binary patch.
[gitweb.git] / diff.c
diff --git a/diff.c b/diff.c
index a14e6644ca9a39d0a0e1702ccefadf7a5bb64081..b14d897f196ddab42d56285ffe14f7f616b01f71 100644 (file)
--- a/diff.c
+++ b/diff.c
@@ -8,7 +8,8 @@
 #include "quote.h"
 #include "diff.h"
 #include "diffcore.h"
-#include "xdiff/xdiff.h"
+#include "delta.h"
+#include "xdiff-interface.h"
 
 static int use_size_cache;
 
@@ -195,6 +196,286 @@ static int fn_out(void *priv, mmbuffer_t *mb, int nbuf)
        return 0;
 }
 
+static char *pprint_rename(const char *a, const char *b)
+{
+       const char *old = a;
+       const char *new = b;
+       char *name = NULL;
+       int pfx_length, sfx_length;
+       int len_a = strlen(a);
+       int len_b = strlen(b);
+
+       /* Find common prefix */
+       pfx_length = 0;
+       while (*old && *new && *old == *new) {
+               if (*old == '/')
+                       pfx_length = old - a + 1;
+               old++;
+               new++;
+       }
+
+       /* Find common suffix */
+       old = a + len_a;
+       new = b + len_b;
+       sfx_length = 0;
+       while (a <= old && b <= new && *old == *new) {
+               if (*old == '/')
+                       sfx_length = len_a - (old - a);
+               old--;
+               new--;
+       }
+
+       /*
+        * pfx{mid-a => mid-b}sfx
+        * {pfx-a => pfx-b}sfx
+        * pfx{sfx-a => sfx-b}
+        * name-a => name-b
+        */
+       if (pfx_length + sfx_length) {
+               name = xmalloc(len_a + len_b - pfx_length - sfx_length + 7);
+               sprintf(name, "%.*s{%.*s => %.*s}%s",
+                       pfx_length, a,
+                       len_a - pfx_length - sfx_length, a + pfx_length,
+                       len_b - pfx_length - sfx_length, b + pfx_length,
+                       a + len_a - sfx_length);
+       }
+       else {
+               name = xmalloc(len_a + len_b + 5);
+               sprintf(name, "%s => %s", a, b);
+       }
+       return name;
+}
+
+struct diffstat_t {
+       struct xdiff_emit_state xm;
+
+       int nr;
+       int alloc;
+       struct diffstat_file {
+               char *name;
+               unsigned is_unmerged:1;
+               unsigned is_binary:1;
+               unsigned is_renamed:1;
+               unsigned int added, deleted;
+       } **files;
+};
+
+static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
+                                         const char *name_a,
+                                         const char *name_b)
+{
+       struct diffstat_file *x;
+       x = xcalloc(sizeof (*x), 1);
+       if (diffstat->nr == diffstat->alloc) {
+               diffstat->alloc = alloc_nr(diffstat->alloc);
+               diffstat->files = xrealloc(diffstat->files,
+                               diffstat->alloc * sizeof(x));
+       }
+       diffstat->files[diffstat->nr++] = x;
+       if (name_b) {
+               x->name = pprint_rename(name_a, name_b);
+               x->is_renamed = 1;
+       }
+       else
+               x->name = strdup(name_a);
+       return x;
+}
+
+static void diffstat_consume(void *priv, char *line, unsigned long len)
+{
+       struct diffstat_t *diffstat = priv;
+       struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
+
+       if (line[0] == '+')
+               x->added++;
+       else if (line[0] == '-')
+               x->deleted++;
+}
+
+static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
+static const char minuses[]= "----------------------------------------------------------------------";
+
+static void show_stats(struct diffstat_t* data)
+{
+       char *prefix = "";
+       int i, len, add, del, total, adds = 0, dels = 0;
+       int max, max_change = 0, max_len = 0;
+       int total_files = data->nr;
+
+       if (data->nr == 0)
+               return;
+
+       for (i = 0; i < data->nr; i++) {
+               struct diffstat_file *file = data->files[i];
+
+               len = strlen(file->name);
+               if (max_len < len)
+                       max_len = len;
+
+               if (file->is_binary || file->is_unmerged)
+                       continue;
+               if (max_change < file->added + file->deleted)
+                       max_change = file->added + file->deleted;
+       }
+
+       for (i = 0; i < data->nr; i++) {
+               char *name = data->files[i]->name;
+               int added = data->files[i]->added;
+               int deleted = data->files[i]->deleted;
+
+               if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
+                       char *qname = xmalloc(len + 1);
+                       quote_c_style(name, qname, NULL, 0);
+                       free(name);
+                       data->files[i]->name = name = qname;
+               }
+
+               /*
+                * "scale" the filename
+                */
+               len = strlen(name);
+               max = max_len;
+               if (max > 50)
+                       max = 50;
+               if (len > max) {
+                       char *slash;
+                       prefix = "...";
