pos]) {
787 803 Py_INCREF(self->cache[pos]);
788 804 return self->cache[pos];
789 805 }
790 806 } else {
791 807 self->cache = calloc(self->raw_length, sizeof(PyObject *));
792 808 if (self->cache == NULL)
793 809 return PyErr_NoMemory();
794 810 }
795 811
796 812 data = index_deref(self, pos);
797 813 if (data == NULL)
798 814 return NULL;
799 815
800 816 offset_flags = getbe32(data + 4);
801 817 if (pos == 0) /* mask out version number for the first entry */
802 818 offset_flags &= 0xFFFF;
803 819 else {
804 820 uint32_t offset_high = getbe32(data);
805 821 offset_flags |= ((uint64_t)offset_high) << 32;
806 822 }
807 823
808 824 comp_len = getbe32(data + 8);
809 825 uncomp_len = getbe32(data + 12);
810 826 base_rev = getbe32(data + 16);
811 827 link_rev = getbe32(data + 20);
812 828 parent_1 = getbe32(data + 24);
813 829 parent_2 = getbe32(data + 28);
814 830 c_node_id = data + 32;
815 831
816 832 entry = Py_BuildValue(tuple_format, offset_flags, comp_len,
817 833 uncomp_len, base_rev, link_rev,
818 834 parent_1, parent_2, c_node_id, 20);
819 835
820 836 if (entry) {
821 837 PyObject_GC_UnTrack(entry);
822 838 Py_INCREF(entry);
823 839 }
824 840
825 841 self->cache[pos] = entry;
826 842
827 843 return entry;
828 844 }
829 845
830 846 /*
831 847 * Return the 20-byte SHA of the node corresponding to the given rev.
832 848 */
833 849 static const char *index_node(indexObject *self, Py_ssize_t pos)
834 850 {
835 851 Py_ssize_t length = index_length(self);
836 852 const char *data;
837 853
838 854 if (pos == length - 1 || pos == INT_MAX)
839 855 return nullid;
840 856
841 857 if (pos >= length)
842 858 return NULL;
843 859
844 860 if (pos >= self->length - 1) {
845 861 PyObject *tuple, *str;
846 862 tuple = PyList_GET_ITEM(self->added, pos - self->length + 1);
847 863 str = PyTuple_GetItem(tuple, 7);
848 864 return str ? PyString_AS_STRING(str) : NULL;
849 865 }
850 866
851 867 data = index_deref(self, pos);
852 868 return data ? data + 32 : NULL;
853 869 }
854 870
855 871 static int nt_insert(indexObject *self, const char *node, int rev);
856 872
857 873 static int node_check(PyObject *obj, char **node, Py_ssize_t *nodelen)
858 874 {
859 875 if (PyString_AsStringAndSize(obj, node, nodelen) == -1)
860 876 return -1;
861 877 if (*nodelen == 20)
862 878 return 0;
863 879 PyErr_SetString(PyExc_ValueError, "20-byte hash required");
864 880 return -1;
865 881 }
866 882
867 883 static PyObject *index_insert(indexObject *self, PyObject *args)
868 884 {
869 885 PyObject *obj;
870 886 char *node;
871 887 int index;
872 888 Py_ssize_t len, nodelen;
873 889
874 890 if (!PyArg_ParseTuple(args, "iO", &index, &obj))
875 891 return NULL;
876 892
877 893 if (!PyTuple_Check(obj) || PyTuple_GET_SIZE(obj) != 8) {
878 894 PyErr_SetString(PyExc_TypeError, "8-tuple required");
879 895 return NULL;
880 896 }
881 897
882 898 if (node_check(PyTuple_GET_ITEM(obj, 7), &node, &nodelen) == -1)
883 899 return NULL;
884 900
885 901 len = index_length(self);
886 902
887 903 if (index < 0)
888 904 index += len;
889 905
890 906 if (index != len - 1) {
891 907 PyErr_SetString(PyExc_IndexError,
892 908 "insert only supported at index -1");
893 909 return NULL;
894 910 }
895 911
896 912 if (self->added == NULL) {
897 913 self->added = PyList_New(0);
898 914 if (self->added == NULL)
899 915 return NULL;
900 916 }
901 917
902 918 if (PyList_Append(self->added, obj) == -1)
903 919 return NULL;
904 920
905 921 if (self->nt)
906 922 nt_insert(self, node, index);
907 923
908 924 Py_CLEAR(self->headrevs);
909 925 Py_RETURN_NONE;
910 926 }
911 927
912 928 static void _index_clearcaches(indexObject *self)
913 929 {
914 930 if (self->cache) {
915 931 Py_ssize_t i;
916 932
917 933 for (i = 0; i < self->raw_length; i++)
918 934 Py_CLEAR(self->cache[i]);
919 935 free(self->cache);
920 936 self->cache = NULL;
921 937 }
922 938 if (self->offsets) {
923 939 free(self->offsets);
924 940 self->offsets = NULL;
925 941 }
926 942 if (self->nt) {
927 943 free(self->nt);
928 944 self->nt = NULL;
929 945 }
930 946 Py_CLEAR(self->headrevs);
931 947 }
932 948
933 949 static PyObject *index_clearcaches(indexObject *self)
934 950 {
935 951 _index_clearcaches(self);
936 952 self->ntlength = self->ntcapacity = 0;
937 953 self->ntdepth = self->ntsplits = 0;
938 954 self->ntrev = -1;
939 955 self->ntlookups = self->ntmisses = 0;
940 956 Py_RETURN_NONE;
941 957 }
942 958
943 959 static PyObject *index_stats(indexObject *self)
944 960 {
945 961 PyObject *obj = PyDict_New();
946 962 PyObject *t = NULL;
947 963
948 964 if (obj == NULL)
949 965 return NULL;
950 966
951 967 #define istat(__n, __d) \
952 968 t = PyInt_FromSsize_t(self->__n); \
953 969 if (!t) \
954 970 goto bail; \
955 971 if (PyDict_SetItemString(obj, __d, t) == -1) \
956 972 goto bail; \
957 973 Py_DECREF(t);
958 974
959 975 if (self->added) {
960 976 Py_ssize_t len = PyList_GET_SIZE(self->added);
961 977 t = PyInt_FromSsize_t(len);
962 978 if (!t)
963 979 goto bail;
964 980 if (PyDict_SetItemString(obj, "index entries added", t) == -1)
965 981 goto bail;
966 982 Py_DECREF(t);
967 983 }
968 984
969 985 if (self->raw_length != self->length - 1)
970 986 istat(raw_length, "revs on disk");
971 987 istat(length, "revs in memory");
972 988 istat(ntcapacity, "node trie capacity");
973 989 istat(ntdepth, "node trie depth");
974 990 istat(ntlength, "node trie count");
975 991 istat(ntlookups, "node trie lookups");
976 992 istat(ntmisses, "node trie misses");
977 993 istat(ntrev, "node trie last rev scanned");
978 994 istat(ntsplits, "node trie splits");
979 995
980 996 #undef istat
981 997
982 998 return obj;
983 999
984 1000 bail:
985 1001 Py_XDECREF(obj);
986 1002 Py_XDECREF(t);
987 1003 return NULL;
988 1004 }
989 1005
990 1006 /*
991 1007 * When we cache a list, we want to be sure the caller can't mutate
992 1008 * the cached copy.
