##// END OF EJS Templates
cext: move braces for control statements to same line...
Gregory Szorc -
r34438:ce26a138 default
parent child Browse files
Show More
@@ -1,189 +1,185
1 1 /*
2 2 base85 codec
3 3
4 4 Copyright 2006 Brendan Cully <brendan@kublai.com>
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 Largely based on git's implementation
10 10 */
11 11
12 12 #define PY_SSIZE_T_CLEAN
13 13 #include <Python.h>
14 14
15 15 #include "util.h"
16 16
17 17 static const char b85chars[] = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"
18 18 "abcdefghijklmnopqrstuvwxyz!#$%&()*+-;<=>?@^_`{|}~";
19 19 static char b85dec[256];
20 20
21 21 static void b85prep(void)
22 22 {
23 23 unsigned i;
24 24
25 25 memset(b85dec, 0, sizeof(b85dec));
26 26 for (i = 0; i < sizeof(b85chars); i++)
27 27 b85dec[(int)(b85chars[i])] = i + 1;
28 28 }
29 29
30 30 static PyObject *b85encode(PyObject *self, PyObject *args)
31 31 {
32 32 const unsigned char *text;
33 33 PyObject *out;
34 34 char *dst;
35 35 Py_ssize_t len, olen, i;
36 36 unsigned int acc, val, ch;
37 37 int pad = 0;
38 38
39 39 if (!PyArg_ParseTuple(args, "s#|i", &text, &len, &pad))
40 40 return NULL;
41 41
42 42 if (pad)
43 43 olen = ((len + 3) / 4 * 5) - 3;
44 44 else {
45 45 olen = len % 4;
46 46 if (olen)
47 47 olen++;
48 48 olen += len / 4 * 5;
49 49 }
50 50 if (!(out = PyBytes_FromStringAndSize(NULL, olen + 3)))
51 51 return NULL;
52 52
53 53 dst = PyBytes_AsString(out);
54 54
55 55 while (len) {
56 56 acc = 0;
57 57 for (i = 24; i >= 0; i -= 8) {
58 58 ch = *text++;
59 59 acc |= ch << i;
60 60 if (--len == 0)
61 61 break;
62 62 }
63 63 for (i = 4; i >= 0; i--) {
64 64 val = acc % 85;
65 65 acc /= 85;
66 66 dst[i] = b85chars[val];
67 67 }
68 68 dst += 5;
69 69 }
70 70
71 71 if (!pad)
72 72 _PyBytes_Resize(&out, olen);
73 73
74 74 return out;
75 75 }
76 76
77 77 static PyObject *b85decode(PyObject *self, PyObject *args)
78 78 {
79 79 PyObject *out;
80 80 const char *text;
81 81 char *dst;
82 82 Py_ssize_t len, i, j, olen, cap;
83 83 int c;
84 84 unsigned int acc;
85 85
86 86 if (!PyArg_ParseTuple(args, "s#", &text, &len))
87 87 return NULL;
88 88
89 89 olen = len / 5 * 4;
90 90 i = len % 5;
91 91 if (i)
92 92 olen += i - 1;
93 93 if (!(out = PyBytes_FromStringAndSize(NULL, olen)))
94 94 return NULL;
95 95
96 96 dst = PyBytes_AsString(out);
97 97
98 98 i = 0;
99 while (i < len)
100 {
99 while (i < len) {
101 100 acc = 0;
102 101 cap = len - i - 1;
103 102 if (cap > 4)
104 103 cap = 4;
105 for (j = 0; j < cap; i++, j++)
106 {
104 for (j = 0; j < cap; i++, j++) {
107 105 c = b85dec[(int)*text++] - 1;
108 106 if (c < 0)
109 107 return PyErr_Format(
110 108 PyExc_ValueError,
111 109 "bad base85 character at position %d",
112 110 (int)i);
113 111 acc = acc * 85 + c;
114 112 }
115 if (i++ < len)
116 {
113 if (i++ < len) {
117 114 c = b85dec[(int)*text++] - 1;
118 115 if (c < 0)
119 116 return PyErr_Format(
120 117 PyExc_ValueError,
121 118 "bad base85 character at position %d",
122 119 (int)i);
123 120 /* overflow detection: 0xffffffff == "|NsC0",
124 121 * "|NsC" == 0x03030303 */
125 122 if (acc > 0x03030303 || (acc *= 85) > 0xffffffff - c)
126 123 return PyErr_Format(
127 124 PyExc_ValueError,
128 125 "bad base85 sequence at position %d",
129 126 (int)i);
130 127 acc += c;
131 128 }
132 129
133 130 cap = olen < 4 ? olen : 4;
134 131 olen -= cap;
135 132 for (j = 0; j < 4 - cap; j++)
136 133 acc *= 85;
137 134 if (cap && cap < 4)
138 135 acc += 0xffffff >> (cap - 1) * 8;
139 for (j = 0; j < cap; j++)
140 {
136 for (j = 0; j < cap; j++) {
141 137 acc = (acc << 8) | (acc >> 24);
142 138 *dst++ = acc;
143 139 }
144 140 }
145 141
146 142 return out;
147 143 }
148 144
149 145 static char base85_doc[] = "Base85 Data Encoding";
150 146
151 147 static PyMethodDef methods[] = {
152 148 {"b85encode", b85encode, METH_VARARGS,
153 149 "Encode text in base85.\n\n"
154 150 "If the second parameter is true, pad the result to a multiple of "
155 151 "five characters.\n"},
156 152 {"b85decode", b85decode, METH_VARARGS, "Decode base85 text.\n"},
157 153 {NULL, NULL}
158 154 };
159 155
160 156 static const int version = 1;
161 157
162 158 #ifdef IS_PY3K
163 159 static struct PyModuleDef base85_module = {
164 160 PyModuleDef_HEAD_INIT,
165 161 "base85",
166 162 base85_doc,
167 163 -1,
168 164 methods
169 165 };
170 166
171 167 PyMODINIT_FUNC PyInit_base85(void)
172 168 {
173 169 PyObject *m;
174 170 b85prep();
175 171
176 172 m = PyModule_Create(&base85_module);
177 173 PyModule_AddIntConstant(m, "version", version);
178 174 return m;
179 175 }
180 176 #else
181 177 PyMODINIT_FUNC initbase85(void)
182 178 {
183 179 PyObject *m;
184 180 m = Py_InitModule3("base85", methods, base85_doc);
185 181
186 182 b85prep();
187 183 PyModule_AddIntConstant(m, "version", version);
188 184 }
189 185 #endif
@@ -1,2090 +1,2089
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 <assert.h>
12 12 #include <ctype.h>
13 13 #include <stddef.h>
14 14 #include <string.h>
15 15
16 16 #include "charencode.h"
17 17 #include "util.h"
18 18 #include "bitmanipulation.h"
19 19
20 20 #ifdef IS_PY3K
21 21 /* The mapping of Python types is meant to be temporary to get Python
22 22 * 3 to compile. We should remove this once Python 3 support is fully
23 23 * supported and proper types are used in the extensions themselves. */
24 24 #define PyInt_Check PyLong_Check
25 25 #define PyInt_FromLong PyLong_FromLong
26 26 #define PyInt_FromSsize_t PyLong_FromSsize_t
27 27 #define PyInt_AS_LONG PyLong_AS_LONG
28 28 #define PyInt_AsLong PyLong_AsLong
29 29 #endif
30 30
31 31 /*
32 32 * A base-16 trie for fast node->rev mapping.
33 33 *
34 34 * Positive value is index of the next node in the trie
35 35 * Negative value is a leaf: -(rev + 1)
36 36 * Zero is empty
37 37 */
38 38 typedef struct {
39 39 int children[16];
40 40 } nodetree;
41 41
42 42 /*
43 43 * This class has two behaviors.
44 44 *
45 45 * When used in a list-like way (with integer keys), we decode an
46 46 * entry in a RevlogNG index file on demand. Our last entry is a
47 47 * sentinel, always a nullid. We have limited support for
48 48 * integer-keyed insert and delete, only at elements right before the
49 49 * sentinel.
50 50 *
51 51 * With string keys, we lazily perform a reverse mapping from node to
52 52 * rev, using a base-16 trie.
53 53 */
54 54 typedef struct {
55 55 PyObject_HEAD
56 56 /* Type-specific fields go here. */
57 57 PyObject *data; /* raw bytes of index */
58 58 Py_buffer buf; /* buffer of data */
59 59 PyObject **cache; /* cached tuples */
60 60 const char **offsets; /* populated on demand */
61 61 Py_ssize_t raw_length; /* original number of elements */
62 62 Py_ssize_t length; /* current number of elements */
63 63 PyObject *added; /* populated on demand */
64 64 PyObject *headrevs; /* cache, invalidated on changes */
65 65 PyObject *filteredrevs;/* filtered revs set */
66 66 nodetree *nt; /* base-16 trie */
67 67 unsigned ntlength; /* # nodes in use */
68 68 unsigned ntcapacity; /* # nodes allocated */
69 69 int ntdepth; /* maximum depth of tree */
70 70 int ntsplits; /* # splits performed */
71 71 int ntrev; /* last rev scanned */
72 72 int ntlookups; /* # lookups */
73 73 int ntmisses; /* # lookups that miss the cache */
74 74 int inlined;
75 75 } indexObject;
76 76
77 77 static Py_ssize_t index_length(const indexObject *self)
78 78 {
79 79 if (self->added == NULL)
80 80 return self->length;
81 81 return self->length + PyList_GET_SIZE(self->added);
82 82 }
83 83
84 84 static PyObject *nullentry;
85 85 static const char nullid[20];
86 86
87 87 static Py_ssize_t inline_scan(indexObject *self, const char **offsets);
88 88
89 89 #if LONG_MAX == 0x7fffffffL
90 90 static char *tuple_format = "Kiiiiiis#";
91 91 #else
92 92 static char *tuple_format = "kiiiiiis#";
93 93 #endif
94 94
95 95 /* A RevlogNG v1 index entry is 64 bytes long. */
96 96 static const long v1_hdrsize = 64;
97 97
98 98 /*
99 99 * Return a pointer to the beginning of a RevlogNG record.