+                       max -= 3;
+                       name += len - max;
+                       slash = strchr(name, '/');
+                       if (slash)
+                               name = slash;
+               }
+               len = max;
+
+               /*
+                * scale the add/delete
+                */
+               max = max_change;
+               if (max + len > 70)
+                       max = 70 - len;
+
+               if (data->files[i]->is_binary) {
+                       printf(" %s%-*s |  Bin\n", prefix, len, name);
+                       goto free_diffstat_file;
+               }
+               else if (data->files[i]->is_unmerged) {
+                       printf(" %s%-*s |  Unmerged\n", prefix, len, name);
+                       goto free_diffstat_file;
+               }
+               else if (!data->files[i]->is_renamed &&
+                        (added + deleted == 0)) {
+                       total_files--;
+                       goto free_diffstat_file;
+               }
+
+               add = added;
+               del = deleted;
+               total = add + del;
+               adds += add;
+               dels += del;
+
+               if (max_change > 0) {
+                       total = (total * max + max_change / 2) / max_change;
+                       add = (add * max + max_change / 2) / max_change;
+                       del = total - add;
+               }
+               printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
+                               len, name, added + deleted,
+                               add, pluses, del, minuses);
+       free_diffstat_file:
+               free(data->files[i]->name);
+               free(data->files[i]);
+       }
+       free(data->files);
+       printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
+                       total_files, adds, dels);
+}
+
+static void *encode_delta_size(void *data, unsigned long size)
+{
+       unsigned char *cp = data;
+       *cp++ = size;
+       size >>= 7;
+       while (size) {
+               cp[-1] |= 0x80;
+               *cp++ = size;
+               size >>= 7;
+       }
+       return cp;
+}
+
+static void *safe_diff_delta(const unsigned char *src, unsigned long src_size,
+                            const unsigned char *dst, unsigned long dst_size,
+                            unsigned long *delta_size)
+{
+       unsigned long bufsize;
+       unsigned char *data;
+       unsigned char *cp;
+
+       if (src_size && dst_size)
+               return diff_delta(src, src_size, dst, dst_size, delta_size, 0);
+
+       /* diff-delta does not like to do delta with empty, so
+        * we do that by hand here.  Sigh...
+        */
+
+       if (!src_size)
+               /* literal copy can be done only 127-byte at a time.
+                */
+               bufsize = dst_size + (dst_size / 127) + 40;
+       else
+               bufsize = 40;
+       data = xmalloc(bufsize);
+       cp = encode_delta_size(data, src_size);
+       cp = encode_delta_size(cp, dst_size);
+
+       if (dst_size) {
+               /* copy out literally */
+               while (dst_size) {
+                       int sz = (127 < dst_size) ? 127 : dst_size;
+                       *cp++ = sz;
+                       dst_size -= sz;
+                       while (sz) {
+                               *cp++ = *dst++;
+                               sz--;
+                       }
+               }
+       }
+       *delta_size = (cp - data);
+       return data;
+}
+
+static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
+{
+       void *delta, *cp;
+       unsigned long delta_size;
+
+       printf("GIT binary patch\n");
+       delta = safe_diff_delta(one->ptr, one->size,
+                               two->ptr, two->size,
+                               &delta_size);
+       if (!delta)
+               die("unable to generate binary diff");
+
+       /* emit delta encoded in base85 */
+       cp = delta;
+       while (delta_size) {
+               int bytes = (52 < delta_size) ? 52 : delta_size;
+               char line[70];
+               delta_size -= bytes;
+               if (bytes <= 26)
+                       line[0] = bytes + 'A' - 1;
+               else
+                       line[0] = bytes - 26 + 'a' - 1;
+               encode_85(line + 1, cp, bytes);
+               cp += bytes;
+               puts(line);
+       }
+       printf("\n");
+       free(delta);
+}
+
 #define FIRST_FEW_BYTES 8000
 static int mmfile_is_binary(mmfile_t *mf)
 {
@@ -211,6 +492,7 @@ static void builtin_diff(const char *name_a,
                         struct diff_filespec *one,
                         struct diff_filespec *two,
                         const char *xfrm_msg,
+                        struct diff_options *o,
                         int complete_rewrite)
 {
        mmfile_t mf1, mf2;
@@ -255,8 +537,13 @@ static void builtin_diff(const char *name_a,
        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
                die("unable to read files to diff");
 
