Organizing an application
The library knows nothing about your entities. A practical layout keeps the views of each entity in one file and gathers them into a plain views object, so every call site reads the same way.
import { fieldsOf } from '@atom-forge/expose';
type PublicUser = { id: string; name: string; email: string };
export const fields = fieldsOf<PublicUser>();
export const privateSalary = (source: { salary: number }) => ({ salary: source.salary });
import * as user from './user';
export const views = { user };
import { expose } from '@atom-forge/expose';
import { views } from '../views';
return rows.map(row => expose(row, views.user.fields('id', 'name')));
Who, and what
Keep two questions apart. The module decides who may get the data, with its usual permission check. A view only says what goes out, and it cannot check anything about the caller. When a view carries sensitive data, make that visible in its name and call it only after the check:
assertCanReadSalary(viewer, row);
return expose(row, views.user.fields('id', 'name'), views.user.privateSalary);
Keeping raw rows out
The library cannot stop code from returning a row without calling expose. If that matters in your project, add a test that scans the server code for the outputs you want to forbid, with an explicit allow-list for the exceptions.
Organizing an application
The library knows nothing about your entities. A practical layout keeps the views of each entity in one file and gathers them into a plain
viewsobject, so every call site reads the same way.// server/views/user.ts import { fieldsOf } from '@atom-forge/expose'; type PublicUser = { id: string; name: string; email: string }; export const fields = fieldsOf<PublicUser>(); export const privateSalary = (source: { salary: number }) => ({ salary: source.salary });// server/views/index.ts import * as user from './user'; export const views = { user };// at a call site import { expose } from '@atom-forge/expose'; import { views } from '../views'; return rows.map(row => expose(row, views.user.fields('id', 'name')));Who, and what
Keep two questions apart. The module decides who may get the data, with its usual permission check. A view only says what goes out, and it cannot check anything about the caller. When a view carries sensitive data, make that visible in its name and call it only after the check:
assertCanReadSalary(viewer, row); // your own permission check return expose(row, views.user.fields('id', 'name'), views.user.privateSalary);Keeping raw rows out
The library cannot stop code from returning a row without calling
expose. If that matters in your project, add a test that scans the server code for the outputs you want to forbid, with an explicit allow-list for the exceptions.