100 100 */
101 101 static const char *index_deref(indexObject *self, Py_ssize_t pos)
102 102 {
103 103 if (self->inlined && pos > 0) {
104 104 if (self->offsets == NULL) {
105 105 self->offsets = PyMem_Malloc(self->raw_length *
106 106 sizeof(*self->offsets));
107 107 if (self->offsets == NULL)
108 108 return (const char *)PyErr_NoMemory();
109 109 inline_scan(self, self->offsets);
110 110 }
111 111 return self->offsets[pos];
112 112 }
113 113
114 114 return (const char *)(self->buf.buf) + pos * v1_hdrsize;
115 115 }
116 116
117 117 static inline int index_get_parents(indexObject *self, Py_ssize_t rev,
118 118 int *ps, int maxrev)
119 119 {
120 120 if (rev >= self->length - 1) {
121 121 PyObject *tuple = PyList_GET_ITEM(self->added,
122 122 rev - self->length + 1);
123 123 ps[0] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 5));
124 124 ps[1] = (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 6));
125 125 } else {
126 126 const char *data = index_deref(self, rev);
127 127 ps[0] = getbe32(data + 24);
128 128 ps[1] = getbe32(data + 28);
129 129 }
130 130 /* If index file is corrupted, ps[] may point to invalid revisions. So
131 131 * there is a risk of buffer overflow to trust them unconditionally. */
132 132 if (ps[0] > maxrev || ps[1] > maxrev) {
133 133 PyErr_SetString(PyExc_ValueError, "parent out of range");
134 134 return -1;
135 135 }
136 136 return 0;
137 137 }
138 138
139 139
140 140 /*
141 141 * RevlogNG format (all in big endian, data may be inlined):
142 142 * 6 bytes: offset
143 143 * 2 bytes: flags
144 144 * 4 bytes: compressed length
145 145 * 4 bytes: uncompressed length
146 146 * 4 bytes: base revision
147 147 * 4 bytes: link revision
148 148 * 4 bytes: parent 1 revision
149 149 * 4 bytes: parent 2 revision
150 150 * 32 bytes: nodeid (only 20 bytes used)
151 151 */
152 152 static PyObject *index_get(indexObject *self, Py_ssize_t pos)
153 153 {
154 154 uint64_t offset_flags;
155 155 int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2;
156 156 const char *c_node_id;
157 157 const char *data;
158 158 Py_ssize_t length = index_length(self);
159 159 PyObject *entry;
160 160
161 161 if (pos < 0)
162 162 pos += length;
163 163
164 164 if (pos < 0 || pos >= length) {
165 165 PyErr_SetString(PyExc_IndexError, "revlog index out of range");
166 166 return NULL;
167 167 }
168 168
169 169 if (pos == length - 1) {
170 170 Py_INCREF(nullentry);
171 171 return nullentry;
172 172 }
173 173
174 174 if (pos >= self->length - 1) {
175 175 PyObject *obj;
176 176 obj = PyList_GET_ITEM(self->added, pos - self->length + 1);
177 177 Py_INCREF(obj);
178 178 return obj;
179 179 }
180 180
181 181 if (self->cache) {
182 182 if (self->cache[pos]) {
183 183 Py_INCREF(self->cache[pos]);
184 184 return self->cache[pos];
185 185 }
186 186 } else {
187 187 self->cache = calloc(self->raw_length, sizeof(PyObject *));
188 188 if (self->cache == NULL)
189 189 return PyErr_NoMemory();
190 190 }
191 191
192 192 data = index_deref(self, pos);
193 193 if (data == NULL)
194 194 return NULL;
195 195
196 196 offset_flags = getbe32(data + 4);
197 197 if (pos == 0) /* mask out version number for the first entry */
198 198 offset_flags &= 0xFFFF;
199 199 else {
200 200 uint32_t offset_high = getbe32(data);
201 201 offset_flags |= ((uint64_t)offset_high) << 32;
202 202 }
203 203
204 204 comp_len = getbe32(data + 8);
205 205 uncomp_len = getbe32(data + 12);
206 206 base_rev = getbe32(data + 16);
207 207 link_rev = getbe32(data + 20);
208 208 parent_1 = getbe32(data + 24);
209 209 parent_2 = getbe32(data + 28);
210 210 c_node_id = data + 32;
211 211
212 212 entry = Py_BuildValue(tuple_format, offset_flags, comp_len,
213 213 uncomp_len, base_rev, link_rev,
214 214 parent_1, parent_2, c_node_id, 20);
215 215
216 216 if (entry) {
217 217 PyObject_GC_UnTrack(entry);
218 218 Py_INCREF(entry);
219 219 }
220 220
221 221 self->cache[pos] = entry;
222 222
223 223 return entry;
224 224 }
225 225
226 226 /*
227 227 * Return the 20-byte SHA of the node corresponding to the given rev.
228 228 */
229 229 static const char *index_node(indexObject *self, Py_ssize_t pos)
230 230 {
231 231 Py_ssize_t length = index_length(self);
232 232 const char *data;
233 233
234 234 if (pos == length - 1 || pos == INT_MAX)
235 235 return nullid;
236 236
237 237 if (pos >= length)
238 238 return NULL;
239 239
240 240 if (pos >= self->length - 1) {
241 241 PyObject *tuple, *str;
242 242 tuple = PyList_GET_ITEM(self->added, pos - self->length + 1);
243 243 str = PyTuple_GetItem(tuple, 7);
244 244 return str ? PyBytes_AS_STRING(str) : NULL;
245 245 }
246 246
247 247 data = index_deref(self, pos);
248 248 return data ? data + 32 : NULL;
249 249 }
250 250
251 251 static int nt_insert(indexObject *self, const char *node, int rev);
252 252
253 253 static int node_check(PyObject *obj, char **node, Py_ssize_t *nodelen)
254 254 {
255 255 if (PyBytes_AsStringAndSize(obj, node, nodelen) == -1)
256 256 return -1;
257 257 if (*nodelen == 20)
258 258 return 0;
259 259 PyErr_SetString(PyExc_ValueError, "20-byte hash required");
260 260 return -1;
261 261 }
262 262
263 263 static PyObject *index_insert(indexObject *self, PyObject *args)
264 264 {
265 265 PyObject *obj;
266 266 char *node;
267 267 int index;
268 268 Py_ssize_t len, nodelen;
269 269
270 270 if (!PyArg_ParseTuple(args, "iO", &index, &obj))
271 271 return NULL;
272 272
273 273 if (!PyTuple_Check(obj) || PyTuple_GET_SIZE(obj) != 8) {
274 274 PyErr_SetString(PyExc_TypeError, "8-tuple required");
275 275 return NULL;
276 276 }
277 277
278 278 if (node_check(PyTuple_GET_ITEM(obj, 7), &node, &nodelen) == -1)
279 279 return NULL;
280 280
281 281 len = index_length(self);
282 282
283 283 if (index < 0)
284 284 index += len;
285 285
286 286 if (index != len - 1) {
287 287 PyErr_SetString(PyExc_IndexError,
288 288 "insert only supported at index -1");
289 289 return NULL;
290 290 }
291 291
292 292 if (self->added == NULL) {
293 293 self->added = PyList_New(0);
294 294 if (self->added == NULL)
295 295 return NULL;
296 296 }
297 297
298 298 if (PyList_Append(self->added, obj) == -1)
299 299 return NULL;
300 300
301 301 if (self->nt)
302 302 nt_insert(self, node, index);
303 303
304 304 Py_CLEAR(self->headrevs);
305 305 Py_RETURN_NONE;
306 306 }
307 307
308 308 static void _index_clearcaches(indexObject *self)
309 309 {
310 310 if (self->cache) {
311 311 Py_ssize_t i;
312 312
313 313 for (i = 0; i < self->raw_length; i++)
314 314 Py_CLEAR(self->cache[i]);
315 315 free(self->cache);
316 316 self->cache = NULL;
317 317 }
318 318 if (self->offsets) {
319 319 PyMem_Free(self->offsets);
320 320 self->offsets = NULL;
321 321 }
322 322 if (self->nt) {
323 323 free(self->nt);
324 324 self->nt = NULL;
325 325 }
326 326 Py_CLEAR(self->headrevs);
327 327 }
328 328
329 329 static PyObject *index_clearcaches(indexObject *self)
330 330 {
331 331 _index_clearcaches(self);
332 332 self->ntlength = self->ntcapacity = 0;
333 333 self->ntdepth = self->ntsplits = 0;
334 334 self->ntrev = -1;
335 335 self->ntlookups = self->ntmisses = 0;
336 336 Py_RETURN_NONE;
337 337 }
338 338
339 339 static PyObject *index_stats(indexObject *self)
340 340 {
341 341 PyObject *obj = PyDict_New();
342 342 PyObject *t = NULL;
343 343
344 344 if (obj == NULL)
345 345 return NULL;
346 346
347 347 #define istat(__n, __d) \
348 348 do { \
349 349 t = PyInt_FromSsize_t(self->__n); \
350 350 if (!t) \
351 351 goto bail; \
352 352 if (PyDict_SetItemString(obj, __d, t) == -1) \
353 353 goto bail; \
354 354 Py_DECREF(t); \
355 355 } while (0)
356 356
357 357 if (self->added) {
358 358 Py_ssize_t len = PyList_GET_SIZE(self->added);
359 359 t = PyInt_FromSsize_t(len);
360 360 if (!t)
361 361 goto bail;
362 362 if (PyDict_SetItemString(obj, "index entries added", t) == -1)
363 363 goto bail;
364 364 Py_DECREF(t);
365 365 }
366 366
367 367 if (self->raw_length != self->length - 1)
368 368 istat(raw_length, "revs on disk");
369 369 istat(length, "revs in memory");
370 370 istat(ntcapacity, "node trie capacity");
371 371 istat(ntdepth, "node trie depth");
372 372 istat(ntlength, "node trie count");
373 373 istat(ntlookups, "node trie lookups");
374 374 istat(ntmisses, "node trie misses");
375 375 istat(ntrev, "node trie last rev scanned");
376 376 istat(ntsplits, "node trie splits");
377 377
378 378 #undef istat
379 379
380 380 return obj;
381 381
382 382 bail:
383 383 Py_XDECREF(obj);
384 384 Py_XDECREF(t);
385 385 return NULL;
386 386 }
387 387
388 388 /*
389 389 * When we cache a list, we want to be sure the caller can't mutate
390 390 * the cached copy.
391 391 */
392 392 static PyObject *list_copy(PyObject *list)
393 393 {
394 394 Py_ssize_t len = PyList_GET_SIZE(list);
395 395 PyObject *newlist = PyList_New(len);
396 396 Py_ssize_t i;
397 397
398 398 if (newlist == NULL)
399 399 return NULL;
400 400
401 401 for (i = 0; i < len; i++) {
402 402 PyObject *obj = PyList_GET_ITEM(list, i);
403 403 Py_INCREF(obj);
404 404 PyList_SET_ITEM(newlist, i, obj);
405 405 }
406 406
407 407 return newlist;
408 408 }
409 409
410 410 static int check_filter(PyObject *filter, Py_ssize_t arg) {
411 411 if (filter) {
412 412 PyObject *arglist, *result;
413 413 int isfiltered;
414 414
415 415 arglist = Py_BuildValue("(n)", arg);
416 416 if (!arglist) {
417 417 return -1;
418 418 }
419 419
420 420 result = PyEval_CallObject(filter, arglist);
421 421 Py_DECREF(arglist);
422 422 if (!result) {
423 423 return -1;
424 424 }
425 425
426 426 /* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error,
427 427 * same as this function, so we can just return it directly.*/
428 428 isfiltered = PyObject_IsTrue(result);
429 429 Py_DECREF(result);
430 430 return isfiltered;
431 431 } else {
432 432 return 0;
433 433 }
434 434 }
435 435
436 436 static Py_ssize_t add_roots_get_min(indexObject *self, PyObject *list,
437 437 Py_ssize_t marker, char *phases)
438 438 {
439 439 PyObject *iter = NULL;
440 440 PyObject *iter_item = NULL;
441 441 Py_ssize_t min_idx = index_length(self) + 1;
442 442 long iter_item_long;
443 443
444 444 if (PyList_GET_SIZE(list) != 0) {
445 445 iter = PyObject_GetIter(list);
446 446 if (iter == NULL)
447 447 return -2;
448 while ((iter_item = PyIter_Next(iter)))
449 {
448 while ((iter_item = PyIter_Next(iter))) {
450 449 iter_item_long = PyInt_AS_LONG(iter_item);
451 450 Py_DECREF(iter_item);
452 451 if (iter_item_long < min_idx)
453 452 min_idx = iter_item_long;
454 453 phases[iter_item_long] = marker;
455 454 }
456 455 Py_DECREF(iter);
457 456 }
458 457
459 458 return min_idx;
460 459 }
461 460
462 461 static inline void set_phase_from_parents(char *phases, int parent_1,
463 462 int parent_2, Py_ssize_t i)
464 463 {
465 464 if (parent_1 >= 0 && phases[parent_1] > phases[i])
466 465 phases[i] = phases[parent_1];
467 466 if (parent_2 >= 0 && phases[parent_2] > phases[i])
468 467 phases[i] = phases[parent_2];
469 468 }
470 469
471 470 static PyObject *reachableroots2(indexObject *self, PyObject *args)
472 471 {
473 472
474 473 /* Input */
475 474 long minroot;
476 475 PyObject *includepatharg = NULL;
477 476 int includepath = 0;
478 477 /* heads and roots are lists */
479 478 PyObject *heads = NULL;
480 479 PyObject *roots = NULL;
481 480 PyObject *reachable = NULL;
482 481
483 482 PyObject *val;
484 483 Py_ssize_t len = index_length(self) - 1;
485 484 long revnum;
486 485 Py_ssize_t k;
487 486 Py_ssize_t i;
488 487 Py_ssize_t l;
489 488 int r;
490 489 int parents[2];
491 490
492 491 /* Internal data structure:
493 492 * tovisit: array of length len+1 (all revs + nullrev), filled upto lentovisit
494 493 * revstates: array of length len+1 (all revs + nullrev) */
495 494 int *tovisit = NULL;
496 495 long lentovisit = 0;
497 496 enum { RS_SEEN = 1, RS_ROOT = 2, RS_REACHABLE = 4 };
498 497 char *revstates = NULL;
499 498
500 499 /* Get arguments */
501 500 if (!PyArg_ParseTuple(args, "lO!O!O!", &minroot, &PyList_Type, &heads,
502 501 &PyList_Type, &roots,
503 502 &PyBool_Type, &includepatharg))
504 503 goto bail;
505 504
506 505 if (includepatharg == Py_True)
507 506 includepath = 1;
508 507
509 508 /* Initialize return set */
510 509 reachable = PyList_New(0);
511 510 if (reachable == NULL)
512 511 goto bail;
513 512
514 513 /* Initialize internal datastructures */
515 514 tovisit = (int *)malloc((len + 1) * sizeof(int));
516 515 if (tovisit == NULL) {
517 516 PyErr_NoMemory();
518 517 goto bail;
519 518 }
520 519
521 520 revstates = (char *)calloc(len + 1, 1);
522 521 if (revstates == NULL) {
523 522 PyErr_NoMemory();
524 523 goto bail;
525 524 }
526 525
527 526 l = PyList_GET_SIZE(roots);
528 527 for (i = 0; i < l; i++) {
529 528 revnum = PyInt_AsLong(PyList_GET_ITEM(roots, i));
530 529 if (revnum == -1 && PyErr_Occurred())
531 530 goto bail;
532 531 /* If root is out of range, e.g. wdir(), it must be unreachable
533 532 * from heads. So we can just ignore it. */
534 533 if (revnum + 1 < 0 || revnum + 1 >= len + 1)
535 534 continue;
536 535 revstates[revnum + 1] |= RS_ROOT;
537 536 }
538 537
539 538 /* Populate tovisit with all the heads */
540 539 l = PyList_GET_SIZE(heads);
541 540 for (i = 0; i < l; i++) {
542 541 revnum = PyInt_AsLong(PyList_GET_ITEM(heads, i));
543 542 if (revnum == -1 && PyErr_Occurred())
544 543 goto bail;
545 544 if (revnum + 1 < 0 || revnum + 1 >= len + 1) {
546 545 PyErr_SetString(PyExc_IndexError, "head out of range");
547 546 goto bail;
548 547 }
549 548 if (!(revstates[revnum + 1] & RS_SEEN)) {
550 549 tovisit[lentovisit++] = (int)revnum;
551 550 revstates[revnum + 1] |= RS_SEEN;
552 551 }
553 552 }
554 553
555 554 /* Visit the tovisit list and find the reachable roots */
556 555 k = 0;
557 556 while (k < lentovisit) {
558 557 /* Add the node to reachable if it is a root*/
559 558 revnum = tovisit[k++];
560 559 if (revstates[revnum + 1] & RS_ROOT) {
561 560 revstates[revnum + 1] |= RS_REACHABLE;
562 561 val = PyInt_FromLong(revnum);
563 562 if (val == NULL)
564 563 goto bail;
565 564 r = PyList_Append(reachable, val);
566 565 Py_DECREF(val);
567 566 if (r < 0)
568 567 goto bail;
569 568 if (includepath == 0)
570 569 continue;
571 570 }
572 571
573 572 /* Add its parents to the list of nodes to visit */
574 573 if (revnum == -1)
575 574 continue;
576 575 r = index_get_parents(self, revnum, parents, (int)len - 1);
577 576 if (r < 0)
578 577 goto bail;
579 578 for (i = 0; i < 2; i++) {
580 579 if (!(revstates[parents[i] + 1] & RS_SEEN)
581 580 && parents[i] >= minroot) {
582 581 tovisit[lentovisit++] = parents[i];
583 582 revstates[parents[i] + 1] |= RS_SEEN;
584 583 }
585 584 }
586 585 }
587 586
588 587 /* Find all the nodes in between the roots we found and the heads
589 588 * and add them to the reachable set */
590 589 if (includepath == 1) {
591 590 long minidx = minroot;
592 591 if (minidx < 0)
593 592 minidx = 0;
594 593 for (i = minidx; i < len; i++) {
595 594 if (!(revstates[i + 1] & RS_SEEN))
596 595 continue;
597 596 r = index_get_parents(self, i, parents, (int)len - 1);
598 597 /* Corrupted index file, error is set from
599 598 * index_get_parents */
600 599 if (r < 0)
601 600 goto bail;
602 601 if (((revstates[parents[0] + 1] |
603 602 revstates[parents[1] + 1]) & RS_REACHABLE)
604 603 && !(revstates[i + 1] & RS_REACHABLE)) {
605 604 revstates[i + 1] |= RS_REACHABLE;
606 605 val = PyInt_FromLong(i);
607 606 if (val == NULL)
608 607 goto bail;
609 608 r = PyList_Append(reachable, val);
610 609 Py_DECREF(val);
611 610 if (r < 0)
612 611 goto bail;
613 612 }
614 613 }
615 614 }
616 615
617 616 free(revstates);
618 617 free(tovisit);
619 618 return reachable;
620 619 bail:
621 620 Py_XDECREF(reachable);
622 621 free(revstates);
623 622 free(tovisit);
624 623 return NULL;
625 624 }
626 625
627 626 static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args)
628 627 {
629 628 PyObject *roots = Py_None;
630 629 PyObject *ret = NULL;
631 630 PyObject *phaseslist = NULL;
632 631 PyObject *phaseroots = NULL;
633 632 PyObject *phaseset = NULL;
634 633 PyObject *phasessetlist = NULL;
635 634 PyObject *rev = NULL;
636 635 Py_ssize_t len = index_length(self) - 1;
637 636 Py_ssize_t numphase = 0;
638 637 Py_ssize_t minrevallphases = 0;
639 638 Py_ssize_t minrevphase = 0;
640 639 Py_ssize_t i = 0;
641 640 char *phases = NULL;
642 641 long phase;
643 642
644 643 if (!PyArg_ParseTuple(args, "O", &roots))
645 644 goto done;
646 645 if (roots == NULL || !PyList_Check(roots))
647 646 goto done;
648 647
649 648 phases = calloc(len, 1); /* phase per rev: {0: public, 1: draft, 2: secret} */
650 649 if (phases == NULL) {
651 650 PyErr_NoMemory();
652 651 goto done;
653 652 }
654 653 /* Put the phase information of all the roots in phases */
655 654 numphase = PyList_GET_SIZE(roots)+1;
656 655 minrevallphases = len + 1;
657 656 phasessetlist = PyList_New(numphase);
658 657 if (phasessetlist == NULL)
659 658 goto done;
660 659
661 660 PyList_SET_ITEM(phasessetlist, 0, Py_None);
662 661 Py_INCREF(Py_None);
663 662
664 663 for (i = 0; i < numphase-1; i++) {
665 664 phaseroots = PyList_GET_ITEM(roots, i);
666 665 phaseset = PySet_New(NULL);
667 666 if (phaseset == NULL)
668 667 goto release;
669 668 PyList_SET_ITEM(phasessetlist, i+1, phaseset);
670 669 if (!PyList_Check(phaseroots))
671 670 goto release;
672 671 minrevphase = add_roots_get_min(self, phaseroots, i+1, phases);
673 672 if (minrevphase == -2) /* Error from add_roots_get_min */
674 673 goto release;
675 674 minrevallphases = MIN(minrevallphases, minrevphase);
676 675 }
677 676 /* Propagate the phase information from the roots to the revs */
678 677 if (minrevallphases != -1) {
679 678 int parents[2];
680 679 for (i = minrevallphases; i < len; i++) {
681 680 if (index_get_parents(self, i, parents,
682 681 (int)len - 1) < 0)
683 682 goto release;
684 683 set_phase_from_parents(phases, parents[0], parents[1], i);
