impress/docs/wardrobe-v2-migration.md

# Wardrobe V2 Migration Status

This document tracks the status of Wardrobe V2, a ground-up rewrite of the outfit editor using Rails + Turbo, replacing the React + GraphQL system embedded from Impress 2020.

## Goal

Replace the complex React outfit editor (`app/javascript/wardrobe-2020/`) with a simpler Rails/Turbo implementation that:
- Eliminates dependency on Impress 2020's GraphQL API
- Uses progressive enhancement (works without JavaScript)
- Leverages Web Components for interactive features
- Reduces frontend complexity and maintenance burden
- Eventually enables full deprecation of the Impress 2020 service

## Current Status

**Wardrobe V2 is in active development** on the `feature/wardrobe-v2` branch. It's accessible at `/wardrobe/v2` but is not yet linked from the main UI.

### What Works

- **Species/color/pose/style selection**: Full picker UI with visual pose thumbnails, availability indicators, canonical fallback, and alt style picker with tabbed UI
- **Item search**: Text search with auto-filtering by pet compatibility, pagination, add/remove items
- **Item display**: Grouped by zone with multi-zone simplification, incompatible items section, NC/NP badges
- **Outfit rendering**: Uses the shared `<outfit-viewer>` web component
- **Progressive enhancement**: Everything works without JS; web components add auto-submit and smoother interactions
- **Smooth navigation**: Idiomorph DOM morphing reuses `<outfit-viewer>` layers across full-page navigations
- **Outfit saving/loading**: Load saved outfits at `/outfits/:id/v2`, save changes (owner) or save copies (non-owner), editable outfit name, unsaved changes warning

### Key implementation files

Code lives in `app/controllers/wardrobe_controller.rb`, `app/views/wardrobe/`, `app/helpers/wardrobe_helper.rb`, and `app/assets/{stylesheets,javascripts}/wardrobe/`.

## Technical Approach

**Simplicity over polish**: Unlike many webapps, we value the simplicity of the codebase very highly, even at the expense of an ideal user experience. Start with simple solutions, even if the UX is clunky; then we'll iterate up as needed.

**URL as single source of truth**: All outfit state lives in URL params (`species`, `color`, `pose`, `style`, `objects[]`, `q[...]`). Every interaction is a GET request that generates a new URL. No client-side state management. Browser back/forward work naturally.

**Server-side rendering + Web Components**: All HTML is generated server-side. Lightweight web components (`<auto-submit-form>`, `<pose-picker-popover>`, `<tab-panel>`, `<outfit-viewer>`) add interactivity without framework overhead.

**Progressive enhancement**: Submit buttons appear when JS is slow/disabled. Web components enhance forms with auto-submit on change.

## Roadmap

### Phase 1: Core Functionality (MVP)

The goal is a basic usable wardrobe. Species/color/pose selection, item search, and add/remove are already done.

**Outfit Saving/Loading** (basic implementation done)
- Save button near editable outfit name, disabled/"Saved!" when state matches saved
- Route to load saved outfits (`GET /outfits/:id/v2`) with redirect-based state initialization
- "Save a copy" for non-owners, login prompt for unauthenticated users
- `beforeunload` warning for unsaved changes via MutationObserver
- Outfit name: For saved outfits, rename is a standalone PATCH operation (not a URL param). For unsaved outfits, name is tracked as a URL param. Progressive enhancement shows static text + pencil icon for renaming.

**Alt Styles Support** (done)
- `Outfit#visible_layers` handles alt styles
- Picker UI with tabbed Expressions/Styles panels, `style` URL param, visual thumbnails, "Default" option
- Stale style params dropped gracefully when switching species
- Search results auto-filtered by alt style compatibility

**Closeted Items** (baseline done, progressive enhancement pending)
- Instead of just wearing/unwearing items, also support a "closeted" state: the user is *considering* this item,
  but it is not displayed on the pet itself right now.
- Wearing an item will stop wearing, but keep in closet, items that are mutually incompatible with it.
- Baseline behavior: separate toggle buttons for worn state and closeted state.
  - Unworn items have "Add" (wear).
  - Worn items have "Hide" (stop wearing, keep in closet) and "Remove" (remove from worn and closet).
  - Closeted items have "Show" (wear) and "Remove" (remove from closet).
- Visual distinction: worn items have green emphasis, closeted items have dashed border and reduced opacity.
- Closeted items appear in zone groups alongside worn items.
- `closet[]` URL params, saving/loading, and search pagination all work.
- Progressive enhancement (done):
  - Each item card is an `<item-card>` custom element with a visually-hidden input inside a `<label>`.
  - In the outfit view, items use radio inputs (mutual exclusivity within zone via `name` attribute). Arrow keys
    navigate between items via native radio behavior.
  - In the search view, items use checkbox inputs (independent toggle).
  - In both views, the `<item-card>` web component delegates clicks to the baseline forms: clicking a closeted or
    absent item submits its Show/Add form (wears it), clicking a worn item submits its Hide form (un-wears it).
  - When the item is closeted or worn, there is a "Remove" button.
  - The radio button and checkbox are visually hidden, and are reflected in the item card worn/closeted styles instead.

### Phase 2: Polish & Parity

Match the quality and usability of Wardrobe 2020 where it matters.

- **Item UX**: Wear/unwear toggle, item info links, zone badges, removal animations, incompatibility tooltips
- **Preview controls**: Download as PNG, copy link, settings (hi-res mode, archive toggle), HTML5 conversion badge
- **Loading & errors**: Spinners, skeleton screens, smooth transitions, error recovery
- **Mobile**: Touch-friendly targets, fix hover-dependent interactions, always-visible controls on touch
- **Accessibility**: ARIA labels, keyboard navigation, semantic headings, skip links, color contrast

### Phase 3: Advanced Features

Feature parity with Wardrobe 2020 where valuable.

- **User features**: Auth integration, closet/ownership/wishlist badges, filter search by owned/wanted
- **Conflict management**: Auto zone conflict resolution, smart item restoration on unwear
- **Pet loading**: "Load my pet" by name, modeling integration
- **Search enhancements**: Inline syntax (`is:nc`, `fits:blue-acara`), advanced filter UI, autocomplete
- **Outfit auto-saving**: Save outfit changes automatically over time, rather than requiring clicking Save

### Phase 4: Migration & Rollout

- Gradual rollout (staff → beta → default)
- Performance measurement and optimization (Turbo Frames for partial updates?)
- Visual polish and design refinement
- Remove wardrobe-2020 code and update Impress 2020 dependencies doc

### Deferred / Maybe Never

- Support mode features (can keep using old wardrobe)
- Known glitches system (complex, low value)
- Appearance version pinning (niche)
- UC pet support (being phased out)

## Open Questions

- Is the URL-as-state approach sustainable as complexity grows (saving, closet, conflicts)?
- Which Wardrobe 2020 features are actually essential vs. nice-to-have? (Need user feedback)
- How to handle the transition period (maintain both? redirect? feature flag?)

## References

- [Impress 2020 Dependencies](./impress-2020-dependencies.md) - What still depends on the Impress 2020 service
- [Customization Architecture](./customization-architecture.md) - Data model deep dive
- Wardrobe 2020 source: `app/javascript/wardrobe-2020/` (the authoritative reference for feature parity comparisons)