Merge branch 'bc/hash-independent-tests-part-5'
[git/git.git] / diffcore-delta.c
CommitLineData
65416758
JH
1#include "cache.h"
2#include "diff.h"
3#include "diffcore.h"
ba23bbc8
JH
4
5/*
6 * Idea here is very simple.
7 *
af3abef9
JH
8 * Almost all data we are interested in are text, but sometimes we have
9 * to deal with binary data. So we cut them into chunks delimited by
10 * LF byte, or 64-byte sequence, whichever comes first, and hash them.
ba23bbc8 11 *
af3abef9
JH
12 * For those chunks, if the source buffer has more instances of it
13 * than the destination buffer, that means the difference are the
14 * number of bytes not copied from source to destination. If the
15 * counts are the same, everything was copied from source to
16 * destination. If the destination has more, everything was copied,
17 * and destination added more.
ba23bbc8
JH
18 *
19 * We are doing an approximation so we do not really have to waste
20 * memory by actually storing the sequence. We just hash them into
21 * somewhere around 2^16 hashbuckets and count the occurrences.
ba23bbc8
JH
22 */
23
c06c7966 24/* Wild guess at the initial hash size */
2821104d 25#define INITIAL_HASH_SIZE 9
fc66d213 26
2821104d
JH
27/* We leave more room in smaller hash but do not let it
28 * grow to have unused hole too much.
29 */
30#define INITIAL_FREE(sz_log2) ((1<<(sz_log2))*(sz_log2-3)/(sz_log2))
ba23bbc8 31
fc66d213
JH
32/* A prime rather carefully chosen between 2^16..2^17, so that
33 * HASHBASE < INITIAL_FREE(17). We want to keep the maximum hashtable
34 * size under the current 2<<17 maximum, which can hold this many
35 * different values before overflowing to hashtable of size 2<<18.
36 */
37#define HASHBASE 107927
38
c06c7966 39struct spanhash {
e31c9f24
LT
40 unsigned int hashval;
41 unsigned int cnt;
c06c7966
JH
42};
43struct spanhash_top {
44 int alloc_log2;
45 int free;
46 struct spanhash data[FLEX_ARRAY];
47};
48
c06c7966
JH
49static struct spanhash_top *spanhash_rehash(struct spanhash_top *orig)
50{
94a5c5d5 51 struct spanhash_top *new_spanhash;
c06c7966
JH
52 int i;
53 int osz = 1 << orig->alloc_log2;
54 int sz = osz << 1;
55
94a5c5d5 56 new_spanhash = xmalloc(st_add(sizeof(*orig),
50a6c8ef 57 st_mult(sizeof(struct spanhash), sz)));
94a5c5d5
BW
58 new_spanhash->alloc_log2 = orig->alloc_log2 + 1;
59 new_spanhash->free = INITIAL_FREE(new_spanhash->alloc_log2);
60 memset(new_spanhash->data, 0, sizeof(struct spanhash) * sz);
c06c7966
JH
61 for (i = 0; i < osz; i++) {
62 struct spanhash *o = &(orig->data[i]);
63 int bucket;
64 if (!o->cnt)
65 continue;
66 bucket = o->hashval & (sz - 1);
67 while (1) {
94a5c5d5 68 struct spanhash *h = &(new_spanhash->data[bucket++]);
c06c7966
JH
69 if (!h->cnt) {
70 h->hashval = o->hashval;
71 h->cnt = o->cnt;
94a5c5d5 72 new_spanhash->free--;
c06c7966
JH
73 break;
74 }
75 if (sz <= bucket)
76 bucket = 0;
77 }
78 }
79 free(orig);
94a5c5d5 80 return new_spanhash;
c06c7966
JH
81}
82
83static struct spanhash_top *add_spanhash(struct spanhash_top *top,
e31c9f24 84 unsigned int hashval, int cnt)
c06c7966
JH
85{
86 int bucket, lim;
87 struct spanhash *h;
88
89 lim = (1 << top->alloc_log2);
90 bucket = hashval & (lim - 1);
91 while (1) {
92 h = &(top->data[bucket++]);
93 if (!h->cnt) {
94 h->hashval = hashval;
3c7ceba4 95 h->cnt = cnt;
c06c7966
JH
96 top->free--;
97 if (top->free < 0)
98 return spanhash_rehash(top);
99 return top;
100 }
101 if (h->hashval == hashval) {
3c7ceba4 102 h->cnt += cnt;
c06c7966
JH
103 return top;
104 }
105 if (lim <= bucket)
106 bucket = 0;
107 }
108}
109
