@cosmicdrift/kumiko-bundled-features 0.23.1 → 0.24.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cosmicdrift/kumiko-bundled-features",
-  "version": "0.23.1",
+  "version": "0.24.0",
   "description": "Built-in features — tenant, user, auth, delivery. The stuff you'd rewrite anyway, already typed.",
   "license": "BUSL-1.1",
   "author": "Marc Frost <marc@cosmicdriftgamestudio.com>",

package/src/user-data-rights/__tests__/file-binary-forget-cleanup.integration.test.ts

@@ -0,0 +1,178 @@
+// Forget-Hook binary-Cleanup Integration-Test.
+//
+// Beweist, dass der `fileRef`-Forget-Hook bei strategy="delete" die
+// Storage-Binaries via `storageProvider.delete()` entfernt, BEVOR die
+// row hard-gelöscht wird — ohne provider leakt sonst jede gelöschte
+// Datei ihre Bytes dauerhaft auf Disk (Issue gefunden im Review zu #177).
+
+import { afterAll, beforeAll, beforeEach, describe, expect, test } from "bun:test";
+import { asRawClient, insertOne } from "@cosmicdrift/kumiko-framework/bun-db";
+import type { DbConnection } from "@cosmicdrift/kumiko-framework/db";
+import {
+  createInMemoryFileProvider,
+  fileRefsTable,
+  type InMemoryFileProvider,
+} from "@cosmicdrift/kumiko-framework/files";
+import {
+  setupTestStack,
+  type TestStack,
+  unsafeCreateEntityTable,
+  unsafePushTables,
+} from "@cosmicdrift/kumiko-framework/stack";
+import { resetTestTables } from "@cosmicdrift/kumiko-framework/testing";
+import { getTemporal } from "@cosmicdrift/kumiko-framework/time";
+import { createComplianceProfilesFeature } from "../../compliance-profiles";
+import { createDataRetentionFeature, tenantRetentionOverrideEntity } from "../../data-retention";
+import { createFilesFeature } from "../../files";
+import { createSessionsFeature } from "../../sessions";
+import { createUserFeature, USER_STATUS, userEntity, userTable } from "../../user";
+import { createUserDataRightsDefaultsFeature } from "../../user-data-rights-defaults";
+import { createUserDataRightsFeature } from "../feature";
+import { runForgetCleanup } from "../run-forget-cleanup";
+
+let stack: TestStack;
+let db: DbConnection;
+let provider: InMemoryFileProvider;
+
+const TENANT = "00000000-0000-4000-8000-00000000000c";
+const TENANT_SYSTEM = "00000000-0000-4000-8000-000000000001";
+
+function uuid(suffix: number): string {
+  return `bbbbbbbb-bbbb-4bbb-8bbb-${suffix.toString(16).padStart(12, "0")}`;
+}
+
+type Instant = InstanceType<ReturnType<typeof getTemporal>["Instant"]>;
+const NOW = (): Instant => getTemporal().Now.instant();
+const pastInstant = (): Instant => getTemporal().Instant.fromEpochMilliseconds(Date.now() - 60_000);
+
+beforeAll(async () => {
+  provider = createInMemoryFileProvider();
+  stack = await setupTestStack({
+    features: [
+      createUserFeature(),
+      createFilesFeature(),
+      createDataRetentionFeature(),
+      createComplianceProfilesFeature(),
+      createSessionsFeature(),
+      createUserDataRightsFeature(),
+      createUserDataRightsDefaultsFeature({ storageProvider: provider }),
+    ],
+    files: { storageProvider: provider },
+  });
+  db = stack.db;
+
+  await unsafeCreateEntityTable(db, userEntity);
+  await unsafeCreateEntityTable(db, tenantRetentionOverrideEntity);
+  await unsafePushTables(db, { fileRefsTable });
+  await asRawClient(db).unsafe(`
+    CREATE TABLE IF NOT EXISTS read_tenant_memberships (
+      id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+      tenant_id UUID NOT NULL,
+      user_id TEXT NOT NULL,
+      version INTEGER NOT NULL DEFAULT 0,
+      inserted_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+      modified_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+      inserted_by_id TEXT,
+      modified_by_id TEXT,
+      is_deleted BOOLEAN NOT NULL DEFAULT false,
+      deleted_at TIMESTAMPTZ,
+      deleted_by_id TEXT,
+      roles TEXT NOT NULL DEFAULT '[]',
+      UNIQUE(user_id, tenant_id)
+    )
