EN Install Upgrade Notes V2 To V3

Upgrade notes: v2 to v3

Français · Reference · Published, reviewed by a person · Applies from: 3.0.0

These are the notes written for 3.0.0, when v3 broke the authoring contract once (authoring schema 2 to 3). They are kept, unchanged, for a game still on authoring schema 2: a game written before 3.0.0, or a 4.1.x game that never opted in. They do not describe a step of a current upgrade.

• Today the automatic part is one command. In a created project, npx quest-n-click migrate sees schema 2 and runs ids --write --map (step 2 below) for you; in the engine repository it is GAME=<id> npm run migrate. The three lines of game.ts (step 2) stay yours. Upgrade gives the whole procedure.
• A game on a hosted Studio is upgraded by its owner through the engine upgrade (POST /api/engine/migrate), see Upgrading the engine of a hosted game.
• The commands below are written as they were in 3.0.0, for a game inside the engine repository (GAME=<id>). Some have changed since: npm run build no longer runs the pixel tests (4.1.6), the names web-scumm became quest-n-click (4.2.5, removed in 4.3.0). The later notes say when: 4.1, 4.2, 4.3.

The original introduction:

v3 deliberately breaks the authoring contract once so saves can become durable afterwards. Migrate on a branch, keep a representative v2 save, and do not change prose or reorder content until the ids below have been assigned.

Audience, context, and the statuses of what this page rests on

Audience: author · Context: created project

1. Establish the baseline

On the last v2 commit, export saves at important checkpoints and keep one complete session. Record:

GAME=<id> npm run validate
GAME=<id> npm run solve
GAME=<id> npm test

Do not commit private game material or exported player saves to this public repository.

2. Opt into the v3 schema

Add schemaVersion: 3 to defineGame({...}). compileGame clones and normalises the source; the compiled v3 value is frozen, so integrations must not rely on engine.game === source or mutate it after construction.

Let the tool write the ids, then review them:

npm run ids                  # dry run: what would be written, what must be done by hand
npm run ids -- --write --map # writes the ids into rooms/*.ts, rules.ts, game.ts; renames locales/*.json keys;
                             # writes ids.migration.json (renameSeen / renameCounter) and ids.paths.json

It names things after their content (house.open-pantry, house.grandma.where-is-the-key, house.open-pantry.once, clock.wait), uniquely across the game, deterministically. What it needs on every room and game Rule, TalkTopic, Choice, EventRule, once / nth / cycle / random block: an id; on every script: one stepIds entry per top-level command. Lists built by code (a spread, a .map(...)) are skipped and reported with the id the engine expects: add it in the generator. Then the three lines by hand in game.ts: schemaVersion: 3, saveVersion + 1, migrations: [idsMigration] with import idsMigration from './ids.migration.json'. npm run validate refuses a v3 game with a missing or duplicate id; npm run ids can run again at any time (it never renames twice). Generated exit rules receive deterministic ids automatically.

3. Migrate saves deliberately

The browser stores a web-scumm-save envelope with schema 3 in IndexedDB. The first v3 run imports the legacy <game>.save localStorage autosave once and removes it only after a verified write. Manual slots (saves.slots) live in the same database, as envelopes, verified the same way; the v2 <game>.slot.<n> entries are imported once too. A file imported from the save menu also lands in the first free slot.

Increase saveVersion for a state change and supply one Migration per integer step. Besides v2 fields, v3 supports:

migrations: [{
  from: 2,
  renameCounter: { 'house:on3.0': 'pantry.first_open' },
  renameSeen: { 'topic.house/grandma[0]': 'topic.grandma.ask_key' },
  renameScript: { old_clock: 'clock' },
  renameScriptStep: { clock: { 'step.1': 'chime' } },
  renamePlayer: { kid: 'laverne' },
  renameCharacter: { old_grandma: 'grandma' },
  dropCounter: ['temporary_gag'],
  dropSeen: ['removed_choice'],
  dropScript: ['removed_patrol'],
}],

Map positional v2 counter/topic/script keys only when the old content and its meaning are known. There is no universal safe conversion after content has already been reordered or translated. An unmapped stale optional reference is pruned with a player-visible warning; malformed JSON, a foreign game, a missing current room or an unknown active player is rejected without replacing the current session.

