perf(dispatch): index Client event bus and Dispatcher by event discriminant

[devin-ai-integration]

Summary

Both the Client subscription bus (Client::on_event / Client::on_any) and the high-level Dispatcher (teamtalk::dispatch::Dispatcher) used to linearly scan every registered subscription/handler on every event, even though the filter check was just mem::discriminant(expected) == mem::discriminant(event). Dispatch cost was therefore O(total_subs) per event, not O(subscribers-actually-watching-this-event).

This PR indexes both stores by mem::Discriminant<Event> so dispatch only walks handlers that can possibly match the current event, while preserving the visible firing order.

Changes:

• crates/teamtalk/src/client/bus/subscription.rs — EventBus now maintains two secondary indexes on top of the single insertion-order Vec<Subscription>:

  • by_event: HashMap<Discriminant<Event>, Vec<usize>> — indices of event-specific subscribers;
  • any_indices: Vec<usize> — indices of wildcard subscribers.
  • subscribe, unsubscribe, unsubscribe_group, and clear keep the indexes consistent (rebuild on removal, rather than tombstones).
  • dispatch merges the specific + wildcard index slices in ascending order, so cross-bucket firing order is identical to the pre-change linear scan.
  • Subscription drops its event: Option<Event> field; the discriminant is now implicit in bucket placement and an explicit event_filter: Option<Discriminant<Event>> is used for index rebuilds after unsubscribe.

• crates/teamtalk/src/dispatch/dispatcher.rs / source.rs — same two-index pattern applied to Dispatcher::handlers. The per-entry filter field is dropped from HandlerEntry (bucket placement is the source of truth), and process_event merges the specific + wildcard slices.

• crates/teamtalk/src/client/core/mod.rs — new mock-only helper Client::mock_dispatch_bus_for_tests(event, source) so integration tests can exercise the bus path synchronously without spinning the full poll loop. Gated by #[cfg(feature = "mock")] alongside the other existing mock_*_for_tests helpers.

• crates/teamtalk/tests/indexed_dispatch_tests.rs (new, 9 tests):

  • specific handler fires only for its discriminant;
  • wildcard fires for every event;
  • insertion order is preserved across interleaved specific + wildcard registrations (the key behavioural invariant of the merge);
  • an unrelated event skips the whole bucket and only fans out to wildcards;
  • DispatchFlow::Stop from a wildcard slot surfaces as the step result;
  • Client::on_event/on_any register through both buckets and dispatch merges in insertion order;
  • unsubscribe_event rebuilds the index and later dispatches skip the removed sub;
  • unsubscribe_event_group removes from specific and wildcard buckets simultaneously;
  • clear_event_subscriptions resets both buckets.

Total test count locally: 305 passed, 0 failed (296 pre-existing + 9 new), cargo fmt / cargo clippy --all-features --all-targets -- -D warnings clean.

Complexity

Before: O(total_subs) per event (every sub checked for discriminant + filter match).
After: O(specific_for_this_event + wildcards) per event. The per-subscription filter-match check (user id, channel id, text type, predicate, etc.) is unchanged — only the discriminant filter moved into the bucket lookup.

Not changed

• bot/router/dispatch.rs is intentionally left linear: route ordering is semantically load-bearing there (the matched_command/matched_command_path / auto-help fallback logic depends on walking routes in registration order, interleaving Any / Event / Command matchers). Indexing routes by event discriminant would change dispatch semantics; that belongs to a separate behavioural PR if wanted.

Review & Testing Checklist for Human

• Verify that the merge loop in EventBus::dispatch and Dispatcher::process_event preserves strict insertion order across the specific + any index lists when a handler is registered as wildcard between two specific handlers for the current event. Test insertion_order_is_preserved_across_specific_and_wildcard and client_bus_specific_and_wildcard_dispatch_in_insertion_order pin this down, but a second pair of eyes on the while si < … || ai < … merge is worth it.
• Spot-check unsubscribe / unsubscribe_group — they rebuild indexes via retain + rebuild_indices. Ensure there's no subtle index-invalidation path I missed (the rebuild is deliberately O(n) on removal rather than tombstones to keep dispatch zero-cost).
• Confirm you're OK with Client::mock_dispatch_bus_for_tests becoming part of the public mock-feature test surface (follows the same pattern as existing mock_apply_event_for_tests / mock_set_connection_state_for_tests).

Notes

• Backward compatible: every public API signature is unchanged (no breaking changes to Client::on_event, on_any, unsubscribe_event, event_subscription_count, Dispatcher::on_event, Dispatcher::on_any, etc.). Handler closure signatures, return types, and visibility are identical to main.
• mem::Discriminant<T> is Hash + Eq since Rust 1.73, so no nightly / no extra dependency is needed.
• The pre-existing semver CI gate will be red on this PR too — same pre-existing red as on feat(types): add StreamTypes newtype for stream-kind bit masks #52–feat(types): introduce SecretString and use it for LoginParams password #56, release-plz resolves it at release time.

Link to Devin session: https://app.devin.ai/sessions/71fdd6196cb74723a2e277bb81993a9c
Requested by: @BlindMaster24

---

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring