apiv2

[bkendall]

Description

Creating a new apiv2 package that's in typescript and will allow us to change out the current api package (based on request). This is built using node-fetch which is fairly light-weight.

The exported Client module has a generic request method for... generic requesting, but also has 'verb' methods matching various HTTP verbs to make most requests very simple to write.

Scenarios Tested

I've kept this to the hosting:channel:* commands and have been using them as my tests. Seems to be working quite well!

• hosting:channel:list works
• hosting:channel:create works (and with --expires)
• hosting:channel:deploy works
• hosting:channel:delete works

[mbleigh]

I think creating a parallel package is a great idea (allows us to migrate gradually without worrying about getting it 100% right all at once). But if we're gonna do that, feels like we can kinda just go nuts on the new implementation and make the interface exactly what we want without worrying about backwards compat. For instance, maybe we should take an instantiated client approach, something like:

import { Client, hostingOrigin } from "../apiv2";

const api = new Client({
  origin: hostingOrigin,
  apiVersion: "v1beta1",
  auth: true
});

const channel = await api.post<CreateChannelRequest, Channel>({
  ttl: `${ttlMillis / 1000}s`
});

that's just an example, but we could try to build a more opinionated client experience and maybe pull in some OP-specific utilities (for instance, imagine a listPaginated() method that auto-paginates using OP parlance).

I'm not opposed to a "the existing thing, but TypeScript", but it looks like this approach is changing a fair amount of the surface but not really as a ground-up reimagining.

Another example here is that we no longer really need to vary requests by scope -- that's not a thing we do anymore since we rely on IAM to be our permissions gate.

[mbleigh]

Isn't substr preferred now?

[bkendall]

Not that I recall. They're two different functions (which can be basically substituted here): the second argument for substr is the number is the length to return, the second argument to substring is the index to stop at. Using 0 for the first argument, either is technically fine.

[mbleigh]

Please manually test codepaths for all swapped implementations before merging.