Dollar: Devlog 2

Published See discussion on Twitter

This post is a continuation of my first Dollar Devlog here.

It's been a bit since the last devlog, life has generally got in the way (work more-so than anything else), however this weekend I was able to spend quite a bit of time working on Dollar again.

I've made some changes since the last devlog, but still nothing publically available to showcase just yet!

Primarily, this Saturday I started off by forking my component library (@ds-pack/components) into a new package that completely swaps out Vanilla Extract for TailwindCSS (using daisyui). For those of you that know me, you might be a bit surprised that I decided to try out tailwind, I've frequently referred to it as "bootstrapcss for the modern age", yet ironically - that's exactly what I've been looking for as of late.

I don't intend to fully dig into that transition in this post (I might blog about it in another post however), but the main reason that I opted to switch was that I wanted to tear down how my css build pipeline works in most of my modern Next.js applications. I've written about that in my opinionated Next.js app setup blog post, but the gist is that having two process running at once to compile styles and run the app was a pain and let to a ton of issues when iteratively updating styles.

Daisyui seemed like a good option, and really all I wanted was a React version of it - which I felt like @ds-pack/components was a good bare bones React component library to fit that need. I've published a very early version of the library as @ds-pack/daisyui, feel free to check it out if you want to experiment with the library in your own side project! (I do plan to refactor my personal site with this version of my components, but that might take a while to fully implement)

Alright, back to my work on Dollar, I haven't made much progress because of the effort required to build the new UI library (although completeing it in the span of an afternoon was pretty sweet).

Outside of the new coat of paint, I've been able to:

  • Fully integrate React Server Components
  • Experiment with Server Actions (more on this below)
  • Begin to wire up supabase as the DB for the project

Server Actions

Alright, I know server actions are still super early, but they're kind of promising. Being able to do remote procedure calls within React is pretty dope!

For those that don't know, that roughly looks like the following:

1// some client code, like a form!
2'use client';
3
4import { login } from './actions';
5
6export default function Form() {
7 return (
8 <form>
9 <input name='email' type='text' />
10 <button
11 type='button'
12 onClick={async () => {
13 let userToken = await login({
14 email: document.getElementsByName('email')[0].value,
15 });
16 // do what you want with userToken here!
17 }}
18 >
19 Login!
20 </button>
21 </form>
22 );
23}
1// some client code, like a form!
2'use client';
3
4import { login } from './actions';
5
6export default function Form() {
7 return (
8 <form>
9 <input name='email' type='text' />
10 <button
11 type='button'
12 onClick={async () => {
13 let userToken = await login({
14 email: document.getElementsByName('email')[0].value,
15 });
16 // do what you want with userToken here!
17 }}
18 >
19 Login!
20 </button>
21 </form>
22 );
23}

You might be thinking, that doesn't seem too cool, we've been doing that for years, login from ./actions could just be a wrapper around a fetch() call!

Here's the cool thing, it is, but I don't have to do all the complext work of wiring up that fetch call, here's the full code of ./actions:

1// new directive!
2'use server';
3
4export async function login({ email }) {
5 // do something to login the user
6 return userToken;
7}
1// new directive!
2'use server';
3
4export async function login({ email }) {
5 // do something to login the user
6 return userToken;
7}

The really neat thing here is that React / Next are doing the heavy lifting for me to do the fetch call in the background!

There are some limitations however, e.g. for logging in a user I'd probably want to store a JWT as a cookie, unfortunately I don't have access to the raw Request and also I'm not able to send a raw Response back to the client - so I'd probably still want to reach for Next.js routes to support some of those use cases.

None the less, this is still a pretty neat pattern that I plan on exploring more as I get deeper into the project!