/src/git/diffcore-delta.c
Line | Count | Source (jump to first uncovered line) |
1 | | #include "git-compat-util.h" |
2 | | #include "diffcore.h" |
3 | | |
4 | | /* |
5 | | * Idea here is very simple. |
6 | | * |
7 | | * Almost all data we are interested in are text, but sometimes we have |
8 | | * to deal with binary data. So we cut them into chunks delimited by |
9 | | * LF byte, or 64-byte sequence, whichever comes first, and hash them. |
10 | | * |
11 | | * For those chunks, if the source buffer has more instances of it |
12 | | * than the destination buffer, that means the difference are the |
13 | | * number of bytes not copied from source to destination. If the |
14 | | * counts are the same, everything was copied from source to |
15 | | * destination. If the destination has more, everything was copied, |
16 | | * and destination added more. |
17 | | * |
18 | | * We are doing an approximation so we do not really have to waste |
19 | | * memory by actually storing the sequence. We just hash them into |
20 | | * somewhere around 2^16 hashbuckets and count the occurrences. |
21 | | */ |
22 | | |
23 | | /* Wild guess at the initial hash size */ |
24 | 0 | #define INITIAL_HASH_SIZE 9 |
25 | | |
26 | | /* We leave more room in smaller hash but do not let it |
27 | | * grow to have unused hole too much. |
28 | | */ |
29 | 0 | #define INITIAL_FREE(sz_log2) ((1<<(sz_log2))*(sz_log2-3)/(sz_log2)) |
30 | | |
31 | | /* A prime rather carefully chosen between 2^16..2^17, so that |
32 | | * HASHBASE < INITIAL_FREE(17). We want to keep the maximum hashtable |
33 | | * size under the current 2<<17 maximum, which can hold this many |
34 | | * different values before overflowing to hashtable of size 2<<18. |
35 | | */ |
36 | 0 | #define HASHBASE 107927 |
37 | | |
38 | | struct spanhash { |
39 | | unsigned int hashval; |
40 | | unsigned int cnt; |
41 | | }; |
42 | | struct spanhash_top { |
43 | | int alloc_log2; |
44 | | int free; |
45 | | struct spanhash data[FLEX_ARRAY]; |
46 | | }; |
47 | | |
48 | | static struct spanhash_top *spanhash_rehash(struct spanhash_top *orig) |
49 | 0 | { |
50 | 0 | struct spanhash_top *new_spanhash; |
51 | 0 | int i; |
52 | 0 | int osz = 1 << orig->alloc_log2; |
53 | 0 | int sz = osz << 1; |
54 | |
|
55 | 0 | new_spanhash = xmalloc(st_add(sizeof(*orig), |
56 | 0 | st_mult(sizeof(struct spanhash), sz))); |
57 | 0 | new_spanhash->alloc_log2 = orig->alloc_log2 + 1; |
58 | 0 | new_spanhash->free = INITIAL_FREE(new_spanhash->alloc_log2); |
59 | 0 | memset(new_spanhash->data, 0, sizeof(struct spanhash) * sz); |
60 | 0 | for (i = 0; i < osz; i++) { |
61 | 0 | struct spanhash *o = &(orig->data[i]); |
62 | 0 | int bucket; |
63 | 0 | if (!o->cnt) |
64 | 0 | continue; |
65 | 0 | bucket = o->hashval & (sz - 1); |
66 | 0 | while (1) { |
67 | 0 | struct spanhash *h = &(new_spanhash->data[bucket++]); |
68 | 0 | if (!h->cnt) { |
69 | 0 | h->hashval = o->hashval; |
70 | 0 | h->cnt = o->cnt; |
71 | 0 | new_spanhash->free--; |
72 | 0 | break; |
73 | 0 | } |
74 | 0 | if (sz <= bucket) |
75 | 0 | bucket = 0; |
76 | 0 | } |
77 | 0 | } |
78 | 0 | free(orig); |
79 | 0 | return new_spanhash; |
80 | 0 | } |
81 | | |
82 | | static struct spanhash_top *add_spanhash(struct spanhash_top *top, |
83 | | unsigned int hashval, int cnt) |
84 | 0 | { |
85 | 0 | int bucket, lim; |
86 | 0 | struct spanhash *h; |
87 | |
|
88 | 0 | lim = (1 << top->alloc_log2); |
89 | 0 | bucket = hashval & (lim - 1); |
90 | 0 | while (1) { |
91 | 0 | h = &(top->data[bucket++]); |
92 | 0 | if (!h->cnt) { |
93 | 0 | h->hashval = hashval; |
94 | 0 | h->cnt = cnt; |
95 | 0 | top->free--; |
96 | 0 | if (top->free < 0) |
97 | 0 | return spanhash_rehash(top); |
98 | 0 | return top; |
99 | 0 | } |
100 | 0 | if (h->hashval == hashval) { |
101 | 0 | h->cnt += cnt; |
102 | 0 | return top; |
103 | 0 | } |
104 | 0 | if (lim <= bucket) |
105 | 0 | bucket = 0; |
106 | 0 | } |
107 | 0 | } |
108 | | |
109 | | static int spanhash_cmp(const void *a_, const void *b_) |
110 | 0 | { |
