diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/sync/filters.rs | 138 |
1 files changed, 130 insertions, 8 deletions
diff --git a/src/sync/filters.rs b/src/sync/filters.rs index c4e20e7..3592489 100644 --- a/src/sync/filters.rs +++ b/src/sync/filters.rs | |||
| @@ -13,15 +13,15 @@ use std::collections::HashSet; | |||
| 13 | 13 | ||
| 14 | use nostr_sdk::prelude::*; | 14 | use nostr_sdk::prelude::*; |
| 15 | 15 | ||
| 16 | /// Layer 1: Announcements filter (kinds 30617 + 30618 + 10317) | 16 | /// Layer 1: Announcements filter (kinds 30617 + 10317) |
| 17 | /// | 17 | /// |
| 18 | /// Subscribed ONCE on connect - NOT included in consolidation rebuilds. | 18 | /// Subscribed ONCE on connect - NOT included in consolidation rebuilds. |
| 19 | /// Note: 30618 is ONLY synced from remote relays, not self-subscribed. | 19 | /// Note: State events (30618) are now subscribed via identifier-based filters in Layer 2 |
| 20 | /// to avoid receiving state events for repositories we don't host. | ||
| 20 | /// Note: 10317 (User Grasp List) is synced for better GRASP discovery. | 21 | /// Note: 10317 (User Grasp List) is synced for better GRASP discovery. |
| 21 | pub fn build_announcement_filter(since: Option<Timestamp>) -> Filter { | 22 | pub fn build_announcement_filter(since: Option<Timestamp>) -> Filter { |
| 22 | let filter = Filter::new().kinds([ | 23 | let filter = Filter::new().kinds([ |
| 23 | Kind::GitRepoAnnouncement, // Repository announcements | 24 | Kind::GitRepoAnnouncement, // Repository announcements |
| 24 | Kind::RepoState, // Repository state | ||
| 25 | Kind::GitUserGraspList, // User Grasp List | 25 | Kind::GitUserGraspList, // User Grasp List |
| 26 | ]); | 26 | ]); |
| 27 | 27 | ||
| @@ -31,6 +31,64 @@ pub fn build_announcement_filter(since: Option<Timestamp>) -> Filter { | |||
| 31 | } | 31 | } |
| 32 | } | 32 | } |
| 33 | 33 | ||
| 34 | /// State event filters for our hosted repositories | ||
| 35 | /// | ||
| 36 | /// Subscribes to kind 30618 state events using #d (identifier) tags. | ||
| 37 | /// This is more efficient than Layer 1 broadcast subscription because: | ||
| 38 | /// - Only receives state events for repos we actually host | ||
| 39 | /// - One filter can include multiple identifiers (batched per 100) | ||
| 40 | /// - Avoids 1000+ rejections for repos we don't care about | ||
| 41 | /// | ||
| 42 | /// # Arguments | ||
| 43 | /// * `repos` - Set of repo addressable refs (format: 30617:pubkey:identifier) | ||
| 44 | /// * `since` - Optional timestamp for incremental sync | ||
| 45 | /// | ||
| 46 | /// # Returns | ||
| 47 | /// Vec of filters, one filter per 100-identifier chunk | ||
| 48 | pub fn state_event_filters_for_our_repos( | ||
| 49 | repos: &HashSet<String>, | ||
| 50 | since: Option<Timestamp>, | ||
| 51 | ) -> Vec<Filter> { | ||
| 52 | if repos.is_empty() { | ||
| 53 | return vec![]; | ||
| 54 | } | ||
| 55 | |||
| 56 | // Extract unique identifiers from addressable refs | ||
| 57 | let mut identifiers: HashSet<String> = HashSet::new(); | ||
| 58 | for repo_ref in repos { | ||
| 59 | // Format: 30617:pubkey:identifier | ||
| 60 | if let Some(identifier) = repo_ref.split(':').nth(2) { | ||
| 61 | identifiers.insert(identifier.to_string()); | ||
| 62 | } | ||
| 63 | } | ||