685 684 }
686 685 }
687 686 /* Transform phase list to a python list */
688 687 phaseslist = PyList_New(len);
689 688 if (phaseslist == NULL)
690 689 goto release;
691 690 for (i = 0; i < len; i++) {
692 691 PyObject *phaseval;
693 692
694 693 phase = phases[i];
695 694 /* We only store the sets of phase for non public phase, the public phase
696 695 * is computed as a difference */
697 696 if (phase != 0) {
698 697 phaseset = PyList_GET_ITEM(phasessetlist, phase);
699 698 rev = PyInt_FromLong(i);
700 699 if (rev == NULL)
701 700 goto release;
702 701 PySet_Add(phaseset, rev);
703 702 Py_XDECREF(rev);
704 703 }
705 704 phaseval = PyInt_FromLong(phase);
706 705 if (phaseval == NULL)
707 706 goto release;
708 707 PyList_SET_ITEM(phaseslist, i, phaseval);
709 708 }
710 709 ret = PyTuple_Pack(2, phaseslist, phasessetlist);
711 710
712 711 release:
713 712 Py_XDECREF(phaseslist);
714 713 Py_XDECREF(phasessetlist);
715 714 done:
716 715 free(phases);
717 716 return ret;
718 717 }
719 718
720 719 static PyObject *index_headrevs(indexObject *self, PyObject *args)
721 720 {
722 721 Py_ssize_t i, j, len;
723 722 char *nothead = NULL;
724 723 PyObject *heads = NULL;
725 724 PyObject *filter = NULL;
726 725 PyObject *filteredrevs = Py_None;
727 726
728 727 if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) {
729 728 return NULL;
730 729 }
731 730
732 731 if (self->headrevs && filteredrevs == self->filteredrevs)
733 732 return list_copy(self->headrevs);
734 733
735 734 Py_DECREF(self->filteredrevs);
736 735 self->filteredrevs = filteredrevs;
737 736 Py_INCREF(filteredrevs);
738 737
739 738 if (filteredrevs != Py_None) {
740 739 filter = PyObject_GetAttrString(filteredrevs, "__contains__");
741 740 if (!filter) {
742 741 PyErr_SetString(PyExc_TypeError,
743 742 "filteredrevs has no attribute __contains__");
744 743 goto bail;
745 744 }
746 745 }
747 746
748 747 len = index_length(self) - 1;
749 748 heads = PyList_New(0);
750 749 if (heads == NULL)
751 750 goto bail;
752 751 if (len == 0) {
753 752 PyObject *nullid = PyInt_FromLong(-1);
754 753 if (nullid == NULL || PyList_Append(heads, nullid) == -1) {
755 754 Py_XDECREF(nullid);
756 755 goto bail;
757 756 }
758 757 goto done;
759 758 }
760 759
761 760 nothead = calloc(len, 1);
762 761 if (nothead == NULL) {
763 762 PyErr_NoMemory();
764 763 goto bail;
765 764 }
766 765
767 766 for (i = len - 1; i >= 0; i--) {
768 767 int isfiltered;
769 768 int parents[2];
770 769
771 770 /* If nothead[i] == 1, it means we've seen an unfiltered child of this
772 771 * node already, and therefore this node is not filtered. So we can skip
773 772 * the expensive check_filter step.
774 773 */
775 774 if (nothead[i] != 1) {
776 775 isfiltered = check_filter(filter, i);
777 776 if (isfiltered == -1) {
778 777 PyErr_SetString(PyExc_TypeError,
779 778 "unable to check filter");
780 779 goto bail;
781 780 }
782 781
783 782 if (isfiltered) {
784 783 nothead[i] = 1;
785 784 continue;
786 785 }
787 786 }
788 787
789 788 if (index_get_parents(self, i, parents, (int)len - 1) < 0)
790 789 goto bail;
791 790 for (j = 0; j < 2; j++) {
792 791 if (parents[j] >= 0)
793 792 nothead[parents[j]] = 1;
794 793 }
795 794 }
796 795
797 796 for (i = 0; i < len; i++) {
798 797 PyObject *head;
799 798
800 799 if (nothead[i])
801 800 continue;
802 801 head = PyInt_FromSsize_t(i);
803 802 if (head == NULL || PyList_Append(heads, head) == -1) {
804 803 Py_XDECREF(head);
805 804 goto bail;
806 805 }
807 806 }
808 807
809 808 done:
810 809 self->headrevs = heads;
811 810 Py_XDECREF(filter);
812 811 free(nothead);
813 812 return list_copy(self->headrevs);
814 813 bail:
815 814 Py_XDECREF(filter);
816 815 Py_XDECREF(heads);
817 816 free(nothead);
818 817 return NULL;
819 818 }
820 819
821 820 /**
822 821 * Obtain the base revision index entry.
823 822 *
824 823 * Callers must ensure that rev >= 0 or illegal memory access may occur.
825 824 */
826 825 static inline int index_baserev(indexObject *self, int rev)
827 826 {
828 827 const char *data;
829 828
830 829 if (rev >= self->length - 1) {
831 830 PyObject *tuple = PyList_GET_ITEM(self->added,
832 831 rev - self->length + 1);
833 832 return (int)PyInt_AS_LONG(PyTuple_GET_ITEM(tuple, 3));
834 833 }
835 834 else {
836 835 data = index_deref(self, rev);
837 836 if (data == NULL) {
838 837 return -2;
839 838 }
840 839
841 840 return getbe32(data + 16);
842 841 }
843 842 }
844 843
845 844 static PyObject *index_deltachain(indexObject *self, PyObject *args)
846 845 {
847 846 int rev, generaldelta;
848 847 PyObject *stoparg;
849 848 int stoprev, iterrev, baserev = -1;
850 849 int stopped;
851 850 PyObject *chain = NULL, *result = NULL;
852 851 const Py_ssize_t length = index_length(self);
853 852
854 853 if (!PyArg_ParseTuple(args, "iOi", &rev, &stoparg, &generaldelta)) {
855 854 return NULL;
856 855 }
857 856
858 857 if (PyInt_Check(stoparg)) {
859 858 stoprev = (int)PyInt_AsLong(stoparg);
860 859 if (stoprev == -1 && PyErr_Occurred()) {
861 860 return NULL;
862 861 }
863 862 }
864 863 else if (stoparg == Py_None) {
865 864 stoprev = -2;
866 865 }
867 866 else {
868 867 PyErr_SetString(PyExc_ValueError,
869 868 "stoprev must be integer or None");
870 869 return NULL;
871 870 }
872 871
873 872 if (rev < 0 || rev >= length - 1) {
874 873 PyErr_SetString(PyExc_ValueError, "revlog index out of range");
875 874 return NULL;
876 875 }
877 876
878 877 chain = PyList_New(0);
879 878 if (chain == NULL) {
880 879 return NULL;
881 880 }
882 881
883 882 baserev = index_baserev(self, rev);
884 883
885 884 /* This should never happen. */
886 885 if (baserev <= -2) {
887 886 /* Error should be set by index_deref() */
888 887 assert(PyErr_Occurred());
889 888 goto bail;
890 889 }
891 890
892 891 iterrev = rev;
893 892
894 893 while (iterrev != baserev && iterrev != stoprev) {
895 894 PyObject *value = PyInt_FromLong(iterrev);
896 895 if (value == NULL) {
897 896 goto bail;
898 897 }
899 898 if (PyList_Append(chain, value)) {
900 899 Py_DECREF(value);
901 900 goto bail;
902 901 }
903 902 Py_DECREF(value);
904 903
905 904 if (generaldelta) {
906 905 iterrev = baserev;
907 906 }
908 907 else {
909 908 iterrev--;
910 909 }
911 910
912 911 if (iterrev < 0) {
913 912 break;
914 913 }
915 914
916 915 if (iterrev >= length - 1) {
917 916 PyErr_SetString(PyExc_IndexError, "revision outside index");
918 917 return NULL;
919 918 }
920 919
921 920 baserev = index_baserev(self, iterrev);
922 921
923 922 /* This should never happen. */
924 923 if (baserev <= -2) {
925 924 /* Error should be set by index_deref() */
926 925 assert(PyErr_Occurred());
927 926 goto bail;
928 927 }
929 928 }
930 929
931 930 if (iterrev == stoprev) {
932 931 stopped = 1;
933 932 }
934 933 else {
935 934 PyObject *value = PyInt_FromLong(iterrev);
936 935 if (value == NULL) {
937 936 goto bail;
938 937 }
939 938 if (PyList_Append(chain, value)) {
940 939 Py_DECREF(value);
941 940 goto bail;
942 941 }
943 942 Py_DECREF(value);
944 943
945 944 stopped = 0;
946 945 }
947 946
948 947 if (PyList_Reverse(chain)) {
949 948 goto bail;
950 949 }
951 950
952 951 result = Py_BuildValue("OO", chain, stopped ? Py_True : Py_False);
953 952 Py_DECREF(chain);
954 953 return result;
955 954
956 955 bail:
957 956 Py_DECREF(chain);
958 957 return NULL;
959 958 }
960 959
961 960 static inline int nt_level(const char *node, Py_ssize_t level)
962 961 {
963 962 int v = node[level>>1];
964 963 if (!(level & 1))
965 964 v >>= 4;
966 965 return v & 0xf;
967 966 }
968 967
969 968 /*
970 969 * Return values:
971 970 *
972 971 * -4: match is ambiguous (multiple candidates)
973 972 * -2: not found
974 973 * rest: valid rev
975 974 */
976 975 static int nt_find(indexObject *self, const char *node, Py_ssize_t nodelen,
977 976 int hex)
978 977 {
979 978 int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level;
980 979 int level, maxlevel, off;
981 980
982 981 if (nodelen == 20 && node[0] == '\0' && memcmp(node, nullid, 20) == 0)
983 982 return -1;
984 983
985 984 if (self->nt == NULL)
986 985 return -2;
987 986
988 987 if (hex)
989 988 maxlevel = nodelen > 40 ? 40 : (int)nodelen;
990 989 else
991 990 maxlevel = nodelen > 20 ? 40 : ((int)nodelen * 2);
992 991
993 992 for (level = off = 0; level < maxlevel; level++) {
994 993 int k = getnybble(node, level);
995 994 nodetree *n = &self->nt[off];
996 995 int v = n->children[k];
997 996
998 997 if (v < 0) {
999 998 const char *n;
1000 999 Py_ssize_t i;
1001 1000
1002 1001 v = -(v + 1);
1003 1002 n = index_node(self, v);
1004 1003 if (n == NULL)
1005 1004 return -2;
1006 1005 for (i = level; i < maxlevel; i++)
1007 1006 if (getnybble(node, i) != nt_level(n, i))
1008 1007 return -2;
1009 1008 return v;
1010 1009 }
1011 1010 if (v == 0)
1012 1011 return -2;
1013 1012 off = v;
1014 1013 }
1015 1014 /* multiple matches against an ambiguous prefix */
1016 1015 return -4;
1017 1016 }
1018 1017
1019 1018 static int nt_new(indexObject *self)
1020 1019 {
1021 1020 if (self->ntlength == self->ntcapacity) {
1022 1021 if (self->ntcapacity >= INT_MAX / (sizeof(nodetree) * 2)) {
1023 1022 PyErr_SetString(PyExc_MemoryError,
1024 1023 "overflow in nt_new");
1025 1024 return -1;
1026 1025 }
1027 1026 self->ntcapacity *= 2;