993 1009 */
994 1010 static PyObject *list_copy(PyObject *list)
995 1011 {
996 1012 Py_ssize_t len = PyList_GET_SIZE(list);
997 1013 PyObject *newlist = PyList_New(len);
998 1014 Py_ssize_t i;
999 1015
1000 1016 if (newlist == NULL)
1001 1017 return NULL;
1002 1018
1003 1019 for (i = 0; i < len; i++) {
1004 1020 PyObject *obj = PyList_GET_ITEM(list, i);
1005 1021 Py_INCREF(obj);
1006 1022 PyList_SET_ITEM(newlist, i, obj);
1007 1023 }
1008 1024
1009 1025 return newlist;
1010 1026 }
1011 1027
1012 1028 /* arg should be Py_ssize_t but Python 2.4 do not support the n format */
1013 1029 static int check_filter(PyObject *filter, unsigned long arg) {
1014 1030 if (filter) {
1015 1031 PyObject *arglist, *result;
1016 1032 int isfiltered;
1017 1033
1018 1034 arglist = Py_BuildValue("(k)", arg);
1019 1035 if (!arglist) {
1020 1036 return -1;
1021 1037 }
1022 1038
1023 1039 result = PyEval_CallObject(filter, arglist);
1024 1040 Py_DECREF(arglist);
1025 1041 if (!result) {
1026 1042 return -1;
1027 1043 }
1028 1044
1029 1045 /* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error,
1030 1046 * same as this function, so we can just return it directly.*/
1031 1047 isfiltered = PyObject_IsTrue(result);
1032 1048 Py_DECREF(result);
1033 1049 return isfiltered;
1034 1050 } else {
1035 1051 return 0;
1036 1052 }
1037 1053 }
1038 1054
1039 1055 static Py_ssize_t add_roots_get_min(indexObject *self, PyObject *list,
1040 1056 Py_ssize_t marker, char *phases)
1041 1057 {
1042 1058 PyObject *iter = NULL;
1043 1059 PyObject *iter_item = NULL;
1044 1060 Py_ssize_t min_idx = index_length(self) + 1;
1045 1061 long iter_item_long;
1046 1062
1047 1063 if (PyList_GET_SIZE(list) != 0) {
1048 1064 iter = PyObject_GetIter(list);
1049 1065 if (iter == NULL)
1050 1066 return -2;
1051 1067 while ((iter_item = PyIter_Next(iter)))
1052 1068 {
1053 1069 iter_item_long = PyInt_AS_LONG(iter_item);
1054 1070 Py_DECREF(iter_item);
1055 1071 if (iter_item_long < min_idx)
1056 1072 min_idx = iter_item_long;
1057 1073 phases[iter_item_long] = marker;
1058 1074 }
1059 1075 Py_DECREF(iter);
1060 1076 }
1061 1077
1062 1078 return min_idx;
1063 1079 }
1064 1080
1065 1081 static inline void set_phase_from_parents(char *phases, int parent_1,
1066 1082 int parent_2, Py_ssize_t i)
1067 1083 {
1068 1084 if (parent_1 >= 0 && phases[parent_1] > phases[i])
1069 1085 phases[i] = phases[parent_1];
1070 1086 if (parent_2 >= 0 && phases[parent_2] > phases[i])
1071 1087 phases[i] = phases[parent_2];
1072 1088 }
1073 1089
1074 1090 static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args)
1075 1091 {
1076 1092 PyObject *roots = Py_None;
1077 1093 PyObject *ret = NULL;
1078 1094 PyObject *phaseslist = NULL;
1079 1095 PyObject *phaseroots = NULL;
1080 1096 PyObject *rev = NULL;
1081 1097 PyObject *p1 = NULL;
1082 1098 PyObject *p2 = NULL;
1083 1099 PyObject *phaseset = NULL;
1084 1100 PyObject *phasessetlist = NULL;
1085 1101 Py_ssize_t addlen = self->added ? PyList_GET_SIZE(self->added) : 0;
1086 1102 Py_ssize_t len = index_length(self) - 1;
1087 1103 Py_ssize_t numphase = 0;
1088 1104 Py_ssize_t minrevallphases = 0;
1089 1105 Py_ssize_t minrevphase = 0;
1090 1106 Py_ssize_t i = 0;
1091 1107 int parent_1, parent_2;
1092 1108 char *phases = NULL;
1093 1109 const char *data;
1094 1110 long phase;
1095 1111
1096 1112 if (!PyArg_ParseTuple(args, "O", &roots))
1097 1113 goto release_none;
1098 1114 if (roots == NULL || !PyList_Check(roots))
1099 1115 goto release_none;
1100 1116
1101 1117 phases = calloc(len, 1); /* phase per rev: {0: public, 1: draft, 2: secret} */
1102 1118 if (phases == NULL)
1103 1119 goto release_none;
1104 1120 /* Put the phase information of all the roots in phases */
1105 1121 numphase = PyList_GET_SIZE(roots)+1;
1106 1122 minrevallphases = len + 1;
1107 1123 phasessetlist = PyList_New(numphase);
1108 1124 if (phasessetlist == NULL)
1109 1125 goto release_none;
1110 1126
1111 1127 PyList_SET_ITEM(phasessetlist, 0, Py_None);
1112 1128 Py_INCREF(Py_None);
1113 1129
1114 1130 for (i = 0; i < numphase-1; i++) {
1115 1131 phaseroots = PyList_GET_ITEM(roots, i);
1116 1132 phaseset = PySet_New(NULL);
1117 1133 if (phaseset == NULL)
1118 1134 goto release_phasesetlist;
1119 1135 PyList_SET_ITEM(phasessetlist, i+1, phaseset);
1120 1136 if (!PyList_Check(phaseroots))
1121 1137 goto release_phasesetlist;
1122 1138 minrevphase = add_roots_get_min(self, phaseroots, i+1, phases);
1123 1139 if (minrevphase == -2) /* Error from add_roots_get_min */
1124 1140 goto release_phasesetlist;
1125 1141 minrevallphases = MIN(minrevallphases, minrevphase);
1126 1142 }
1127 1143 /* Propagate the phase information from the roots to the revs */
1128 1144 if (minrevallphases != -1) {
1129 1145 for (i = minrevallphases; i < self->raw_length; i++) {
1130 1146 data = index_deref(self, i);
1131 1147 set_phase_from_parents(phases, getbe32(data+24), getbe32(data+28), i);
1132 1148 }
1133 1149 for (i = 0; i < addlen; i++) {
1134 1150 rev = PyList_GET_ITEM(self->added, i);
1135 1151 p1 = PyTuple_GET_ITEM(rev, 5);
1136 1152 p2 = PyTuple_GET_ITEM(rev, 6);
1137 1153 if (!PyInt_Check(p1) || !PyInt_Check(p2)) {
1138 1154 PyErr_SetString(PyExc_TypeError, "revlog parents are invalid");
1139 1155 goto release_phasesetlist;
1140 1156 }
1141 1157 parent_1 = (int)PyInt_AS_LONG(p1);
1142 1158 parent_2 = (int)PyInt_AS_LONG(p2);
1143 1159 set_phase_from_parents(phases, parent_1, parent_2, i+self->raw_length);
1144 1160 }
1145 1161 }
1146 1162 /* Transform phase list to a python list */
1147 1163 phaseslist = PyList_New(len);
1148 1164 if (phaseslist == NULL)
1149 1165 goto release_phasesetlist;
1150 1166 for (i = 0; i < len; i++) {
1151 1167 phase = phases[i];
1152 1168 /* We only store the sets of phase for non public phase, the public phase
1153 1169 * is computed as a difference */
1154 1170 if (phase != 0) {
1155 1171 phaseset = PyList_GET_ITEM(phasessetlist, phase);
1156 1172 PySet_Add(phaseset, PyInt_FromLong(i));
1157 1173 }
1158 1174 PyList_SET_ITEM(phaseslist, i, PyInt_FromLong(phase));
1159 1175 }
1160 1176 ret = PyList_New(2);
1161 1177 if (ret == NULL)
1162 1178 goto release_phaseslist;
1163 1179
1164 1180 PyList_SET_ITEM(ret, 0, phaseslist);
1165 1181 PyList_SET_ITEM(ret, 1, phasessetlist);
1166 1182 /* We don't release phaseslist and phasessetlist as we return them to
1167 1183 * python */
1168 1184 goto release_phases;
1169 1185
1170 1186 release_phaseslist:
1171 1187 Py_XDECREF(phaseslist);
1172 1188 release_phasesetlist:
1173 1189 Py_XDECREF(phasessetlist);
1174 1190 release_phases:
1175 1191 free(phases);
1176 1192 release_none:
1177 1193 return ret;
1178 1194 }
1179 1195
1180 static inline void index_get_parents(indexObject *self, Py_ssize_t rev,
1181 int *ps)
1182 {
1183 if (rev >= self->length - 1) {
1184 PyObject *tuple = PyList_GET_ITEM(self->added,
1185 rev - self->length + 1);
1186 ps[0] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 5));
1187 ps[1] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 6));
1188 } else {
1189 const char *data = index_deref(self, rev);
1190 ps[0] = getbe32(data + 24);
1191 ps[1] = getbe32(data + 28);
1192 }
1193 }
1194
1195 1196 static PyObject *index_headrevs(indexObject *self, PyObject *args)
1196 1197 {
1197 1198 Py_ssize_t i, j, len;
1198 1199 char *nothead = NULL;
1199 1200 PyObject *heads = NULL;
1200 1201 PyObject *filter = NULL;
1201 1202 PyObject *filteredrevs = Py_None;
1202 1203
1203 1204 if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) {
1204 1205 return NULL;
1205 1206 }
1206 1207
