Consider switching from dependencies.method(...) to inject(method)(...)

Overview

It’s a little annoying how the dependency injection pattern here means you can’t F12 / right-click > go-to-definition on injected methods…

Actual Behavior

dependencies.method(/* ... */);

Expected Behavior

What if an inject method were used instead:

inject(method)(/* ... */)

…where inject.ts looks something like:

export const inject = process.env !== 'test'
    ? <Method>(method: Method) => method
    : require('./spies').inject;

… and test-inject.ts uses a global Map of jest.fns that are reset before each test?

const spies = new Map<[Function, jest.Mock]>();

export const inject = (method: Function) => {
    const existingSpy = spies.get(method);
    if (existingSpy) {
        return spy;
    }
    
    const newSpy = jest.fn();
    spies.add(method, newSpy);
    return newSpy;
};

export const getSpy = <Method>(method: Method): jest.Mock<Method> =>
    spies.get(method) as Method;

Tests would then be able to getSpy(method) for any injected method…

[JoshuaKGoldberg]

Aha, I have it! They’ll still take in a first parameter, but it’ll be inject: Inject. Now they won’t have to explicitly declare dependencies.

export type Inject = <Value>(raw: Value) =>
    Value extends (inject: Inject, ...args: infer Args) => infer Return
    ? (...args: Args) => Return
    : Value;

import { logger } from '../../adapters/logger';

// ...

inject(logger).stdout.write('...');

[KingDarBoja]

💀 Rest in Peace I though you solved it and got lazy 😆