@matheusbsbs / matheus-bagliano / commits / dd4bc01c0a

revision: use priority queue for non-limited streaming walks

The streaming (non-limited) walk in get_revision_1() inserts newly discovered parent commits into a date-sorted queue via commit_list_insert_by_date(), which scans the linked list to find the insertion point -- O(w) per insert, where w is the width of the active walk frontier. Replace this with an O(log w) priority queue. Add a commit_queue field to rev_info alongside the existing commits linked list. The two representations are mutually exclusive: setup and external callers that need list access use the linked list, then get_revision_1() lazily drains it into the priority queue on first call. Add a REV_WALK_NO_WALK enum value to distinguish the no_walk case (which still uses the commit list) from the streaming case. The conversion function rev_info_commit_list_to_queue() is public so callers that know they will iterate can convert early. Combined with the limit_list() priority queue change already in master, this eliminates all O(w) sorted linked-list insertion from the revision walk machinery. Signed-off-by: Kristofer Karlsson <krka@spotify.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>

Kristofer Karlsson committed May 27, 2026 at 15:50 UTC dd4bc01c0a8fc871a68a5027ed5ac953fa47fc6e

4 files changed +41 -41

commit.c

-13

@@ -729,19 +729,6 @@ void commit_list_free(struct commit_list *list)

729 pop_commit(&list);

730 }

731

732 -struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)

733 -{

734 - struct commit_list **pp = list;

735 - struct commit_list *p;

736 - while ((p = *pp) != NULL) {

737 - if (p->item->date < item->date) {

738 - break;

739 - }

740 - pp = &p->next;

741 - }

742 - return commit_list_insert(item, pp);

743 -}

744 -

732 static int commit_list_compare_by_date(const struct commit_list *a,

733 const struct commit_list *b)

734 {

commit.h

-2

@@ -191,8 +191,6 @@ int commit_list_contains(struct commit *item,

191 struct commit_list **commit_list_append(struct commit *commit,

192 struct commit_list **next);

193 unsigned commit_list_count(const struct commit_list *l);

194 -struct commit_list *commit_list_insert_by_date(struct commit *item,

195 - struct commit_list **list);

194 void commit_list_sort_by_date(struct commit_list **list);

195

196 /* Shallow copy of the input list */

revision.c

+30 -25

@@ -1116,7 +1116,7 @@ static void try_to_simplify_commit(struct rev_info *revs, struct commit *commit)

1116 }

1117

1118 static int process_parents(struct rev_info *revs, struct commit *commit,

1119 - struct commit_list **list, struct prio_queue *queue)

1119 + struct prio_queue *queue)

1120 {

1121 struct commit_list *parent = commit->parents;

1122 unsigned pass_flags;

@@ -1158,8 +1158,6 @@ static int process_parents(struct rev_info *revs, struct commit *commit,

1158 if (p->object.flags & SEEN)

1159 continue;

1160 p->object.flags |= (SEEN | NOT_USER_GIVEN);

1161 - if (list)

1162 - commit_list_insert_by_date(p, list);

1161 if (queue)

1162 prio_queue_put(queue, p);

1163 if (revs->exclude_first_parent_only)

@@ -1207,8 +1205,6 @@ static int process_parents(struct rev_info *revs, struct commit *commit,

1205 p->object.flags |= pass_flags | CHILD_VISITED;

1206 if (!(p->object.flags & SEEN)) {

1207 p->object.flags |= (SEEN | NOT_USER_GIVEN);

1210 - if (list)

1211 - commit_list_insert_by_date(p, list);

1208 if (queue)

1209 prio_queue_put(queue, p);

1210 }

@@ -1470,7 +1466,7 @@ static int limit_list(struct rev_info *revs)

1466

1467 if (revs->max_age != -1 && (commit->date < revs->max_age))

1468 obj->flags |= UNINTERESTING;