1207 1208 if (self->headrevs && filteredrevs == self->filteredrevs)
1208 1209 return list_copy(self->headrevs);
1209 1210
1210 1211 Py_DECREF(self->filteredrevs);
1211 1212 self->filteredrevs = filteredrevs;
1212 1213 Py_INCREF(filteredrevs);
1213 1214
1214 1215 if (filteredrevs != Py_None) {
1215 1216 filter = PyObject_GetAttrString(filteredrevs, "__contains__");
1216 1217 if (!filter) {
1217 1218 PyErr_SetString(PyExc_TypeError,
1218 1219 "filteredrevs has no attribute __contains__");
1219 1220 goto bail;
1220 1221 }
1221 1222 }
1222 1223
1223 1224 len = index_length(self) - 1;
1224 1225 heads = PyList_New(0);
1225 1226 if (heads == NULL)
1226 1227 goto bail;
1227 1228 if (len == 0) {
1228 1229 PyObject *nullid = PyInt_FromLong(-1);
1229 1230 if (nullid == NULL || PyList_Append(heads, nullid) == -1) {
1230 1231 Py_XDECREF(nullid);
1231 1232 goto bail;
1232 1233 }
1233 1234 goto done;
1234 1235 }
1235 1236
1236 1237 nothead = calloc(len, 1);
1237 1238 if (nothead == NULL)
1238 1239 goto bail;
1239 1240
1240 1241 for (i = 0; i < len; i++) {
1241 1242 int isfiltered;
1242 1243 int parents[2];
1243 1244
1244 1245 isfiltered = check_filter(filter, i);
1245 1246 if (isfiltered == -1) {
1246 1247 PyErr_SetString(PyExc_TypeError,
1247 1248 "unable to check filter");
1248 1249 goto bail;
1249 1250 }
1250 1251
1251 1252 if (isfiltered) {
1252 1253 nothead[i] = 1;
1253 1254 continue;
1254 1255 }
1255 1256
1256 1257 index_get_parents(self, i, parents);
1257 1258 for (j = 0; j < 2; j++) {
1258 1259 if (parents[j] >= 0)
1259 1260 nothead[parents[j]] = 1;
1260 1261 }
1261 1262 }
1262 1263
1263 1264 for (i = 0; i < len; i++) {
1264 1265 PyObject *head;
1265 1266
1266 1267 if (nothead[i])
1267 1268 continue;
1268 1269 head = PyInt_FromSsize_t(i);
1269 1270 if (head == NULL || PyList_Append(heads, head) == -1) {
1270 1271 Py_XDECREF(head);
1271 1272 goto bail;
1272 1273 }
1273 1274 }
1274 1275
1275 1276 done:
1276 1277 self->headrevs = heads;
1277 1278 Py_XDECREF(filter);
1278 1279 free(nothead);
1279 1280 return list_copy(self->headrevs);
1280 1281 bail:
1281 1282 Py_XDECREF(filter);
1282 1283 Py_XDECREF(heads);
1283 1284 free(nothead);
1284 1285 return NULL;
1285 1286 }
1286 1287
1287 1288 static inline int nt_level(const char *node, Py_ssize_t level)
1288 1289 {
1289 1290 int v = node[level>>1];
1290 1291 if (!(level & 1))
1291 1292 v >>= 4;
1292 1293 return v & 0xf;
1293 1294 }
1294 1295
1295 1296 /*
1296 1297 * Return values:
1297 1298 *
1298 1299 * -4: match is ambiguous (multiple candidates)
1299 1300 * -2: not found
1300 1301 * rest: valid rev
1301 1302 */
1302 1303 static int nt_find(indexObject *self, const char *node, Py_ssize_t nodelen,
1303 1304 int hex)
1304 1305 {
1305 1306 int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level;
1306 1307 int level, maxlevel, off;
1307 1308
1308 1309 if (nodelen == 20 && node[0] == '\0' && memcmp(node, nullid, 20) == 0)
1309 1310 return -1;
1310 1311
1311 1312 if (self->nt == NULL)
1312 1313 return -2;
1313 1314
1314 1315 if (hex)
1315 1316 maxlevel = nodelen > 40 ? 40 : (int)nodelen;
1316 1317 else
1317 1318 maxlevel = nodelen > 20 ? 40 : ((int)nodelen * 2);
1318 1319
1319 1320 for (level = off = 0; level < maxlevel; level++) {
1320 1321 int k = getnybble(node, level);
1321 1322 nodetree *n = &self->nt[off];
1322 1323 int v = n->children[k];
1323 1324
1324 1325 if (v < 0) {
1325 1326 const char *n;
1326 1327 Py_ssize_t i;
1327 1328
1328 1329 v = -(v + 1);
1329 1330 n = index_node(self, v);
1330 1331 if (n == NULL)
1331 1332 return -2;
1332 1333 for (i = level; i < maxlevel; i++)
1333 1334 if (getnybble(node, i) != nt_level(n, i))
1334 1335 return -2;
1335 1336 return v;
1336 1337 }
1337 1338 if (v == 0)
1338 1339 return -2;
1339 1340 off = v;
1340 1341 }
1341 1342 /* multiple matches against an ambiguous prefix */
1342 1343 return -4;
1343 1344 }
1344 1345
1345 1346 static int nt_new(indexObject *self)
1346 1347 {
1347 1348 if (self->ntlength == self->ntcapacity) {
1348 1349 if (self->ntcapacity >= INT_MAX / (sizeof(nodetree) * 2)) {
1349 1350 PyErr_SetString(PyExc_MemoryError,
1350 1351 "overflow in nt_new");
1351 1352 return -1;
1352 1353 }
1353 1354 self->ntcapacity *= 2;
1354 1355 self->nt = realloc(self->nt,
1355 1356 self->ntcapacity * sizeof(nodetree));
1356 1357 if (self->nt == NULL) {
1357 1358 PyErr_SetString(PyExc_MemoryError, "out of memory");
1358 1359 return -1;
1359 1360 }
1360 1361 memset(&self->nt[self->ntlength], 0,
1361 1362 sizeof(nodetree) * (self->ntcapacity - self->ntlength));
1362 1363 }
1363 1364 return self->ntlength++;
1364 1365 }
1365 1366
1366 1367 static int nt_insert(indexObject *self, const char *node, int rev)
1367 1368 {
1368 1369 int level = 0;
1369 1370 int off = 0;
1370 1371
1371 1372 while (level < 40) {
1372 1373 int k = nt_level(node, level);
1373 1374 nodetree *n;
1374 1375 int v;
1375 1376
1376 1377 n = &self->nt[off];
1377 1378 v = n->children[k];
1378 1379
1379 1380 if (v == 0) {
1380 1381 n->children[k] = -rev - 1;
1381 1382 return 0;
1382 1383 }
1383 1384 if (v < 0) {
1384 1385 const char *oldnode = index_node(self, -(v + 1));
1385 1386 int noff;
1386 1387
1387 1388 if (!oldnode || !memcmp(oldnode, node, 20)) {
1388 1389 n->children[k] = -rev - 1;
1389 1390 return 0;
1390 1391 }
1391 1392 noff = nt_new(self);
1392 1393 if (noff == -1)
1393 1394 return -1;
1394 1395 /* self->nt may have been changed by realloc */
1395 1396 self->nt[off].children[k] = noff;
1396 1397 off = noff;
1397 1398 n = &self->nt[off];
1398 1399 n->children[nt_level(oldnode, ++level)] = v;
1399 1400 if (level > self->ntdepth)
1400 1401 self->ntdepth = level;
1401 1402 self->ntsplits += 1;
1402 1403 } else {
1403 1404 level += 1;
1404 1405 off = v;
1405 1406 }
1406 1407 }
1407 1408
1408 1409 return -1;
1409 1410 }
1410 1411
1411 1412 static int nt_init(indexObject *self)
1412 1413 {
1413 1414 if (self->nt == NULL) {
1414 1415 if (self->raw_length > INT_MAX / sizeof(nodetree)) {
1415 1416 PyErr_SetString(PyExc_ValueError, "overflow in nt_init");
1416 1417 return -1;
1417 1418 }
1418 1419 self->ntcapacity = self->raw_length < 4
1419 1420 ? 4 : (int)self->raw_length / 2;
1420 1421
1421 1422 self->nt = calloc(self->ntcapacity, sizeof(nodetree));
1422 1423 if (self->nt == NULL) {
1423 1424 PyErr_NoMemory();
1424 1425 return -1;
1425 1426 }
1426 1427 self->ntlength = 1;
1427 1428 self->ntrev = (int)index_length(self) - 1;
1428 1429 self->ntlookups = 1;
1429 1430 self->ntmisses = 0;
1430 1431 if (nt_insert(self, nullid, INT_MAX) == -1)
1431 1432 return -1;
1432 1433 }
1433 1434 return 0;
1434 1435 }
1435 1436
1436 1437 /*
1437 1438 * Return values:
1438 1439 *
1439 1440 * -3: error (exception set)
1440 1441 * -2: not found (no exception set)
1441 1442 * rest: valid rev
1442 1443 */
1443 1444 static int index_find_node(indexObject *self,
1444 1445 const char *node, Py_ssize_t nodelen)
1445 1446 {
1446 1447 int rev;
1447 1448
1448 1449 self->ntlookups++;
1449 1450 rev = nt_find(self, node, nodelen, 0);
1450 1451 if (rev >= -1)
1451 1452 return rev;
1452 1453
1453 1454 if (nt_init(self) == -1)
1454 1455 return -3;
1455 1456
1456 1457 /*
1457 1458 * For the first handful of lookups, we scan the entire index,
1458 1459 * and cache only the matching nodes. This optimizes for cases
1459 1460 * like "hg tip", where only a few nodes are accessed.
1460 1461 *
1461 1462 * After that, we cache every node we visit, using a single
1462 1463 * scan amortized over multiple lookups. This gives the best
1463 1464 * bulk performance, e.g. for "hg log".
1464 1465 */
1465 1466 if (self->ntmisses++ < 4) {
1466 1467 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1467 1468 const char *n = index_node(self, rev);
1468 1469 if (n == NULL)
1469 1470 return -2;
1470 1471 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1471 1472 if (nt_insert(self, n, rev) == -1)
1472 1473 return -3;
1473 1474 break;
1474 1475 }
1475 1476 }
1476 1477 } else {
1477 1478 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1478 1479 const char *n = index_node(self, rev);
1479 1480 if (n == NULL) {
1480 1481 self->ntrev = rev + 1;
1481 1482 return -2;
1482 1483 }
1483 1484 if (nt_insert(self, n, rev) == -1) {
1484 1485 self->ntrev = rev + 1;
1485 1486 return -3;
1486 1487 }
1487 1488 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1488 1489 break;
1489 1490 }
1490 1491 }
1491 1492 self->ntrev = rev;
1492 1493 }
1493 1494
1494 1495 if (rev >= 0)
1495 1496 return rev;
1496 1497 return -2;
1497 1498 }
1498 1499
1499 1500 static PyObject *raise_revlog_error(void)
1500 1501 {
1501 1502 static PyObject *errclass;
1502 1503 PyObject *mod = NULL, *errobj;
1503 1504
1504 1505 if (errclass == NULL) {
1505 1506 PyObject *dict;
1506 1507
1507 1508 mod = PyImport_ImportModule("mercurial.error");
1508 1509 if (mod == NULL)
1509 1510 goto classfail;
1510 1511
1511 1512 dict = PyModule_GetDict(mod);
1512 1513 if (dict == NULL)
1513 1514 goto classfail;
1514 1515
1515 1516 errclass = PyDict_GetItemString(dict, "RevlogError");
1516 1517 if (errclass == NULL) {
1517 1518 PyErr_SetString(PyExc_SystemError,
1518 1519 "could not find RevlogError");
1519 1520 goto classfail;
1520 1521 }
1521 1522 Py_INCREF(errclass);
1522 1523 Py_DECREF(mod);
1523 1524 }
1524 1525
1525 1526 errobj = PyObject_CallFunction(errclass, NULL);
1526 1527 if (errobj == NULL)
1527 1528 return NULL;
1528 1529 PyErr_SetObject(errclass, errobj);
1529 1530 return errobj;
1530 1531
1531 1532 classfail:
1532 1533 Py_XDECREF(mod);
1533 1534 return NULL;
1534 1535 }
1535 1536
1536 1537 static PyObject *index_getitem(indexObject *self, PyObject *value)
1537 1538 {
1538 1539 char *node;
1539 1540 Py_ssize_t nodelen;
1540 1541 int rev;
1541 1542
1542 1543 if (PyInt_Check(value))
1543 1544 return index_get(self, PyInt_AS_LONG(value));
1544 1545
1545 1546 if (node_check(value, &node, &nodelen) == -1)
1546 1547 return NULL;
1547 1548 rev = index_find_node(self, node, nodelen);
1548 1549 if (rev >= -1)
1549 1550 return PyInt_FromLong(rev);
1550 1551 if (rev == -2)
1551 1552 raise_revlog_error();
1552 1553 return NULL;
1553 1554 }
1554 1555
1555 1556 static int nt_partialmatch(indexObject *self, const char *node,
1556 1557 Py_ssize_t nodelen)
1557 1558 {
1558 1559 int rev;
1559 1560
1560 1561 if (nt_init(self) == -1)
1561 1562 return -3;
1562 1563
1563 1564 if (self->ntrev > 0) {
1564 1565 /* ensure that the radix tree is fully populated */
1565 1566 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1566 1567 const char *n = index_node(self, rev);
1567 1568 if (n == NULL)
1568 1569 return -2;
1569 1570 if (nt_insert(self, n, rev) == -1)
1570 1571 return -3;
1571 1572 }
1572 1573 self->ntrev = rev;
1573 1574 }
1574 1575
1575 1576 return nt_find(self, node, nodelen, 1);
1576 1577 }
1577 1578
1578 1579 static PyObject *index_partialmatch(indexObject *self, PyObject *args)
1579 1580 {
1580 1581 const char *fullnode;
1581 1582 int nodelen;
1582 1583 char *node;
1583 1584 int rev, i;
1584 1585
1585 1586 if (!PyArg_ParseTuple(args, "s#", &node, &nodelen))
1586 1587 return NULL;
1587 1588
1588 1589 if (nodelen < 4) {
1589 1590 PyErr_SetString(PyExc_ValueError, "key too short");
1590 1591 return NULL;
1591 1592 }
1592 1593
1593 1594 if (nodelen > 40) {
1594 1595 PyErr_SetString(PyExc_ValueError, "key too long");
1595 1596 return NULL;
1596 1597 }
1597 1598
1598 1599 for (i = 0; i < nodelen; i++)
1599 1600 hexdigit(node, i);
1600 1601 if (PyErr_Occurred()) {
1601 1602 /* input contains non-hex characters */
1602 1603 PyErr_Clear();
1603 1604 Py_RETURN_NONE;
1604 1605 }
1605 1606
1606 1607 rev = nt_partialmatch(self, node, nodelen);
1607 1608
1608 1609 switch (rev) {
1609 1610 case -4:
1610 1611 raise_revlog_error();
1611 1612 case -3:
1612 1613 return NULL;
1613 1614 case -2:
1614 1615 Py_RETURN_NONE;
1615 1616 case -1:
1616 1617 return PyString_FromStringAndSize(nullid, 20);
1617 1618 }
1618 1619
1619 1620 fullnode = index_node(self, rev);
1620 1621 if (fullnode == NULL) {
1621 1622 PyErr_Format(PyExc_IndexError,
1622 1623 "could not access rev %d", rev);
1623 1624 return NULL;
1624 1625 }
1625 1626 return PyString_FromStringAndSize(fullnode, 20);
1626 1627 }
1627 1628
1628 1629 static PyObject *index_m_get(indexObject *self, PyObject *args)
1629 1630 {
1630 1631 Py_ssize_t nodelen;
1631 1632 PyObject *val;
1632 1633 char *node;
1633 1634 int rev;
1634 1635
1635 1636 if (!PyArg_ParseTuple(args, "O", &val))
1636 1637 return NULL;
1637 1638 if (node_check(val, &node, &nodelen) == -1)
1638 1639 return NULL;
1639 1640 rev = index_find_node(self, node, nodelen);
1640 1641 if (rev == -3)
1641 1642 return NULL;
1642 1643 if (rev == -2)
1643 1644 Py_RETURN_NONE;
1644 1645 return PyInt_FromLong(rev);
1645 1646 }
1646 1647
1647 1648 static int index_contains(indexObject *self, PyObject *value)
1648 1649 {
1649 1650 char *node;
1650 1651 Py_ssize_t nodelen;
1651 1652
1652 1653 if (PyInt_Check(value)) {
1653 1654 long rev = PyInt_AS_LONG(value);
1654 1655 return rev >= -1 && rev < index_length(self);
1655 1656 }
1656 1657
1657 1658 if (node_check(value, &node, &nodelen) == -1)
1658 1659 return -1;
1659 1660
1660 1661 switch (index_find_node(self, node, nodelen)) {
1661 1662 case -3:
1662 1663 return -1;
1663 1664 case -2:
1664 1665 return 0;
1665 1666 default:
1666 1667 return 1;
1667 1668 }
1668 1669 }
1669 1670
1670 1671 typedef uint64_t bitmask;
1671 1672
1672 1673 /*
1673 1674 * Given a disjoint set of revs, return all candidates for the
1674 1675 * greatest common ancestor. In revset notation, this is the set
1675 1676 * "heads(::a and ::b and ...)"
1676 1677 */
1677 1678 static PyObject *find_gca_candidates(indexObject *self, const int *revs,
1678 1679 int revcount)
1679 1680 {
1680 1681 const bitmask allseen = (1ull << revcount) - 1;
1681 1682 const bitmask poison = 1ull << revcount;
1682 1683 PyObject *gca = PyList_New(0);
1683 1684 int i, v, interesting;
1684 1685 int maxrev = -1;
1685 1686 bitmask sp;
1686 1687 bitmask *seen;
1687 1688
1688 1689 if (gca == NULL)
1689 1690 return PyErr_NoMemory();
1690 1691
1691 1692 for (i = 0; i < revcount; i++) {
1692 1693 if (revs[i] > maxrev)
1693 1694 maxrev = revs[i];
1694 1695 }
1695 1696
1696 1697 seen = calloc(sizeof(*seen), maxrev + 1);
1697 1698 if (seen == NULL) {
1698 1699 Py_DECREF(gca);
1699 1700 return PyErr_NoMemory();
1700 1701 }
1701 1702
1702 1703 for (i = 0; i < revcount; i++)
1703 1704 seen[revs[i]] = 1ull << i;
1704 1705
1705 1706 interesting = revcount;
1706 1707
1707 1708 for (v = maxrev; v >= 0 && interesting; v--) {
1708 1709 bitmask sv = seen[v];
1709 1710 int parents[2];
1710 1711
1711 1712 if (!sv)
1712 1713 continue;
1713 1714
1714 1715 if (sv < poison) {
1715 1716 interesting -= 1;
1716 1717 if (sv == allseen) {
1717 1718 PyObject *obj = PyInt_FromLong(v);
1718 1719 if (obj == NULL)
1719 1720 goto bail;
1720 1721 if (PyList_Append(gca, obj) == -1) {
1721 1722 Py_DECREF(obj);
1722 1723 goto bail;
1723 1724 }
1724 1725 sv |= poison;
1725 1726 for (i = 0; i < revcount; i++) {
1726 1727 if (revs[i] == v)
1727 1728 goto done;
1728 1729 }
1729 1730 }
1730 1731 }
1731 1732 index_get_parents(self, v, parents);
1732 1733
1733 1734 for (i = 0; i < 2; i++) {
1734 1735 int p = parents[i];
1735 1736 if (p == -1)
1736 1737 continue;
1737 1738 sp = seen[p];
1738 1739 if (sv < poison) {
1739 1740 if (sp == 0) {
1740 1741 seen[p] = sv;
1741 1742 interesting++;
1742 1743 }
1743 1744 else if (sp != sv)
1744 1745 seen[p] |= sv;
1745 1746 } else {
1746 1747 if (sp && sp < poison)
1747 1748 interesting--;
1748 1749 seen[p] = sv;
1749 1750 }
1750 1751 }
1751 1752 }
1752 1753
1753 1754 done:
1754 1755 free(seen);
1755 1756 return gca;
1756 1757 bail:
1757 1758 free(seen);
1758 1759 Py_XDECREF(gca);
1759 1760 return NULL;
1760 1761 }
1761 1762
1762 1763 /*
1763 1764 * Given a disjoint set of revs, return the subset with the longest
1764 1765 * path to the root.
