bdiff.c
356 lines
| 7.4 KiB
| text/x-c
|
CLexer
/ mercurial / bdiff.c
|
r400 | /* | ||
bdiff.c - efficient binary diff extension for Mercurial | ||||
Copyright 2005 Matt Mackall <mpm@selenic.com> | ||||
This software may be used and distributed according to the terms of | ||||
the GNU General Public License, incorporated herein by reference. | ||||
Based roughly on Python difflib | ||||
*/ | ||||
#include <Python.h> | ||||
#include <stdlib.h> | ||||
#include <string.h> | ||||
|
r867 | |||
#ifdef __hpux | ||||
#define inline | ||||
#endif | ||||
|
r411 | #ifdef _WIN32 | ||
|
r551 | #ifdef _MSC_VER | ||
#define inline __inline | ||||
typedef unsigned long uint32_t; | ||||
#else | ||||
|
r510 | #include <stdint.h> | ||
|
r551 | #endif | ||
|
r411 | static uint32_t htonl(uint32_t x) | ||
{ | ||||
return ((x & 0x000000ffUL) << 24) | | ||||
((x & 0x0000ff00UL) << 8) | | ||||
((x & 0x00ff0000UL) >> 8) | | ||||
((x & 0xff000000UL) >> 24); | ||||
} | ||||
#else | ||||
|
r510 | #include <sys/types.h> | ||
|
r597 | #include <arpa/inet.h> | ||
|
r411 | #endif | ||
|
r400 | |||
struct line { | ||||
|
r433 | int h, len, n, e; | ||
|
r400 | const char *l; | ||
}; | ||||
|
r474 | struct pos { | ||
int pos, len; | ||||
}; | ||||
|
r400 | struct hunk { | ||
int a1, a2, b1, b2; | ||||
}; | ||||
struct hunklist { | ||||
struct hunk *base, *head; | ||||
}; | ||||
|
r1397 | static inline uint32_t rol32(uint32_t word, unsigned int shift) | ||
|
r400 | { | ||
return (word << shift) | (word >> (32 - shift)); | ||||
} | ||||
int splitlines(const char *a, int len, struct line **lr) | ||||
{ | ||||
int h, i; | ||||
const char *p, *b = a; | ||||
struct line *l; | ||||
/* count the lines */ | ||||
i = 1; /* extra line for sentinel */ | ||||
for (p = a; p < a + len; p++) | ||||
if (*p == '\n' || p == a + len - 1) | ||||
i++; | ||||
*lr = l = malloc(sizeof(struct line) * i); | ||||
if (!l) | ||||
return -1; | ||||
/* build the line array and calculate hashes */ | ||||
h = 0; | ||||
for (p = a; p < a + len; p++) { | ||||
h = *p + rol32(h, 7); /* a simple hash from GNU diff */ | ||||
if (*p == '\n' || p == a + len - 1) { | ||||
l->len = p - b + 1; | ||||
|
r433 | l->h = h * l->len; | ||
|
r400 | l->l = b; | ||
l->n = -1; | ||||
l++; | ||||
b = p + 1; | ||||
h = 0; | ||||
} | ||||
} | ||||
/* set up a sentinel */ | ||||
l->h = l->len = 0; | ||||
l->l = a + len; | ||||
return i - 1; | ||||
} | ||||
int inline cmp(struct line *a, struct line *b) | ||||
{ | ||||
|
r433 | return a->h != b->h || a->len != b->len || memcmp(a->l, b->l, a->len); | ||
|
r400 | } | ||
static int equatelines(struct line *a, int an, struct line *b, int bn) | ||||
{ | ||||
|
r474 | int i, j, buckets = 1, t; | ||
struct pos *h; | ||||
|
r400 | |||
/* build a hash table of the next highest power of 2 */ | ||||
while (buckets < bn + 1) | ||||
buckets *= 2; | ||||
|
r474 | h = malloc(buckets * sizeof(struct pos)); | ||
|
r400 | buckets = buckets - 1; | ||
|
r474 | if (!h) | ||
|
r400 | return 0; | ||
/* clear the hash table */ | ||||
|
r474 | for (i = 0; i <= buckets; i++) { | ||
h[i].pos = -1; | ||||
h[i].len = 0; | ||||
} | ||||
|
r400 | |||
/* add lines to the hash table chains */ | ||||
for (i = bn - 1; i >= 0; i--) { | ||||
/* find the equivalence class */ | ||||
|
r474 | for (j = b[i].h & buckets; h[j].pos != -1; | ||
j = (j + 1) & buckets) | ||||