1473 - if (process_parents(revs, commit, NULL, &queue) < 0) {

1469 + if (process_parents(revs, commit, &queue) < 0) {

1470 clear_prio_queue(&queue);

1471 return -1;

1472 }

@@ -3257,6 +3253,7 @@ static void free_void_commit_list(void *list)

3253 void release_revisions(struct rev_info *revs)

3254 {

3255 commit_list_free(revs->commits);

3256 + clear_prio_queue(&revs->commit_queue);

3257 commit_list_free(revs->ancestry_path_bottoms);

3258 release_display_notes(&revs->notes_opt);

3259 object_array_clear(&revs->pending);

@@ -3726,7 +3723,7 @@ static void explore_walk_step(struct rev_info *revs)

3723 if (revs->max_age != -1 && (c->date < revs->max_age))

3724 c->object.flags |= UNINTERESTING;

3725

3729 - if (process_parents(revs, c, NULL, NULL) < 0)

3726 + if (process_parents(revs, c, NULL) < 0)

3727 return;

3728

3729 if (c->object.flags & UNINTERESTING)

@@ -3902,7 +3899,7 @@ static void expand_topo_walk(struct rev_info *revs, struct commit *commit)

3899 {

3900 struct commit_list *p;

3901 struct topo_walk_info *info = revs->topo_walk_info;

3905 - if (process_parents(revs, commit, NULL, NULL) < 0) {

3902 + if (process_parents(revs, commit, NULL) < 0) {

3903 if (!revs->ignore_missing_links)

3904 die("Failed to traverse parents of commit %s",

3905 oid_to_hex(&commit->object.oid));

@@ -3938,6 +3935,13 @@ static void expand_topo_walk(struct rev_info *revs, struct commit *commit)

3935 }

3936 }

3937

3938 +void rev_info_commit_list_to_queue(struct rev_info *revs)

3939 +{

3940 + while (revs->commits)

3941 + prio_queue_put(&revs->commit_queue, pop_commit(&revs->commits));

3942 +}

3943 +

3944 +

3945 int prepare_revision_walk(struct rev_info *revs)

3946 {

3947 int i;

@@ -4006,7 +4010,7 @@ static enum rewrite_result rewrite_one_1(struct rev_info *revs,

4010 for (;;) {

4011 struct commit *p = *pp;

4012 if (!revs->limited)

4009 - if (process_parents(revs, p, NULL, queue) < 0)

4013 + if (process_parents(revs, p, queue) < 0)

4014 return rewrite_one_error;

4015 if (p->object.flags & UNINTERESTING)

4016 return rewrite_one_ok;

@@ -4020,27 +4024,18 @@ static enum rewrite_result rewrite_one_1(struct rev_info *revs,

4024 }

4025 }

4026

4023 -static void merge_queue_into_list(struct prio_queue *q, struct commit_list **list)

4027 +static void merge_queue_into_prio_queue(struct prio_queue *from,

4028 + struct prio_queue *to)

4029 {

4025 - while (q->nr) {

4026 - struct commit *item = prio_queue_peek(q);

4027 - struct commit_list *p = *list;

4028 -

4029 - if (p && p->item->date >= item->date)

4030 - list = &p->next;

4031 - else {

4032 - p = commit_list_insert(item, list);

4033 - list = &p->next; /* skip newly added item */

4034 - prio_queue_get(q); /* pop item */

4035 - }

4036 - }

4030 + while (from->nr)

4031 + prio_queue_put(to, prio_queue_get(from));

4032 }

4033

4034 static enum rewrite_result rewrite_one(struct rev_info *revs, struct commit **pp)

4035 {

4036 struct prio_queue queue = { compare_commits_by_commit_date };

4037 enum rewrite_result ret = rewrite_one_1(revs, pp, &queue);

4043 - merge_queue_into_list(&queue, &revs->commits);

4038 + merge_queue_into_prio_queue(&queue, &revs->commit_queue);