1028 1027 self->nt = realloc(self->nt,
1029 1028 self->ntcapacity * sizeof(nodetree));
1030 1029 if (self->nt == NULL) {
1031 1030 PyErr_SetString(PyExc_MemoryError, "out of memory");
1032 1031 return -1;
1033 1032 }
1034 1033 memset(&self->nt[self->ntlength], 0,
1035 1034 sizeof(nodetree) * (self->ntcapacity - self->ntlength));
1036 1035 }
1037 1036 return self->ntlength++;
1038 1037 }
1039 1038
1040 1039 static int nt_insert(indexObject *self, const char *node, int rev)
1041 1040 {
1042 1041 int level = 0;
1043 1042 int off = 0;
1044 1043
1045 1044 while (level < 40) {
1046 1045 int k = nt_level(node, level);
1047 1046 nodetree *n;
1048 1047 int v;
1049 1048
1050 1049 n = &self->nt[off];
1051 1050 v = n->children[k];
1052 1051
1053 1052 if (v == 0) {
1054 1053 n->children[k] = -rev - 1;
1055 1054 return 0;
1056 1055 }
1057 1056 if (v < 0) {
1058 1057 const char *oldnode = index_node(self, -(v + 1));
1059 1058 int noff;
1060 1059
1061 1060 if (!oldnode || !memcmp(oldnode, node, 20)) {
1062 1061 n->children[k] = -rev - 1;
1063 1062 return 0;
1064 1063 }
1065 1064 noff = nt_new(self);
1066 1065 if (noff == -1)
1067 1066 return -1;
1068 1067 /* self->nt may have been changed by realloc */
1069 1068 self->nt[off].children[k] = noff;
1070 1069 off = noff;
1071 1070 n = &self->nt[off];
1072 1071 n->children[nt_level(oldnode, ++level)] = v;
1073 1072 if (level > self->ntdepth)
1074 1073 self->ntdepth = level;
1075 1074 self->ntsplits += 1;
1076 1075 } else {
1077 1076 level += 1;
1078 1077 off = v;
1079 1078 }
1080 1079 }
1081 1080
1082 1081 return -1;
1083 1082 }
1084 1083
1085 1084 static int nt_init(indexObject *self)
1086 1085 {
1087 1086 if (self->nt == NULL) {
1088 1087 if ((size_t)self->raw_length > INT_MAX / sizeof(nodetree)) {
1089 1088 PyErr_SetString(PyExc_ValueError, "overflow in nt_init");
1090 1089 return -1;
1091 1090 }
1092 1091 self->ntcapacity = self->raw_length < 4
1093 1092 ? 4 : (int)self->raw_length / 2;
1094 1093
1095 1094 self->nt = calloc(self->ntcapacity, sizeof(nodetree));
1096 1095 if (self->nt == NULL) {
1097 1096 PyErr_NoMemory();
1098 1097 return -1;
1099 1098 }
1100 1099 self->ntlength = 1;
1101 1100 self->ntrev = (int)index_length(self) - 1;
1102 1101 self->ntlookups = 1;
1103 1102 self->ntmisses = 0;
1104 1103 if (nt_insert(self, nullid, INT_MAX) == -1)
1105 1104 return -1;
1106 1105 }
1107 1106 return 0;
1108 1107 }
1109 1108
1110 1109 /*
1111 1110 * Return values:
1112 1111 *
1113 1112 * -3: error (exception set)
1114 1113 * -2: not found (no exception set)
1115 1114 * rest: valid rev
1116 1115 */
1117 1116 static int index_find_node(indexObject *self,
1118 1117 const char *node, Py_ssize_t nodelen)
1119 1118 {
1120 1119 int rev;
1121 1120
1122 1121 self->ntlookups++;
1123 1122 rev = nt_find(self, node, nodelen, 0);
1124 1123 if (rev >= -1)
1125 1124 return rev;
1126 1125
1127 1126 if (nt_init(self) == -1)
1128 1127 return -3;
1129 1128
1130 1129 /*
1131 1130 * For the first handful of lookups, we scan the entire index,
1132 1131 * and cache only the matching nodes. This optimizes for cases
1133 1132 * like "hg tip", where only a few nodes are accessed.
1134 1133 *
1135 1134 * After that, we cache every node we visit, using a single
1136 1135 * scan amortized over multiple lookups. This gives the best
1137 1136 * bulk performance, e.g. for "hg log".
1138 1137 */
1139 1138 if (self->ntmisses++ < 4) {
1140 1139 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1141 1140 const char *n = index_node(self, rev);
1142 1141 if (n == NULL)
1143 1142 return -2;
1144 1143 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1145 1144 if (nt_insert(self, n, rev) == -1)
1146 1145 return -3;
1147 1146 break;
1148 1147 }
1149 1148 }
1150 1149 } else {
1151 1150 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1152 1151 const char *n = index_node(self, rev);
1153 1152 if (n == NULL) {
1154 1153 self->ntrev = rev + 1;
1155 1154 return -2;
1156 1155 }
1157 1156 if (nt_insert(self, n, rev) == -1) {
1158 1157 self->ntrev = rev + 1;
1159 1158 return -3;
1160 1159 }
1161 1160 if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
1162 1161 break;
1163 1162 }
1164 1163 }
1165 1164 self->ntrev = rev;
1166 1165 }
1167 1166
1168 1167 if (rev >= 0)
1169 1168 return rev;
1170 1169 return -2;
1171 1170 }
1172 1171
1173 1172 static void raise_revlog_error(void)
1174 1173 {
1175 1174 PyObject *mod = NULL, *dict = NULL, *errclass = NULL;
1176 1175
1177 1176 mod = PyImport_ImportModule("mercurial.error");
1178 1177 if (mod == NULL) {
1179 1178 goto cleanup;
1180 1179 }
1181 1180
1182 1181 dict = PyModule_GetDict(mod);
1183 1182 if (dict == NULL) {
1184 1183 goto cleanup;
1185 1184 }
1186 1185 Py_INCREF(dict);
1187 1186
1188 1187 errclass = PyDict_GetItemString(dict, "RevlogError");
1189 1188 if (errclass == NULL) {
1190 1189 PyErr_SetString(PyExc_SystemError,
1191 1190 "could not find RevlogError");
1192 1191 goto cleanup;
1193 1192 }
1194 1193
1195 1194 /* value of exception is ignored by callers */
1196 1195 PyErr_SetString(errclass, "RevlogError");
1197 1196
1198 1197 cleanup:
1199 1198 Py_XDECREF(dict);
1200 1199 Py_XDECREF(mod);
1201 1200 }
1202 1201
1203 1202 static PyObject *index_getitem(indexObject *self, PyObject *value)
1204 1203 {
1205 1204 char *node;
1206 1205 Py_ssize_t nodelen;
1207 1206 int rev;
1208 1207
1209 1208 if (PyInt_Check(value))
1210 1209 return index_get(self, PyInt_AS_LONG(value));
1211 1210
1212 1211 if (node_check(value, &node, &nodelen) == -1)
1213 1212 return NULL;
1214 1213 rev = index_find_node(self, node, nodelen);
1215 1214 if (rev >= -1)
1216 1215 return PyInt_FromLong(rev);
1217 1216 if (rev == -2)
1218 1217 raise_revlog_error();
1219 1218 return NULL;
1220 1219 }
1221 1220
1222 1221 static int nt_partialmatch(indexObject *self, const char *node,
1223 1222 Py_ssize_t nodelen)
1224 1223 {
1225 1224 int rev;
1226 1225
1227 1226 if (nt_init(self) == -1)
1228 1227 return -3;
1229 1228
1230 1229 if (self->ntrev > 0) {
1231 1230 /* ensure that the radix tree is fully populated */
1232 1231 for (rev = self->ntrev - 1; rev >= 0; rev--) {
1233 1232 const char *n = index_node(self, rev);
1234 1233 if (n == NULL)
1235 1234 return -2;
1236 1235 if (nt_insert(self, n, rev) == -1)
1237 1236 return -3;
1238 1237 }
1239 1238 self->ntrev = rev;
1240 1239 }
1241 1240
1242 1241 return nt_find(self, node, nodelen, 1);
1243 1242 }
1244 1243
1245 1244 static PyObject *index_partialmatch(indexObject *self, PyObject *args)
1246 1245 {
1247 1246 const char *fullnode;
1248 1247 int nodelen;
1249 1248 char *node;
1250 1249 int rev, i;
1251 1250
1252 1251 if (!PyArg_ParseTuple(args, "s#", &node, &nodelen))
1253 1252 return NULL;
1254 1253
1255 1254 if (nodelen < 4) {
1256 1255 PyErr_SetString(PyExc_ValueError, "key too short");
1257 1256 return NULL;
1258 1257 }
1259 1258
1260 1259 if (nodelen > 40) {
1261 1260 PyErr_SetString(PyExc_ValueError, "key too long");
1262 1261 return NULL;
1263 1262 }
1264 1263
1265 1264 for (i = 0; i < nodelen; i++)
1266 1265 hexdigit(node, i);
1267 1266 if (PyErr_Occurred()) {
1268 1267 /* input contains non-hex characters */
1269 1268 PyErr_Clear();
1270 1269 Py_RETURN_NONE;
1271 1270 }
1272 1271
1273 1272 rev = nt_partialmatch(self, node, nodelen);
1274 1273
1275 1274 switch (rev) {
1276 1275 case -4:
1277 1276 raise_revlog_error();
1278 1277 case -3:
1279 1278 return NULL;
1280 1279 case -2:
1281 1280 Py_RETURN_NONE;
1282 1281 case -1:
1283 1282 return PyBytes_FromStringAndSize(nullid, 20);
1284 1283 }
1285 1284
1286 1285 fullnode = index_node(self, rev);
1287 1286 if (fullnode == NULL) {
1288 1287 PyErr_Format(PyExc_IndexError,
1289 1288 "could not access rev %d", rev);
1290 1289 return NULL;
1291 1290 }
1292 1291 return PyBytes_FromStringAndSize(fullnode, 20);
1293 1292 }
1294 1293
1295 1294 static PyObject *index_m_get(indexObject *self, PyObject *args)
1296 1295 {
1297 1296 Py_ssize_t nodelen;
1298 1297 PyObject *val;
1299 1298 char *node;
1300 1299 int rev;
1301 1300
1302 1301 if (!PyArg_ParseTuple(args, "O", &val))
1303 1302 return NULL;
1304 1303 if (node_check(val, &node, &nodelen) == -1)
1305 1304 return NULL;
1306 1305 rev = index_find_node(self, node, nodelen);
1307 1306 if (rev == -3)
1308 1307 return NULL;
1309 1308 if (rev == -2)
1310 1309 Py_RETURN_NONE;
1311 1310 return PyInt_FromLong(rev);
1312 1311 }
1313 1312
1314 1313 static int index_contains(indexObject *self, PyObject *value)
1315 1314 {
1316 1315 char *node;
1317 1316 Py_ssize_t nodelen;
1318 1317
1319 1318 if (PyInt_Check(value)) {
1320 1319 long rev = PyInt_AS_LONG(value);
1321 1320 return rev >= -1 && rev < index_length(self);
1322 1321 }
1323 1322
1324 1323 if (node_check(value, &node, &nodelen) == -1)
1325 1324 return -1;
1326 1325
1327 1326 switch (index_find_node(self, node, nodelen)) {
1328 1327 case -3:
1329 1328 return -1;
1330 1329 case -2:
1331 1330 return 0;
1332 1331 default:
1333 1332 return 1;
1334 1333 }
1335 1334 }
1336 1335
1337 1336 typedef uint64_t bitmask;
1338 1337
1339 1338 /*
1340 1339 * Given a disjoint set of revs, return all candidates for the
1341 1340 * greatest common ancestor. In revset notation, this is the set
1342 1341 * "heads(::a and ::b and ...)"