1765 1766 */
1766 1767 static PyObject *find_deepest(indexObject *self, PyObject *revs)
1767 1768 {
1768 1769 const Py_ssize_t revcount = PyList_GET_SIZE(revs);
1769 1770 static const Py_ssize_t capacity = 24;
1770 1771 int *depth, *interesting = NULL;
1771 1772 int i, j, v, ninteresting;
1772 1773 PyObject *dict = NULL, *keys = NULL;
1773 1774 long *seen = NULL;
1774 1775 int maxrev = -1;
1775 1776 long final;
1776 1777
1777 1778 if (revcount > capacity) {
1778 1779 PyErr_Format(PyExc_OverflowError,
1779 1780 "bitset size (%ld) > capacity (%ld)",
1780 1781 (long)revcount, (long)capacity);
1781 1782 return NULL;
1782 1783 }
1783 1784
1784 1785 for (i = 0; i < revcount; i++) {
1785 1786 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1786 1787 if (n > maxrev)
1787 1788 maxrev = n;
1788 1789 }
1789 1790
1790 1791 depth = calloc(sizeof(*depth), maxrev + 1);
1791 1792 if (depth == NULL)
1792 1793 return PyErr_NoMemory();
1793 1794
1794 1795 seen = calloc(sizeof(*seen), maxrev + 1);
1795 1796 if (seen == NULL) {
1796 1797 PyErr_NoMemory();
1797 1798 goto bail;
1798 1799 }
1799 1800
1800 1801 interesting = calloc(sizeof(*interesting), 2 << revcount);
1801 1802 if (interesting == NULL) {
1802 1803 PyErr_NoMemory();
1803 1804 goto bail;
1804 1805 }
1805 1806
1806 1807 if (PyList_Sort(revs) == -1)
1807 1808 goto bail;
1808 1809
1809 1810 for (i = 0; i < revcount; i++) {
1810 1811 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1811 1812 long b = 1l << i;
1812 1813 depth[n] = 1;
1813 1814 seen[n] = b;
1814 1815 interesting[b] = 1;
1815 1816 }
1816 1817
1817 1818 ninteresting = (int)revcount;
1818 1819
1819 1820 for (v = maxrev; v >= 0 && ninteresting > 1; v--) {
1820 1821 int dv = depth[v];
1821 1822 int parents[2];
1822 1823 long sv;
1823 1824
1824 1825 if (dv == 0)
1825 1826 continue;
1826 1827
1827 1828 sv = seen[v];
1828 1829 index_get_parents(self, v, parents);
1829 1830
1830 1831 for (i = 0; i < 2; i++) {
1831 1832 int p = parents[i];
1832 1833 long nsp, sp;
1833 1834 int dp;
1834 1835
1835 1836 if (p == -1)
1836 1837 continue;
1837 1838
1838 1839 dp = depth[p];
1839 1840 nsp = sp = seen[p];
1840 1841 if (dp <= dv) {
1841 1842 depth[p] = dv + 1;
1842 1843 if (sp != sv) {
1843 1844 interesting[sv] += 1;
1844 1845 nsp = seen[p] = sv;
1845 1846 if (sp) {
1846 1847 interesting[sp] -= 1;
1847 1848 if (interesting[sp] == 0)
1848 1849 ninteresting -= 1;
1849 1850 }
1850 1851 }
1851 1852 }
1852 1853 else if (dv == dp - 1) {
1853 1854 nsp = sp | sv;
1854 1855 if (nsp == sp)
1855 1856 continue;
1856 1857 seen[p] = nsp;
1857 1858 interesting[sp] -= 1;
1858 1859 if (interesting[sp] == 0 && interesting[nsp] > 0)
1859 1860 ninteresting -= 1;
1860 1861 interesting[nsp] += 1;
1861 1862 }
1862 1863 }
1863 1864 interesting[sv] -= 1;
1864 1865 if (interesting[sv] == 0)
1865 1866 ninteresting -= 1;
1866 1867 }
1867 1868
1868 1869 final = 0;
1869 1870 j = ninteresting;
1870 1871 for (i = 0; i < (int)(2 << revcount) && j > 0; i++) {
1871 1872 if (interesting[i] == 0)
1872 1873 continue;
1873 1874 final |= i;
1874 1875 j -= 1;
1875 1876 }
1876 1877 if (final == 0) {
1877 1878 keys = PyList_New(0);
1878 1879 goto bail;
1879 1880 }
1880 1881
1881 1882 dict = PyDict_New();
1882 1883 if (dict == NULL)
1883 1884 goto bail;
1884 1885
1885 1886 for (i = 0; i < revcount; i++) {
1886 1887 PyObject *key;
1887 1888
1888 1889 if ((final & (1 << i)) == 0)
1889 1890 continue;
1890 1891
1891 1892 key = PyList_GET_ITEM(revs, i);
1892 1893 Py_INCREF(key);
1893 1894 Py_INCREF(Py_None);
1894 1895 if (PyDict_SetItem(dict, key, Py_None) == -1) {
1895 1896 Py_DECREF(key);
1896 1897 Py_DECREF(Py_None);
1897 1898 goto bail;
1898 1899 }
1899 1900 }
1900 1901
1901 1902 keys = PyDict_Keys(dict);
1902 1903
1903 1904 bail:
1904 1905 free(depth);
1905 1906 free(seen);
1906 1907 free(interesting);
1907 1908 Py_XDECREF(dict);
1908 1909
1909 1910 return keys;
1910 1911 }
1911 1912
1912 1913 /*
1913 1914 * Given a (possibly overlapping) set of revs, return all the
1914 1915 * common ancestors heads: heads(::args[0] and ::a[1] and ...)
1915 1916 */
1916 1917 static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args)
1917 1918 {
1918 1919 PyObject *ret = NULL;
1919 1920 Py_ssize_t argcount, i, len;
1920 1921 bitmask repeat = 0;
1921 1922 int revcount = 0;
1922 1923 int *revs;
1923 1924
1924 1925 argcount = PySequence_Length(args);
1925 1926 revs = malloc(argcount * sizeof(*revs));
1926 1927 if (argcount > 0 && revs == NULL)
1927 1928 return PyErr_NoMemory();
1928 1929 len = index_length(self) - 1;
1929 1930
1930 1931 for (i = 0; i < argcount; i++) {
1931 1932 static const int capacity = 24;
1932 1933 PyObject *obj = PySequence_GetItem(args, i);
1933 1934 bitmask x;
1934 1935 long val;
1935 1936
1936 1937 if (!PyInt_Check(obj)) {
1937 1938 PyErr_SetString(PyExc_TypeError,
1938 1939 "arguments must all be ints");
1939 1940 Py_DECREF(obj);
1940 1941 goto bail;
1941 1942 }
1942 1943 val = PyInt_AsLong(obj);
1943 1944 Py_DECREF(obj);
1944 1945 if (val == -1) {
1945 1946 ret = PyList_New(0);
1946 1947 goto done;
1947 1948 }
1948 1949 if (val < 0 || val >= len) {
1949 1950 PyErr_SetString(PyExc_IndexError,
1950 1951 "index out of range");
1951 1952 goto bail;
1952 1953 }
1953 1954 /* this cheesy bloom filter lets us avoid some more
1954 1955 * expensive duplicate checks in the common set-is-disjoint
1955 1956 * case */
1956 1957 x = 1ull << (val & 0x3f);
1957 1958 if (repeat & x) {
1958 1959 int k;
1959 1960 for (k = 0; k < revcount; k++) {
1960 1961 if (val == revs[k])
1961 1962 goto duplicate;
1962 1963 }
1963 1964 }
1964 1965 else repeat |= x;
1965 1966 if (revcount >= capacity) {
1966 1967 PyErr_Format(PyExc_OverflowError,
1967 1968 "bitset size (%d) > capacity (%d)",
1968 1969 revcount, capacity);
1969 1970 goto bail;
1970 1971 }
1971 1972 revs[revcount++] = (int)val;
1972 1973 duplicate:;
1973 1974 }
1974 1975
1975 1976 if (revcount == 0) {
1976 1977 ret = PyList_New(0);
1977 1978 goto done;
1978 1979 }
1979 1980 if (revcount == 1) {
1980 1981 PyObject *obj;
1981 1982 ret = PyList_New(1);
1982 1983 if (ret == NULL)
1983 1984 goto bail;
1984 1985 obj = PyInt_FromLong(revs[0]);
1985 1986 if (obj == NULL)
1986 1987 goto bail;
1987 1988 PyList_SET_ITEM(ret, 0, obj);
1988 1989 goto done;
1989 1990 }
1990 1991
1991 1992 ret = find_gca_candidates(self, revs, revcount);
1992 1993 if (ret == NULL)
1993 1994 goto bail;
1994 1995
1995 1996 done:
1996 1997 free(revs);
1997 1998 return ret;
1998 1999
1999 2000 bail:
2000 2001 free(revs);
2001 2002 Py_XDECREF(ret);
2002 2003 return NULL;
2003 2004 }
2004 2005
2005 2006 /*
2006 2007 * Given a (possibly overlapping) set of revs, return the greatest
2007 2008 * common ancestors: those with the longest path to the root.