4039 clear_prio_queue(&queue);

4040 return ret;

4041 }

@@ -4331,6 +4326,7 @@ enum rev_walk_mode {

4326 REV_WALK_REFLOG,

4327 REV_WALK_TOPO,

4328 REV_WALK_LIMITED,

4329 + REV_WALK_NO_WALK,

4330 REV_WALK_STREAMING,

4331 };

4332

@@ -4342,6 +4338,8 @@ static enum rev_walk_mode get_walk_mode(struct rev_info *revs)

4338 return REV_WALK_TOPO;

4339 if (revs->limited)

4340 return REV_WALK_LIMITED;

4341 + if (revs->no_walk)

4342 + return REV_WALK_NO_WALK;

4343 return REV_WALK_STREAMING;

4344 }

4345

@@ -4349,6 +4347,9 @@ static struct commit *get_revision_1(struct rev_info *revs)

4347 {

4348 enum rev_walk_mode mode = get_walk_mode(revs);

4349

4350 + if (mode == REV_WALK_STREAMING && revs->commits)

4351 + rev_info_commit_list_to_queue(revs);

4352 +

4353 while (1) {

4354 struct commit *commit;

4355

@@ -4360,9 +4361,12 @@ static struct commit *get_revision_1(struct rev_info *revs)

4361 commit = next_topo_commit(revs);

4362 break;

4363 case REV_WALK_LIMITED:

4363 - case REV_WALK_STREAMING:

4364 + case REV_WALK_NO_WALK:

4365 commit = pop_commit(&revs->commits);

4366 break;

4367 + case REV_WALK_STREAMING:

4368 + commit = prio_queue_get(&revs->commit_queue);

4369 + break;

4370 }

4371

4372 if (!commit)

@@ -4390,12 +4394,13 @@ static struct commit *get_revision_1(struct rev_info *revs)

4394 break;

4395 case REV_WALK_STREAMING:

4396 if (process_parents(revs, commit,

4393 - &revs->commits, NULL) < 0) {

4397 + &revs->commit_queue) < 0) {

4398 if (!revs->ignore_missing_links)

4399 die("Failed to traverse parents of commit %s",

4400 oid_to_hex(&commit->object.oid));

4401 }

4402 break;

4403 + case REV_WALK_NO_WALK:

4404 case REV_WALK_LIMITED:

4405 break;

4406 }

revision.h

+11 -1

@@ -12,6 +12,7 @@

12 #include "decorate.h"

13 #include "ident.h"

14 #include "list-objects-filter-options.h"

15 +#include "prio-queue.h"

16 #include "strvec.h"

18 /**

@@ -122,8 +123,14 @@ struct oidset;

123 struct topo_walk_info;

124

125 struct rev_info {

125 - /* Starting list */

126 + /*

127 + * Work queue of commits, stored as either a linked list or a

128 + * priority queue, but never both at the same time.

129 + * rev_info_commit_list_to_queue() converts list to queue.

130 + */

131 struct commit_list *commits;

132 + struct prio_queue commit_queue;

133 +

134 struct object_array pending;

135 struct repository *repo;

136

@@ -400,6 +407,7 @@ struct rev_info {

407 * uninitialized.

408 */

409 #define REV_INFO_INIT { \

410 + .commit_queue = { .compare = compare_commits_by_commit_date }, \

411 .abbrev = DEFAULT_ABBREV, \

412 .simplify_history = 1, \

413 .pruning.flags.recursive = 1, \

@@ -478,6 +486,8 @@ void reset_revision_walk(void);

486 */

487 int prepare_revision_walk(struct rev_info *revs);

488

489 +/* Drain the commits linked list into the priority queue. */

490 +void rev_info_commit_list_to_queue(struct rev_info *revs);

491 /**

492 * Takes a pointer to a `rev_info` structure and iterates over it, returning a

493 * `struct commit *` each time you call it. The end of the revision list is