Merge branch 'cc/delta-islands'
[git/debian.git] / diffcore-delta.c
blobc83d45a0470d412c29f8c392e84bb010150dbb2a
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_spanhash;
52 int i;
53 int osz = 1 << orig->alloc_log2;
54 int sz = osz << 1;
56 new_spanhash = xmalloc(st_add(sizeof(*orig),
57 st_mult(sizeof(struct spanhash), sz)));
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);
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_spanhash->data[bucket++]);
69 if (!h->cnt) {
70 h->hashval = o->hashval;
71 h->cnt = o->cnt;
72 new_spanhash->free--;
73 break;
75 if (sz <= bucket)
76 bucket = 0;
79 free(orig);
80 return new_spanhash;
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, 1ul << hash->alloc_log2, spanhash_cmp);
162 return hash;
165 int diffcore_count_changes(struct diff_filespec *src,
166 struct diff_filespec *dst,
167 void **src_count_p,
168 void **dst_count_p,
169 unsigned long *src_copied,
170 unsigned long *literal_added)
172 struct spanhash *s, *d;
173 struct spanhash_top *src_count, *dst_count;
174 unsigned long sc, la;
176 src_count = dst_count = NULL;
177 if (src_count_p)
178 src_count = *src_count_p;
179 if (!src_count) {
180 src_count = hash_chars(src);
181 if (src_count_p)
182 *src_count_p = src_count;
184 if (dst_count_p)
185 dst_count = *dst_count_p;
186 if (!dst_count) {
187 dst_count = hash_chars(dst);
188 if (dst_count_p)
189 *dst_count_p = dst_count;
191 sc = la = 0;
193 s = src_count->data;
194 d = dst_count->data;
195 for (;;) {
196 unsigned dst_cnt, src_cnt;
197 if (!s->cnt)
198 break; /* we checked all in src */
199 while (d->cnt) {
200 if (d->hashval >= s->hashval)
201 break;
202 la += d->cnt;
203 d++;
205 src_cnt = s->cnt;
206 dst_cnt = 0;
207 if (d->cnt && d->hashval == s->hashval) {
208 dst_cnt = d->cnt;
209 d++;
211 if (src_cnt < dst_cnt) {
212 la += dst_cnt - src_cnt;
213 sc += src_cnt;
215 else
216 sc += dst_cnt;
217 s++;
219 while (d->cnt) {
220 la += d->cnt;
221 d++;
224 if (!src_count_p)
225 free(src_count);
226 if (!dst_count_p)
227 free(dst_count);
228 *src_copied = sc;
229 *literal_added = la;
230 return 0;