+  `);
+});
+
+afterAll(async () => {
+  await stack.cleanup();
+});
+
+beforeEach(async () => {
+  provider.clear();
+  await resetTestTables(db, [userTable, "read_tenant_memberships", fileRefsTable]);
+});
+
+async function seedForgetUser(id: string): Promise<void> {
+  await insertOne(db, userTable, {
+    id,
+    tenantId: TENANT_SYSTEM,
+    email: `user-${id}@example.com`,
+    passwordHash: "hashed",
+    displayName: `User ${id}`,
+    locale: "de",
+    emailVerified: true,
+    roles: '["Member"]',
+    status: USER_STATUS.DeletionRequested,
+    gracePeriodEnd: pastInstant(),
+  });
+}
+
+async function seedMembership(userId: string, tenantId: string): Promise<void> {
+  await asRawClient(db).unsafe(
+    `INSERT INTO read_tenant_memberships (tenant_id, user_id, roles)
+     VALUES ($1, $2, '["Member"]') ON CONFLICT (user_id, tenant_id) DO NOTHING`,
+    [tenantId, userId],
+  );
+}
+
+async function seedFile(id: string, tenantId: string, insertedById: string): Promise<string> {
+  const storageKey = `storage/${id}`;
+  await provider.write(storageKey, new Uint8Array([1, 2, 3, 4]), "application/pdf");
+  await asRawClient(db).unsafe(
+    `INSERT INTO file_refs (id, tenant_id, storage_key, file_name, mime_type, size, inserted_by_id)
+     VALUES ($1, $2, $3, $4, 'application/pdf', 4, $5) ON CONFLICT (id) DO NOTHING`,
+    [id, tenantId, storageKey, `${id}.pdf`, insertedById],
+  );
+  return storageKey;
+}
+
+describe("forget-binary-cleanup :: storage.delete fires before row hard-delete", () => {
+  test("Forget deletes the binary from the storage provider", async () => {
+    const userId = uuid(1);
+    await seedForgetUser(userId);
+    await seedMembership(userId, TENANT);
+    const key = await seedFile(uuid(101), TENANT, userId);
+    expect(await provider.exists(key)).toBe(true);
+
+    const result = await runForgetCleanup({ db, registry: stack.registry, now: NOW() });
+
+    expect(result.processedUserIds).toContain(userId);
+    expect(await provider.exists(key)).toBe(false);
+    expect(provider.keys()).not.toContain(key);
+  });
+
+  test("Multiple files from the same user — all binaries cleaned up", async () => {
+    const userId = uuid(2);
+    await seedForgetUser(userId);
+    await seedMembership(userId, TENANT);
+    const keys = await Promise.all([
+      seedFile(uuid(201), TENANT, userId),
+      seedFile(uuid(202), TENANT, userId),
+      seedFile(uuid(203), TENANT, userId),
+    ]);
+    expect(provider.keys()).toHaveLength(3);
+
+    await runForgetCleanup({ db, registry: stack.registry, now: NOW() });
+
+    for (const key of keys) {
+      expect(await provider.exists(key)).toBe(false);
+    }
+    expect(provider.keys()).toHaveLength(0);
+  });
+
+  test("Other tenants' files stay untouched", async () => {
+    const userId = uuid(3);
+    const otherTenant = "00000000-0000-4000-8000-00000000000d";
+    await seedForgetUser(userId);
+    await seedMembership(userId, TENANT);
+    const myKey = await seedFile(uuid(301), TENANT, userId);
+    const otherKey = await seedFile(uuid(302), otherTenant, "another-user");
+    // The other-tenant file is owned by a different user; the forget run for
+    // userId must NOT touch it.
+
+    await runForgetCleanup({ db, registry: stack.registry, now: NOW() });
+
+    expect(await provider.exists(myKey)).toBe(false);
+    expect(await provider.exists(otherKey)).toBe(true);
+  });
+});

package/src/user-data-rights-defaults/feature.ts

@@ -3,9 +3,20 @@ import {
   EXT_USER_DATA,
   type FeatureDefinition,
 } from "@cosmicdrift/kumiko-framework/engine";
-import { fileRefDeleteHook, fileRefExportHook } from "./hooks/file-ref.userdata-hook";
+import type { FileStorageProvider } from "@cosmicdrift/kumiko-framework/files";
+import { createFileRefDeleteHook, fileRefExportHook } from "./hooks/file-ref.userdata-hook";
 import { userDeleteHook, userExportHook } from "./hooks/user.userdata-hook";
 
