diffcore-delta.c

   1 #include "cache.h"
   2 #include "diff.h"
   3 #include "diffcore.h"
   4
   5 /*
   6  * Idea here is very simple.
   7  *
   8  * We have total of (sz-N+1) N-byte overlapping sequences in buf whose
   9  * size is sz.  If the same N-byte sequence appears in both source and
  10  * destination, we say the byte that starts that sequence is shared
  11  * between them (i.e. copied from source to destination).
  12  *
  13  * For each possible N-byte sequence, if the source buffer has more
  14  * instances of it than the destination buffer, that means the
  15  * difference are the number of bytes not copied from source to
  16  * destination.  If the counts are the same, everything was copied
  17  * from source to destination.  If the destination has more,
  18  * everything was copied, and destination added more.
  19  *
  20  * We are doing an approximation so we do not really have to waste
  21  * memory by actually storing the sequence.  We just hash them into
  22  * somewhere around 2^16 hashbuckets and count the occurrences.
  23  *
  24  * The length of the sequence is arbitrarily set to 8 for now.
  25  */
  26
  27 /* Wild guess at the initial hash size */
  28 #define INITIAL_HASH_SIZE 9
  29
  30 /* We leave more room in smaller hash but do not let it
  31  * grow to have unused hole too much.
  32  */
  33 #define INITIAL_FREE(sz_log2) ((1<<(sz_log2))*(sz_log2-3)/(sz_log2))
  34
  35 /* A prime rather carefully chosen between 2^16..2^17, so that
  36  * HASHBASE < INITIAL_FREE(17).  We want to keep the maximum hashtable
  37  * size under the current 2<<17 maximum, which can hold this many
  38  * different values before overflowing to hashtable of size 2<<18.
  39  */
  40 #define HASHBASE 107927
  41
  42 struct spanhash {
  43         unsigned long hashval;
  44         unsigned long cnt;
  45 };
  46 struct spanhash_top {
  47         int alloc_log2;
  48         int free;
  49         struct spanhash data[FLEX_ARRAY];
  50 };
  51
  52 static struct spanhash *spanhash_find(struct spanhash_top *top,
  53                                       unsigned long hashval)
  54 {
  55         int sz = 1 << top->alloc_log2;
  56         int bucket = hashval & (sz - 1);
  57         while (1) {
  58                 struct spanhash *h = &(top->data[bucket++]);
  59                 if (!h->cnt)
  60                         return NULL;
  61                 if (h->hashval == hashval)
  62                         return h;
  63                 if (sz <= bucket)
  64                         bucket = 0;
  65         }
  66 }
  67
  68 static struct spanhash_top *spanhash_rehash(struct spanhash_top *orig)
  69 {
  70         struct spanhash_top *new;
  71         int i;
  72         int osz = 1 << orig->alloc_log2;
  73         int sz = osz << 1;
  74
  75         new = xmalloc(sizeof(*orig) + sizeof(struct spanhash) * sz);
  76         new->alloc_log2 = orig->alloc_log2 + 1;
  77         new->free = INITIAL_FREE(new->alloc_log2);
  78         memset(new->data, 0, sizeof(struct spanhash) * sz);
  79         for (i = 0; i < osz; i++) {
  80                 struct spanhash *o = &(orig->data[i]);
  81                 int bucket;
  82                 if (!o->cnt)
  83                         continue;
  84                 bucket = o->hashval & (sz - 1);
  85                 while (1) {
  86                         struct spanhash *h = &(new->data[bucket++]);
  87                         if (!h->cnt) {
  88                                 h->hashval = o->hashval;
  89                                 h->cnt = o->cnt;
  90                                 new->free--;
  91                                 break;
  92                         }
  93                         if (sz <= bucket)
  94                                 bucket = 0;
  95                 }
  96         }
  97         free(orig);
  98         return new;
  99 }
 100
 101 static struct spanhash_top *add_spanhash(struct spanhash_top *top,
 102                                          unsigned long hashval, int cnt)
 103 {
 104         int bucket, lim;
 105         struct spanhash *h;
 106
 107         lim = (1 << top->alloc_log2);
 108         bucket = hashval & (lim - 1);
 109         while (1) {
 110                 h = &(top->data[bucket++]);
 111                 if (!h->cnt) {
 112                         h->hashval = hashval;
 113                         h->cnt = cnt;
 114                         top->free--;
 115                         if (top->free < 0)
 116                                 return spanhash_rehash(top);
 117                         return top;
 118                 }
 119                 if (h->hashval == hashval) {
 120                         h->cnt += cnt;
 121                         return top;
 122                 }
 123                 if (lim <= bucket)
 124                         bucket = 0;
 125         }
 126 }
 127
 128 static struct spanhash_top *hash_chars(unsigned char *buf, unsigned long sz)
 129 {
 130         int i, n;
 131         unsigned long accum1, accum2, hashval;
 132         struct spanhash_top *hash;
 133
 134         i = INITIAL_HASH_SIZE;
 135         hash = xmalloc(sizeof(*hash) + 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));
 139
 140         n = 0;
 141         accum1 = accum2 = 0;
 142         while (sz) {
 143                 unsigned long c = *buf++;
 144                 sz--;
 145                 accum1 = (accum1 << 7) | (accum2 >> 25);
 146                 accum2 = (accum2 << 7) | (accum1 >> 25);
 147                 accum1 += c;
 148                 if (++n < 64 && c != '\n')
 149                         continue;
 150                 hashval = (accum1 + accum2 * 0x61) % HASHBASE;
 151                 hash = add_spanhash(hash, hashval, n);
 152                 n = 0;
 153                 accum1 = accum2 = 0;
 154         }
 155         return hash;
 156 }
 157
 158 int diffcore_count_changes(void *src, unsigned long src_size,
 159                            void *dst, unsigned long dst_size,
 160                            void **src_count_p,
 161                            void **dst_count_p,
 162                            unsigned long delta_limit,
 163                            unsigned long *src_copied,
 164                            unsigned long *literal_added)
 165 {
 166         int i, ssz;
 167         struct spanhash_top *src_count, *dst_count;
 168         unsigned long sc, la;
 169
 170         src_count = dst_count = NULL;
 171         if (src_count_p)
 172                 src_count = *src_count_p;
 173         if (!src_count) {
 174                 src_count = hash_chars(src, src_size);
 175                 if (src_count_p)
 176                         *src_count_p = src_count;
 177         }
 178         if (dst_count_p)
 179                 dst_count = *dst_count_p;
 180         if (!dst_count) {
 181                 dst_count = hash_chars(dst, dst_size);
 182                 if (dst_count_p)
 183                         *dst_count_p = dst_count;
 184         }
 185         sc = la = 0;
 186
 187         ssz = 1 << src_count->alloc_log2;
 188         for (i = 0; i < ssz; i++) {
 189                 struct spanhash *s = &(src_count->data[i]);
 190                 struct spanhash *d;
 191                 unsigned dst_cnt, src_cnt;
 192                 if (!s->cnt)
 193                         continue;
 194                 src_cnt = s->cnt;
 195                 d = spanhash_find(dst_count, s->hashval);
 196                 dst_cnt = d ? d->cnt : 0;
 197                 if (src_cnt < dst_cnt) {
 198                         la += dst_cnt - src_cnt;
 199                         sc += src_cnt;
 200                 }
 201                 else
 202                         sc += dst_cnt;
 203         }
 204
 205         if (!src_count_p)
 206                 free(src_count);
 207         if (!dst_count_p)
 208                 free(dst_count);
 209         *src_copied = sc;
 210         *literal_added = la;
 211         return 0;
 212 }