4. Update translations and minigames

In v3 the translation paths name things by id (room:house/on.house.open-pantry.do[1], talk.grandma.<id>.topic, .choice.<id>.text, events.<id>.do); npm run ids -- --write renamed the existing tables (ids.paths.json holds the map). Run npm run i18n -- extract --lang <xx> for every shipped language. v3 also extracts verb labels/join words and the text paths declared by each minigame's textParams. Then run npm run i18n -- status and translate every new entry.

Custom minigames should declare required and textParams so validation and translation tools understand their parameters. Asset ids keep the existing sheet/cell convention and are discovered by the asset tools.

5. Update development and CI commands

• npm run dev and npm run studio are loopback-only.
• Use dev:lan / studio:lan for a phone and open the printed token URL.
• npm test is the Node suite; npm run test:assets is the Python-backed asset suite.
• npm run build requires validation and a winning witness, but not an unbounded state-space proof.
• npm run prove:game proves global and chapter state graphs and fails on softlocks or truncation.
• npm run release-check includes prerequisites, build, exhaustive proof and dependency audit.

Do not enable a partial-order reduction in proof mode until equivalence with the reference exploration is demonstrated for the relevant game class.

6. Verify the migrated game

GAME=<id> npm run doctor
GAME=<id> npm run build
GAME=<id> npm run prove:game
GAME=<id> npm run e2e -- http://127.0.0.1:5173/ --prod

Load every representative v2 save and verify its room, active player, inventories, persistent choices and script step. Reorder one rule, topic, choice and script step in a test fixture; the same v3 save must retain its meaning. Finally, test the production PWA online, offline in a room never visited (npm run e2e:pwa does it), and through an update. The whole game is cached after the first visit unless the game says offline: 'nearby' (decision D5).

7. At a glance

What migrates on its own. The v2 autosave is imported into IndexedDB on the first v3 launch and verified before the old key goes. A save that names something the content no longer has is pruned with one toast (ui.saveAdjusted), never refused. Without an id, a once / nth / cycle / random block keeps its positional v2 key, so v2 counters survive. The solver's witness is unchanged (the private reference game: 59 actions, 355 states, 0.3 s on both).

v2	v3
npm run dev listens on the network	npm run dev listens on 127.0.0.1; npm run dev:lan / studio:lan print a one-session token URL for the phone
npm test runs everything	npm test = Node tests; npm run test:assets = the Python-backed image tests; npm run check = tsc + Node tests
npm run build = tsc + tests + vite	npm run build = check + test:assets + verify:game (validate, solve, --chapters, i18n status) + vite + spoilers + asset audit
—	npm run prove:game = --prove and --prove --chapters, the exhaustive softlock gate, run by release-check, never by build
npm run audit	npm run audit (assets, private names) + npm run audit:deps (npm audit, production deps)
—	npm run doctor, npm run e2e:smoke, npm run e2e:pwa, npm run release-check

Solver exit codes. 0 solved and no broken invariant; 1 unsolved, softlocks, errors or a broken invariant; 2 truncated. --json gains status, mode, softlocks, assumptions; the human output gains one line after the verdict (Witness status: / Proof status:), everything else is unchanged.

Where	Field	Used by
a rule in on (room or game)	id	saves, the puzzle graph, the solver's heatmap (rule:<id>)
a choice option	id	once choices in saves, translations, voice files
a talk topic	id	seen topics in saves, translations
a listener in events	id	once listeners in saves
once / nth / cycle / random	id	their counters in saves (v2 key still read, deprecated)
a script	stepIds, one per command of do	a save resumes at the named step after a reorder

New ui keys (English defaults when absent): saveFailed, saveAdjusted, updateAvailable, updateNow, advance, offlineStatus, offlineComplete, offlineRetry (3.1.1: the pause menu's offline row) (3.1, the "tap to continue" marker for screen readers), shareSession (3.1, the playtest row of the pause menu).

---

Edit the source of this page · Report an error