| 64 | |||
| 65 | if identifiers.is_empty() { | ||
| 66 | return vec![]; | ||
| 67 | } | ||
| 68 | |||
| 69 | let mut filters = Vec::new(); | ||
| 70 | let identifier_vec: Vec<_> = identifiers.iter().collect(); | ||
| 71 | |||
| 72 | // Batch by 100 identifiers per filter | ||
| 73 | for chunk in identifier_vec.chunks(100) { | ||
| 74 | let mut filter = Filter::new().kind(Kind::RepoState); // kind 30618 | ||
| 75 | |||
| 76 | // Add #d tags for all identifiers in this chunk | ||
| 77 | for identifier in chunk { | ||
| 78 | filter = | ||
| 79 | filter.custom_tag(SingleLetterTag::lowercase(Alphabet::D), identifier.as_str()); | ||
| 80 | } | ||
| 81 | |||
| 82 | if let Some(ts) = since { | ||
| 83 | filter = filter.since(ts); | ||
| 84 | } | ||
| 85 | |||
| 86 | filters.push(filter); | ||
| 87 | } | ||
| 88 | |||
| 89 | filters | ||
| 90 | } | ||
| 91 | |||
| 34 | /// Layer 2: Events tagging one of our repos | 92 | /// Layer 2: Events tagging one of our repos |
| 35 | /// | 93 | /// |
| 36 | /// Uses lowercase a, uppercase A, and q tags for comprehensive coverage. | 94 | /// Uses lowercase a, uppercase A, and q tags for comprehensive coverage. |
| @@ -166,6 +224,11 @@ pub fn tagged_one_of_our_root_event_filters( | |||
| 166 | /// - compute_actions for incremental subscriptions | 224 | /// - compute_actions for incremental subscriptions |
| 167 | /// - consolidation rebuilds (Layer 1 remains active) | 225 | /// - consolidation rebuilds (Layer 1 remains active) |
| 168 | /// | 226 | /// |
| 227 | /// Includes: | ||
| 228 | /// - State event filters (kind 30618 with #d tags for our repo identifiers) | ||
| 229 | /// - Repo-tagging filters (a/A/q tags) | ||
| 230 | /// - Root event filters (e/E/q tags) | ||
| 231 | /// | ||
| 169 | /// # Arguments | 232 | /// # Arguments |
| 170 | /// * `repos` - Set of repo addressable refs | 233 | /// * `repos` - Set of repo addressable refs |
| 171 | /// * `root_events` - Set of root event IDs | 234 | /// * `root_events` - Set of root event IDs |
| @@ -176,6 +239,7 @@ pub fn build_layer2_and_layer3_filters( | |||
| 176 | since: Option<Timestamp>, | 239 | since: Option<Timestamp>, |
| 177 | ) -> Vec<Filter> { | 240 | ) -> Vec<Filter> { |
| 178 | let mut filters = Vec::new(); | 241 | let mut filters = Vec::new(); |
| 242 | filters.extend(state_event_filters_for_our_repos(repos, since)); | ||
| 179 | filters.extend(tagged_one_of_our_repo_event_filters(repos, since)); | 243 | filters.extend(tagged_one_of_our_repo_event_filters(repos, since)); |
| 180 | filters.extend(tagged_one_of_our_root_event_filters(root_events, since)); | 244 | filters.extend(tagged_one_of_our_root_event_filters(root_events, since)); |
| 181 | filters | 245 | filters |
| @@ -315,8 +379,8 @@ mod tests { | |||
| 315 | 379 | ||
| 316 | let filters = build_layer2_and_layer3_filters(&repos, &root_events, None); | 380 | let filters = build_layer2_and_layer3_filters(&repos, &root_events, None); |
| 317 | 381 | ||
| 318 | // Should have 6 filters (3 for repos + 3 for root events) | 382 | // Should have 7 filters (1 state + 3 for repos + 3 for root events) |
| 319 | assert_eq!(filters.len(), 6); | 383 | assert_eq!(filters.len(), 7); |
| 320 | } | 384 | } |
| 321 | 385 | ||
| 322 | #[test] | 386 | #[test] |
| @@ -328,8 +392,8 @@ mod tests { | |||