+export interface UserDataRightsDefaultsOptions {
+  /**
+   * Wired into the fileRef delete-hook: on strategy="delete" the hook
+   * calls `storageProvider.delete(key)` per row before hard-deleting
+   * the row. Without it, file binaries leak on forget (Art. 17) — the
+   * hook logs a one-shot warning so misconfiguration stays visible.
+   */
+  readonly storageProvider?: FileStorageProvider;
+}
+
 // user-data-rights-defaults — Default-Hooks für die Core-Entities
 // `user` (S2.H1) und `fileRef` (S2.H2).
 //
@@ -23,7 +34,10 @@ import { userDeleteHook, userExportHook } from "./hooks/user.userdata-hook";
 // Pattern matched file-foundation + file-provider-s3 (separate Plugin-
 // Feature), nicht user/files schreiben ihre eigenen Hooks selbst weil
 // das circular-requires waere.
-export function createUserDataRightsDefaultsFeature(): FeatureDefinition {
+export function createUserDataRightsDefaultsFeature(
+  options: UserDataRightsDefaultsOptions = {},
+): FeatureDefinition {
+  const fileRefDeleteHook = createFileRefDeleteHook(options.storageProvider);
   return defineFeature("user-data-rights-defaults", (r) => {
     r.requires("user", "files", "user-data-rights");
 

package/src/user-data-rights-defaults/hooks/file-ref.userdata-hook.ts

@@ -1,6 +1,6 @@
 import { deleteMany, selectMany, updateMany } from "@cosmicdrift/kumiko-framework/bun-db";
 import type { UserDataDeleteHook, UserDataExportHook } from "@cosmicdrift/kumiko-framework/engine";
-import { fileRefsTable } from "@cosmicdrift/kumiko-framework/files";
+import { type FileStorageProvider, fileRefsTable } from "@cosmicdrift/kumiko-framework/files";
 
 // userData-Hook fuer fileRef-entity (S2.H2).
 //
@@ -9,10 +9,9 @@ import { fileRefsTable } from "@cosmicdrift/kumiko-framework/files";
 // NICHT direkt — sie werden via signed-Download-URLs separat ins ZIP
 // gepackt (S2.U3 Export-Job-Pipeline orchestriert das).
 //
-// Delete-Hook entfernt FileRef-Zeile + Storage-Binary. Plan-Roadmap
-// docs/plans/datenschutz/storage-encryption.md hat das Subject-
-// Resolver-Pattern fuer File-Encryption als Sprint 4 — bis dahin:
-//   "delete":    Row hard-delete + storageProvider.delete() pro File
+// Delete-Hook entfernt FileRef-Zeile via factory
+// `createFileRefDeleteHook(storageProvider)`:
+//   "delete":    storageProvider.delete() pro File (best-effort) + Row hard-delete
 //   "anonymize": insertedById=null, Row + binary bleiben (FK-Refs
 //                koennen weiter zeigen; Personenbezug raus)
 //
@@ -22,12 +21,17 @@ import { fileRefsTable } from "@cosmicdrift/kumiko-framework/files";
 // idempotent, KEIN globaler Rollback — wenn ein File-Delete failt,
 // bleibt der User-Row trotzdem anonymisiert.
 //
-// Storage-Provider kommt aus dem App-Bootstrap (createBunServer-
-// options.files.storageProvider). Wir greifen darauf via ctx — der
-// Hook-ctx hat aktuell nur db/tenantId/userId, also fuer Storage-
-// Calls braucht es eine Erweiterung. S2.U3 Export-Job-Pipeline regelt
-// das (Job-ctx hat ctx.files.ref(key)). Hier lassen wir Storage-
-// Cleanup als TODO und faellen das in S2.U5 nochmal an.
+// `storageProvider` ist optional. App-Author wired es beim
+// Feature-Mount rein (`createUserDataRightsDefaultsFeature({
+// storageProvider })`). Ohne Provider macht der Hook row-only-delete,
+// die Bytes leaken — der Caller bekommt EINEN Warn beim ersten Lauf
+// pro Process, damit die Konfiguration sichtbar fehlerhaft ist.
+//
+// Caveat: hard-delete via deleteMany emittiert KEIN fileRef.deleted —
+// die storage-tracking-MSP dekrementiert nicht. Wenn die zu loeschenden
+// Files vorher nicht soft-deleted waren, bleibt `tenant_storage_usage`
+// inflated. Forget-Flows sind selten (per-User-Art.-17) und damit
+// bounded; ein executor.purge-API folgt mit dem trashed-files-GC.
 
