annotate mercurial/bdiff.c @ 5363:058e93c3d07d

I have spotted the biggest bottleneck in "bdiff.c". Actually it was pretty easy to find after I recompiled the python interpreter and mercurial for profiling. In "bdiff.c" function "equatelines" allocates the minimum hash table size, which can lead to tons of collisions. I introduced an "overcommit" factor of 16, this is, I allocate 16 times more memory than the minimum value. Overcommiting 128 times does not improve the performance over the 16-times case.
author Christoph Spiel <cspiel@freenet.de>
date Thu, 27 Sep 2007 23:57:57 -0500
parents 9dc64c8414ca
children 5737845fd974
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
1 /*
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
2 bdiff.c - efficient binary diff extension for Mercurial
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
3
2858
345bac2bc4ec update copyrights.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2600
diff changeset
4 Copyright 2005, 2006 Matt Mackall <mpm@selenic.com>
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
5
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
6 This software may be used and distributed according to the terms of
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
7 the GNU General Public License, incorporated herein by reference.
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
8
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
9 Based roughly on Python difflib
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
10 */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
11
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
12 #include <Python.h>
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
13 #include <stdlib.h>
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
14 #include <string.h>
867
0cd2ee61b10a Allow Mercurial to build on HP-UX 11
tksoh@users.sourceforge.net
parents: 839
diff changeset
15
3547
8c617d48564a add AIX to the list of compilers that don't have inline keyword
Benoit Boissinot <benoit.boissinot@ens-lyon.org>
parents: 3369
diff changeset
16 #if defined __hpux || defined __SUNPRO_C || defined _AIX
1759
5afd459db177 Sunpro compiler patch
Fabian Otto <sigsegv@alchiba.ni.cs.tu-berlin.de>
parents: 1542
diff changeset
17 # define inline
2600
c4325f0a9b91 clean up trailing white space.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2577
diff changeset
18 #endif
1759
5afd459db177 Sunpro compiler patch
Fabian Otto <sigsegv@alchiba.ni.cs.tu-berlin.de>
parents: 1542
diff changeset
19
411
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
20 #ifdef _WIN32
551
b460a2fd8bb7 [PATCH] bdiff/mpatch under MSVC
mpm@selenic.com
parents: 510
diff changeset
21 #ifdef _MSC_VER
b460a2fd8bb7 [PATCH] bdiff/mpatch under MSVC
mpm@selenic.com
parents: 510
diff changeset
22 #define inline __inline
b460a2fd8bb7 [PATCH] bdiff/mpatch under MSVC
mpm@selenic.com
parents: 510
diff changeset
23 typedef unsigned long uint32_t;
b460a2fd8bb7 [PATCH] bdiff/mpatch under MSVC
mpm@selenic.com
parents: 510
diff changeset
24 #else
510
7f3fc8fd427e More fiddling with uint32_t includes for extensions
mpm@selenic.com
parents: 495
diff changeset
25 #include <stdint.h>
551
b460a2fd8bb7 [PATCH] bdiff/mpatch under MSVC
mpm@selenic.com
parents: 510
diff changeset
26 #endif
411
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
27 static uint32_t htonl(uint32_t x)
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
28 {
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
29 return ((x & 0x000000ffUL) << 24) |
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
30 ((x & 0x0000ff00UL) << 8) |
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
31 ((x & 0x00ff0000UL) >> 8) |
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
32 ((x & 0xff000000UL) >> 24);
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
33 }
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
34 #else
510
7f3fc8fd427e More fiddling with uint32_t includes for extensions
mpm@selenic.com
parents: 495
diff changeset
35 #include <sys/types.h>
4073
95ffa36d1d2a BeOS compatibility support
Andrew Bachmann <andrewbachmann@gmail.com>
parents: 3547
diff changeset
36 #ifdef __BEOS__
95ffa36d1d2a BeOS compatibility support
Andrew Bachmann <andrewbachmann@gmail.com>
parents: 3547
diff changeset
37 #include <ByteOrder.h>
95ffa36d1d2a BeOS compatibility support
Andrew Bachmann <andrewbachmann@gmail.com>
parents: 3547
diff changeset
38 #else
597
e530637ea060 [PATCH] use <arpa/inet.h> instead of <netinet/in.h> for ntohl/htonl
mpm@selenic.com
parents: 553
diff changeset
39 #include <arpa/inet.h>
4073
95ffa36d1d2a BeOS compatibility support
Andrew Bachmann <andrewbachmann@gmail.com>
parents: 3547
diff changeset
40 #endif
2543
860e9c83fc59 Include inttypes.h instead of stdint.h (fixes issue299)
Thomas Arendsen Hein <thomas@intevation.de>
parents: 2483
diff changeset
41 #include <inttypes.h>
411
9e9f7ab43ce2 Add 'other OS' bits to bdiff.c / style cleanups
mpm@selenic.com
parents: 400
diff changeset
42 #endif
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
43
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
44 struct line {
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
45 int h, len, n, e;
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
46 const char *l;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
47 };
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
48
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
49 struct pos {
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
50 int pos, len;
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
51 };
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
52
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
53 struct hunk {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
54 int a1, a2, b1, b2;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
55 };
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
56
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
57 struct hunklist {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
58 struct hunk *base, *head;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
59 };
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
60
1397
66fd3bc1cfcf bdiff: change spurious __inline to inline
Matt Mackall <mpm@selenic.com>
parents: 981
diff changeset
61 static inline uint32_t rol32(uint32_t word, unsigned int shift)
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
62 {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
63 return (word << shift) | (word >> (32 - shift));
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
