Re: [PATCH v2 07/13] merge-ort: populate caches of rename detection results

On Mon, May 17, 2021 at 6:51 AM Derrick Stolee [off-list ref] wrote:

On 5/3/21 10:12 PM, Elijah Newren via GitGitGadget wrote:

quoted

From: Elijah Newren <redacted>

Fill in cache_pairs, cached_target_names, and cached_irrelevant based on
rename detection results.  Future commits will make use of these values.

Thank you for continuing to break this down into nice-sized pieces.

quoted

+static void possibly_cache_new_pair(struct rename_info *renames,
+                                 struct diff_filepair *p,
+                                 unsigned side,
+                                 char *new_path)
+{
+     char *old_value;
+     int dir_renamed_side = 0;
+
+     if (new_path) {
+             /*
+              * Directory renames happen on the other side of history from
+              * the side that adds new files to the old directory.
+              */
+             dir_renamed_side = 3 - side;

Neat trick. Side is in { 1, 2 } so this makes sense.

quoted

+     } else {
+             int val = strintmap_get(&renames->relevant_sources[side],
+                                     p->one->path);
+             if (val == RELEVANT_NO_MORE) {
+                     assert(p->status == 'D');
+                     strset_add(&renames->cached_irrelevant[side],
+                                p->one->path);

Ok, I see a transition here from a relevant side to an
irrelevant one.

quoted

+             }
+             if (val <= 0)
+                     return;
+     }
+
+     if (p->status == 'D') {
+             /*
+              * If we already had this delete, we'll just set it's value
+              * to NULL again, so no harm.
+              */
+             strmap_put(&renames->cached_pairs[side], p->one->path, NULL);
+     } else if (p->status == 'R') {
+             if (new_path) {
+                     new_path = xstrdup(new_path);
+                     old_value = strmap_put(&renames->cached_pairs[dir_renamed_side],
+                                            p->two->path, new_path);
+                     strset_add(&renames->cached_target_names[dir_renamed_side],
+                                new_path);
+                     assert(!old_value);

This assert implies that p->status == 'R' only if this is the
first side (and first commit) to show a rename, right?

Um, this assert implies that p->two->path was not already found in
renames->cached_pairs[dir_renamed_side].

quoted

+             }
+             if (!new_path)
+                     new_path = p->two->path;
+             new_path = xstrdup(new_path);

If new_path was provided as non-NULL, then this is the second
time we are dup-ing it. However, that seems correct because we
want a different copy or every time we add it to the cached_pairs
and cached_target_names data.

quoted

+             old_value = strmap_put(&renames->cached_pairs[side],
+                                    p->one->path, new_path);
+             strset_add(&renames->cached_target_names[side],
+                        new_path);

Since we appear to be doing this in multiple places, would this
be a good place for a helper method? We could have it take a
`const char *new_path` and have the helper manage the `xstrdup()`
so we never forget to do that exactly once per insert to these
sets.

Makes sense.

quoted

+             free(old_value);
+     } else if (p->status == 'A' && new_path) {
+             new_path = xstrdup(new_path);
+             old_value = strmap_put(&renames->cached_pairs[dir_renamed_side],
+                                    p->two->path, new_path);
+             strset_add(&renames->cached_target_names[dir_renamed_side],
+                        new_path);
+             assert(!old_value);

And here's the third instance, making the "three is many" rule
kick in. A helper method would help make this easier. You can
also have a parameter corresponding to whether you need to
free() the old_value or assert it is NULL.

Yep, I'll add a helper.

quoted

+     }
+}
+
 static int compare_pairs(const void *a_, const void *b_)
 {
      const struct diff_filepair *a = *((const struct diff_filepair **)a_);

@@ -2415,6 +2474,7 @@ static int collect_renames(struct merge_options *opt,
              char *new_path; /* non-NULL only with directory renames */

              if (p->status != 'A' && p->status != 'R') {
+                     possibly_cache_new_pair(renames, p, side_index, NULL);
                      diff_free_filepair(p);
                      continue;
              }

@@ -2426,11 +2486,11 @@ static int collect_renames(struct merge_options *opt,
                                                    &collisions,
                                                    &clean);

+             possibly_cache_new_pair(renames, p, side_index, new_path);
              if (p->status != 'R' && !new_path) {
                      diff_free_filepair(p);
                      continue;
              }
-

nit: this deletion seems unnecessary.

Will fix.

quoted

              if (new_path)
                      apply_directory_rename_modifications(opt, p, new_path);

@@ -3701,8 +3761,16 @@ static void merge_start(struct merge_options *opt, struct merge_result *result)
                                       NULL, 1);
              strmap_init_with_options(&renames->dir_renames[i],
                                       NULL, 0);
+             /*
+              * relevant_sources uses -1 for the default, because we need
+              * to be able to distinguish not-in-strintmap from valid
+              * relevant_source values from enum file_rename_relevance.
+              * In particular, possibly_cache_new_pair() expects a negative
+              * value for not-found entries.
+              */
              strintmap_init_with_options(&renames->relevant_sources[i],
-                                         0, NULL, 0);
+                                         -1 /* explicitly invalid */,
+                                         NULL, 0);
              strmap_init_with_options(&renames->cached_pairs[i],
                                       NULL, 1);
              strset_init_with_options(&renames->cached_irrelevant[i],

Functionally looks good. I just had some nits about organization.

As always, thanks for the review and the helpful suggestions!