NodeJS - Lazy Init Patterns

One of the main challenges when dealing w/ the async nature of NodeJS is initializing classes/clients that requires some sort of side effect - such as database connection, disk reads or whatsoever. Even the simple idea of waiting for the first use-case to connect/initialize a resource.

Besides Dependency Injection - I like to use two approaches for this:

1) Leaving it up to the client to call connect or any other synonym - easy as creating an async function as the example below

const redis = require('redis');
const crypto = require('crypto');
//PROS: Damn easy, simple and straight-forward

//CONS: This leaves the entire responsibility to the client
class DistributedDataStructure {
    constructor(){
        this.client = redis.createClient();
    }

    async connect(){
        return this.client.connect();
    }

    async add(staffName, reviewId){
        //Do some business here - idk,
        const accountName = await this.client.get(key);
        return this.client.sAdd(`v1:${accountName}:pending-reviews`, reviewId);
    }
}

(async () => {
    const ds = new DistributedDataStructure();
    await ds.connect();
    ds.add('Jerome', crypto.randomBytes(12).toString('hex'));
})()

2) Proxying the access

In the real and wild-world we know that we have to deal w/ legacy code, legacy initialization methods and much more unexpected stuff - for this we have a second use-case which leverages the Proxy API for JS

Using Proxy it would look poorly-like

const redis = require('redis');
const { once } = require('events');
const crypto = require('crypto');

//PROS: No client responsibility - makes it easy for the client
//CONS: More complex and error prone
class ProxiedDistributedDataStructure {
    constructor(){
        this.client = redis.createClient();
        this.client.connect();
        return new Proxy(this, {
            get(target, property){
                const descriptor = target[property];
                if(!descriptor){
                    return;
                }
                if(target.isReady){
                    return descriptor;
                }
                return async function(){
                    await once(target.client, 'ready');
                    return descriptor.apply(target, arguments);
                }
            }
        });
    }

    async add(staffName, reviewId){
        //Do some business here - idk - like below
        const accountName = await this.client.get(staffName);
        return this.client.sAdd(`v1:${accountName}:pending-reviews`, reviewId);
    }
}

const client = new ProxiedDistributedDataStructure();
client.add('Jerome', crypto.randomBytes(12).toString('hex'));

The main benefit for the second approach is that we can instantiate the objects in sync contexts and only treat the method calls as async - instead of needing to play around some dirty gimmicks to call connect and chain promises - even worse, callbackifying.

Dev.to Code Examples

• First Approach
• Second Approach

Note: AFAIC from Redis V3^ we have an option legacyMode whenever creating the client which we can keep this lazy nature of Redis - doing client buffering of calls.