64 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
65
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
66 int splitlines(const char *a, int len, struct line **lr)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
67 {
2577
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
68 int g, h, i;
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
69 const char *p, *b = a;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
70 struct line *l;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
71
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
72 /* count the lines */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
73 i = 1; /* extra line for sentinel */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
74 for (p = a; p < a + len; p++)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
75 if (*p == '\n' || p == a + len - 1)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
76 i++;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
77
1978
10606ee61107 do proper typecasting on malloc() and calloc() calls
TK Soh <teekaysoh@yahoo.com>
parents: 1759
diff changeset
78 *lr = l = (struct line *)malloc(sizeof(struct line) * i);
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
79 if (!l)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
80 return -1;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
81
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
82 /* build the line array and calculate hashes */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
83 h = 0;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
84 for (p = a; p < a + len; p++) {
2577
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
85 /*
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
86 * a simple hash from GNU diff, with better collision
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
87 * resistance from hashpjw. this slows down common
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
88 * case by 10%, but speeds up worst case by 100x.
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
89 */
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
90 h = *p + rol32(h, 7);
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
91 if ((g = h & 0xf0000000)) {
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
92 h ^= g >> 24;
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
93 h ^= g;
fa76c5d609c9 bdiff: improve worst case behavior by 100x.
Vadim Gelfer <vadim.gelfer@gmail.com>
parents: 2543
diff changeset
94 }
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
95 if (*p == '\n' || p == a + len - 1) {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
96 l->len = p - b + 1;
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
97 l->h = h * l->len;
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
98 l->l = b;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
99 l->n = -1;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
100 l++;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
101 b = p + 1;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
102 h = 0;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
103 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
104 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
105
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
106 /* set up a sentinel */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
107 l->h = l->len = 0;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
108 l->l = a + len;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
109 return i - 1;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
110 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
111
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
112 int inline cmp(struct line *a, struct line *b)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
113 {
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
114 return a->h != b->h || a->len != b->len || memcmp(a->l, b->l, a->len);
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
115 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
116
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
117 static int equatelines(struct line *a, int an, struct line *b, int bn)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
118 {
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
119 int i, j, buckets = 1, t;
5363
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
120 int scale = 32;
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
121 struct pos *h;
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
122
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
123 /* build a hash table of the next highest power of 2 */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
124 while (buckets < bn + 1)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
125 buckets *= 2;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
126
5363
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
127 /* try to allocate a large hash table to avoid collisions */
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
128 do {
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
129 scale /= 2;
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
130 h = (struct pos *)malloc(scale * buckets * sizeof(struct pos));
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
131 } while (!h && scale != 1);
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
132
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
133 if (!h)
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
134 return 0;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
135
5363
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
136 buckets = buckets * scale - 1;
058e93c3d07d I have spotted the biggest bottleneck in "bdiff.c". Actually it was
Christoph Spiel <cspiel@freenet.de>
parents: 4134
diff changeset
137
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
138 /* clear the hash table */
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
139 for (i = 0; i <= buckets; i++) {
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
140 h[i].pos = -1;
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
141 h[i].len = 0;
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
142 }
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
143
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
144 /* add lines to the hash table chains */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
145 for (i = bn - 1; i >= 0; i--) {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
146 /* find the equivalence class */
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
