2
Copyright (C) 2007, 2010 Canonical Ltd
4
This program is free software; you can redistribute it and/or modify
5
it under the terms of the GNU General Public License as published by
6
the Free Software Foundation; either version 2 of the License, or
7
(at your option) any later version.
9
This program is distributed in the hope that it will be useful,
10
but WITHOUT ANY WARRANTY; without even the implied warranty of
11
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12
GNU General Public License for more details.
14
You should have received a copy of the GNU General Public License
15
along with this program; if not, write to the Free Software
16
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18
Function equate_lines based on bdiff.c from Mercurial.
19
Copyright (C) 2005, 2006 Matt Mackall <mpm@selenic.com>
21
Functions unique_lcs/recurse_matches based on _patiencediff_py.py.
22
Copyright (C) 2005 Bram Cohen, Copyright (C) 2005, 2006 Canonical Ltd
30
#include "python-compat.h"
34
# define inline __inline__
35
#elif defined(_MSC_VER)
36
# define inline __inline
42
#define MIN(a, b) (((a) > (b)) ? (b) : (a))
43
#define MAX(a, b) (((a) > (b)) ? (a) : (b))
49
/* malloc returns NULL on some platforms if you try to allocate nothing,
50
* causing <https://bugs.launchpad.net/bzr/+bug/511267> and
51
* <https://bugs.launchpad.net/bzr/+bug/331095>. On glibc it passes, but
52
* let's make it fail to aid testing. */
53
#define guarded_malloc(x) ( (x) ? malloc(x) : NULL )
63
/* values from this array need to correspont to the order of the enum above */
64
static char *opcode_names[] = {
73
long hash; /* hash code of the string/object */
74
Py_ssize_t next; /* next line from the same equivalence class */
75
Py_ssize_t equiv; /* equivalence class */
81
Py_ssize_t a_head; /* first item in `a` from this equivalence class */
83
Py_ssize_t b_head; /* first item in `b` from this equivalence class */
91
Py_ssize_t last_a_pos;
92
Py_ssize_t last_b_pos;
97
struct matching_line {
98
Py_ssize_t a; /* index of the line in `a` */
99
Py_ssize_t b; /* index of the line in `b` */
103
struct matching_block {
104
Py_ssize_t a; /* index of the first line in `a` */
105
Py_ssize_t b; /* index of the first line in `b` */
106
Py_ssize_t len; /* length of the block */
110
struct matching_blocks {
111
struct matching_block *matches;
131
struct hashtable hashtable;
132
Py_ssize_t *backpointers;
133
} PatienceSequenceMatcher;
136
static inline Py_ssize_t
137
bisect_left(Py_ssize_t *list, Py_ssize_t item, Py_ssize_t lo, Py_ssize_t hi)
140
Py_ssize_t mid = lo / 2 + hi / 2 + (lo % 2 + hi % 2) / 2;
141
if (list[mid] < item)
151
compare_lines(struct line *a, struct line *b)
153
return ((a->hash != b->hash)
154
|| PyObject_Compare(a->data, b->data));
159
find_equivalence_class(struct bucket *hashtable, Py_ssize_t hsize,
160
struct line *lines, struct line *ref_lines,
164
for (j = lines[i].hash & hsize; hashtable[j].b_head != SENTINEL; j = (j + 1) & hsize) {
165
if (!compare_lines(lines + i, ref_lines + hashtable[j].b_head)) {
174
equate_lines(struct hashtable *result,
175
struct line *lines_a, struct line *lines_b,
176
Py_ssize_t asize, Py_ssize_t bsize)
178
Py_ssize_t i, j, hsize;
179
struct bucket *hashtable;
181
/* check for overflow, we need the table to be at least bsize+1 */
182
if (bsize == PY_SSIZE_T_MAX) {
183
PyErr_SetNone(PyExc_OverflowError);
187
/* build a hash table of the next highest power of 2 */
189
while (hsize < bsize + 1)
193
hashtable = (struct bucket *) guarded_malloc(sizeof(struct bucket) * hsize);
194
if (hashtable == NULL) {
199
/* initialise the hashtable */
200
for (i = 0; i < hsize; i++) {
201
hashtable[i].a_count = 0;
202
hashtable[i].b_count = 0;
203
hashtable[i].a_head = SENTINEL;
204
hashtable[i].b_head = SENTINEL;
208
/* add lines from lines_b to the hash table chains. iterating
209
backwards so the matching lines are sorted to the linked list
210
by the line number (because we are adding new lines to the
212
for (i = bsize - 1; i >= 0; i--) {
213
/* find the first hashtable entry, which is either empty or contains
214
the same line as lines_b[i] */
215
j = find_equivalence_class(hashtable, hsize, lines_b, lines_b, i);
217
/* set the equivalence class */
218
lines_b[i].equiv = j;
220
/* add to the head of the equivalence class */
221
lines_b[i].next = hashtable[j].b_head;
222
hashtable[j].b_head = i;
223
hashtable[j].b_count++;
226
/* match items from lines_a to their equivalence class in lines_b.
