Implement pull-to-refresh and infinite scroll pagination
epic-in-app-notification-centre-ui-task-009 — Add pull-to-refresh to the NotificationCentreScreen using RefreshIndicator that dispatches RefreshNotifications to the BLoC. Implement infinite scroll pagination using a ScrollController that detects when the user is within 200px of the bottom of the list, then dispatches LoadNextPage. The BLoC must manage a page cursor and append new items to the existing list. Show a bottom loading indicator (CircularProgressIndicator) while a next page is being fetched. Ensure pagination resets when the active filter changes.
Acceptance Criteria
Technical Requirements
Execution Context
Tier 4 - 323 tasks
Can start after Tier 3 completes
Implementation Notes
Attach the ScrollController in initState and add a listener: `if (_scrollController.position.pixels >= _scrollController.position.maxScrollExtent - 200 && !state.isLoadingNextPage && state.hasMore) { context.read
On filter change, emit a state with cursor = null, items = [], then immediately fetch page 1. Avoid calling `setState` in the scroll listener — let BLoC state drive rebuilds via BlocBuilder. Use a `GlobalKey
Testing Requirements
Write flutter_test widget tests covering: (1) RefreshIndicator callback dispatches RefreshNotifications and resets cursor, (2) ScrollController triggers LoadNextPage when scrolled to within 200px of the bottom, (3) BLoC state transitions — loading → appended success → no more pages, (4) filter change clears list and resets cursor before fetching, (5) duplicate LoadNextPage guard — rapid scroll does not fire multiple events, (6) error state during pagination shows snackbar without clearing existing list. Use bloc_test package to test BLoC event/state sequences in isolation. Target 80%+ branch coverage for the BLoC and the screen widget.
If a referenced entity (contact, certification, activity) has been deleted or its RLS policy now excludes the current user, the deep link handler may navigate to a screen that renders in an error state or throws an unhandled exception.
Mitigation & Contingency
Mitigation: The deep link handler must perform a lightweight existence check (HEAD request or minimal SELECT) before pushing the route. Define a contract with each destination screen for how to handle a not-found entity ID passed as a route parameter.
Contingency: If the existence check itself fails (network error), navigate to the destination screen anyway and let it handle the error gracefully with its own error state; do not block navigation for network timeouts.
If the tab badge widget triggers a full rebuild of the bottom navigation bar on every unread count change, it will cause visible jank on devices with many active Realtime events (e.g., org admins receiving org-wide alerts).
Mitigation & Contingency
Mitigation: Scope the badge widget to a dedicated BlocSelector that rebuilds only when the unread count value changes, not on any BLoC state emission. Use RepaintBoundary to isolate the badge from the rest of the nav bar.
Contingency: If performance issues persist, debounce badge updates to a maximum of one rebuild per 500ms and display the last known count during the debounce window.
Complex swipe-to-mark-read gestures and dynamic list updates may conflict with VoiceOver/TalkBack navigation patterns, particularly for Blindeforbundet users who rely exclusively on screen readers.
Mitigation & Contingency
Mitigation: Provide a dedicated accessibility action (Semantics.onTap / CustomSemanticsAction) for mark-as-read on each list item so screen reader users do not need the swipe gesture. Test with VoiceOver on iOS and TalkBack on Android before each release.
Contingency: If the swipe gesture proves incompatible with assistive technologies, disable it when a screen reader is detected (via ScreenReaderDetectionService) and rely solely on the tap-to-read and accessible action pathways.