t0006: skip "far in the future" test when unsigned long is not long enough
[git.git] / diffcore-delta.c
blob4159748a70ccc0ddee7cf2fcf82976629dfa6867
1 #include "cache.h"
2 #include "diff.h"
3 #include "diffcore.h"
5 /*
6 * Idea here is very simple.
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.
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.
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.
24 /* Wild guess at the initial hash size */
25 #define INITIAL_HASH_SIZE 9
27 /* We leave more room in smaller hash but do not let it
28 * grow to have unused hole too much.
30 #define INITIAL_FREE(sz_log2) ((1<<(sz_log2))*(sz_log2-3)/(sz_log2))
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.
37 #define HASHBASE 107927
39 struct spanhash {
40 unsigned int hashval;
41 unsigned int cnt;
43 struct spanhash_top {
44 int alloc_log2;
45 int free;
46 struct spanhash data[FLEX_ARRAY];
49 static struct spanhash_top *spanhash_rehash(struct spanhash_top *orig)
51 struct spanhash_top *new;
52 int i;
53 int osz = 1 << orig->alloc_log2;
54 int sz = osz << 1;
56 new = xmalloc(st_add(sizeof(*orig),
57 st_mult(sizeof(struct spanhash), sz)));
58 new->alloc_log2 = orig->alloc_log2 + 1;
59 new->free = INITIAL_FREE(new->alloc_log2);
60 memset(new->data, 0, sizeof(struct spanhash) * sz);
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) {
68 struct spanhash *h = &(new->data[bucket++]);
69 if (!h->cnt) {
70 h->hashval = o->hashval;
71 h->cnt = o->cnt;
72 new->free--;
73 break;
75 if (sz <= bucket)
76 bucket = 0;
79 free(orig);
80 return new;
83 static struct spanhash_top *add_spanhash(struct spanhash_top *top,
84 unsigned int hashval, int cnt)
86 int bucket, lim;
87 struct spanhash *h;
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;
95 h->cnt = cnt;
96 top->free--;
97 if (top->free < 0)
98 return spanhash_rehash(top);
99 return top;
101 if (h->hashval == hashval) {
102 h->cnt += cnt;
103 return top;
105 if (lim <= bucket)
106 bucket = 0;
110 static int spanhash_cmp(const void *a_, const void *b_)
112 const struct spanhash *a = a_;
113 const struct spanhash *b = b_;
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;
124 static struct spanhash_top *hash_chars(struct diff_filespec *one)
126 int i, n;
127 unsigned int accum1, accum2, hashval;
128 struct spanhash_top *hash;
129 unsigned char *buf = one->data;
130 unsigned int sz = one->size;
131 int is_text = !diff_filespec_is_binary(one);
133 i = INITIAL_HASH_SIZE;
134 hash = xmalloc(st_add(sizeof(*hash),
135 st_mult(sizeof(struct spanhash), 1<<i)));
136 hash->alloc_log2 = i;
137 hash->free = INITIAL_FREE(i);
138 memset(hash->data, 0, sizeof(struct spanhash) * (1<<i));
140 n = 0;
141 accum1 = accum2 = 0;
142 while (sz) {
143 unsigned int c = *buf++;
144 unsigned int old_1 = accum1;
145 sz--;
147 /* Ignore CR in CRLF sequence if text */
148 if (is_text && c == '\r' && sz && *buf == '\n')
149 continue;
151 accum1 = (accum1 << 7) ^ (accum2 >> 25);
152 accum2 = (accum2 << 7) ^ (old_1 >> 25);
153 accum1 += c;
154 if (++n < 64 && c != '\n')
155 continue;
156 hashval = (accum1 + accum2 * 0x61) % HASHBASE;
157 hash = add_spanhash(hash, hashval, n);
158 n = 0;
159 accum1 = accum2 = 0;
161 qsort(hash->data,
162 1ul << hash->alloc_log2,
163 sizeof(hash->data[0]),
164 spanhash_cmp);
165 return hash;
168 int diffcore_count_changes(struct diff_filespec *src,
169 struct diff_filespec *dst,
170 void **src_count_p,
171 void **dst_count_p,
172 unsigned long delta_limit,
173 unsigned long *src_copied,
174 unsigned long *literal_added)
176 struct spanhash *s, *d;
177 struct spanhash_top *src_count, *dst_count;
178 unsigned long sc, la;
180 src_count = dst_count = NULL;
181 if (src_count_p)
182 src_count = *src_count_p;
183 if (!src_count) {
184 src_count = hash_chars(src);
185 if (src_count_p)
186 *src_count_p = src_count;
188 if (dst_count_p)
189 dst_count = *dst_count_p;
190 if (!dst_count) {
191 dst_count = hash_chars(dst);
192 if (dst_count_p)
193 *dst_count_p = dst_count;
195 sc = la = 0;
197 s = src_count->data;
198 d = dst_count->data;
199 for (;;) {
200 unsigned dst_cnt, src_cnt;
201 if (!s->cnt)
202 break; /* we checked all in src */
203 while (d->cnt) {
204 if (d->hashval >= s->hashval)
205 break;
206 la += d->cnt;
207 d++;
209 src_cnt = s->cnt;
210 dst_cnt = 0;
211 if (d->cnt && d->hashval == s->hashval) {
212 dst_cnt = d->cnt;
213 d++;
215 if (src_cnt < dst_cnt) {
216 la += dst_cnt - src_cnt;
217 sc += src_cnt;
219 else
220 sc += dst_cnt;
221 s++;
223 while (d->cnt) {
224 la += d->cnt;
225 d++;
228 if (!src_count_p)
229 free(src_count);
230 if (!dst_count_p)
231 free(dst_count);
232 *src_copied = sc;
233 *literal_added = la;
234 return 0;