111 | 0 | const struct spanhash *a = a_; |
112 | 0 | const struct spanhash *b = b_; |
113 | | |
114 | | /* A count of zero compares at the end.. */ |
115 | 0 | if (!a->cnt) |
116 | 0 | return !b->cnt ? 0 : 1; |
117 | 0 | if (!b->cnt) |
118 | 0 | return -1; |
119 | 0 | return a->hashval < b->hashval ? -1 : |
120 | 0 | a->hashval > b->hashval ? 1 : 0; |
121 | 0 | } |
122 | | |
123 | | static struct spanhash_top *hash_chars(struct repository *r, |
124 | | struct diff_filespec *one) |
125 | 0 | { |
126 | 0 | int i, n; |
127 | 0 | unsigned int accum1, accum2, hashval; |
128 | 0 | struct spanhash_top *hash; |
129 | 0 | unsigned char *buf = one->data; |
130 | 0 | unsigned int sz = one->size; |
131 | 0 | int is_text = !diff_filespec_is_binary(r, one); |
132 | |
|
133 | 0 | i = INITIAL_HASH_SIZE; |
134 | 0 | hash = xmalloc(st_add(sizeof(*hash), |
135 | 0 | st_mult(sizeof(struct spanhash), (size_t)1 << i))); |
136 | 0 | hash->alloc_log2 = i; |
137 | 0 | hash->free = INITIAL_FREE(i); |
138 | 0 | memset(hash->data, 0, sizeof(struct spanhash) * ((size_t)1 << i)); |
139 | |
|
140 | 0 | n = 0; |
141 | 0 | accum1 = accum2 = 0; |
142 | 0 | while (sz) { |
143 | 0 | unsigned int c = *buf++; |
144 | 0 | unsigned int old_1 = accum1; |
145 | 0 | sz--; |
146 | | |
147 | | /* Ignore CR in CRLF sequence if text */ |
148 | 0 | if (is_text && c == '\r' && sz && *buf == '\n') |
149 | 0 | continue; |
150 | | |
151 | 0 | accum1 = (accum1 << 7) ^ (accum2 >> 25); |
152 | 0 | accum2 = (accum2 << 7) ^ (old_1 >> 25); |
153 | 0 | accum1 += c; |
154 | 0 | if (++n < 64 && c != '\n') |
155 | 0 | continue; |
156 | 0 | hashval = (accum1 + accum2 * 0x61) % HASHBASE; |
157 | 0 | hash = add_spanhash(hash, hashval, n); |
158 | 0 | n = 0; |
159 | 0 | accum1 = accum2 = 0; |
160 | 0 | } |
161 | 0 | if (n > 0) { |
162 | 0 | hashval = (accum1 + accum2 * 0x61) % HASHBASE; |
163 | 0 | hash = add_spanhash(hash, hashval, n); |
164 | 0 | } |
165 | 0 | QSORT(hash->data, (size_t)1ul << hash->alloc_log2, spanhash_cmp); |
166 | 0 | return hash; |
167 | 0 | } |
168 | | |
169 | | int diffcore_count_changes(struct repository *r, |
170 | | struct diff_filespec *src, |
171 | | struct diff_filespec *dst, |
172 | | void **src_count_p, |
173 | | void **dst_count_p, |
174 | | unsigned long *src_copied, |
175 | | unsigned long *literal_added) |
176 | 0 | { |
177 | 0 | struct spanhash *s, *d; |
178 | 0 | struct spanhash_top *src_count, *dst_count; |
179 | 0 | unsigned long sc, la; |
180 | |
|
181 | 0 | src_count = dst_count = NULL; |
182 | 0 | if (src_count_p) |
183 | 0 | src_count = *src_count_p; |
184 | 0 | if (!src_count) { |
185 | 0 | src_count = hash_chars(r, src); |
186 | 0 | if (src_count_p) |
187 | 0 | *src_count_p = src_count; |
188 | 0 | } |
189 | 0 | if (dst_count_p) |
190 | 0 | dst_count = *dst_count_p; |
191 | 0 | if (!dst_count) { |
192 | 0 | dst_count = hash_chars(r, dst); |
193 | 0 | if (dst_count_p) |
194 | 0 | *dst_count_p = dst_count; |
195 | 0 | } |
196 | 0 | sc = la = 0; |
197 | |
|
198 | 0 | s = src_count->data; |
199 | 0 | d = dst_count->data; |
200 | 0 | for (;;) { |
201 | 0 | unsigned dst_cnt, src_cnt; |
202 | 0 | if (!s->cnt) |
203 | 0 | break; /* we checked all in src */ |
204 | 0 | while (d->cnt) { |
205 | 0 | if (d->hashval >= s->hashval) |
206 | 0 | break; |
207 | 0 | la += d->cnt; |
208 | 0 | d++; |
209 | 0 | } |
210 | 0 | src_cnt = s->cnt; |
211 | 0 | dst_cnt = 0; |
212 | 0 | if (d->cnt && d->hashval == s->hashval) { |
213 | 0 | dst_cnt = d->cnt; |
214 | 0 | d++; |
215 | 0 | } |
216 | 0 | if (src_cnt < dst_cnt) { |
217 | 0 | la += dst_cnt - src_cnt; |
218 | 0 | sc += src_cnt; |
219 | 0 | } |
220 | 0 | else |
221 | 0 | sc += dst_cnt; |
222 | 0 | s++; |
223 | 0 | } |
224 | 0 | while (d->cnt) { |
225 | 0 | la += d->cnt; |
226 | 0 | d++; |
227 | 0 | } |
228 | |
|
229 | 0 | if (!src_count_p) |
230 | 0 | free(src_count); |
231 | 0 | if (!dst_count_p) |
232 | 0 | free(dst_count); |
233 | 0 | *src_copied = sc; |
234 | 0 | *literal_added = la; |
235 | 0 | return 0; |
236 | 0 | } |