dc93745841
The react-beautiful-dnd → Pragmatic Drag and Drop migration in list_table landed without unit coverage for the new hooks (`use_list_table_dnd`, `use_list_table_row_dnd`), the new `row_drop_indicator` component, or the keyboard-reorder path in `list_table` itself. Add Jest coverage for all four, plus a regression block in `user_properties_table.test.tsx` that confirms the arrow-key reorder path on an existing list_table consumer still calls back into `reorderField` after the migration. - use_list_table_dnd: monitor registration lifecycle, canMonitor kind filtering, full getDropIndex truth table (top/bottom × source position), no-op when dropIndex == sourceIndex, useLatest pickup of onReorder without re-registration, dragKind change tears down and re-registers - use_list_table_row_dnd: initial null edge, no registration when rowElement is null or enabled is false, draggable + dropTarget registration, canDrop kind + rowId filter, getData edge payload, onDrag/onDragLeave/onDrop edge state transitions, cleanup on unmount, custom preview wiring - row_drop_indicator: render structure, edge prop forwarding, edge-change rerender (wrapped in act to settle floating-ui's flushSync autoUpdate) - list_table: drag-handle render variants (no onReorder, enabled, isRowDragDisabled), keyboard ArrowUp/ArrowDown dispatch with bounds, non-arrow no-op, drag-handle click does not bubble into row onRowClick - user_properties_table: regression coverage that the keyboard reorder path still resolves the correct UserPropertyField and next index for `reorderField` Co-authored-by: Cursor <cursoragent@cursor.com> |
||
|---|---|---|
| .. | ||
| channels | ||
| patches | ||
| platform | ||
| scripts | ||
| .gitignore | ||
| .npmrc | ||
| AGENTS.md | ||
| CLAUDE.OPTIONAL.md | ||
| config.mk | ||
| Makefile | ||
| package-lock.json | ||
| package.json | ||
| README.md | ||
| STYLE_GUIDE.md | ||
Mattermost Web App
This folder contains the client code for the Mattermost web app. It's broken up into multiple packages each of which either contains an area of the app (such as playbooks) or shared logic used across other packages (such as the packages located in the platform directory). For anyone who's used to working in the mattermost/mattermost-webapp repo, most of that is now located in channels.
npm Workspaces
To interact with a workspace using npm, such as to add a dependency or run a script, use the --workspace (or --workspaces) flag. This can be done when using built-in npm commands such as npm add or when running scripts. Those commands should be run from this directory.
# Add a dependency to a single package
npm add react --workspace=playbooks
# Build multiple packages
npm run build --workspace=platform/client --workspace=platform/components
# Test all workspaces
npm test --workspaces
# Clean all workspaces that have a clean script defined
npm run clean --workspaces --if-present
To install dependencies for a workspace, simply run npm install from this folder as you would do normally. Most packages' dependencies will be included in the root node_modules, and all packages' dependencies will appear in the package-lock.json. A node_modules will only be created inside a package if one of its dependencies conflicts with that of another package.
Useful Links
- Developer setup, now included with the Mattermost server developer setup
- Web app developer documentation
Dependency Changes
Any PR that modifies package.json or package-lock.json needs extra scrutiny:
- No duplicate libraries. Before adding a new dependency, check whether an existing one already covers the same use case. Multiple libraries for the same purpose (e.g., two different date pickers, or Bootstrap 3 and Bootstrap 4 simultaneously) create long-term upgrade pain.
- License check. New dependencies must not use GPL or similarly restrictive licenses. Dependencies with no license at all should also be flagged.
- Justify the addition. A new dependency should solve a real problem that existing code or dependencies don't already address. Push back on adding packages for trivial functionality.
- Version conflicts. Check whether the new dependency introduces conflicting peer dependency versions. Cascading version conflicts are expensive to untangle later and have historically blocked upgrades for months.