if (!cmp(b + i, b + h[j].pos)) | ||||
|
r400 | break; | ||
/* add to the head of the equivalence class */ | ||||
|
r474 | b[i].n = h[j].pos; | ||
|
r433 | b[i].e = j; | ||
|
r474 | h[j].pos = i; | ||
h[j].len++; /* keep track of popularity */ | ||||
|
r400 | } | ||
/* compute popularity threshold */ | ||||
t = (bn >= 200) ? bn / 100 : bn + 1; | ||||
/* match items in a to their equivalence class in b */ | ||||
for (i = 0; i < an; i++) { | ||||
/* find the equivalence class */ | ||||
|
r474 | for (j = a[i].h & buckets; h[j].pos != -1; | ||
j = (j + 1) & buckets) | ||||
if (!cmp(a + i, b + h[j].pos)) | ||||
|
r400 | break; | ||
|
r433 | a[i].e = j; /* use equivalence class for quick compare */ | ||
|
r1542 | if (h[j].len <= t) | ||
|
r474 | a[i].n = h[j].pos; /* point to head of match list */ | ||
|
r400 | else | ||
a[i].n = -1; /* too popular */ | ||||
} | ||||
/* discard hash tables */ | ||||
free(h); | ||||
return 1; | ||||
} | ||||
|
r474 | static int longest_match(struct line *a, struct line *b, struct pos *pos, | ||
|
r400 | int a1, int a2, int b1, int b2, int *omi, int *omj) | ||
{ | ||||
int mi = a1, mj = b1, mk = 0, mb = 0, i, j, k; | ||||
for (i = a1; i < a2; i++) { | ||||
/* skip things before the current block */ | ||||
for (j = a[i].n; j != -1 && j < b1; j = b[j].n) | ||||
; | ||||
/* loop through all lines match a[i] in b */ | ||||
for (; j != -1 && j < b2; j = b[j].n) { | ||||
/* does this extend an earlier match? */ | ||||
|
r474 | if (i > a1 && j > b1 && pos[j - 1].pos == i - 1) | ||
k = pos[j - 1].len + 1; | ||||
|
r400 | else | ||
k = 1; | ||||
|
r474 | pos[j].pos = i; | ||
pos[j].len = k; | ||||
|
r400 | |||
/* best match so far? */ | ||||
if (k > mk) { | ||||
mi = i; | ||||
mj = j; | ||||
mk = k; | ||||
} | ||||
} | ||||
} | ||||
if (mk) { | ||||
mi = mi - mk + 1; | ||||
mj = mj - mk + 1; | ||||
} | ||||
/* expand match to include neighboring popular lines */ | ||||
while (mi - mb > a1 && mj - mb > b1 && | ||||
|
r433 | a[mi - mb - 1].e == b[mj - mb - 1].e) | ||
|
r400 | mb++; | ||
while (mi + mk < a2 && mj + mk < b2 && | ||||
|
r433 | a[mi + mk].e == b[mj + mk].e) | ||
|
r400 | mk++; | ||
*omi = mi - mb; | ||||
*omj = mj - mb; | ||||
return mk + mb; | ||||
} | ||||
|
r474 | static void recurse(struct line *a, struct line *b, struct pos *pos, | ||
|
r400 | int a1, int a2, int b1, int b2, struct hunklist *l) | ||
{ | ||||
int i, j, k; | ||||
/* find the longest match in this chunk */ | ||||
|
r474 | k = longest_match(a, b, pos, a1, a2, b1, b2, &i, &j); | ||
|
r400 | if (!k) | ||
return; | ||||
/* and recurse on the remaining chunks on either side */ | ||||
|
r474 | recurse(a, b, pos, a1, i, b1, j, l); | ||
|
r400 | l->head->a1 = i; | ||
l->head->a2 = i + k; | ||||
l->head->b1 = j; | ||||
l->head->b2 = j + k; | ||||
l->head++; | ||||
|
r474 | recurse(a, b, pos, i + k, a2, j + k, b2, l); | ||
|
r400 | } | ||
|
r433 | static struct hunklist diff(struct line *a, int an, struct line *b, int bn) | ||
|
r400 | { | ||
|
r433 | struct hunklist l; | ||
|
r474 | struct pos *pos; | ||
int t; | ||||
|
r433 | |||
/* allocate and fill arrays */ | ||||
t = equatelines(a, an, b, bn); | ||||
|
r474 | pos = calloc(bn, sizeof(struct pos)); | ||
|
r827 | /* we can't have more matches than lines in the shorter file */ | ||
l.head = l.base = malloc(sizeof(struct hunk) * ((an<bn ? an:bn) + 1)); | ||||