227
again, iterating backwards for the right order of the linked lists */
228
for (i = asize - 1; i >= 0; i--) {
229
/* find the first hash entry, which is either empty or contains
230
the same line as lines_a[i] */
231
j = find_equivalence_class(hashtable, hsize, lines_a, lines_b, i);
233
/* set the equivalence class, even if we are not interested in this
234
line, because the values are not pre-filled */
235
lines_a[i].equiv = j;
237
/* we are not interested in lines which are not also in lines_b */
238
if (hashtable[j].b_head == SENTINEL)
241
/* add to the head of the equivalence class */
242
lines_a[i].next = hashtable[j].a_head;
243
hashtable[j].a_head = i;
244
hashtable[j].a_count++;
247
result->last_a_pos = -1;
248
result->last_b_pos = -1;
249
result->size = hsize + 1;
250
result->table = hashtable;
257
/* Finds longest common subsequence of unique lines in a[alo:ahi] and
259
Parameter backpointers must have allocated memory for at least
260
4 * (bhi - blo) ints. */
262
unique_lcs(struct matching_line *answer,
263
struct hashtable *hashtable, Py_ssize_t *backpointers,
264
struct line *lines_a, struct line *lines_b,
265
Py_ssize_t alo, Py_ssize_t blo, Py_ssize_t ahi, Py_ssize_t bhi)
267
Py_ssize_t i, k, equiv, apos, bpos, norm_apos, norm_bpos, bsize, stacksize;
268
Py_ssize_t *stacks, *lasts, *btoa;
274
h = hashtable->table;
276
/* "unpack" the allocated memory */
277
stacks = backpointers + bsize;
278
lasts = stacks + bsize;
279
btoa = lasts + bsize;
281
/* initialise the backpointers */
282
for (i = 0; i < bsize; i++)
283
backpointers[i] = SENTINEL;
285
if (hashtable->last_a_pos == -1 || hashtable->last_a_pos > alo)
286
for (i = 0; i < hashtable->size; i++)
287
h[i].a_pos = h[i].a_head;
288
hashtable->last_a_pos = alo;
290
if (hashtable->last_b_pos == -1 || hashtable->last_b_pos > blo)
291
for (i = 0; i < hashtable->size; i++)
292
h[i].b_pos = h[i].b_head;
293
hashtable->last_b_pos = blo;
295
for (bpos = blo; bpos < bhi; bpos++) {
296
equiv = lines_b[bpos].equiv;
298
/* no lines in a or b */
299
if (h[equiv].a_count == 0 || h[equiv].b_count == 0)
302
/* find an unique line in lines_a that matches lines_b[bpos]
303
if we find more than one line within the range alo:ahi,
304
jump to the next line from lines_b immediately */
306
/* loop through all lines in the linked list */
307
for (i = h[equiv].a_pos; i != SENTINEL; i = lines_a[i].next) {
308
/* the index is lower than alo, continue to the next line */
313
/* the index is higher than ahi, stop searching */
316
/* if the line is within our range, check if it's a duplicate */
317
if (apos != SENTINEL)
319
/* save index to the line */
322
/* this line has no equivalent in lines_a[alo:ahi] */
323
if (apos == SENTINEL)
326
/* check for duplicates of this line in lines_b[blo:bhi] */
327
/* loop through all lines in the linked list */
328
for (i = h[equiv].b_pos; i != SENTINEL; i = lines_b[i].next) {
329
/* the index is lower than blo, continue to the next line */
334
/* the index is higher than bhi, stop searching */
337
/* if this isn't the line with started with and it's within
338
our range, it's a duplicate */
343
/* use normalised indexes ([0,ahi-alo) instead of [alo,ahi))
344
for the patience sorting algorithm */
345
norm_bpos = bpos - blo;
346
norm_apos = apos - alo;
347
btoa[norm_bpos] = norm_apos;
350
Ok, how does this work...