2008 2009 */
2009 2010 static PyObject *index_ancestors(indexObject *self, PyObject *args)
2010 2011 {
2011 2012 PyObject *gca = index_commonancestorsheads(self, args);
2012 2013 if (gca == NULL)
2013 2014 return NULL;
2014 2015
2015 2016 if (PyList_GET_SIZE(gca) <= 1) {
2016 2017 Py_INCREF(gca);
2017 2018 return gca;
2018 2019 }
2019 2020
2020 2021 return find_deepest(self, gca);
2021 2022 }
2022 2023
2023 2024 /*
2024 2025 * Invalidate any trie entries introduced by added revs.
2025 2026 */
2026 2027 static void nt_invalidate_added(indexObject *self, Py_ssize_t start)
2027 2028 {
2028 2029 Py_ssize_t i, len = PyList_GET_SIZE(self->added);
2029 2030
2030 2031 for (i = start; i < len; i++) {
2031 2032 PyObject *tuple = PyList_GET_ITEM(self->added, i);
2032 2033 PyObject *node = PyTuple_GET_ITEM(tuple, 7);
2033 2034
2034 2035 nt_insert(self, PyString_AS_STRING(node), -1);
2035 2036 }
2036 2037
2037 2038 if (start == 0)
2038 2039 Py_CLEAR(self->added);
2039 2040 }
2040 2041
2041 2042 /*
2042 2043 * Delete a numeric range of revs, which must be at the end of the
2043 2044 * range, but exclude the sentinel nullid entry.
2044 2045 */
2045 2046 static int index_slice_del(indexObject *self, PyObject *item)
2046 2047 {
2047 2048 Py_ssize_t start, stop, step, slicelength;
2048 2049 Py_ssize_t length = index_length(self);
2049 2050 int ret = 0;
2050 2051
2051 2052 if (PySlice_GetIndicesEx((PySliceObject*)item, length,
2052 2053 &start, &stop, &step, &slicelength) < 0)
2053 2054 return -1;
2054 2055
2055 2056 if (slicelength <= 0)
2056 2057 return 0;
2057 2058
2058 2059 if ((step < 0 && start < stop) || (step > 0 && start > stop))
2059 2060 stop = start;
2060 2061
2061 2062 if (step < 0) {
2062 2063 stop = start + 1;
2063 2064 start = stop + step*(slicelength - 1) - 1;
2064 2065 step = -step;
2065 2066 }
2066 2067
2067 2068 if (step != 1) {
2068 2069 PyErr_SetString(PyExc_ValueError,
2069 2070 "revlog index delete requires step size of 1");
2070 2071 return -1;
2071 2072 }
2072 2073
2073 2074 if (stop != length - 1) {
2074 2075 PyErr_SetString(PyExc_IndexError,
2075 2076 "revlog index deletion indices are invalid");
2076 2077 return -1;
2077 2078 }
2078 2079
2079 2080 if (start < self->length - 1) {
2080 2081 if (self->nt) {
2081 2082 Py_ssize_t i;
2082 2083
2083 2084 for (i = start + 1; i < self->length - 1; i++) {
2084 2085 const char *node = index_node(self, i);
2085 2086
2086 2087 if (node)
2087 2088 nt_insert(self, node, -1);
2088 2089 }
2089 2090 if (self->added)
2090 2091 nt_invalidate_added(self, 0);
2091 2092 if (self->ntrev > start)
2092 2093 self->ntrev = (int)start;
2093 2094 }
2094 2095 self->length = start + 1;
2095 2096 if (start < self->raw_length) {
2096 2097 if (self->cache) {
2097 2098 Py_ssize_t i;
2098 2099 for (i = start; i < self->raw_length; i++)
2099 2100 Py_CLEAR(self->cache[i]);
2100 2101 }
2101 2102 self->raw_length = start;
2102 2103 }
2103 2104 goto done;
2104 2105 }
2105 2106
2106 2107 if (self->nt) {
2107 2108 nt_invalidate_added(self, start - self->length + 1);
2108 2109 if (self->ntrev > start)
2109 2110 self->ntrev = (int)start;
2110 2111 }
2111 2112 if (self->added)
2112 2113 ret = PyList_SetSlice(self->added, start - self->length + 1,
2113 2114 PyList_GET_SIZE(self->added), NULL);
2114 2115 done:
2115 2116 Py_CLEAR(self->headrevs);
2116 2117 return ret;
2117 2118 }
2118 2119
2119 2120 /*
2120 2121 * Supported ops:
2121 2122 *
2122 2123 * slice deletion
2123 2124 * string assignment (extend node->rev mapping)
2124 2125 * string deletion (shrink node->rev mapping)
2125 2126 */
2126 2127 static int index_assign_subscript(indexObject *self, PyObject *item,
2127 2128 PyObject *value)
2128 2129 {
2129 2130 char *node;
2130 2131 Py_ssize_t nodelen;
2131 2132 long rev;
2132 2133
2133 2134 if (PySlice_Check(item) && value == NULL)
2134 2135 return index_slice_del(self, item);
2135 2136
2136 2137 if (node_check(item, &node, &nodelen) == -1)
2137 2138 return -1;
2138 2139
2139 2140 if (value == NULL)
2140 2141 return self->nt ? nt_insert(self, node, -1) : 0;
2141 2142 rev = PyInt_AsLong(value);
2142 2143 if (rev > INT_MAX || rev < 0) {
2143 2144 if (!PyErr_Occurred())
2144 2145 PyErr_SetString(PyExc_ValueError, "rev out of range");
2145 2146 return -1;
2146 2147 }
2147 2148
2148 2149 if (nt_init(self) == -1)
2149 2150 return -1;
2150 2151 return nt_insert(self, node, (int)rev);
2151 2152 }
2152 2153
2153 2154 /*
2154 2155 * Find all RevlogNG entries in an index that has inline data. Update
2155 2156 * the optional "offsets" table with those entries.
2156 2157 */
2157 2158 static Py_ssize_t inline_scan(indexObject *self, const char **offsets)
2158 2159 {
2159 2160 const char *data = PyString_AS_STRING(self->data);
2160 2161 Py_ssize_t pos = 0;
2161 2162 Py_ssize_t end = PyString_GET_SIZE(self->data);
2162 2163 long incr = v1_hdrsize;
2163 2164 Py_ssize_t len = 0;
2164 2165
2165 2166 while (pos + v1_hdrsize <= end && pos >= 0) {
2166 2167 uint32_t comp_len;
2167 2168 /* 3rd element of header is length of compressed inline data */
2168 2169 comp_len = getbe32(data + pos + 8);
2169 2170 incr = v1_hdrsize + comp_len;
2170 2171 if (offsets)
2171 2172 offsets[len] = data + pos;
2172 2173 len++;
2173 2174 pos += incr;
2174 2175 }
2175 2176
2176 2177 if (pos != end) {
2177 2178 if (!PyErr_Occurred())
2178 2179 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2179 2180 return -1;
2180 2181 }
2181 2182
2182 2183 return len;
2183 2184 }
2184 2185
2185 2186 static int index_init(indexObject *self, PyObject *args)
2186 2187 {
2187 2188 PyObject *data_obj, *inlined_obj;
2188 2189 Py_ssize_t size;
2189 2190
2190 2191 /* Initialize before argument-checking to avoid index_dealloc() crash. */
2191 2192 self->raw_length = 0;
2192 2193 self->added = NULL;
2193 2194 self->cache = NULL;
2194 2195 self->data = NULL;
2195 2196 self->headrevs = NULL;
2196 2197 self->filteredrevs = Py_None;
2197 2198 Py_INCREF(Py_None);
2198 2199 self->nt = NULL;
2199 2200 self->offsets = NULL;
2200 2201
2201 2202 if (!PyArg_ParseTuple(args, "OO", &data_obj, &inlined_obj))
2202 2203 return -1;
2203 2204 if (!PyString_Check(data_obj)) {
2204 2205 PyErr_SetString(PyExc_TypeError, "data is not a string");
2205 2206 return -1;
2206 2207 }
2207 2208 size = PyString_GET_SIZE(data_obj);
2208 2209
2209 2210 self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj);
2210 2211 self->data = data_obj;
2211 2212
2212 2213 self->ntlength = self->ntcapacity = 0;
2213 2214 self->ntdepth = self->ntsplits = 0;
2214 2215 self->ntlookups = self->ntmisses = 0;
2215 2216 self->ntrev = -1;
2216 2217 Py_INCREF(self->data);
2217 2218
2218 2219 if (self->inlined) {
2219 2220 Py_ssize_t len = inline_scan(self, NULL);
2220 2221 if (len == -1)
2221 2222 goto bail;
2222 2223 self->raw_length = len;
2223 2224 self->length = len + 1;
2224 2225 } else {
2225 2226 if (size % v1_hdrsize) {
2226 2227 PyErr_SetString(PyExc_ValueError, "corrupt index file");
2227 2228 goto bail;
2228 2229 }
2229 2230 self->raw_length = size / v1_hdrsize;
2230 2231 self->length = self->raw_length + 1;
2231 2232 }
2232 2233
2233 2234 return 0;
2234 2235 bail:
2235 2236 return -1;
2236 2237 }
2237 2238