1343 1342 */
1344 1343 static PyObject *find_gca_candidates(indexObject *self, const int *revs,
1345 1344 int revcount)
1346 1345 {
1347 1346 const bitmask allseen = (1ull << revcount) - 1;
1348 1347 const bitmask poison = 1ull << revcount;
1349 1348 PyObject *gca = PyList_New(0);
1350 1349 int i, v, interesting;
1351 1350 int maxrev = -1;
1352 1351 bitmask sp;
1353 1352 bitmask *seen;
1354 1353
1355 1354 if (gca == NULL)
1356 1355 return PyErr_NoMemory();
1357 1356
1358 1357 for (i = 0; i < revcount; i++) {
1359 1358 if (revs[i] > maxrev)
1360 1359 maxrev = revs[i];
1361 1360 }
1362 1361
1363 1362 seen = calloc(sizeof(*seen), maxrev + 1);
1364 1363 if (seen == NULL) {
1365 1364 Py_DECREF(gca);
1366 1365 return PyErr_NoMemory();
1367 1366 }
1368 1367
1369 1368 for (i = 0; i < revcount; i++)
1370 1369 seen[revs[i]] = 1ull << i;
1371 1370
1372 1371 interesting = revcount;
1373 1372
1374 1373 for (v = maxrev; v >= 0 && interesting; v--) {
1375 1374 bitmask sv = seen[v];
1376 1375 int parents[2];
1377 1376
1378 1377 if (!sv)
1379 1378 continue;
1380 1379
1381 1380 if (sv < poison) {
1382 1381 interesting -= 1;
1383 1382 if (sv == allseen) {
1384 1383 PyObject *obj = PyInt_FromLong(v);
1385 1384 if (obj == NULL)
1386 1385 goto bail;
1387 1386 if (PyList_Append(gca, obj) == -1) {
1388 1387 Py_DECREF(obj);
1389 1388 goto bail;
1390 1389 }
1391 1390 sv |= poison;
1392 1391 for (i = 0; i < revcount; i++) {
1393 1392 if (revs[i] == v)
1394 1393 goto done;
1395 1394 }
1396 1395 }
1397 1396 }
1398 1397 if (index_get_parents(self, v, parents, maxrev) < 0)
1399 1398 goto bail;
1400 1399
1401 1400 for (i = 0; i < 2; i++) {
1402 1401 int p = parents[i];
1403 1402 if (p == -1)
1404 1403 continue;
1405 1404 sp = seen[p];
1406 1405 if (sv < poison) {
1407 1406 if (sp == 0) {
1408 1407 seen[p] = sv;
1409 1408 interesting++;
1410 1409 }
1411 1410 else if (sp != sv)
1412 1411 seen[p] |= sv;
1413 1412 } else {
1414 1413 if (sp && sp < poison)
1415 1414 interesting--;
1416 1415 seen[p] = sv;
1417 1416 }
1418 1417 }
1419 1418 }
1420 1419
1421 1420 done:
1422 1421 free(seen);
1423 1422 return gca;
1424 1423 bail:
1425 1424 free(seen);
1426 1425 Py_XDECREF(gca);
1427 1426 return NULL;
1428 1427 }
1429 1428
1430 1429 /*
1431 1430 * Given a disjoint set of revs, return the subset with the longest
1432 1431 * path to the root.
1433 1432 */
1434 1433 static PyObject *find_deepest(indexObject *self, PyObject *revs)
1435 1434 {
1436 1435 const Py_ssize_t revcount = PyList_GET_SIZE(revs);
1437 1436 static const Py_ssize_t capacity = 24;
1438 1437 int *depth, *interesting = NULL;
1439 1438 int i, j, v, ninteresting;
1440 1439 PyObject *dict = NULL, *keys = NULL;
1441 1440 long *seen = NULL;
1442 1441 int maxrev = -1;
1443 1442 long final;
1444 1443
1445 1444 if (revcount > capacity) {
1446 1445 PyErr_Format(PyExc_OverflowError,
1447 1446 "bitset size (%ld) > capacity (%ld)",
1448 1447 (long)revcount, (long)capacity);
1449 1448 return NULL;
1450 1449 }
1451 1450
1452 1451 for (i = 0; i < revcount; i++) {
1453 1452 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1454 1453 if (n > maxrev)
1455 1454 maxrev = n;
1456 1455 }
1457 1456
1458 1457 depth = calloc(sizeof(*depth), maxrev + 1);
1459 1458 if (depth == NULL)
1460 1459 return PyErr_NoMemory();
1461 1460
1462 1461 seen = calloc(sizeof(*seen), maxrev + 1);
1463 1462 if (seen == NULL) {
1464 1463 PyErr_NoMemory();
1465 1464 goto bail;
1466 1465 }
1467 1466
1468 1467 interesting = calloc(sizeof(*interesting), 1 << revcount);
1469 1468 if (interesting == NULL) {
1470 1469 PyErr_NoMemory();
1471 1470 goto bail;
1472 1471 }
1473 1472
1474 1473 if (PyList_Sort(revs) == -1)
1475 1474 goto bail;
1476 1475
1477 1476 for (i = 0; i < revcount; i++) {
1478 1477 int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
1479 1478 long b = 1l << i;
1480 1479 depth[n] = 1;
1481 1480 seen[n] = b;
1482 1481 interesting[b] = 1;
1483 1482 }
1484 1483
1485 1484 /* invariant: ninteresting is the number of non-zero entries in
1486 1485 * interesting. */
1487 1486 ninteresting = (int)revcount;
1488 1487
1489 1488 for (v = maxrev; v >= 0 && ninteresting > 1; v--) {
1490 1489 int dv = depth[v];
1491 1490 int parents[2];
1492 1491 long sv;
1493 1492
1494 1493 if (dv == 0)
1495 1494 continue;
1496 1495
1497 1496 sv = seen[v];
1498 1497 if (index_get_parents(self, v, parents, maxrev) < 0)
1499 1498 goto bail;
1500 1499
1501 1500 for (i = 0; i < 2; i++) {
1502 1501 int p = parents[i];
1503 1502 long sp;
1504 1503 int dp;
1505 1504
1506 1505 if (p == -1)
1507 1506 continue;
1508 1507
1509 1508 dp = depth[p];
1510 1509 sp = seen[p];
1511 1510 if (dp <= dv) {
1512 1511 depth[p] = dv + 1;
1513 1512 if (sp != sv) {
1514 1513 interesting[sv] += 1;
1515 1514 seen[p] = sv;
1516 1515 if (sp) {
1517 1516 interesting[sp] -= 1;
1518 1517 if (interesting[sp] == 0)
1519 1518 ninteresting -= 1;
1520 1519 }
1521 1520 }
1522 1521 }
1523 1522 else if (dv == dp - 1) {
1524 1523 long nsp = sp | sv;
1525 1524 if (nsp == sp)
1526 1525 continue;
1527 1526 seen[p] = nsp;
1528 1527 interesting[sp] -= 1;
1529 1528 if (interesting[sp] == 0)
1530 1529 ninteresting -= 1;
1531 1530 if (interesting[nsp] == 0)
1532 1531 ninteresting += 1;
1533 1532 interesting[nsp] += 1;
1534 1533 }
1535 1534 }
1536 1535 interesting[sv] -= 1;
1537 1536 if (interesting[sv] == 0)
1538 1537 ninteresting -= 1;
1539 1538 }
1540 1539
1541 1540 final = 0;
1542 1541 j = ninteresting;
1543 1542 for (i = 0; i < (int)(2 << revcount) && j > 0; i++) {
1544 1543 if (interesting[i] == 0)
1545 1544 continue;
1546 1545 final |= i;
1547 1546 j -= 1;
1548 1547 }
1549 1548 if (final == 0) {
1550 1549 keys = PyList_New(0);
1551 1550 goto bail;
1552 1551 }
1553 1552
1554 1553 dict = PyDict_New();
1555 1554 if (dict == NULL)
1556 1555 goto bail;
1557 1556
1558 1557 for (i = 0; i < revcount; i++) {
1559 1558 PyObject *key;
1560 1559
1561 1560 if ((final & (1 << i)) == 0)
1562 1561 continue;
1563 1562
1564 1563 key = PyList_GET_ITEM(revs, i);
1565 1564 Py_INCREF(key);
1566 1565 Py_INCREF(Py_None);
1567 1566 if (PyDict_SetItem(dict, key, Py_None) == -1) {
1568 1567 Py_DECREF(key);
1569 1568 Py_DECREF(Py_None);
1570 1569 goto bail;
1571 1570 }
1572 1571 }
1573 1572
1574 1573 keys = PyDict_Keys(dict);
1575 1574
1576 1575 bail:
1577 1576 free(depth);
1578 1577 free(seen);
1579 1578 free(interesting);
1580 1579 Py_XDECREF(dict);
1581 1580
1582 1581 return keys;
1583 1582 }
1584 1583
1585 1584 /*
1586 1585 * Given a (possibly overlapping) set of revs, return all the
1587 1586 * common ancestors heads: heads(::args[0] and ::a[1] and ...)
