ML-190: fix redundant unsubscribe (plan)

This commit is contained in:
Claudio Ortolina
2026-05-19 10:38:41 +01:00
parent 0f529daa4e
commit 2fca0bb106
@@ -4,6 +4,7 @@ title: Skip redundant unsubscribe+resubscribe on same-record reconnect
status: To Do status: To Do
assignee: [] assignee: []
created_date: "2026-05-19 08:42" created_date: "2026-05-19 08:42"
updated_date: "2026-05-19 09:26"
labels: labels:
- audit - audit
- pubsub - pubsub
@@ -39,3 +40,72 @@ ordinal: 25000
- [ ] #2 Same-record reconnect skips unsubscribe+resubscribe entirely - [ ] #2 Same-record reconnect skips unsubscribe+resubscribe entirely
- [ ] #3 All existing tests pass - [ ] #3 All existing tests pass
<!-- AC:END --> <!-- AC:END -->
## Implementation Plan
<!-- SECTION:PLAN:BEGIN -->
## Overview
Micro-optimization to skip the redundant `Records.unsubscribe/1` call when a LiveView reconnects to the same record. The change is a single guard addition in `RecordActions.manage_subscription/2`. No other modules are affected.
**Nature:** Code hygiene, not a measurable performance win. The skip eliminates at most two ETS operations on reconnect, which Phoenix.PubSub already deduplicates by PID internally.
## Callers (no changes needed)
| LiveView | File | Line | Hook |
| ------------------- | ---------------------------------------------------- | ---- | ----------------- |
| CollectionLive.Show | `lib/music_library_web/live/collection_live/show.ex` | 361 | `handle_params/3` |
| WishlistLive.Show | `lib/music_library_web/live/wishlist_live/show.ex` | 298 | `handle_params/3` |
Both call `manage_subscription(socket, id)` at the top of `handle_params/3` before loading the record. No caller changes are needed because the function signature and return value (`:ok`) remain unchanged.
## Code change
**File:** `lib/music_library_web/live_helpers/record_actions.ex`
Change the unsubscribe guard from:
```elixir
if socket.assigns[:record], do: Records.unsubscribe(socket.assigns.record.id)
```
to:
```elixir
if socket.assigns[:record] && socket.assigns.record.id != new_id,
do: Records.unsubscribe(socket.assigns.record.id)
```
The `Records.subscribe(new_id)` call is intentionally left unconditional: on same-record reconnect it's a no-op (PubSub deduplicates by PID), and on first mount it's required.
## Edge-case coverage
| Scenario | `socket.assigns[:record]` | old vs new | Unsub? | Sub? | Correct? |
| ---------------------------- | ------------------------- | -------------- | ------ | ----------- | -------- |
| First mount | `nil` | short-circuits | No | Yes | ✅ |
| Navigate to different record | record A | A ≠ B | Yes | Yes | ✅ |
| Reconnect same record | record X | X == X | No | Yes (no-op) | ✅ |
## Verification steps
1. **Run existing tests:**
```bash
mix test test/music_library_web/live_helpers/record_actions_test.exs
mix test test/music_library_web/live/collection_live/show_test.exs
mix test test/music_library_web/live/wishlist_live/show_test.exs
```
All should pass. These cover `manage_subscription/2` indirectly through PubSub broadcast handling and full page navigation tests.
2. **Manual reconnect smoke test (optional):** Navigate to a record show page, kill the server, restart it, and confirm the page reconnects and continues to receive PubSub updates (e.g., trigger a background cover refresh and verify the toast appears).
3. **No new test is strictly required.** The change is a pure optimization — it removes a redundant call that was already harmless. Existing PubSub broadcast tests in `record_actions_test.exs` (the `handle_record_updated` describe block) confirm the subscription lifecycle works end-to-end.
## Documentation
- No user-facing docs need updating.
- The `@doc` for `manage_subscription/2` is already accurate and does not need amendment.
- No production infrastructure changes, environment variables, or migrations are required.
<!-- SECTION:PLAN:END -->