352
We have a list of matching lines from two lists, a and b. These
353
matches are stored in variable `btoa`. As we are iterating over this
354
table by bpos, the lines from b already form an increasing sequence.
355
We need to "sort" also the lines from a using the patience sorting
356
algorithm, ignoring the lines which would need to be swapped.
358
http://en.wikipedia.org/wiki/Patience_sorting
360
For each pair of lines, we need to place the line from a on either
361
an existing pile that has higher value on the top or create a new
362
pile. Variable `stacks` represents the tops of these piles and in
363
variable `lasts` we store the lines from b, that correspond to the
364
lines from a in `stacks`.
366
Whenever we place a new line on top of a pile, we store a
367
backpointer to the line (b) from top of the previous pile. This means
368
that after the loop, variable `backpointers` will contain an index
369
to the previous matching lines that forms an increasing sequence
370
(over both indexes a and b) with the current matching lines. If
371
either index a or b of the previous matching lines would be higher
372
than indexes of the current one or if the indexes of the current
373
one are 0, it will contain SENTINEL.
375
To construct the LCS, we will just need to follow these backpointers
376
from the top of the last pile and stop when we reach SENTINEL.
379
/* as an optimization, check if the next line comes at the end,
380
because it usually does */
381
if (stacksize && stacks[stacksize - 1] < norm_apos)
383
/* as an optimization, check if the next line comes right after
384
the previous line, because usually it does */
385
else if (stacksize && (stacks[k] < norm_apos) &&
386
(k == stacksize - 1 || stacks[k + 1] > norm_apos))
389
k = bisect_left(stacks, norm_apos, 0, stacksize);
392
backpointers[norm_bpos] = lasts[k - 1];
395
stacks[k] = norm_apos;
396
lasts[k] = norm_bpos;
399
stacks[stacksize] = norm_apos;
400
lasts[stacksize] = norm_bpos;
412
/* backtrace the structures to find the LCS */
414
k = lasts[stacksize - 1];
415
while (k != SENTINEL) {
416
answer[i].a = btoa[k];
425
/* Adds a new line to the list of matching blocks, either extending the
426
current block or adding a new one. */
428
add_matching_line(struct matching_blocks *answer, Py_ssize_t a, Py_ssize_t b)
430
Py_ssize_t last_index = answer->count - 1;
431
if ((last_index >= 0) &&
432
(a == answer->matches[last_index].a +
433
answer->matches[last_index].len) &&
434
(b == answer->matches[last_index].b +
435
answer->matches[last_index].len)) {
436
/* enlarge the last block */
437
answer->matches[last_index].len++;
440
/* create a new block */
442
answer->matches[last_index].a = a;
443
answer->matches[last_index].b = b;
444
answer->matches[last_index].len = 1;
451
recurse_matches(struct matching_blocks *answer, struct hashtable *hashtable,
452
Py_ssize_t *backpointers, struct line *a, struct line *b,
453
Py_ssize_t alo, Py_ssize_t blo, Py_ssize_t ahi, Py_ssize_t bhi,
457
Py_ssize_t new, last_a_pos, last_b_pos, lcs_size, nahi, nbhi, i, apos, bpos;
458
struct matching_line *lcs;
460
if (maxrecursion < 0)