147 for (j = b[i].h & buckets; h[j].pos != -1;
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
148 j = (j + 1) & buckets)
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
149 if (!cmp(b + i, b + h[j].pos))
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
150 break;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
151
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
152 /* add to the head of the equivalence class */
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
153 b[i].n = h[j].pos;
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
154 b[i].e = j;
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
155 h[j].pos = i;
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
156 h[j].len++; /* keep track of popularity */
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
157 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
158
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
159 /* compute popularity threshold */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
160 t = (bn >= 200) ? bn / 100 : bn + 1;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
161
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
162 /* match items in a to their equivalence class in b */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
163 for (i = 0; i < an; i++) {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
164 /* find the equivalence class */
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
165 for (j = a[i].h & buckets; h[j].pos != -1;
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
166 j = (j + 1) & buckets)
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
167 if (!cmp(a + i, b + h[j].pos))
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
168 break;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
169
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
170 a[i].e = j; /* use equivalence class for quick compare */
1542
8e80eefb3de6 made C src formatting more consistent
twaldmann@thinkmo.de
parents: 1397
diff changeset
171 if (h[j].len <= t)
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
172 a[i].n = h[j].pos; /* point to head of match list */
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
173 else
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
174 a[i].n = -1; /* too popular */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
175 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
176
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
177 /* discard hash tables */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
178 free(h);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
179 return 1;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
180 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
181
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
182 static int longest_match(struct line *a, struct line *b, struct pos *pos,
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
183 int a1, int a2, int b1, int b2, int *omi, int *omj)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
184 {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
185 int mi = a1, mj = b1, mk = 0, mb = 0, i, j, k;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
186
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
187 for (i = a1; i < a2; i++) {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
188 /* skip things before the current block */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
189 for (j = a[i].n; j != -1 && j < b1; j = b[j].n)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
190 ;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
191
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
192 /* loop through all lines match a[i] in b */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
193 for (; j != -1 && j < b2; j = b[j].n) {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
194 /* does this extend an earlier match? */
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
195 if (i > a1 && j > b1 && pos[j - 1].pos == i - 1)
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
196 k = pos[j - 1].len + 1;
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
197 else
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
198 k = 1;
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
199 pos[j].pos = i;
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
200 pos[j].len = k;
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
201
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
202 /* best match so far? */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
203 if (k > mk) {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
204 mi = i;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
205 mj = j;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
206 mk = k;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
207 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
208 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
209 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
210
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
211 if (mk) {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
212 mi = mi - mk + 1;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
213 mj = mj - mk + 1;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
214 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
215
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
216 /* expand match to include neighboring popular lines */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
217 while (mi - mb > a1 && mj - mb > b1 &&
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
218 a[mi - mb - 1].e == b[mj - mb - 1].e)
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
219 mb++;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
220 while (mi + mk < a2 && mj + mk < b2 &&
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
221 a[mi + mk].e == b[mj + mk].e)
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
222 mk++;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
223
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
224 *omi = mi - mb;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
225 *omj = mj - mb;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
226 return mk + mb;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
227 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
228
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
229 static void recurse(struct line *a, struct line *b, struct pos *pos,
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