-       if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))
-               printf("Binary files %s and %s differ\n", lbl[0], lbl[1]);
+       if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2)) {
+               if (o->full_index)
+                       emit_binary_diff(&mf1, &mf2);
+               else
+                       printf("Binary files %s and %s differ\n",
+                              lbl[0], lbl[1]);
+       }
        else {
                /* Crazy xdl interfaces.. */
                const char *diffopts = getenv("GIT_DIFF_OPTS");
@@ -286,6 +573,48 @@ static void builtin_diff(const char *name_a,
        return;
 }
 
+static void builtin_diffstat(const char *name_a, const char *name_b,
+                            struct diff_filespec *one,
+                            struct diff_filespec *two,
+                            struct diffstat_t *diffstat,
+                            int complete_rewrite)
+{
+       mmfile_t mf1, mf2;
+       struct diffstat_file *data;
+
+       data = diffstat_add(diffstat, name_a, name_b);
+
+       if (!one || !two) {
+               data->is_unmerged = 1;
+               return;
+       }
+       if (complete_rewrite) {
+               diff_populate_filespec(one, 0);
+               diff_populate_filespec(two, 0);
+               data->deleted = count_lines(one->data, one->size);
+               data->added = count_lines(two->data, two->size);
+               return;
+       }
+       if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
+               die("unable to read files to diff");
+
+       if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))
+               data->is_binary = 1;
+       else {
+               /* Crazy xdl interfaces.. */
+               xpparam_t xpp;
+               xdemitconf_t xecfg;
+               xdemitcb_t ecb;
+
+               xpp.flags = XDF_NEED_MINIMAL;
+               xecfg.ctxlen = 0;
+               xecfg.flags = 0;
+               ecb.outf = xdiff_outf;
+               ecb.priv = diffstat;
+               xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
+       }
+}
+
 struct diff_filespec *alloc_filespec(const char *path)
 {
        int namelen = strlen(path);
@@ -690,6 +1019,7 @@ static void run_diff_cmd(const char *pgm,
                         struct diff_filespec *one,
                         struct diff_filespec *two,
                         const char *xfrm_msg,
+                        struct diff_options *o,
                         int complete_rewrite)
 {
        if (pgm) {
@@ -699,7 +1029,7 @@ static void run_diff_cmd(const char *pgm,
        }
        if (one && two)
                builtin_diff(name, other ? other : name,
-                            one, two, xfrm_msg, complete_rewrite);
+                            one, two, xfrm_msg, o, complete_rewrite);
        else
                printf("* Unmerged path %s\n", name);
 }
@@ -733,7 +1063,7 @@ static void run_diff(struct diff_filepair *p, struct diff_options *o)
 
        if (DIFF_PAIR_UNMERGED(p)) {
                /* unmerged */
-               run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, 0);
+               run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
                return;
        }
 
@@ -780,14 +1110,12 @@ static void run_diff(struct diff_filepair *p, struct diff_options *o)
        }
 
        if (memcmp(one->sha1, two->sha1, 20)) {
-               char one_sha1[41];
                int abbrev = o->full_index ? 40 : DEFAULT_ABBREV;
-               memcpy(one_sha1, sha1_to_hex(one->sha1), 41);
 
                len += snprintf(msg + len, sizeof(msg) - len,
                                "index %.*s..%.*s",
-                               abbrev, one_sha1, abbrev,
-                               sha1_to_hex(two->sha1));
+                               abbrev, sha1_to_hex(one->sha1),
+                               abbrev, sha1_to_hex(two->sha1));
                if (one->mode == two->mode)
                        len += snprintf(msg + len, sizeof(msg) - len,
                                        " %06o", one->mode);
@@ -805,20 +1133,44 @@ static void run_diff(struct diff_filepair *p, struct diff_options *o)
                 * needs to be split into deletion and creation.
                 */
                struct diff_filespec *null = alloc_filespec(two->path);
-               run_diff_cmd(NULL, name, other, one, null, xfrm_msg, 0);
+               run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
                free(null);
                null = alloc_filespec(one->path);
-               run_diff_cmd(NULL, name, other, null, two, xfrm_msg, 0);
+               run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
                free(null);
        }
        else
-               run_diff_cmd(pgm, name, other, one, two, xfrm_msg,
+               run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
                             complete_rewrite);
 