|
r433 | |||
|
r474 | if (pos && l.base && t) { | ||
|
r433 | /* generate the matching block list */ | ||
|
r474 | recurse(a, b, pos, 0, an, 0, bn, &l); | ||
|
r433 | l.head->a1 = an; | ||
l.head->b1 = bn; | ||||
l.head++; | ||||
} | ||||
|
r474 | free(pos); | ||
|
r433 | return l; | ||
} | ||||
static PyObject *blocks(PyObject *self, PyObject *args) | ||||
{ | ||||
|
r435 | PyObject *sa, *sb, *rl = NULL, *m; | ||
|
r433 | struct line *a, *b; | ||
|
r970 | struct hunklist l = {NULL, NULL}; | ||
|
r400 | struct hunk *h; | ||
|
r433 | int an, bn, pos = 0; | ||
|
r400 | |||
if (!PyArg_ParseTuple(args, "SS:bdiff", &sa, &sb)) | ||||
return NULL; | ||||
|
r433 | an = splitlines(PyString_AsString(sa), PyString_Size(sa), &a); | ||
bn = splitlines(PyString_AsString(sb), PyString_Size(sb), &b); | ||||
if (!a || !b) | ||||
goto nomem; | ||||
l = diff(a, an, b, bn); | ||||
rl = PyList_New(l.head - l.base); | ||||
if (!l.head || !rl) | ||||
goto nomem; | ||||
|
r1542 | for (h = l.base; h != l.head; h++) { | ||
|
r433 | m = Py_BuildValue("iiii", h->a1, h->a2, h->b1, h->b2); | ||
PyList_SetItem(rl, pos, m); | ||||
pos++; | ||||
} | ||||
nomem: | ||||
free(a); | ||||
free(b); | ||||
free(l.base); | ||||
return rl ? rl : PyErr_NoMemory(); | ||||
} | ||||
static PyObject *bdiff(PyObject *self, PyObject *args) | ||||
{ | ||||
PyObject *sa, *sb, *result = NULL; | ||||
struct line *al, *bl; | ||||
|
r970 | struct hunklist l = {NULL, NULL}; | ||
|
r433 | struct hunk *h; | ||
char encode[12], *rb; | ||||
int an, bn, len = 0, la = 0, lb = 0; | ||||
if (!PyArg_ParseTuple(args, "SS:bdiff", &sa, &sb)) | ||||
return NULL; | ||||
|
r400 | an = splitlines(PyString_AsString(sa), PyString_Size(sa), &al); | ||
bn = splitlines(PyString_AsString(sb), PyString_Size(sb), &bl); | ||||
|
r433 | if (!al || !bl) | ||
|
r400 | goto nomem; | ||
|
r433 | l = diff(al, an, bl, bn); | ||
if (!l.head) | ||||
goto nomem; | ||||
|
r400 | |||
/* calculate length of output */ | ||||
|
r1542 | for (h = l.base; h != l.head; h++) { | ||
|
r400 | if (h->a1 != la || h->b1 != lb) | ||
len += 12 + bl[h->b1].l - bl[lb].l; | ||||
la = h->a2; | ||||
lb = h->b2; | ||||
} | ||||
result = PyString_FromStringAndSize(NULL, len); | ||||
if (!result) | ||||
goto nomem; | ||||
/* build binary patch */ | ||||
rb = PyString_AsString(result); | ||||
la = lb = 0; | ||||
|
r1542 | for (h = l.base; h != l.head; h++) { | ||
|
r400 | if (h->a1 != la || h->b1 != lb) { | ||
len = bl[h->b1].l - bl[lb].l; | ||||
*(uint32_t *)(encode) = htonl(al[la].l - al->l); | ||||
*(uint32_t *)(encode + 4) = htonl(al[h->a1].l - al->l); | ||||
*(uint32_t *)(encode + 8) = htonl(len); | ||||
memcpy(rb, encode, 12); | ||||
memcpy(rb + 12, bl[lb].l, len); | ||||
rb += 12 + len; | ||||
} | ||||
la = h->a2; | ||||
lb = h->b2; | ||||
} | ||||
nomem: | ||||
free(al); | ||||
free(bl); | ||||
free(l.base); | ||||
return result ? result : PyErr_NoMemory(); | ||||
} | ||||
static char mdiff_doc[] = "Efficient binary diff."; | ||||
static PyMethodDef methods[] = { | ||||
{"bdiff", bdiff, METH_VARARGS, "calculate a binary diff\n"}, | ||||
|
r433 | {"blocks", blocks, METH_VARARGS, "find a list of matching lines\n"}, | ||
|
r400 | {NULL, NULL} | ||
}; | ||||
PyMODINIT_FUNC initbdiff(void) | ||||
{ | ||||
Py_InitModule3("bdiff", methods, mdiff_doc); | ||||
} | ||||
|
r1542 | |||