230 int a1, int a2, int b1, int b2, struct hunklist *l)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
231 {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
232 int i, j, k;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
233
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
234 /* find the longest match in this chunk */
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
235 k = longest_match(a, b, pos, a1, a2, b1, b2, &i, &j);
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
236 if (!k)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
237 return;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
238
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
239 /* and recurse on the remaining chunks on either side */
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
240 recurse(a, b, pos, a1, i, b1, j, l);
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
241 l->head->a1 = i;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
242 l->head->a2 = i + k;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
243 l->head->b1 = j;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
244 l->head->b2 = j + k;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
245 l->head++;
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
246 recurse(a, b, pos, i + k, a2, j + k, b2, l);
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
247 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
248
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
249 static struct hunklist diff(struct line *a, int an, struct line *b, int bn)
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
250 {
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
251 struct hunklist l;
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
252 struct pos *pos;
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
253 int t;
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
254
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
255 /* allocate and fill arrays */
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
256 t = equatelines(a, an, b, bn);
1978
10606ee61107 do proper typecasting on malloc() and calloc() calls
TK Soh <teekaysoh@yahoo.com>
parents: 1759
diff changeset
257 pos = (struct pos *)calloc(bn, sizeof(struct pos));
827
a61728b58dc0 Fix array overflow bug in bdiff
"Wallace, Eric S" <eric.s.wallace@intel.com>
parents: 597
diff changeset
258 /* we can't have more matches than lines in the shorter file */
1978
10606ee61107 do proper typecasting on malloc() and calloc() calls
TK Soh <teekaysoh@yahoo.com>
parents: 1759
diff changeset
259 l.head = l.base = (struct hunk *)malloc(sizeof(struct hunk) *
10606ee61107 do proper typecasting on malloc() and calloc() calls
TK Soh <teekaysoh@yahoo.com>
parents: 1759
diff changeset
260 ((an<bn ? an:bn) + 1));
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
261
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
262 if (pos && l.base && t) {
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
263 /* generate the matching block list */
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
264 recurse(a, b, pos, 0, an, 0, bn, &l);
4131
1ca664c964e0 don't return uninitialized memory from bdiff.blocks()
Erling Ellingsen <erlingalf@gmail.com>
parents: 3547
diff changeset
265 l.head->a1 = l.head->a2 = an;
1ca664c964e0 don't return uninitialized memory from bdiff.blocks()
Erling Ellingsen <erlingalf@gmail.com>
parents: 3547
diff changeset
266 l.head->b1 = l.head->b2 = bn;
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
267 l.head++;
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
268 }
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
269
474
b2ae8283d1a6 Minor speed improvements for bdiff
mpm@selenic.com
parents: 472
diff changeset
270 free(pos);
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
271 return l;
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
272 }
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
273
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
274 static PyObject *blocks(PyObject *self, PyObject *args)
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
275 {
435
e731d25ddab2 Fix a compile warning for bdiff
mpm@selenic.com
parents: 433
diff changeset
276 PyObject *sa, *sb, *rl = NULL, *m;
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
277 struct line *a, *b;
970
bf375e93073b Fix possible unitialized variable warnings
mpm@selenic.com
parents: 896
diff changeset
278 struct hunklist l = {NULL, NULL};
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
279 struct hunk *h;
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
280 int an, bn, pos = 0;
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
281
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
282 if (!PyArg_ParseTuple(args, "SS:bdiff", &sa, &sb))
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
283 return NULL;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
284
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
285 an = splitlines(PyString_AsString(sa), PyString_Size(sa), &a);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
286 bn = splitlines(PyString_AsString(sb), PyString_Size(sb), &b);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
287 if (!a || !b)
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
288 goto nomem;
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
289
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
290 l = diff(a, an, b, bn);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
291 rl = PyList_New(l.head - l.base);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
292 if (!l.head || !rl)
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
293 goto nomem;
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
294
1542
8e80eefb3de6 made C src formatting more consistent
twaldmann@thinkmo.de
parents: 1397
diff changeset
295 for (h = l.base; h != l.head; h++) {
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
296 m = Py_BuildValue("iiii", h->a1, h->a2, h->b1, h->b2);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
297 PyList_SetItem(rl, pos, m);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
298 pos++;
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
299 }
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
300
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
301 nomem:
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
302 free(a);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
303 free(b);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
304 free(l.base);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