463
if (alo == ahi || blo == bhi)
467
last_a_pos = alo - 1;
468
last_b_pos = blo - 1;
470
lcs = (struct matching_line *)guarded_malloc(sizeof(struct matching_line) * (bhi - blo));
474
lcs_size = unique_lcs(lcs, hashtable, backpointers, a, b, alo, blo, ahi, bhi);
476
/* recurse between lines which are unique in each file and match */
477
for (i = lcs_size - 1; i >= 0; i--) {
478
apos = alo + lcs[i].a;
479
bpos = blo + lcs[i].b;
480
if (last_a_pos + 1 != apos || last_b_pos + 1 != bpos) {
481
res = recurse_matches(answer, hashtable,
483
last_a_pos + 1, last_b_pos + 1,
484
apos, bpos, maxrecursion - 1);
490
add_matching_line(answer, apos, bpos);
497
/* find matches between the last match and the end */
499
res = recurse_matches(answer, hashtable,
501
last_a_pos + 1, last_b_pos + 1,
502
ahi, bhi, maxrecursion - 1);
508
/* find matching lines at the very beginning */
509
else if (a[alo].equiv == b[blo].equiv) {
510
while (alo < ahi && blo < bhi && a[alo].equiv == b[blo].equiv)
511
add_matching_line(answer, alo++, blo++);
512
res = recurse_matches(answer, hashtable,
514
alo, blo, ahi, bhi, maxrecursion - 1);
519
/* find matching lines at the very end */
520
else if (a[ahi - 1].equiv == b[bhi - 1].equiv) {
523
while (nahi > alo && nbhi > blo && a[nahi - 1].equiv == b[nbhi - 1].equiv) {
527
res = recurse_matches(answer, hashtable,
529
last_a_pos + 1, last_b_pos + 1,
530
nahi, nbhi, maxrecursion - 1);
533
for (i = 0; i < ahi - nahi; i++)
534
add_matching_line(answer, nahi + i, nbhi + i);
546
delete_lines(struct line *lines, Py_ssize_t size)
548
struct line *line = lines;
550
Py_XDECREF(line->data);
558
load_lines(PyObject *orig, struct line **lines)
562
PyObject *seq, *item;
564
seq = PySequence_Fast(orig, "sequence expected");
569
size = PySequence_Fast_GET_SIZE(seq);
575
/* Allocate a memory block for line data, initialized to 0 */
576
line = *lines = (struct line *)calloc(size, sizeof(struct line));
583
for (i = 0; i < size; i++) {
584
item = PySequence_Fast_GET_ITEM(seq, i);
587
line->hash = PyObject_Hash(item);
588
if (line->hash == (-1)) {
589
/* Propogate the hash exception */
593
line->next = SENTINEL;
600
/* Error -- cleanup unused object references */
601
delete_lines(*lines, i);
609
py_unique_lcs(PyObject *self, PyObject *args)
611
PyObject *aseq, *bseq, *res, *item;
612
Py_ssize_t asize, bsize, i, nmatches, *backpointers = NULL;
613
struct line *a = NULL, *b = NULL;
614
struct matching_line *matches = NULL;
615
struct hashtable hashtable;
617
if (!PyArg_ParseTuple(args, "OO", &aseq, &bseq))
620
hashtable.table = NULL;
622
asize = load_lines(aseq, &a);
623
bsize = load_lines(bseq, &b);
624
if (asize == -1 || bsize == -1)
627
if (!equate_lines(&hashtable, a, b, asize, bsize))
631
matches = (struct matching_line *)guarded_malloc(sizeof(struct matching_line) * bsize);
635
backpointers = (Py_ssize_t *)guarded_malloc(sizeof(Py_ssize_t) * bsize * 4);
636
if (backpointers == NULL)
640
nmatches = unique_lcs(matches, &hashtable, backpointers, a, b, 0, 0, asize, bsize);
642
res = PyList_New(nmatches);
643
for (i = 0; i < nmatches; i++) {
644