| 328 | 392 | ||
| 329 | let filters = build_layer2_and_layer3_filters(&repos, &root_events, None); | 393 | let filters = build_layer2_and_layer3_filters(&repos, &root_events, None); |
| 330 | 394 | ||
| 331 | // Should have 3 filters (3 for repos only) | 395 | // Should have 4 filters (1 state + 3 for repos only) |
| 332 | assert_eq!(filters.len(), 3); | 396 | assert_eq!(filters.len(), 4); |
| 333 | } | 397 | } |
| 334 | 398 | ||
| 335 | #[test] | 399 | #[test] |
| @@ -356,6 +420,64 @@ mod tests { | |||
| 356 | let since = Timestamp::from(1700000000); | 420 | let since = Timestamp::from(1700000000); |
| 357 | let filters = build_layer2_and_layer3_filters(&repos, &root_events, Some(since)); | 421 | let filters = build_layer2_and_layer3_filters(&repos, &root_events, Some(since)); |
| 358 | 422 | ||
| 359 | assert_eq!(filters.len(), 6); | 423 | // Should have 7 filters (1 state + 3 for repos + 3 for root events) |
| 424 | assert_eq!(filters.len(), 7); | ||
| 425 | } | ||
| 426 | |||
| 427 | #[test] | ||
| 428 | fn test_state_event_filters_empty() { | ||
| 429 | let repos: HashSet<String> = HashSet::new(); | ||
| 430 | let filters = state_event_filters_for_our_repos(&repos, None); | ||
| 431 | |||
| 432 | assert!(filters.is_empty()); | ||
| 433 | } | ||
| 434 | |||
| 435 | #[test] | ||
| 436 | fn test_state_event_filters_single_repo() { | ||
| 437 | let mut repos = HashSet::new(); | ||
| 438 | repos.insert("30617:abc123:test-repo".to_string()); | ||
| 439 | |||
| 440 | let filters = state_event_filters_for_our_repos(&repos, None); | ||
| 441 | |||
| 442 | // Should create 1 filter with kind 30618 and #d tag | ||
| 443 | assert_eq!(filters.len(), 1); | ||
| 444 | } | ||
| 445 | |||
| 446 | #[test] | ||
| 447 | fn test_state_event_filters_batching() { | ||
| 448 | let mut repos = HashSet::new(); | ||
| 449 | for i in 0..250 { | ||
| 450 | repos.insert(format!("30617:pubkey{}:repo{}", i, i)); | ||
| 451 | } | ||
| 452 | |||
| 453 | let filters = state_event_filters_for_our_repos(&repos, None); | ||
| 454 | |||
| 455 | // Should create 3 filters (250 identifiers = 100 + 100 + 50 = 3 chunks) | ||
| 456 | assert_eq!(filters.len(), 3); | ||
| 457 | } | ||
| 458 | |||
| 459 | #[test] | ||
| 460 | fn test_state_event_filters_deduplicates_identifiers() { | ||
| 461 | let mut repos = HashSet::new(); | ||
| 462 | // Same identifier with different pubkeys | ||
| 463 | repos.insert("30617:pubkey1:same-repo".to_string()); | ||
| 464 | repos.insert("30617:pubkey2:same-repo".to_string()); | ||
| 465 | repos.insert("30617:pubkey3:same-repo".to_string()); | ||
| 466 | |||
| 467 | let filters = state_event_filters_for_our_repos(&repos, None); | ||
| 468 | |||
| 469 | // Should create 1 filter with deduplicated identifier | ||
| 470 | assert_eq!(filters.len(), 1); | ||
| 471 | } | ||
| 472 | |||
| 473 | #[test] | ||
| 474 | fn test_state_event_filters_with_since() { | ||
| 475 | let mut repos = HashSet::new(); | ||
| 476 | repos.insert("30617:abc123:test-repo".to_string()); | ||
| 477 | |||
| 478 | let since = Timestamp::from(1700000000); | ||
| 479 | let filters = state_event_filters_for_our_repos(&repos, Some(since)); | ||
| 480 | |||
| 481 | assert_eq!(filters.len(), 1); | ||
| 360 | } | 482 | } |
| 361 | } | 483 | } |