2238 2239 static PyObject *index_nodemap(indexObject *self)
2239 2240 {
2240 2241 Py_INCREF(self);
2241 2242 return (PyObject *)self;
2242 2243 }
2243 2244
2244 2245 static void index_dealloc(indexObject *self)
2245 2246 {
2246 2247 _index_clearcaches(self);
2247 2248 Py_XDECREF(self->filteredrevs);
2248 2249 Py_XDECREF(self->data);
2249 2250 Py_XDECREF(self->added);
2250 2251 PyObject_Del(self);
2251 2252 }
2252 2253
2253 2254 static PySequenceMethods index_sequence_methods = {
2254 2255 (lenfunc)index_length, /* sq_length */
2255 2256 0, /* sq_concat */
2256 2257 0, /* sq_repeat */
2257 2258 (ssizeargfunc)index_get, /* sq_item */
2258 2259 0, /* sq_slice */
2259 2260 0, /* sq_ass_item */
2260 2261 0, /* sq_ass_slice */
2261 2262 (objobjproc)index_contains, /* sq_contains */
2262 2263 };
2263 2264
2264 2265 static PyMappingMethods index_mapping_methods = {
2265 2266 (lenfunc)index_length, /* mp_length */
2266 2267 (binaryfunc)index_getitem, /* mp_subscript */
2267 2268 (objobjargproc)index_assign_subscript, /* mp_ass_subscript */
2268 2269 };
2269 2270
2270 2271 static PyMethodDef index_methods[] = {
2271 2272 {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS,
2272 2273 "return the gca set of the given revs"},
2273 2274 {"commonancestorsheads", (PyCFunction)index_commonancestorsheads,
2274 2275 METH_VARARGS,
2275 2276 "return the heads of the common ancestors of the given revs"},
2276 2277 {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS,
2277 2278 "clear the index caches"},
2278 2279 {"get", (PyCFunction)index_m_get, METH_VARARGS,
2279 2280 "get an index entry"},
2280 2281 {"computephasesmapsets", (PyCFunction)compute_phases_map_sets,
2281 2282 METH_VARARGS, "compute phases"},
2282 2283 {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS,
2283 2284 "get head revisions"}, /* Can do filtering since 3.2 */
2284 2285 {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS,
2285 2286 "get filtered head revisions"}, /* Can always do filtering */
2286 2287 {"insert", (PyCFunction)index_insert, METH_VARARGS,
2287 2288 "insert an index entry"},
2288 2289 {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS,
2289 2290 "match a potentially ambiguous node ID"},
2290 2291 {"stats", (PyCFunction)index_stats, METH_NOARGS,
2291 2292 "stats for the index"},
2292 2293 {NULL} /* Sentinel */
2293 2294 };
2294 2295
2295 2296 static PyGetSetDef index_getset[] = {
2296 2297 {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL},
2297 2298 {NULL} /* Sentinel */
2298 2299 };
2299 2300
2300 2301 static PyTypeObject indexType = {
2301 2302 PyObject_HEAD_INIT(NULL)
2302 2303 0, /* ob_size */
2303 2304 "parsers.index", /* tp_name */
2304 2305 sizeof(indexObject), /* tp_basicsize */
2305 2306 0, /* tp_itemsize */
2306 2307 (destructor)index_dealloc, /* tp_dealloc */
2307 2308 0, /* tp_print */
2308 2309 0, /* tp_getattr */
2309 2310 0, /* tp_setattr */
2310 2311 0, /* tp_compare */
2311 2312 0, /* tp_repr */
2312 2313 0, /* tp_as_number */
2313 2314 &index_sequence_methods, /* tp_as_sequence */
2314 2315 &index_mapping_methods, /* tp_as_mapping */
2315 2316 0, /* tp_hash */
2316 2317 0, /* tp_call */
2317 2318 0, /* tp_str */
2318 2319 0, /* tp_getattro */
2319 2320 0, /* tp_setattro */
2320 2321 0, /* tp_as_buffer */
2321 2322 Py_TPFLAGS_DEFAULT, /* tp_flags */
2322 2323 "revlog index", /* tp_doc */
2323 2324 0, /* tp_traverse */
2324 2325 0, /* tp_clear */
2325 2326 0, /* tp_richcompare */
2326 2327 0, /* tp_weaklistoffset */
2327 2328 0, /* tp_iter */
2328 2329 0, /* tp_iternext */
2329 2330 index_methods, /* tp_methods */
2330 2331 0, /* tp_members */
2331 2332 index_getset, /* tp_getset */
2332 2333 0, /* tp_base */
2333 2334 0, /* tp_dict */
2334 2335 0, /* tp_descr_get */
2335 2336 0, /* tp_descr_set */
2336 2337 0, /* tp_dictoffset */
2337 2338 (initproc)index_init, /* tp_init */
2338 2339 0, /* tp_alloc */
2339 2340 };
2340 2341
2341 2342 /*
2342 2343 * returns a tuple of the form (index, index, cache) with elements as
2343 2344 * follows:
2344 2345 *
2345 2346 * index: an index object that lazily parses RevlogNG records
2346 2347 * cache: if data is inlined, a tuple (index_file_content, 0), else None
2347 2348 *
2348 2349 * added complications are for backwards compatibility
2349 2350 */
2350 2351 static PyObject *parse_index2(PyObject *self, PyObject *args)
2351 2352 {
2352 2353 PyObject *tuple = NULL, *cache = NULL;
2353 2354 indexObject *idx;
2354 2355 int ret;
2355 2356
2356 2357 idx = PyObject_New(indexObject, &indexType);
2357 2358 if (idx == NULL)
2358 2359 goto bail;
2359 2360
2360 2361 ret = index_init(idx, args);
2361 2362 if (ret == -1)
2362 2363 goto bail;
2363 2364
2364 2365 if (idx->inlined) {
2365 2366 cache = Py_BuildValue("iO", 0, idx->data);
2366 2367 if (cache == NULL)
2367 2368 goto bail;
2368 2369 } else {
2369 2370 cache = Py_None;
2370 2371 Py_INCREF(cache);
2371 2372 }
2372 2373
2373 2374 tuple = Py_BuildValue("NN", idx, cache);
2374 2375 if (!tuple)
2375 2376 goto bail;
2376 2377 return tuple;
2377 2378
2378 2379 bail:
2379 2380 Py_XDECREF(idx);
2380 2381 Py_XDECREF(cache);
2381 2382 Py_XDECREF(tuple);
2382 2383 return NULL;
2383 2384 }
2384 2385
2385 2386 #define BUMPED_FIX 1
2386 2387 #define USING_SHA_256 2
2387 2388
2388 2389 static PyObject *readshas(
2389 2390 const char *source, unsigned char num, Py_ssize_t hashwidth)
2390 2391 {
2391 2392 int i;
2392 2393 PyObject *list = PyTuple_New(num);
2393 2394 if (list == NULL) {
2394 2395 return NULL;
2395 2396 }
2396 2397 for (i = 0; i < num; i++) {
2397 2398 PyObject *hash = PyString_FromStringAndSize(source, hashwidth);
2398 2399 if (hash == NULL) {
2399 2400 Py_DECREF(list);
2400 2401 return NULL;
2401 2402 }
2402 2403 PyTuple_SetItem(list, i, hash);
2403 2404 source += hashwidth;
2404 2405 }
2405 2406 return list;
2406 2407 }
2407 2408
2408 2409 static PyObject *fm1readmarker(const char *data, uint32_t *msize)
2409 2410 {
2410 2411 const char *meta;
2411 2412
2412 2413 double mtime;
2413 2414 int16_t tz;
2414 2415 uint16_t flags;
2415 2416 unsigned char nsuccs, nparents, nmetadata;
2416 2417 Py_ssize_t hashwidth = 20;
2417 2418
2418 2419 PyObject *prec = NULL, *parents = NULL, *succs = NULL;
2419 2420 PyObject *metadata = NULL, *ret = NULL;
2420 2421 int i;
2421 2422
2422 2423 *msize = getbe32(data);
2423 2424 data += 4;
2424 2425 mtime = getbefloat64(data);
2425 2426 data += 8;
2426 2427 tz = getbeint16(data);
2427 2428 data += 2;
2428 2429 flags = getbeuint16(data);
2429 2430 data += 2;
2430 2431
2431 2432 if (flags & USING_SHA_256) {
2432 2433 hashwidth = 32;
2433 2434 }
2434 2435
2435 2436 nsuccs = (unsigned char)(*data++);
2436 2437 nparents = (unsigned char)(*data++);
2437 2438 nmetadata = (unsigned char)(*data++);
2438 2439
2439 2440 prec = PyString_FromStringAndSize(data, hashwidth);
2440 2441 data += hashwidth;
2441 2442 if (prec == NULL) {
2442 2443 goto bail;
2443 2444 }
2444 2445
2445 2446 succs = readshas(data, nsuccs, hashwidth);
2446 2447 if (succs == NULL) {
2447 2448 goto bail;
2448 2449 }
2449 2450 data += nsuccs * hashwidth;
2450 2451
2451 2452 if (nparents == 1 || nparents == 2) {
2452 2453 parents = readshas(data, nparents, hashwidth);
2453 2454 if (parents == NULL) {