item = Py_BuildValue("nn", matches[nmatches - i - 1].a, matches[nmatches - i - 1].b);
647
if (PyList_SetItem(res, i, item) != 0)
653
free(hashtable.table);
654
delete_lines(b, bsize);
655
delete_lines(a, asize);
661
free(hashtable.table);
662
delete_lines(b, bsize);
663
delete_lines(a, asize);
669
py_recurse_matches(PyObject *self, PyObject *args)
671
PyObject *aseq, *bseq, *item, *answer;
672
int maxrecursion, res;
673
Py_ssize_t i, j, asize, bsize, alo, blo, ahi, bhi;
674
Py_ssize_t *backpointers = NULL;
675
struct line *a = NULL, *b = NULL;
676
struct hashtable hashtable;
677
struct matching_blocks matches;
679
if (!PyArg_ParseTuple(args, "OOnnnnOi", &aseq, &bseq, &alo, &blo,
680
&ahi, &bhi, &answer, &maxrecursion))
683
hashtable.table = NULL;
684
matches.matches = NULL;
686
asize = load_lines(aseq, &a);
687
bsize = load_lines(bseq, &b);
688
if (asize == -1 || bsize == -1)
691
if (!equate_lines(&hashtable, a, b, asize, bsize))
697
matches.matches = (struct matching_block *)guarded_malloc(sizeof(struct matching_block) * bsize);
698
if (matches.matches == NULL)
701
backpointers = (Py_ssize_t *)guarded_malloc(sizeof(Py_ssize_t) * bsize * 4);
702
if (backpointers == NULL)
705
matches.matches = NULL;
709
res = recurse_matches(&matches, &hashtable, backpointers,
710
a, b, alo, blo, ahi, bhi, maxrecursion);
714
for (i = 0; i < matches.count; i++) {
715
for (j = 0; j < matches.matches[i].len; j++) {
716
item = Py_BuildValue("nn", matches.matches[i].a + j,
717
matches.matches[i].b + j);
720
if (PyList_Append(answer, item) != 0)
726
free(matches.matches);
727
free(hashtable.table);
728
delete_lines(b, bsize);
729
delete_lines(a, asize);
734
free(matches.matches);
735
free(hashtable.table);
736
delete_lines(b, bsize);
737
delete_lines(a, asize);
743
PatienceSequenceMatcher_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
745
PyObject *junk, *a, *b;
746
PatienceSequenceMatcher *self;
748
self = (PatienceSequenceMatcher *)type->tp_alloc(type, 0);
751
if (!PyArg_ParseTuple(args, "OOO", &junk, &a, &b)) {
756
self->asize = load_lines(a, &(self->a));
757
self->bsize = load_lines(b, &(self->b));
759
if (self->asize == -1 || self->bsize == -1) {
764
if (!equate_lines(&self->hashtable, self->a, self->b, self->asize, self->bsize)) {
769
if (self->bsize > 0) {
770
self->backpointers = (Py_ssize_t *)guarded_malloc(sizeof(Py_ssize_t) * self->bsize * 4);
771
if (self->backpointers == NULL) {
777
self->backpointers = NULL;
782
return (PyObject *)self;
787
PatienceSequenceMatcher_dealloc(PatienceSequenceMatcher* self)
789
free(self->backpointers);
790
free(self->hashtable.table);
791
delete_lines(self->b, self->bsize);
792
delete_lines(self->a, self->asize);
793
self->ob_type->tp_free((PyObject *)self);
797
static char PatienceSequenceMatcher_get_matching_blocks_doc[] =
798
"Return list of triples describing matching subsequences.\n"
800
"Each triple is of the form (i, j, n), and means that\n"
801
"a[i:i+n] == b[j:j+n]. The triples are monotonically increasing in\n"
804
"The last triple is a dummy, (len(a), len(b), 0), and is the only\n"
805
"triple with n==0.\n"
807
">>> s = PatienceSequenceMatcher(None, \"abxcd\", \"abcd\")\n"