1588 1587 */
1589 1588 static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args)
1590 1589 {
1591 1590 PyObject *ret = NULL;
1592 1591 Py_ssize_t argcount, i, len;
1593 1592 bitmask repeat = 0;
1594 1593 int revcount = 0;
1595 1594 int *revs;
1596 1595
1597 1596 argcount = PySequence_Length(args);
1598 1597 revs = PyMem_Malloc(argcount * sizeof(*revs));
1599 1598 if (argcount > 0 && revs == NULL)
1600 1599 return PyErr_NoMemory();
1601 1600 len = index_length(self) - 1;
1602 1601
1603 1602 for (i = 0; i < argcount; i++) {
1604 1603 static const int capacity = 24;
1605 1604 PyObject *obj = PySequence_GetItem(args, i);
1606 1605 bitmask x;
1607 1606 long val;
1608 1607
1609 1608 if (!PyInt_Check(obj)) {
1610 1609 PyErr_SetString(PyExc_TypeError,
1611 1610 "arguments must all be ints");
1612 1611 Py_DECREF(obj);
1613 1612 goto bail;
1614 1613 }
1615 1614 val = PyInt_AsLong(obj);
1616 1615 Py_DECREF(obj);
1617 1616 if (val == -1) {
1618 1617 ret = PyList_New(0);
1619 1618 goto done;
1620 1619 }
1621 1620 if (val < 0 || val >= len) {
1622 1621 PyErr_SetString(PyExc_IndexError,
1623 1622 "index out of range");
1624 1623 goto bail;
1625 1624 }
1626 1625 /* this cheesy bloom filter lets us avoid some more
1627 1626 * expensive duplicate checks in the common set-is-disjoint
1628 1627 * case */
1629 1628 x = 1ull << (val & 0x3f);
1630 1629 if (repeat & x) {
1631 1630 int k;
1632 1631 for (k = 0; k < revcount; k++) {
1633 1632 if (val == revs[k])
1634 1633 goto duplicate;
1635 1634 }
1636 1635 }
1637 1636 else repeat |= x;
1638 1637 if (revcount >= capacity) {
1639 1638 PyErr_Format(PyExc_OverflowError,
1640 1639 "bitset size (%d) > capacity (%d)",
1641 1640 revcount, capacity);
1642 1641 goto bail;
1643 1642 }
1644 1643 revs[revcount++] = (int)val;
1645 1644 duplicate:;
1646 1645 }
1647 1646
1648 1647 if (revcount == 0) {
1649 1648 ret = PyList_New(0);
1650 1649 goto done;
1651 1650 }
1652 1651 if (revcount == 1) {
1653 1652 PyObject *obj;
1654 1653 ret = PyList_New(1);
1655 1654 if (ret == NULL)
1656 1655 goto bail;
1657 1656 obj = PyInt_FromLong(revs[0]);
1658 1657 if (obj == NULL)
1659 1658 goto bail;
1660 1659 PyList_SET_ITEM(ret, 0, obj);
1661 1660 goto done;
1662 1661 }
1663 1662
1664 1663 ret = find_gca_candidates(self, revs, revcount);
1665 1664 if (ret == NULL)
1666 1665 goto bail;
1667 1666
1668 1667 done:
1669 1668 PyMem_Free(revs);
1670 1669 return ret;
1671 1670
1672 1671 bail:
1673 1672 PyMem_Free(revs);
1674 1673 Py_XDECREF(ret);
1675 1674 return NULL;
1676 1675 }
1677 1676
1678 1677 /*
1679 1678 * Given a (possibly overlapping) set of revs, return the greatest
1680 1679 * common ancestors: those with the longest path to the root.
1681 1680 */
1682 1681 static PyObject *index_ancestors(indexObject *self, PyObject *args)
1683 1682 {
1684 1683 PyObject *ret;
1685 1684 PyObject *gca = index_commonancestorsheads(self, args);
1686 1685 if (gca == NULL)
1687 1686 return NULL;
1688 1687
1689 1688 if (PyList_GET_SIZE(gca) <= 1) {
1690 1689 return gca;
1691 1690 }
1692 1691
1693 1692 ret = find_deepest(self, gca);
1694 1693 Py_DECREF(gca);
1695 1694 return ret;
1696 1695 }
1697 1696
1698 1697 /*
1699 1698 * Invalidate any trie entries introduced by added revs.
1700 1699 */
1701 1700 static void nt_invalidate_added(indexObject *self, Py_ssize_t start)
1702 1701 {
1703 1702 Py_ssize_t i, len = PyList_GET_SIZE(self->added);
1704 1703
1705 1704 for (i = start; i < len; i++) {
1706 1705 PyObject *tuple = PyList_GET_ITEM(self->added, i);
1707 1706 PyObject *node = PyTuple_GET_ITEM(tuple, 7);
1708 1707
1709 1708 nt_insert(self, PyBytes_AS_STRING(node), -1);
1710 1709 }
1711 1710
1712 1711 if (start == 0)
1713 1712 Py_CLEAR(self->added);
1714 1713 }
1715 1714
1716 1715 /*
1717 1716 * Delete a numeric range of revs, which must be at the end of the
1718 1717 * range, but exclude the sentinel nullid entry.
1719 1718 */
1720 1719 static int index_slice_del(indexObject *self, PyObject *item)
1721 1720 {
1722 1721 Py_ssize_t start, stop, step, slicelength;
1723 1722 Py_ssize_t length = index_length(self);
1724 1723 int ret = 0;
1725 1724
1726 1725 /* Argument changed from PySliceObject* to PyObject* in Python 3. */
1727 1726 #ifdef IS_PY3K
1728 1727 if (PySlice_GetIndicesEx(item, length,
1729 1728 #else
1730 1729 if (PySlice_GetIndicesEx((PySliceObject*)item, length,
1731 1730 #endif
1732 1731 &start, &stop, &step, &slicelength) < 0)
1733 1732 return -1;
1734 1733
1735 1734 if (slicelength <= 0)
1736 1735 return 0;
1737 1736
1738 1737 if ((step < 0 && start < stop) || (step > 0 && start > stop))
1739 1738 stop = start;
1740 1739
1741 1740 if (step < 0) {
1742 1741 stop = start + 1;
1743 1742 start = stop + step*(slicelength - 1) - 1;
1744 1743 step = -step;
1745 1744 }
1746 1745
1747 1746 if (step != 1) {
1748 1747 PyErr_SetString(PyExc_ValueError,
1749 1748 "revlog index delete requires step size of 1");
1750 1749 return -1;
1751 1750 }
1752 1751
1753 1752 if (stop != length - 1) {
1754 1753 PyErr_SetString(PyExc_IndexError,
1755 1754 "revlog index deletion indices are invalid");
1756 1755 return -1;
1757 1756 }
1758 1757
1759 1758 if (start < self->length - 1) {
1760 1759 if (self->nt) {
1761 1760 Py_ssize_t i;
1762 1761
1763 1762 for (i = start + 1; i < self->length - 1; i++) {
1764 1763 const char *node = index_node(self, i);
1765 1764
1766 1765 if (node)
1767 1766 nt_insert(self, node, -1);
1768 1767 }
1769 1768 if (self->added)
1770 1769 nt_invalidate_added(self, 0);
1771 1770 if (self->ntrev > start)
1772 1771 self->ntrev = (int)start;
1773 1772 }
1774 1773 self->length = start + 1;
1775 1774 if (start < self->raw_length) {
1776 1775 if (self->cache) {
1777 1776 Py_ssize_t i;
1778 1777 for (i = start; i < self->raw_length; i++)
1779 1778 Py_CLEAR(self->cache[i]);
1780 1779 }
1781 1780 self->raw_length = start;
1782 1781 }
1783 1782 goto done;
1784 1783 }
1785 1784
1786 1785 if (self->nt) {
1787 1786 nt_invalidate_added(self, start - self->length + 1);
1788 1787 if (self->ntrev > start)
1789 1788 self->ntrev = (int)start;
1790 1789 }
1791 1790 if (self->added)
1792 1791 ret = PyList_SetSlice(self->added, start - self->length + 1,
1793 1792 PyList_GET_SIZE(self->added), NULL);
1794 1793 done:
1795 1794 Py_CLEAR(self->headrevs);
1796 1795 return ret;
1797 1796 }
1798 1797
1799 1798 /*
1800 1799 * Supported ops:
1801 1800 *
1802 1801 * slice deletion
1803 1802 * string assignment (extend node->rev mapping)
1804 1803 * string deletion (shrink node->rev mapping)
1805 1804 */
1806 1805 static int index_assign_subscript(indexObject *self, PyObject *item,
1807 1806 PyObject *value)
1808 1807 {
1809 1808 char *node;
1810 1809 Py_ssize_t nodelen;
1811 1810 long rev;
1812 1811
1813 1812 if (PySlice_Check(item) && value == NULL)
1814 1813 return index_slice_del(self, item);
1815 1814
1816 1815 if (node_check(item, &node, &nodelen) == -1)
1817 1816 return -1;
1818 1817
1819 1818 if (value == NULL)
1820 1819 return self->nt ? nt_insert(self, node, -1) : 0;
1821 1820 rev = PyInt_AsLong(value);
1822 1821 if (rev > INT_MAX || rev < 0) {
1823 1822 if (!PyErr_Occurred())
1824 1823 PyErr_SetString(PyExc_ValueError, "rev out of range");
1825 1824 return -1;
1826 1825 }
1827 1826
1828 1827 if (nt_init(self) == -1)
1829 1828 return -1;
1830 1829 return nt_insert(self, node, (int)rev);
1831 1830 }
1832 1831
1833 1832 /*
1834 1833 * Find all RevlogNG entries in an index that has inline data. Update
1835 1834 * the optional "offsets" table with those entries.
