This should be considered an advanced topic and creating a custom HttpClientImpl is not something you will likely need to do. Also, we don't offer support beyond this article for writing your own implementation.

It is possible you may need complete control over the sending and receiving of requests beyond what is provided by the NodeFetchClient, SPRequestExecutorClient, or the default FetchClient - however these cases are likely rare. If you are looking for ways to support alternative authentication please see this blog post on using the excellent node-sp-auth library with sp-pnp-js as a first step.

Before you get started read and understand the fetch specification as you are essentially writing a custom fetch implementation.

The first step (second if you read the fetch spec as mentioned just above) is to understand the interface you need to implement, HttpClientImpl.

export interface HttpClientImpl {
    fetch(url: string, options: FetchOptions): Promise<Response>;
}

There is a single method "fetch" which takes a url string and a set of options. These options can be just about anything but are constrained within the library to the FetchOptions interface.

export interface FetchOptions {
    method?: string;
    headers?: HeadersInit | { [index: string]: string };
    body?: BodyInit;
    mode?: string | RequestMode;
    credentials?: string | RequestCredentials;
    cache?: string | RequestCache;
}

So you will need to handle any of those options along with the provided url when sending your request. The library will expect your implementation to return a Promise that resolves to a Response defined by the fetch specification - which you've already read 👍.

Once you have written your implementation using it on all your requests is done by setting it in the global library configuration:

import { setup, Web, NodeFetchClient } from "sp-pnp-js";
import { MyAwesomeClient } from "./awesomeclient";

setup({
    fetchClientFactory: () => {
        return new MyAwesomeClient();
    }
});

let w = new Web("{site url}");

// this request will use your client.
w.select("Title").get().then(w => {
    console.log(w);
});

You can of course inherit from one of the provided implementations listed at the start of this article if you just need to say add a header or need to do something to every request sent. Perhaps some advanced logging. This approach will save you from needing to fully write a fetch implementation.

Whatever you do, do not write a client that uses a client id and secret and exposes them on the client side. Client Id and Secret should only ever be used on a server, never exposed to clients as anyone with those values has the full permissions granted to that id and secret.