808
">>> s.get_matching_blocks()\n"
809
"[(0, 0, 2), (3, 2, 2), (5, 4, 0)]\n";
812
PatienceSequenceMatcher_get_matching_blocks(PatienceSequenceMatcher* self)
814
PyObject *answer, *item;
817
struct matching_blocks matches;
820
if (self->bsize > 0) {
821
matches.matches = (struct matching_block *)
822
guarded_malloc(sizeof(struct matching_block) * self->bsize);
823
if (matches.matches == NULL)
824
return PyErr_NoMemory();
826
matches.matches = NULL;
828
res = recurse_matches(&matches, &self->hashtable, self->backpointers,
829
self->a, self->b, 0, 0,
830
self->asize, self->bsize, 10);
832
free(matches.matches);
833
return PyErr_NoMemory();
836
answer = PyList_New(matches.count + 1);
837
if (answer == NULL) {
838
free(matches.matches);
842
for (i = 0; i < matches.count; i++) {
843
item = Py_BuildValue("nnn", matches.matches[i].a,
844
matches.matches[i].b, matches.matches[i].len);
847
if (PyList_SetItem(answer, i, item) != 0)
850
item = Py_BuildValue("nnn", self->asize, self->bsize, 0);
853
if (PyList_SetItem(answer, i, item) != 0)
856
free(matches.matches);
860
free(matches.matches);
866
static char PatienceSequenceMatcher_get_opcodes_doc[] =
867
"Return list of 5-tuples describing how to turn a into b.\n"
869
"Each tuple is of the form (tag, i1, i2, j1, j2). The first tuple\n"
870
"has i1 == j1 == 0, and remaining tuples have i1 == the i2 from the\n"
871
"tuple preceding it, and likewise for j1 == the previous j2.\n"
873
"The tags are strings, with these meanings:\n"
875
"'replace': a[i1:i2] should be replaced by b[j1:j2]\n"
876
"'delete': a[i1:i2] should be deleted.\n"
877
" Note that j1==j2 in this case.\n"
878
"'insert': b[j1:j2] should be inserted at a[i1:i1].\n"
879
" Note that i1==i2 in this case.\n"
880
"'equal': a[i1:i2] == b[j1:j2]\n"
882
">>> a = \"qabxcd\"\n"
883
">>> b = \"abycdf\"\n"
884
">>> s = PatienceSequenceMatcher(None, a, b)\n"
885
">>> for tag, i1, i2, j1, j2 in s.get_opcodes():\n"
886
"... print (\"%7s a[%d:%d] (%s) b[%d:%d] (%s)\" %\n"
887
"... (tag, i1, i2, a[i1:i2], j1, j2, b[j1:j2]))\n"
888
" delete a[0:1] (q) b[0:0] ()\n"
889
" equal a[1:3] (ab) b[0:2] (ab)\n"
890
"replace a[3:4] (x) b[2:3] (y)\n"
891
" equal a[4:6] (cd) b[3:5] (cd)\n"
892
" insert a[6:6] () b[5:6] (f)\n";
895
PatienceSequenceMatcher_get_opcodes(PatienceSequenceMatcher* self)
897
PyObject *answer, *item;
898
Py_ssize_t i, j, k, ai, bj;
900
struct matching_blocks matches;
903
matches.matches = (struct matching_block *)guarded_malloc(sizeof(struct matching_block) * (self->bsize + 1));
904
if (matches.matches == NULL)
905
return PyErr_NoMemory();
907
res = recurse_matches(&matches, &self->hashtable, self->backpointers,
908
self->a, self->b, 0, 0,
909
self->asize, self->bsize, 10);
911
free(matches.matches);
912
return PyErr_NoMemory();
915
matches.matches[matches.count].a = self->asize;
916
matches.matches[matches.count].b = self->bsize;
917
matches.matches[matches.count].len = 0;
920
answer = PyList_New(0);
921
if (answer == NULL) {
922
free(matches.matches);
927
for (k = 0; k < matches.count; k++) {
928
ai = matches.matches[k].a;
929
bj = matches.matches[k].b;
932
