Feature: debug button to wipe all campaigns/encounters/logs in dev builds.
Previously bulk delete fetched all logs per campaign, client-filtered,
batchWrite — 30s+/campaign. Now SQL bulk DELETE, no fetch.
Server (server/db.js, server/index.js):
- deleteCollection(collPath, {where}) — SQL DELETE FROM docs WHERE parent=?,
optional where-filter. Broadcasts deletions to WS subscribers.
- DELETE /api/collection endpoint
- Gate: ALLOW_DEV_ENDPOINTS=1 env OR createServer({allowDevEndpoints:true})
- createServer accepts allowDevEndpoints param (tests bypass env)
Storage (src/storage/server.js, src/storage/firebase.js):
- deleteCollection(path, whereField, whereValue) both adapters
- Firebase: fetch matching + batch-delete (firestore no bulk), 500-chunk
- Gate: throws if NODE_ENV not development/test
- Contract-tested both backends
App (src/App.js):
- deleteCampaignCascade refactored (reusable, no try/catch split)
- handleDeleteAllCampaigns: Promise.all per campaign, deleteCollection for
encounters (no fetch), deleteCollection logs once globally, parallel
- Button dev-gated (NODE_ENV), confirm modal, hidden when no campaigns
Mock fixes (surfaced by new tests):
- firebase firestore mock: added where() export, getDocs applies constraints
(was returning all docs ignoring query constraints — pre-existing gap)
Tests:
- contract: deleteCollection (bulk, where-filter, empty) both backends
- server-contract: live deleteCollection (bulk, where, 403 gate)
- runStorageContract via makeStorage({allowDevEndpoints:true})
Safety (3 layers):
- UI button hidden in prod (NODE_ENV gate)
- storage method throws in prod (NODE_ENV gate)
- HTTP endpoint 403 in prod (env/param gate)
scripts/
Dev orchestration + manual demo tool. NOT test.
dev-start.sh / dev-stop.sh
Local dev stack: backend (:4001, sqlite) + frontend (:3999, server mode).
One command. Writes env vars for you. Logs to tmp/server.log, tmp/fe.log.
./scripts/dev-start.sh # start (idempotent: leaves running ports as-is)
./scripts/dev-stop.sh # stop both
replay-combat.js
Live backend demo. Drives full combat via server adapter (same contract as App). Player display live-updates. Watch UI react to state changes.
# start backend + frontend first (see docs/DEVELOPMENT.md)
node scripts/replay-combat.js [rounds] [delayMs]
# defaults: 100 rounds, 200ms/step
Coverage per round: damage, heal, all 22 conditions, toggleActive, removeParticipant, addParticipant (reinforcements), updateParticipant, pause/resume, reorderParticipants, endEncounter. Revives dead each round to sustain full round count.
See also
tests/audit/— exploratory bug-finders (manual run, non-deterministic){shared,server,src}/tests/— jest unit/integration/characterizationscratch/— gitignored throwaway