fetch: Speed up fetch by rewriting find_non_local_tags
When trying to get a list of remote tags to see if we need to fetch any we were doing a linear search for the matching tag ref for the tag^{} commit entries. This proves to be incredibly slow for large numbers of tags. Rewrite the function so that we build up a string_list of refs to fetch and then process that instead. As an extreme example, for a repository with 50000 tags (and just a single commit on a single branch), a fetch that does nothing goes from ~1m50s to ~4.1s. Signed-off-by: Julian Phillips <julian@quantumfyre.co.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Julian Phillips committed
Sep 17, 2009 at 08:33 UTC
e984c54ada086e7676419b6b1b8a2d7a2429da48
1 file changed
+74
-33
builtin-fetch.c
+74
-33
@@ -504,57 +504,98 @@ static int will_fetch(struct ref **head, const unsigned char *sha1)
504
return 0;
505
}
506
507
+struct tag_data {
508
+ struct ref **head;
509
+ struct ref ***tail;
510
+};
511
+
512
+static int add_to_tail(struct string_list_item *item, void *cb_data)
513
+{
514
+ struct tag_data *data = (struct tag_data *)cb_data;
515
+ struct ref *rm = NULL;
516
+
517
+ /* We have already decided to ignore this item */
518
+ if (!item->util)
519
+ return 0;
520
+
521
+ rm = alloc_ref(item->string);
522
+ rm->peer_ref = alloc_ref(item->string);
523
+ hashcpy(rm->old_sha1, item->util);
524
+
525
+ **data->tail = rm;
526
+ *data->tail = &rm->next;
527
+
528
+ return 0;
529
+}
530
+
531
static void find_non_local_tags(struct transport *transport,
532
struct ref **head,
533
struct ref ***tail)
534
{
535
struct string_list existing_refs = { NULL, 0, 0, 0 };
512
- struct string_list new_refs = { NULL, 0, 0, 1 };
513
- char *ref_name;
514
- int ref_name_len;
515
- const unsigned char *ref_sha1;
516
- const struct ref *tag_ref;
517
- struct ref *rm = NULL;
536
+ struct string_list remote_refs = { NULL, 0, 0, 0 };
537
+ struct tag_data data = {head, tail};
538
const struct ref *ref;
539
+ struct string_list_item *item = NULL;
540
541
for_each_ref(add_existing, &existing_refs);
542
for (ref = transport_get_remote_refs(transport); ref; ref = ref->next) {
543
if (prefixcmp(ref->name, "refs/tags"))
544
continue;
545
525
- ref_name = xstrdup(ref->name);
526
- ref_name_len = strlen(ref_name);
527
- ref_sha1 = ref->old_sha1;
528
-
529
- if (!strcmp(ref_name + ref_name_len - 3, "^{}")) {
530
- ref_name[ref_name_len - 3] = 0;
531
- tag_ref = transport_get_remote_refs(transport);
532
- while (tag_ref) {
533
- if (!strcmp(tag_ref->name, ref_name)) {
534
- ref_sha1 = tag_ref->old_sha1;
535
- break;
536
- }
537
- tag_ref = tag_ref->next;
538
- }
546
+ /*
547
+ * The peeled ref always follows the matching base
548
+ * ref, so if we see a peeled ref that we don't want
549
+ * to fetch then we can mark the ref entry in the list
550
+ * as one to ignore by setting util to NULL.
551
+ */
552
+ if (!strcmp(ref->name + strlen(ref->name) - 3, "^{}")) {
553
+ if (item && !has_sha1_file(ref->old_sha1) &&
554
+ !will_fetch(head, ref->old_sha1) &&
555
+ !has_sha1_file(item->util) &&
556
+ !will_fetch(head, item->util))
557
+ item->util = NULL;
558
+ item = NULL;
559
+ continue;
560
}
561
541
- if (!string_list_has_string(&existing_refs, ref_name) &&
542
- !string_list_has_string(&new_refs, ref_name) &&
543
- (has_sha1_file(ref->old_sha1) ||
544
- will_fetch(head, ref->old_sha1))) {
545
- string_list_insert(ref_name, &new_refs);
562
+ /*
563
+ * If item is non-NULL here, then we previously saw a
564
+ * ref not followed by a peeled reference, so we need
565
+ * to check if it is a lightweight tag that we want to
566
+ * fetch.
567
+ */
568
+ if (item && !has_sha1_file(item->util) &&
569
+ !will_fetch(head, item->util))
570
+ item->util = NULL;
571
547
- rm = alloc_ref(ref_name);
548
- rm->peer_ref = alloc_ref(ref_name);
549
- hashcpy(rm->old_sha1, ref_sha1);
572
+ item = NULL;
573
551
- **tail = rm;
552
- *tail = &rm->next;
553
- }
554
- free(ref_name);
574
+ /* skip duplicates and refs that we already have */
575
+ if (string_list_has_string(&remote_refs, ref->name) ||
576
+ string_list_has_string(&existing_refs, ref->name))
577
+ continue;
578
+
579
+ item = string_list_insert(ref->name, &remote_refs);
580
+ item->util = (void *)ref->old_sha1;
581
}
582
string_list_clear(&existing_refs, 0);
557
- string_list_clear(&new_refs, 0);
583
+
584
+ /*
585
+ * We may have a final lightweight tag that needs to be
586
+ * checked to see if it needs fetching.
587
+ */
588
+ if (item && !has_sha1_file(item->util) &&
589
+ !will_fetch(head, item->util))
590
+ item->util = NULL;
591
+
592
+ /*
593
+ * For all the tags in the remote_refs string list, call
594
+ * add_to_tail to add them to the list of refs to be fetched
595
+ */
596
+ for_each_string_list(add_to_tail, &remote_refs, &data);
597
+
598
+ string_list_clear(&remote_refs, 0);
599
}
600
601
static void check_not_current_branch(struct ref *ref_map)