if (i < ai && j < bj)
940
item = Py_BuildValue("snnnn", opcode_names[tag], i, ai, j, bj);
943
if (PyList_Append(answer, item) != 0)
947
i = ai + matches.matches[k].len;
948
j = bj + matches.matches[k].len;
950
if (matches.matches[k].len > 0) {
951
item = Py_BuildValue("snnnn", opcode_names[OP_EQUAL], ai, i, bj, j);
954
if (PyList_Append(answer, item) != 0)
959
free(matches.matches);
963
free(matches.matches);
969
static char PatienceSequenceMatcher_get_grouped_opcodes_doc[] =
970
"Isolate change clusters by eliminating ranges with no changes.\n"
972
"Return a list of groups with upto n lines of context.\n"
973
"Each group is in the same format as returned by get_opcodes().\n"
975
">>> from pprint import pprint\n"
976
">>> a = map(str, range(1,40))\n"
978
">>> b[8:8] = ['i'] # Make an insertion\n"
979
">>> b[20] += 'x' # Make a replacement\n"
980
">>> b[23:28] = [] # Make a deletion\n"
981
">>> b[30] += 'y' # Make another replacement\n"
982
">>> pprint(PatienceSequenceMatcher(None,a,b).get_grouped_opcodes())\n"
983
"[[('equal', 5, 8, 5, 8), ('insert', 8, 8, 8, 9), ('equal', 8, 11, 9, 12)],\n"
984
" [('equal', 16, 19, 17, 20),\n"
985
" ('replace', 19, 20, 20, 21),\n"
986
" ('equal', 20, 22, 21, 23),\n"
987
" ('delete', 22, 27, 23, 23),\n"
988
" ('equal', 27, 30, 23, 26)],\n"
989
" [('equal', 31, 34, 27, 30),\n"
990
" ('replace', 34, 35, 30, 31),\n"
991
" ('equal', 35, 38, 31, 34)]]\n";
994
PatienceSequenceMatcher_get_grouped_opcodes(PatienceSequenceMatcher* self,
997
PyObject *answer, *group, *item;
998
Py_ssize_t i, j, k, ai, bj, size, ncodes, tag;
999
Py_ssize_t i1, i2, j1, j2;
1001
struct matching_blocks matches;
1002
struct opcode *codes;
1004
if (!PyArg_ParseTuple(args, "|i", &n))
1008
matches.matches = (struct matching_block *)guarded_malloc(sizeof(struct matching_block) * (self->bsize + 1));
1009
if (matches.matches == NULL)
1010
return PyErr_NoMemory();
1012
res = recurse_matches(&matches, &self->hashtable, self->backpointers,
1013
self->a, self->b, 0, 0,
1014
self->asize, self->bsize, 10);
1016
free(matches.matches);
1017
return PyErr_NoMemory();
1020
matches.matches[matches.count].a = self->asize;
1021
matches.matches[matches.count].b = self->bsize;
1022
matches.matches[matches.count].len = 0;
1026
codes = (struct opcode *)guarded_malloc(sizeof(struct opcode) * matches.count * 2);
1027
if (codes == NULL) {
1028
free(matches.matches);
1029
return PyErr_NoMemory();
1033
for (k = 0; k < matches.count; k++) {
1034
ai = matches.matches[k].a;
1035
bj = matches.matches[k].b;
1038
if (i < ai && j < bj)
1046
codes[ncodes].tag = tag;
1047
codes[ncodes].i1 = i;
1048
codes[ncodes].i2 = ai;
1049
codes[ncodes].j1 = j;
1050
codes[ncodes].j2 = bj;
1054
i = ai + matches.matches[k].len;
1055
j = bj + matches.matches[k].len;
1057
if (matches.matches[k].len > 0) {
1058
codes[ncodes].tag = OP_EQUAL;
1059
codes[ncodes].i1 = ai;
1060
codes[ncodes].i2 = i;
1061
codes[ncodes].j1 = bj;
1062
codes[ncodes].j2 = j;
1068
codes[ncodes].tag = OP_EQUAL;
1069
codes[ncodes].i1 = 0;
1070
codes[ncodes].i2 = 1;
1071
codes[ncodes].j1 = 0;
1072
codes[ncodes].j2 = 1;
1076