eb4d0e3f
LT
110static int spanhash_cmp(const void *a_, const void *b_)
111{
112 const struct spanhash *a = a_;
113 const struct spanhash *b = b_;
114
115 /* A count of zero compares at the end.. */
116 if (!a->cnt)
117 return !b->cnt ? 0 : 1;
118 if (!b->cnt)
119 return -1;
120 return a->hashval < b->hashval ? -1 :
121 a->hashval > b->hashval ? 1 : 0;
122}
123
b78ea5fc
NTND
124static struct spanhash_top *hash_chars(struct repository *r,
125 struct diff_filespec *one)
65416758 126{
3c7ceba4 127 int i, n;
e31c9f24 128 unsigned int accum1, accum2, hashval;
c06c7966 129 struct spanhash_top *hash;
b9905fed
JH
130 unsigned char *buf = one->data;
131 unsigned int sz = one->size;
b78ea5fc 132 int is_text = !diff_filespec_is_binary(r, one);
c06c7966
JH
133
134 i = INITIAL_HASH_SIZE;
50a6c8ef
JK
135 hash = xmalloc(st_add(sizeof(*hash),
136 st_mult(sizeof(struct spanhash), 1<<i)));
c06c7966 137 hash->alloc_log2 = i;
2821104d 138 hash->free = INITIAL_FREE(i);
c06c7966 139 memset(hash->data, 0, sizeof(struct spanhash) * (1<<i));
65416758 140
3c7ceba4
LT
141 n = 0;
142 accum1 = accum2 = 0;
ba23bbc8 143 while (sz) {
e31c9f24
LT
144 unsigned int c = *buf++;
145 unsigned int old_1 = accum1;
ba23bbc8 146 sz--;
b9905fed
JH
147
148 /* Ignore CR in CRLF sequence if text */
149 if (is_text && c == '\r' && sz && *buf == '\n')
150 continue;
151
e31c9f24
LT
152 accum1 = (accum1 << 7) ^ (accum2 >> 25);
153 accum2 = (accum2 << 7) ^ (old_1 >> 25);
3c7ceba4
LT
154 accum1 += c;
155 if (++n < 64 && c != '\n')
156 continue;
157 hashval = (accum1 + accum2 * 0x61) % HASHBASE;
158 hash = add_spanhash(hash, hashval, n);
159 n = 0;
160 accum1 = accum2 = 0;
65416758 161 }
9ed0d8d6 162 QSORT(hash->data, 1ul << hash->alloc_log2, spanhash_cmp);
c06c7966 163 return hash;
65416758
JH
164}
165
b78ea5fc
NTND
166int diffcore_count_changes(struct repository *r,
167 struct diff_filespec *src,
d8c3d03a 168 struct diff_filespec *dst,
c06c7966
JH
169 void **src_count_p,
170 void **dst_count_p,
65416758
JH
171 unsigned long *src_copied,
172 unsigned long *literal_added)
173{
eb4d0e3f 174 struct spanhash *s, *d;
c06c7966 175 struct spanhash_top *src_count, *dst_count;
ba23bbc8
JH
176 unsigned long sc, la;
177
c06c7966
JH
178 src_count = dst_count = NULL;
179 if (src_count_p)
180 src_count = *src_count_p;
181 if (!src_count) {
b78ea5fc 182 src_count = hash_chars(r, src);
c06c7966
JH
183 if (src_count_p)
184 *src_count_p = src_count;
185 }
186 if (dst_count_p)
187 dst_count = *dst_count_p;
188 if (!dst_count) {
b78ea5fc 189 dst_count = hash_chars(r, dst);
c06c7966
JH
190 if (dst_count_p)
191 *dst_count_p = dst_count;
192 }
ba23bbc8 193 sc = la = 0;
c06c7966 194
eb4d0e3f
LT
195 s = src_count->data;
196 d = dst_count->data;
197 for (;;) {
c06c7966
JH
198 unsigned dst_cnt, src_cnt;
199 if (!s->cnt)
eb4d0e3f
LT
200 break; /* we checked all in src */
201 while (d->cnt) {
202 if (d->hashval >= s->hashval)
203 break;
77cd6ab6 204 la += d->cnt;
eb4d0e3f
LT
205 d++;
206 }
c06c7966 207 src_cnt = s->cnt;
77cd6ab6
LT
208 dst_cnt = 0;
209 if (d->cnt && d->hashval == s->hashval) {
210 dst_cnt = d->cnt;
211 d++;
212 }
c06c7966
JH
213 if (src_cnt < dst_cnt) {
214 la += dst_cnt - src_cnt;
215 sc += src_cnt;
ba23bbc8 216 }
c06c7966
JH
217 else
218 sc += dst_cnt;
eb4d0e3f 219 s++;
ba23bbc8 220 }
77cd6ab6
LT
221 while (d->cnt) {
222 la += d->cnt;
223 d++;
224 }
c06c7966
JH
225
226 if (!src_count_p)
227 free(src_count);
228 if (!dst_count_p)
229 free(dst_count);
ba23bbc8
JH
230 *src_copied = sc;
231 *literal_added = la;
ba23bbc8 232 return 0;
65416758 233}