GitLab Incubation Engineering, 6 Jul 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Incubation Engineering Jamstack -- Update July 6th, 2022

Description

API Generator Repo: https://gitlab.com/gitlab-org/incubation-engineering/jamstack/apigen

TL;DR: https://gitlab.com/gitlab-org/incubation-engineering/jamstack/meta/-/issues/47

A

Hi everyone welcome to the jumpstack incubation engineering update, I'm, let's talk about what happened since the last update. I've been trying to get these two, mrs merged, that will introduce the pipeline wizard to pages they're, all just sitting there and moving very slowly through the review cycle. Since I need all of the back-end front-end and documentation reviews, the review cycle moves very very slowly. We just need a little more patience and I hope this will be merged soon anyway.

A

Now, for the actually exciting update I've been working on my api generator in the meantime, if you don't know what I'm talking about check out my last video, where I'll explain the larger idea?

A

Okay, so I started first with generating a prototype api that does two things a it's a graphql api and b. It runs on cloudflare workers as a database. It will at some point use cloudflare's new d1.

A

I don't have a lot of information yet on how that d1 interface is going to look like and they're understandably quiet about it, but from what I gather from cloudflare's one single blog post about it, I'm I'm assuming that the interface will likely be similar to the sqlite3 package for node.js.

A

So what I did for the time being, is I created a mock, d1 object that exposes an api. I assume to be similar to what d1 is going to offer.

A

This, for now is routing the requests to a local sqlite instance. So I can go ahead and develop my app and hopefully switch to a d1 instance once cloudflare releases their first beta of it.

A

Now with that d1 object mocked. I also needed to ensure that the libraries I use for the api don't rely on node.js core modules that can't easily be polyfilled.

A

The app will ultimately run on a cloudflare worker, which is not a node.js environment but closer to a browser lucky. For me, this also means I can actually use ecma script 2022. So that's really nice.

A

Now for the app I've decided to use knex as a sql or m on top of sqlite, I'm hoping this will make it easier to auto, generate code pieces and database migrations.

A

It should also make it easier for users that at some point, would like to start manually, maintaining the generated api and maybe switch to a different database engine. While doing so, I don't have a ton of experience with necks at the moment, but I'm learning as I go along um also for next. I couldn't directly use the sqlite dialect, because that requires direct access to the sqlite process with transactions and connections, uh connection management.

A

All these things are already abstracted when using d1. I assume so what I did is. I wrote the d1 dialect as a package which is an extension on the sqlite dialect um that takes care of the slightly different connection, interface, it'll, ultimately, transpile the same way sqlite3 does, but just the connection management is a bit simpler.

A

Now, with this prototype setup working, I've then started to create an app that will create this prototype, so I've abstracted the prototype app into a hydrogen template hygiene is a javascript based code generator, which makes a lot of sense to use when generating javascript apps.

A

So with the prototype transformed into a template, I can now run this command hygiene init repo, and it will generate a basic application structure with no models in it. Yet now I can go and add a model with hygiene model new, let's call it pokemon and it will generate a new pokemon data source and inject it.

A

Now I have this completely auto generated app with just two commands: let's go and test it.

A

I can now cd into the generated, app and run wrangler to dev and use the local flags so that it runs locally and look. It generates a server, and if we open the graphql endpoint, we can get the playground that will actually ship with every app and in here we can run. This query see I actually can't remember a lot of from pokemon, but I think pikachu wasn't the first one. Anyway, you just saw a completely auto-generated graphql api that runs on a cloudflare worker.

A

Right now, the app can only do the read function. The next steps are to add full quad support and this generator actually lives in a repo. If you want to follow its development, it's linked below the video, but beware it's under heavy development and so no docs lots of bugs and dead-ends.

A

Ultimately, I'd love to release this as a package, but that's still miles out. If you have any feedback, do let me know send me an issue ping me on slack anything until then. I hope you enjoyed this week's update. I hope to see you next time. Bye.