/* fixup leading and trailing groups if they show no changes. */
1077
if (codes[0].tag == OP_EQUAL) {
1078
codes[0].i1 = MAX(codes[0].i1, codes[0].i2 - n);
1079
codes[0].j1 = MAX(codes[0].j1, codes[0].j2 - n);
1081
if (codes[ncodes - 1].tag == OP_EQUAL) {
1082
codes[ncodes - 1].i2 = MIN(codes[ncodes - 1].i2,
1083
codes[ncodes - 1].i1 + n);
1084
codes[ncodes - 1].j2 = MIN(codes[ncodes - 1].j2,
1085
codes[ncodes - 1].j1 + n);
1090
answer = PyList_New(0);
1094
group = PyList_New(0);
1100
for (i = 0; i < ncodes; i++) {
1106
/* end the current group and start a new one whenever
1107
there is a large range with no changes. */
1108
if (tag == OP_EQUAL && i2 - i1 > nn) {
1109
item = Py_BuildValue("snnnn", opcode_names[tag],
1110
i1, MIN(i2, i1 + n), j1, MIN(j2, j1 + n));
1113
if (PyList_Append(group, item) != 0)
1115
if (PyList_Append(answer, group) != 0)
1117
group = PyList_New(0);
1120
i1 = MAX(i1, i2 - n);
1121
j1 = MAX(j1, j2 - n);
1123
item = Py_BuildValue("snnnn", opcode_names[tag], i1, i2, j1 ,j2);
1126
if (PyList_Append(group, item) != 0)
1129
size = PyList_Size(group);
1130
if (size > 0 && !(size == 1 && tag == OP_EQUAL)) {
1131
if (PyList_Append(answer, group) != 0)
1138
free(matches.matches);
1143
free(matches.matches);
1150
static PyMethodDef PatienceSequenceMatcher_methods[] = {
1151
{"get_matching_blocks",
1152
(PyCFunction)PatienceSequenceMatcher_get_matching_blocks,
1154
PatienceSequenceMatcher_get_matching_blocks_doc},
1156
(PyCFunction)PatienceSequenceMatcher_get_opcodes,
1158
PatienceSequenceMatcher_get_opcodes_doc},
1159
{"get_grouped_opcodes",
1160
(PyCFunction)PatienceSequenceMatcher_get_grouped_opcodes,
1162
PatienceSequenceMatcher_get_grouped_opcodes_doc},
1167
static char PatienceSequenceMatcher_doc[] =
1168
"C implementation of PatienceSequenceMatcher";
1171
static PyTypeObject PatienceSequenceMatcherType = {
1172
PyObject_HEAD_INIT(NULL)
1174
"PatienceSequenceMatcher", /* tp_name */
1175
sizeof(PatienceSequenceMatcher), /* tp_basicsize */
1176
0, /* tp_itemsize */
1177
(destructor)PatienceSequenceMatcher_dealloc, /* tp_dealloc */
1183
0, /* tp_as_number */
1184
0, /* tp_as_sequence */
1185
0, /* tp_as_mapping */
1189
0, /* tp_getattro */
1190
0, /* tp_setattro */
1191
0, /* tp_as_buffer */
1192
Py_TPFLAGS_DEFAULT, /* tp_flags*/
1193
PatienceSequenceMatcher_doc, /* tp_doc */
1194
0, /* tp_traverse */
1196
0, /* tp_richcompare */
1197
0, /* tp_weaklistoffset */
1199
0, /* tp_iternext */
1200
PatienceSequenceMatcher_methods, /* tp_methods */
1205
0, /* tp_descr_get */
1206
0, /* tp_descr_set */
1207
0, /* tp_dictoffset */
1210
PatienceSequenceMatcher_new, /* tp_new */
1214
static PyMethodDef cpatiencediff_methods[] = {
1215
{"unique_lcs_c", py_unique_lcs, METH_VARARGS},
1216
{"recurse_matches_c", py_recurse_matches, METH_VARARGS},
1222
init_patiencediff_c(void)
1226
if (PyType_Ready(&PatienceSequenceMatcherType) < 0)
1229
m = Py_InitModule3("_patiencediff_c", cpatiencediff_methods,
1230
"C implementation of PatienceSequenceMatcher");
1234
Py_INCREF(&PatienceSequenceMatcherType);
1235
PyModule_AddObject(m, "PatienceSequenceMatcher_c",
1236
(PyObject *)&PatienceSequenceMatcherType);