optimize diffcore-delta by sorting hash entries.
[gitweb.git] / diffcore-rename.c
index 6bde4396f212833cc1d411e723d5215c086e7c2d..142e5376dd741377c311075816f139a0949ee82f 100644 (file)
@@ -184,7 +184,8 @@ static int estimate_similarity(struct diff_filespec *src,
        if (base_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE)
                return 0;
 
-       if (diff_populate_filespec(src, 0) || diff_populate_filespec(dst, 0))
+       if ((!src->cnt_data && diff_populate_filespec(src, 0))
+               || (!dst->cnt_data && diff_populate_filespec(dst, 0)))
                return 0; /* error but caught downstream */
 
 
@@ -298,10 +299,25 @@ void diffcore_rename(struct diff_options *options)
                else if (detect_rename == DIFF_DETECT_COPY)
                        register_rename_src(p->one, 1, p->score);
        }
-       if (rename_dst_nr == 0 || rename_src_nr == 0 ||
-           (0 < rename_limit && rename_limit < rename_dst_nr))
+       if (rename_dst_nr == 0 || rename_src_nr == 0)
                goto cleanup; /* nothing to do */
 
+       /*
+        * This basically does a test for the rename matrix not
+        * growing larger than a "rename_limit" square matrix, ie:
+        *
+        *    rename_dst_nr * rename_src_nr > rename_limit * rename_limit
+        *
+        * but handles the potential overflow case specially (and we
+        * assume at least 32-bit integers)
+        */
+       if (rename_limit <= 0 || rename_limit > 32767)
+               rename_limit = 32767;
+       if (rename_dst_nr > rename_limit && rename_src_nr > rename_limit)
+               goto cleanup;
+       if (rename_dst_nr * rename_src_nr > rename_limit * rename_limit)
+               goto cleanup;
+
        /* We really want to cull the candidates list early
         * with cheap tests in order to avoid doing deltas.
         * The first round matches up the up-to-date entries,
@@ -362,10 +378,10 @@ void diffcore_rename(struct diff_options *options)
                        m->score = estimate_similarity(one, two,
                                                       minimum_score);
                        m->name_score = basename_same(one, two);
-                       diff_free_filespec_data(one);
+                       diff_free_filespec_blob(one);
                }
                /* We do not need the text anymore */
-               diff_free_filespec_data(two);
+               diff_free_filespec_blob(two);
                dst_cnt++;
        }
        /* cost matrix sorted by most to least similar pair */