2454 2455 goto bail;
2455 2456 }
2456 2457 data += nparents * hashwidth;
2457 2458 } else {
2458 2459 parents = Py_None;
2459 2460 }
2460 2461
2461 2462 meta = data + (2 * nmetadata);
2462 2463 metadata = PyTuple_New(nmetadata);
2463 2464 if (metadata == NULL) {
2464 2465 goto bail;
2465 2466 }
2466 2467 for (i = 0; i < nmetadata; i++) {
2467 2468 PyObject *tmp, *left = NULL, *right = NULL;
2468 2469 Py_ssize_t metasize = (unsigned char)(*data++);
2469 2470 left = PyString_FromStringAndSize(meta, metasize);
2470 2471 meta += metasize;
2471 2472 metasize = (unsigned char)(*data++);
2472 2473 right = PyString_FromStringAndSize(meta, metasize);
2473 2474 meta += metasize;
2474 2475 if (!left || !right) {
2475 2476 Py_XDECREF(left);
2476 2477 Py_XDECREF(right);
2477 2478 goto bail;
2478 2479 }
2479 2480 tmp = PyTuple_Pack(2, left, right);
2480 2481 Py_DECREF(left);
2481 2482 Py_DECREF(right);
2482 2483 if (!tmp) {
2483 2484 goto bail;
2484 2485 }
2485 2486 PyTuple_SetItem(metadata, i, tmp);
2486 2487 }
2487 2488 ret = Py_BuildValue("(OOHO(di)O)", prec, succs, flags,
2488 2489 metadata, mtime, (int)tz * 60, parents);
2489 2490 bail:
2490 2491 Py_XDECREF(prec);
2491 2492 Py_XDECREF(succs);
2492 2493 Py_XDECREF(metadata);
2493 2494 if (parents != Py_None)
2494 2495 Py_XDECREF(parents);
2495 2496 return ret;
2496 2497 }
2497 2498
2498 2499
2499 2500 static PyObject *fm1readmarkers(PyObject *self, PyObject *args) {
2500 2501 const char *data;
2501 2502 Py_ssize_t datalen;
2502 2503 /* only unsigned long because python 2.4, should be Py_ssize_t */
2503 2504 unsigned long offset, stop;
2504 2505 PyObject *markers = NULL;
2505 2506
2506 2507 /* replace kk with nn when we drop Python 2.4 */
2507 2508 if (!PyArg_ParseTuple(args, "s#kk", &data, &datalen, &offset, &stop)) {
2508 2509 return NULL;
2509 2510 }
2510 2511 data += offset;
2511 2512 markers = PyList_New(0);
2512 2513 if (!markers) {
2513 2514 return NULL;
2514 2515 }
2515 2516 while (offset < stop) {
2516 2517 uint32_t msize;
2517 2518 int error;
2518 2519 PyObject *record = fm1readmarker(data, &msize);
2519 2520 if (!record) {
2520 2521 goto bail;
2521 2522 }
2522 2523 error = PyList_Append(markers, record);
2523 2524 Py_DECREF(record);
2524 2525 if (error) {
2525 2526 goto bail;
2526 2527 }
2527 2528 data += msize;
2528 2529 offset += msize;
2529 2530 }
2530 2531 return markers;
2531 2532 bail:
2532 2533 Py_DECREF(markers);
2533 2534 return NULL;
2534 2535 }
2535 2536
2536 2537 static char parsers_doc[] = "Efficient content parsing.";
2537 2538
2538 2539 PyObject *encodedir(PyObject *self, PyObject *args);
2539 2540 PyObject *pathencode(PyObject *self, PyObject *args);
2540 2541 PyObject *lowerencode(PyObject *self, PyObject *args);
2541 2542
2542 2543 static PyMethodDef methods[] = {
2543 2544 {"pack_dirstate", pack_dirstate, METH_VARARGS, "pack a dirstate\n"},
2544 2545 {"parse_manifest", parse_manifest, METH_VARARGS, "parse a manifest\n"},
2545 2546 {"parse_dirstate", parse_dirstate, METH_VARARGS, "parse a dirstate\n"},
2546 2547 {"parse_index2", parse_index2, METH_VARARGS, "parse a revlog index\n"},
2547 2548 {"asciilower", asciilower, METH_VARARGS, "lowercase an ASCII string\n"},
2548 2549 {"asciiupper", asciiupper, METH_VARARGS, "uppercase an ASCII string\n"},
2549 2550 {"make_file_foldmap", make_file_foldmap, METH_VARARGS,
2550 2551 "make file foldmap\n"},
2551 2552 {"encodedir", encodedir, METH_VARARGS, "encodedir a path\n"},
2552 2553 {"pathencode", pathencode, METH_VARARGS, "fncache-encode a path\n"},
2553 2554 {"lowerencode", lowerencode, METH_VARARGS, "lower-encode a path\n"},
2554 2555 {"fm1readmarkers", fm1readmarkers, METH_VARARGS,
2555 2556 "parse v1 obsolete markers\n"},
2556 2557 {NULL, NULL}
2557 2558 };
2558 2559
2559 2560 void dirs_module_init(PyObject *mod);
2560 2561 void manifest_module_init(PyObject *mod);
2561 2562
2562 2563 static void module_init(PyObject *mod)
2563 2564 {
2564 2565 /* This module constant has two purposes. First, it lets us unit test
2565 2566 * the ImportError raised without hard-coding any error text. This
2566 2567 * means we can change the text in the future without breaking tests,
2567 2568 * even across changesets without a recompile. Second, its presence
2568 2569 * can be used to determine whether the version-checking logic is
2569 2570 * present, which also helps in testing across changesets without a
2570 2571 * recompile. Note that this means the pure-Python version of parsers
2571 2572 * should not have this module constant. */
2572 2573 PyModule_AddStringConstant(mod, "versionerrortext", versionerrortext);
2573 2574
2574 2575 dirs_module_init(mod);
2575 2576 manifest_module_init(mod);
2576 2577
2577 2578 indexType.tp_new = PyType_GenericNew;
2578 2579 if (PyType_Ready(&indexType) < 0 ||
2579 2580 PyType_Ready(&dirstateTupleType) < 0)
2580 2581 return;
2581 2582 Py_INCREF(&indexType);
2582 2583 PyModule_AddObject(mod, "index", (PyObject *)&indexType);
2583 2584 Py_INCREF(&dirstateTupleType);
2584 2585 PyModule_AddObject(mod, "dirstatetuple",
2585 2586 (PyObject *)&dirstateTupleType);
2586 2587
2587 2588 nullentry = Py_BuildValue("iiiiiiis#", 0, 0, 0,
2588 2589 -1, -1, -1, -1, nullid, 20);
2589 2590 if (nullentry)
2590 2591 PyObject_GC_UnTrack(nullentry);
2591 2592 }
2592 2593
2593 2594 static int check_python_version(void)
2594 2595 {
2595 2596 PyObject *sys = PyImport_ImportModule("sys"), *ver;
2596 2597 long hexversion;
2597 2598 if (!sys)
2598 2599 return -1;
2599 2600 ver = PyObject_GetAttrString(sys, "hexversion");
2600 2601 Py_DECREF(sys);
2601 2602 if (!ver)
2602 2603 return -1;
2603 2604 hexversion = PyInt_AsLong(ver);
2604 2605 Py_DECREF(ver);
2605 2606 /* sys.hexversion is a 32-bit number by default, so the -1 case
2606 2607 * should only occur in unusual circumstances (e.g. if sys.hexversion
2607 2608 * is manually set to an invalid value). */
2608 2609 if ((hexversion == -1) || (hexversion >> 16 != PY_VERSION_HEX >> 16)) {
2609 2610 PyErr_Format(PyExc_ImportError, "%s: The Mercurial extension "
2610 2611 "modules were compiled with Python " PY_VERSION ", but "
2611 2612 "Mercurial is currently using Python with sys.hexversion=%ld: "
2612 2613 "Python %s\n at: %s", versionerrortext, hexversion,
2613 2614 Py_GetVersion(), Py_GetProgramFullPath());
2614 2615 return -1;
2615 2616 }
2616 2617 return 0;
2617 2618 }
2618 2619
2619 2620 #ifdef IS_PY3K
2620 2621 static struct PyModuleDef parsers_module = {
2621 2622 PyModuleDef_HEAD_INIT,
2622 2623 "parsers",
2623 2624 parsers_doc,
2624 2625 -1,
2625 2626 methods
2626 2627 };
2627 2628
2628 2629 PyMODINIT_FUNC PyInit_parsers(void)
2629 2630 {
2630 2631 PyObject *mod;
2631 2632
2632 2633 if (check_python_version() == -1)
2633 2634 return;
2634 2635 mod = PyModule_Create(&parsers_module);
2635 2636 module_init(mod);
2636 2637 return mod;
2637 2638 }
2638 2639 #else
2639 2640 PyMODINIT_FUNC initparsers(void)
2640 2641 {
2641 2642 PyObject *mod;
2642 2643
2643 2644 if (check_python_version() == -1)
2644 2645 return;
2645 2646 mod = Py_InitModule3("parsers", methods, parsers_doc);
2646 2647 module_init(mod);
2647 2648 }
2648 2649 #endif
General Comments 0
You need to be logged in to leave comments. Login now