Integration test: full dashboard user flow
epic-coordinator-statistics-dashboard-ui-task-016 — Write integration tests (flutter_test integration_test package) covering the full coordinator dashboard flow: open dashboard, change period filter, verify charts and cards update, toggle to PersonalStatsView, verify personal bar chart renders, switch org-admin scope and verify updated data. Use a seeded test Supabase environment or golden mock to ensure deterministic chart output.
Acceptance Criteria
Technical Requirements
Execution Context
Tier 8 - 48 tasks
Can start after Tier 7 completes
Implementation Notes
Use `IntegrationTestWidgetsFlutterBinding.ensureInitialized()` at test entry. Provide a dedicated Supabase test project URL and anon key via `--dart-define` to isolate test data. Seed the database with a fixed dataset before each test using a `setUp` helper that calls Supabase RPC or direct REST inserts, then tears down in `tearDown`. For chart determinism, prefer golden mock providers (Riverpod overrides or BLoC test stubs) that return fixed data rather than relying on network.
Use `find.byKey(ValueKey('period-filter'))` style finders — ensure all interactive widgets have `Key` values set in the production widget tree. Avoid `pump(Duration)` with arbitrary durations; always use `pumpAndSettle` after async operations. Tag tests with `@Tags(['integration'])` so they can be excluded from fast unit test runs.
Testing Requirements
Integration tests only (flutter_test integration_test package). All tests run against a seeded Supabase test environment or deterministic golden mocks — no live production data. Cover: (1) initial render, (2) period filter change → data reload, (3) PersonalStatsView toggle → bar chart render, (4) org-admin scope switch → aggregated data. Each scenario asserts UI state after async completion using `pumpAndSettle`.
CI must execute the suite on every PR targeting the dashboard feature branch. Minimum: 1 happy-path test per major interaction; 1 error-state test (e.g., empty data period).
PeerMentorStatsList must handle rosters of 40+ mentors efficiently. A naive ListView implementation that re-renders all rows on filter change may cause frame drops on mid-range devices, degrading the experience for coordinators managing large chapters.
Mitigation & Contingency
Mitigation: Use ListView.builder with const constructors for row widgets. Profile the list with Flutter DevTools on a release build against a 60-mentor dataset before submitting for review. Implement sort in the BLoC layer, not in the widget.
Contingency: If frame drops persist, introduce pagination (load 20 mentors, scroll to load more) and add a search filter to reduce visible row count in practice.
The ActivityTypeDonutChart must render org-configured activity type labels from the org-labels system. If the terminology provider is not available or returns stale labels, chart segments will display raw key strings instead of human-readable organisation-specific names, confusing coordinators.
Mitigation & Contingency
Mitigation: Always resolve labels through the OrganizationLabelsProvider before passing data to the donut chart widget. Implement a fallback that formats the raw key as a readable string (e.g., 'peer_support' → 'Peer Support') if the provider returns null.
Contingency: If the org-labels system is unavailable, display the formatted fallback label and log a warning. Do not block chart rendering on label resolution — render with fallbacks immediately.
fl_chart widgets do not natively expose semantic labels for individual bars and donut segments. Without explicit Semantics wrappers, VoiceOver and TalkBack users will receive no meaningful chart information, failing accessibility requirements critical for Blindeforbundet and NHF deployments.
Mitigation & Contingency
Mitigation: Wrap each fl_chart widget in a Semantics widget with a descriptive label summarising the chart data. Implement the data table fallback toggle from the start, not as an afterthought. Validate with VoiceOver on iOS and TalkBack on Android during development.
Contingency: If fl_chart's rendering pipeline prevents semantic overlay from working reliably, replace chart widgets with a custom Canvas-based implementation that has full semantic control, or use a different charting library with better accessibility support.
The period filter state must be preserved when navigating away from the dashboard (e.g., drilling into a mentor's detail screen and pressing back). If the BLoC is re-created on navigation, the filter resets to default, forcing coordinators to re-select their period every time they drill down, degrading operational workflow efficiency.
Mitigation & Contingency
Mitigation: Provide the StatsBloc at a route level above the dashboard screen using a BlocProvider scoped to the statistics navigation shell, not inside the screen widget itself. Verify persistence with a widget integration test that navigates away and back.
Contingency: If route-level scoping is not achievable within the current navigation architecture, persist the last-used filter to a local session store and restore it on screen re-entry.