1836 1835 */
1837 1836 static Py_ssize_t inline_scan(indexObject *self, const char **offsets)
1838 1837 {
1839 1838 const char *data = (const char *)self->buf.buf;
1840 1839 Py_ssize_t pos = 0;
1841 1840 Py_ssize_t end = self->buf.len;
1842 1841 long incr = v1_hdrsize;
1843 1842 Py_ssize_t len = 0;
1844 1843
1845 1844 while (pos + v1_hdrsize <= end && pos >= 0) {
1846 1845 uint32_t comp_len;
1847 1846 /* 3rd element of header is length of compressed inline data */
1848 1847 comp_len = getbe32(data + pos + 8);
1849 1848 incr = v1_hdrsize + comp_len;
1850 1849 if (offsets)
1851 1850 offsets[len] = data + pos;
1852 1851 len++;
1853 1852 pos += incr;
1854 1853 }
1855 1854
1856 1855 if (pos != end) {
1857 1856 if (!PyErr_Occurred())
1858 1857 PyErr_SetString(PyExc_ValueError, "corrupt index file");
1859 1858 return -1;
1860 1859 }
1861 1860
1862 1861 return len;
1863 1862 }
1864 1863
1865 1864 static int index_init(indexObject *self, PyObject *args)
1866 1865 {
1867 1866 PyObject *data_obj, *inlined_obj;
1868 1867 Py_ssize_t size;
1869 1868
1870 1869 /* Initialize before argument-checking to avoid index_dealloc() crash. */
1871 1870 self->raw_length = 0;
1872 1871 self->added = NULL;
1873 1872 self->cache = NULL;
1874 1873 self->data = NULL;
1875 1874 memset(&self->buf, 0, sizeof(self->buf));
1876 1875 self->headrevs = NULL;
1877 1876 self->filteredrevs = Py_None;
1878 1877 Py_INCREF(Py_None);
1879 1878 self->nt = NULL;
1880 1879 self->offsets = NULL;
1881 1880
1882 1881 if (!PyArg_ParseTuple(args, "OO", &data_obj, &inlined_obj))
1883 1882 return -1;
1884 1883 if (!PyObject_CheckBuffer(data_obj)) {
1885 1884 PyErr_SetString(PyExc_TypeError,
1886 1885 "data does not support buffer interface");
1887 1886 return -1;
1888 1887 }
1889 1888
1890 1889 if (PyObject_GetBuffer(data_obj, &self->buf, PyBUF_SIMPLE) == -1)
1891 1890 return -1;
1892 1891 size = self->buf.len;
1893 1892
1894 1893 self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj);
1895 1894 self->data = data_obj;
1896 1895
1897 1896 self->ntlength = self->ntcapacity = 0;
1898 1897 self->ntdepth = self->ntsplits = 0;
1899 1898 self->ntlookups = self->ntmisses = 0;
1900 1899 self->ntrev = -1;
1901 1900 Py_INCREF(self->data);
1902 1901
1903 1902 if (self->inlined) {
1904 1903 Py_ssize_t len = inline_scan(self, NULL);
1905 1904 if (len == -1)
1906 1905 goto bail;
1907 1906 self->raw_length = len;
1908 1907 self->length = len + 1;
1909 1908 } else {
1910 1909 if (size % v1_hdrsize) {
1911 1910 PyErr_SetString(PyExc_ValueError, "corrupt index file");
1912 1911 goto bail;
1913 1912 }
1914 1913 self->raw_length = size / v1_hdrsize;
1915 1914 self->length = self->raw_length + 1;
1916 1915 }
1917 1916
1918 1917 return 0;
1919 1918 bail:
1920 1919 return -1;
1921 1920 }
1922 1921
1923 1922 static PyObject *index_nodemap(indexObject *self)
1924 1923 {
1925 1924 Py_INCREF(self);
1926 1925 return (PyObject *)self;
1927 1926 }
1928 1927
1929 1928 static void index_dealloc(indexObject *self)
1930 1929 {
1931 1930 _index_clearcaches(self);
1932 1931 Py_XDECREF(self->filteredrevs);
1933 1932 if (self->buf.buf) {
1934 1933 PyBuffer_Release(&self->buf);
1935 1934 memset(&self->buf, 0, sizeof(self->buf));
1936 1935 }
1937 1936 Py_XDECREF(self->data);
1938 1937 Py_XDECREF(self->added);
1939 1938 PyObject_Del(self);
1940 1939 }
1941 1940
1942 1941 static PySequenceMethods index_sequence_methods = {
1943 1942 (lenfunc)index_length, /* sq_length */
1944 1943 0, /* sq_concat */
1945 1944 0, /* sq_repeat */
1946 1945 (ssizeargfunc)index_get, /* sq_item */
1947 1946 0, /* sq_slice */
1948 1947 0, /* sq_ass_item */
1949 1948 0, /* sq_ass_slice */
1950 1949 (objobjproc)index_contains, /* sq_contains */
1951 1950 };
1952 1951
1953 1952 static PyMappingMethods index_mapping_methods = {
1954 1953 (lenfunc)index_length, /* mp_length */
1955 1954 (binaryfunc)index_getitem, /* mp_subscript */
1956 1955 (objobjargproc)index_assign_subscript, /* mp_ass_subscript */
1957 1956 };
1958 1957
1959 1958 static PyMethodDef index_methods[] = {
1960 1959 {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS,
1961 1960 "return the gca set of the given revs"},
1962 1961 {"commonancestorsheads", (PyCFunction)index_commonancestorsheads,
1963 1962 METH_VARARGS,
1964 1963 "return the heads of the common ancestors of the given revs"},
1965 1964 {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS,
1966 1965 "clear the index caches"},
1967 1966 {"get", (PyCFunction)index_m_get, METH_VARARGS,
1968 1967 "get an index entry"},
1969 1968 {"computephasesmapsets", (PyCFunction)compute_phases_map_sets,
1970 1969 METH_VARARGS, "compute phases"},
1971 1970 {"reachableroots2", (PyCFunction)reachableroots2, METH_VARARGS,
1972 1971 "reachableroots"},
1973 1972 {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS,
1974 1973 "get head revisions"}, /* Can do filtering since 3.2 */
1975 1974 {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS,
1976 1975 "get filtered head revisions"}, /* Can always do filtering */
1977 1976 {"deltachain", (PyCFunction)index_deltachain, METH_VARARGS,
1978 1977 "determine revisions with deltas to reconstruct fulltext"},
1979 1978 {"insert", (PyCFunction)index_insert, METH_VARARGS,
1980 1979 "insert an index entry"},
1981 1980 {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS,
1982 1981 "match a potentially ambiguous node ID"},
1983 1982 {"stats", (PyCFunction)index_stats, METH_NOARGS,
1984 1983 "stats for the index"},
1985 1984 {NULL} /* Sentinel */
1986 1985 };
1987 1986
1988 1987 static PyGetSetDef index_getset[] = {
1989 1988 {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL},
1990 1989 {NULL} /* Sentinel */
1991 1990 };
1992 1991
1993 1992 static PyTypeObject indexType = {
1994 1993 PyVarObject_HEAD_INIT(NULL, 0)
1995 1994 "parsers.index", /* tp_name */
1996 1995 sizeof(indexObject), /* tp_basicsize */
1997 1996 0, /* tp_itemsize */
1998 1997 (destructor)index_dealloc, /* tp_dealloc */
1999 1998 0, /* tp_print */
2000 1999 0, /* tp_getattr */
2001 2000 0, /* tp_setattr */
2002 2001 0, /* tp_compare */
2003 2002 0, /* tp_repr */
2004 2003 0, /* tp_as_number */
2005 2004 &index_sequence_methods, /* tp_as_sequence */
2006 2005 &index_mapping_methods, /* tp_as_mapping */
2007 2006 0, /* tp_hash */
2008 2007 0, /* tp_call */
2009 2008 0, /* tp_str */
2010 2009 0, /* tp_getattro */
2011 2010 0, /* tp_setattro */
2012 2011 0, /* tp_as_buffer */
2013 2012 Py_TPFLAGS_DEFAULT, /* tp_flags */
2014 2013 "revlog index", /* tp_doc */
2015 2014 0, /* tp_traverse */
2016 2015 0, /* tp_clear */
2017 2016 0, /* tp_richcompare */
2018 2017 0, /* tp_weaklistoffset */
2019 2018 0, /* tp_iter */
2020 2019 0, /* tp_iternext */
2021 2020 index_methods, /* tp_methods */
2022 2021 0, /* tp_members */
2023 2022 index_getset, /* tp_getset */
2024 2023 0, /* tp_base */
2025 2024 0, /* tp_dict */
2026 2025 0, /* tp_descr_get */
2027 2026 0, /* tp_descr_set */
2028 2027 0, /* tp_dictoffset */
2029 2028 (initproc)index_init, /* tp_init */
2030 2029 0, /* tp_alloc */
2031 2030 };
2032 2031
2033 2032 /*
2034 2033 * returns a tuple of the form (index, index, cache) with elements as
2035 2034 * follows:
2036 2035 *
2037 2036 * index: an index object that lazily parses RevlogNG records
2038 2037 * cache: if data is inlined, a tuple (0, index_file_content), else None
2039 2038 * index_file_content could be a string, or a buffer
2040 2039 *
2041 2040 * added complications are for backwards compatibility
2042 2041 */
2043 2042 PyObject *parse_index2(PyObject *self, PyObject *args)
2044 2043 {
2045 2044 PyObject *tuple = NULL, *cache = NULL;
2046 2045 indexObject *idx;
2047 2046 int ret;
2048 2047
2049 2048 idx = PyObject_New(indexObject, &indexType);
2050 2049 if (idx == NULL)
2051 2050 goto bail;
2052 2051
2053 2052 ret = index_init(idx, args);
2054 2053 if (ret == -1)
2055 2054 goto bail;
2056 2055
2057 2056 if (idx->inlined) {
2058 2057 cache = Py_BuildValue("iO", 0, idx->data);
2059 2058 if (cache == NULL)
2060 2059 goto bail;
2061 2060 } else {
2062 2061 cache = Py_None;
2063 2062 Py_INCREF(cache);
2064 2063 }
2065 2064
2066 2065 tuple = Py_BuildValue("NN", idx, cache);
2067 2066 if (!tuple)
2068 2067 goto bail;
2069 2068 return tuple;
2070 2069
2071 2070 bail:
2072 2071 Py_XDECREF(idx);
2073 2072 Py_XDECREF(cache);
2074 2073 Py_XDECREF(tuple);
2075 2074 return NULL;
2076 2075 }
2077 2076
2078 2077 void revlog_module_init(PyObject *mod)
2079 2078 {
2080 2079 indexType.tp_new = PyType_GenericNew;
2081 2080 if (PyType_Ready(&indexType) < 0)
2082 2081 return;
2083 2082 Py_INCREF(&indexType);
2084 2083 PyModule_AddObject(mod, "index", (PyObject *)&indexType);
2085 2084
2086 2085 nullentry = Py_BuildValue("iiiiiiis#", 0, 0, 0,
2087 2086 -1, -1, -1, -1, nullid, 20);
2088 2087 if (nullentry)
2089 2088 PyObject_GC_UnTrack(nullentry);
2090 2089 }
General Comments 0
You need to be logged in to leave comments. Login now