        free(name_munged);
        free(other_munged);
 }
 
+static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
+                        struct diffstat_t *diffstat)
+{
+       const char *name;
+       const char *other;
+       int complete_rewrite = 0;
+
+       if (DIFF_PAIR_UNMERGED(p)) {
+               /* unmerged */
+               builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, 0);
+               return;
+       }
+
+       name = p->one->path;
+       other = (strcmp(name, p->two->path) ? p->two->path : NULL);
+
+       diff_fill_sha1_info(p->one);
+       diff_fill_sha1_info(p->two);
+
+       if (p->status == DIFF_STATUS_MODIFIED && p->score)
+               complete_rewrite = 1;
+       builtin_diffstat(name, other, p->one, p->two, diffstat, complete_rewrite);
+}
+
 void diff_setup(struct diff_options *options)
 {
        memset(options, 0, sizeof(*options));
@@ -837,6 +1189,15 @@ int diff_setup_done(struct diff_options *options)
             options->detect_rename != DIFF_DETECT_COPY) ||
            (0 <= options->rename_limit && !options->detect_rename))
                return -1;
+
+       /*
+        * These cases always need recursive; we do not drop caller-supplied
+        * recursive bits for other formats here.
+        */
+       if ((options->output_format == DIFF_FORMAT_PATCH) ||
+           (options->output_format == DIFF_FORMAT_DIFFSTAT))
+               options->recursive = 1;
+
        if (options->detect_rename && options->rename_limit < 0)
                options->rename_limit = diff_rename_limit_default;
        if (options->setup & DIFF_SETUP_USE_CACHE) {
@@ -866,6 +1227,12 @@ int diff_opt_parse(struct diff_options *options, const char **av, int ac)
                options->output_format = DIFF_FORMAT_PATCH;
                options->with_raw = 1;
        }
+       else if (!strcmp(arg, "--stat"))
+               options->output_format = DIFF_FORMAT_DIFFSTAT;
+       else if (!strcmp(arg, "--patch-with-stat")) {
+               options->output_format = DIFF_FORMAT_PATCH;
+               options->with_stat = 1;
+       }
        else if (!strcmp(arg, "-z"))
                options->line_termination = 0;
        else if (!strncmp(arg, "-l", 2))
@@ -1160,11 +1527,24 @@ static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
 
        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
-               return; /* no tree diffs in patch format */ 
+               return; /* no tree diffs in patch format */
 
        run_diff(p, o);
 }
 
+static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
+                           struct diffstat_t *diffstat)
+{
+       if (diff_unmodified_pair(p))
+               return;
+
+       if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
+           (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
+               return; /* no tree diffs in patch format */
+
+       run_diffstat(p, o, diffstat);
+}
+
 int diff_queue_is_empty(void)
 {
        struct diff_queue_struct *q = &diff_queued_diff;
@@ -1276,7 +1656,8 @@ static void diff_resolve_rename_copy(void)
 
 static void flush_one_pair(struct diff_filepair *p,
                           int diff_output_format,
-                          struct diff_options *options)
+                          struct diff_options *options,
+                          struct diffstat_t *diffstat)
 {
        int inter_name_termination = '\t';
        int line_termination = options->line_termination;
@@ -1291,6 +1672,9 @@ static void flush_one_pair(struct diff_filepair *p,
                break;
        default:
                switch (diff_output_format) {
+               case DIFF_FORMAT_DIFFSTAT:
+                       diff_flush_stat(p, options, diffstat);
+                       break;
                case DIFF_FORMAT_PATCH:
                        diff_flush_patch(p, options);
                        break;
@@ -1316,19 +1700,42 @@ void diff_flush(struct diff_options *options)
        struct diff_queue_struct *q = &diff_queued_diff;
        int i;
        int diff_output_format = options->output_format;
+       struct diffstat_t *diffstat = NULL;
+
+       if (diff_output_format == DIFF_FORMAT_DIFFSTAT || options->with_stat) {
+               diffstat = xcalloc(sizeof (struct diffstat_t), 1);
+               diffstat->xm.consume = diffstat_consume;
+       }
 
        if (options->with_raw) {
                for (i = 0; i < q->nr; i++) {
                        struct diff_filepair *p = q->queue[i];
-                       flush_one_pair(p, DIFF_FORMAT_RAW, options);
+                       flush_one_pair(p, DIFF_FORMAT_RAW, options, NULL);
                }
                putchar(options->line_termination);
        }
+       if (options->with_stat) {
+               for (i = 0; i < q->nr; i++) {
+                       struct diff_filepair *p = q->queue[i];
+                       flush_one_pair(p, DIFF_FORMAT_DIFFSTAT, options,
+                                      diffstat);
+               }
+               show_stats(diffstat);
+               free(diffstat);
+               diffstat = NULL;
+               putchar(options->line_termination);
+       }
        for (i = 0; i < q->nr; i++) {
                struct diff_filepair *p = q->queue[i];
-               flush_one_pair(p, diff_output_format, options);
+               flush_one_pair(p, diff_output_format, options, diffstat);
                diff_free_filepair(p);
        }
+
+       if (diffstat) {
+               show_stats(diffstat);
+               free(diffstat);
+       }
+
        free(q->queue);
        q->queue = NULL;
        q->nr = q->alloc = 0;