 export const fileRefExportHook: UserDataExportHook = async (ctx) => {
   // isDeleted:false — soft-deleted (trashed) Files gehören nicht ins
@@ -76,22 +80,54 @@ export const fileRefExportHook: UserDataExportHook = async (ctx) => {
   };
 };
 
-export const fileRefDeleteHook: UserDataDeleteHook = async (ctx, strategy) => {
-  if (strategy === "delete") {
-    // Hard-delete der FileRef-Rows fuer diesen User in diesem Tenant.
-    // Storage-Binary-Cleanup folgt in S2.U5 wenn der Forget-Job-Ctx
-    // den Storage-Provider exposed.
-    await deleteMany(ctx.db, fileRefsTable, { tenantId: ctx.tenantId, insertedById: ctx.userId });
-  } else {
-    // anonymize: insertedById=null, FileRef + binary bleiben.
-    // Use-case: shared chat-Attachment in einem Multi-User-Channel —
-    // Author-Identifikation raus, Datei bleibt fuer andere User
-    // sichtbar.
-    await updateMany(
-      ctx.db,
-      fileRefsTable,
-      { insertedById: null },
-      { tenantId: ctx.tenantId, insertedById: ctx.userId },
-    );
-  }
-};
+let missingStorageWarned = false;
+
+export function createFileRefDeleteHook(
+  storageProvider: FileStorageProvider | undefined,
+): UserDataDeleteHook {
+  return async (ctx, strategy) => {
+    if (strategy === "delete") {
+      if (storageProvider) {
+        const rows = await selectMany(ctx.db, fileRefsTable, {
+          tenantId: ctx.tenantId,
+          insertedById: ctx.userId,
+        });
+        for (const row of rows) {
+          const key = (row as Record<string, unknown>)["storageKey"]; // @cast-boundary db-row
+          if (typeof key !== "string" || key.length === 0) continue;
+          try {
+            await storageProvider.delete(key);
+          } catch (err) {
+            // biome-ignore lint/suspicious/noConsole: operator-visibility for binary-cleanup-failure
+            console.warn(
+              `[user-data-rights-defaults:fileRef] storage delete failed key=${key} err=${err instanceof Error ? err.message : String(err)}`,
+            );
+          }
+        }
+      } else if (!missingStorageWarned) {
+        missingStorageWarned = true;
+        // biome-ignore lint/suspicious/noConsole: misconfiguration visibility — disk-leak in forget-flow
+        console.warn(
+          "[user-data-rights-defaults:fileRef] no storageProvider configured — file binaries are NOT deleted on forget. Pass createUserDataRightsDefaultsFeature({ storageProvider }) to fix.",
+        );
+      }
+      await deleteMany(ctx.db, fileRefsTable, { tenantId: ctx.tenantId, insertedById: ctx.userId });
+    } else {
+      // anonymize: insertedById=null, FileRef + binary bleiben.
+      // Use-case: shared chat-Attachment in einem Multi-User-Channel —
+      // Author-Identifikation raus, Datei bleibt fuer andere User
+      // sichtbar.
+      await updateMany(
+        ctx.db,
+        fileRefsTable,
+        { insertedById: null },
+        { tenantId: ctx.tenantId, insertedById: ctx.userId },
+      );
+    }
+  };
+}
+
+// Legacy export: storage-less hook for callers that haven't migrated.
+// Binaries are NOT cleaned up — disk leak. Migrate to
+// createUserDataRightsDefaultsFeature({ storageProvider }).
+export const fileRefDeleteHook: UserDataDeleteHook = createFileRefDeleteHook(undefined);