feat: add 'PREACT_APP_' prefixed env vars automatically & pickup .env file

[rschristian]

What kind of change does this PR introduce?

feature

Did you add tests for your changes?

Yes

Summary

This is something that's asked about fairly often and I figure it should become better supported for how low maintenance it is for us.

Any env vars prefixed with PREACT_APP_ will now automatically be added to DefinePlugin. This matches what CRA does and also (mostly) matches what preact-cli-plugin-env-vars does. This is still a fairly popular plugin, getting ~1/10 of the downloads per week that preact-cli itself is.

I added a note in the changeset to give credit for this functionality in the past. Hopefully that's appropriate; no code is actually shared.

.env files in the user's project root are also now automatically consumed. Added dotenv for handling this.

Does this PR introduce a breaking change?

Shouldn't be, no.

[changeset-bot]

🦋 Changeset detected

Latest commit: 4fe11a8

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
preact-cli	Minor

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[rschristian]

One thing that came up is the question of whether the .env file path should be a flag or not. Can't say I have much of an opinion either way.

[ForsakenHarmony]

Suggested change

	let cwd = resolve(argv.cwd);
	require('dotenv').config({ path: resolve(cwd, '.env') });
	require('dotenv').config();

Wouldn't this work as well?
https://github.com/motdotla/dotenv#path

[rschristian]

In normal circumstances, sure.

However, that would break in our test suite (and probably mono-repos too): procces.cwd() (which dotenv uses) would differ from the project root.

I ran into that issue and moved the dotenv setup here accordingly.

Sorry, that could've done with a comment.

Edit: Added comments to both usages.

[ForsakenHarmony]

One thing that came up is the question of whether the .env file path should be a flag or not.

Could always add that later on

[rschristian]

Hm something in that watch test is badly misbehaving and getting the CI stuck in a loop. Will need to look into that tomorrow, though nothing stands out at the moment.