305 return rl ? rl : PyErr_NoMemory();
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
306 }
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
307
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
308 static PyObject *bdiff(PyObject *self, PyObject *args)
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
309 {
3333
9061613c1593 Teach bdiff to support buffer objects
Brendan Cully <brendan@kublai.com>
parents: 2858
diff changeset
310 char *sa, *sb;
9061613c1593 Teach bdiff to support buffer objects
Brendan Cully <brendan@kublai.com>
parents: 2858
diff changeset
311 PyObject *result = NULL;
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
312 struct line *al, *bl;
970
bf375e93073b Fix possible unitialized variable warnings
mpm@selenic.com
parents: 896
diff changeset
313 struct hunklist l = {NULL, NULL};
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
314 struct hunk *h;
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
315 char encode[12], *rb;
3333
9061613c1593 Teach bdiff to support buffer objects
Brendan Cully <brendan@kublai.com>
parents: 2858
diff changeset
316 int an, bn, len = 0, la, lb;
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
317
3369
4bad632913d8 python2.5 PyArg_ParseTuple fix
Alexis S. L. Carvalho <alexis@cecm.usp.br>
parents: 3333
diff changeset
318 if (!PyArg_ParseTuple(args, "s#s#:bdiff", &sa, &la, &sb, &lb))
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
319 return NULL;
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
320
3333
9061613c1593 Teach bdiff to support buffer objects
Brendan Cully <brendan@kublai.com>
parents: 2858
diff changeset
321 an = splitlines(sa, la, &al);
9061613c1593 Teach bdiff to support buffer objects
Brendan Cully <brendan@kublai.com>
parents: 2858
diff changeset
322 bn = splitlines(sb, lb, &bl);
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
323 if (!al || !bl)
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
324 goto nomem;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
325
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
326 l = diff(al, an, bl, bn);
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
327 if (!l.head)
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
328 goto nomem;
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
329
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
330 /* calculate length of output */
3333
9061613c1593 Teach bdiff to support buffer objects
Brendan Cully <brendan@kublai.com>
parents: 2858
diff changeset
331 la = lb = 0;
1542
8e80eefb3de6 made C src formatting more consistent
twaldmann@thinkmo.de
parents: 1397
diff changeset
332 for (h = l.base; h != l.head; h++) {
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
333 if (h->a1 != la || h->b1 != lb)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
334 len += 12 + bl[h->b1].l - bl[lb].l;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
335 la = h->a2;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
336 lb = h->b2;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
337 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
338
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
339 result = PyString_FromStringAndSize(NULL, len);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
340 if (!result)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
341 goto nomem;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
342
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
343 /* build binary patch */
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
344 rb = PyString_AsString(result);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
345 la = lb = 0;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
346
1542
8e80eefb3de6 made C src formatting more consistent
twaldmann@thinkmo.de
parents: 1397
diff changeset
347 for (h = l.base; h != l.head; h++) {
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
348 if (h->a1 != la || h->b1 != lb) {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
349 len = bl[h->b1].l - bl[lb].l;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
350 *(uint32_t *)(encode) = htonl(al[la].l - al->l);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
351 *(uint32_t *)(encode + 4) = htonl(al[h->a1].l - al->l);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
352 *(uint32_t *)(encode + 8) = htonl(len);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
353 memcpy(rb, encode, 12);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
354 memcpy(rb + 12, bl[lb].l, len);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
355 rb += 12 + len;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
356 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
357 la = h->a2;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
358 lb = h->b2;
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
359 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
360
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
361 nomem:
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
362 free(al);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
363 free(bl);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
364 free(l.base);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
365 return result ? result : PyErr_NoMemory();
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
366 }
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
367
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
368 static char mdiff_doc[] = "Efficient binary diff.";
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
369
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
370 static PyMethodDef methods[] = {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
371 {"bdiff", bdiff, METH_VARARGS, "calculate a binary diff\n"},
433
79c694462294 Add bdiff.blocks / minor performance tweaks
mpm@selenic.com
parents: 411
diff changeset
372 {"blocks", blocks, METH_VARARGS, "find a list of matching lines\n"},
400
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
373 {NULL, NULL}
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
374 };
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
375
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
376 PyMODINIT_FUNC initbdiff(void)
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
377 {
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
378 Py_InitModule3("bdiff", methods, mdiff_doc);
8b067bde6679 Add a fast binary diff extension (not yet used)
mpm@selenic.com
parents:
diff changeset
379 }
1542
8e80eefb3de6 made C src formatting more consistent
twaldmann@thinkmo.de
parents: 1397
diff changeset
380