Install formulae from JSON files

[Rylan12]

• Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't other open Pull Requests for the same change?
• Have you added an explanation of what your changes do and why you'd like us to include them?
• Have you written new tests for your changes? Here's an example.
• Have you successfully run brew style with your changes locally?
• Have you successfully run brew typecheck with your changes locally?
• Have you successfully run brew tests with your changes locally?

---

This PR incorporates the JSON functionality from homebrew/json into this repo. Now, if HOMEBREW_JSON_CORE is set, you can untap homebrew/core and install/reinstall/upgrade formulae using the default install, reinstall, and upgrade commands.

Additionally, brew outdated is now setup to work using the new versions API proposed in Homebrew/formulae.brew.sh#497 (so this PR should not be merged until the formulae.brew.sh one is merged). This will also mean that brew update will show the appropriate message if a formula is out of date, even without the homebrew/core tap installed.

[BrewTestBot]

Review period will end on 2021-07-06 at 00:00:00 UTC.

[BrewTestBot]

Review period will end on 2021-07-06 at 00:00:00 UTC.

[dawidd6]

Could you explain the reasoning for this? Are we going to switch from git based taps to REST API ones? What about third-party taps?

[Rylan12]

The idea is that users shouldn't need to have homebrew/core (and probably homebrew/cask, eventually) tapped in order to install a formula (or cask). These repos are huge and annoying to update, so it slows down brew update a lot to have them tapped. Instead, we can use the /api/bottle/{FORMULA}.json API from https://formulae.brew.sh to download formula bottles and use the embedded formula files to install the formula.

Third party taps would still need to be installed locally, and this wouldn't remove the need for HOMEBREW_DEVELOPERs and devcmdrun users to have a full clone, but for the average user it's designed to reduce Homebrew's disk usage and speed up Homebrew updates.

I don't know about the future, I don't think we're planning on totally scrapping git-based taps any time soon (although @MikeMcQuaid may have a different vision). I could see us creating a way for third-party taps to use this API-based install process but for the moment, its designed just for homebrew/core.

[MikeMcQuaid]

I don't know about the future, I don't think we're planning on totally scrapping git-based taps any time soon (although @MikeMcQuaid may have a different vision).

No, I don't have any plans personally to ever do this. As you've mentioned, this approach is pretty much just for homebrew-core and homebrew-cask (and would be for linuxbrew-core were we not so close to the migration) because:

1. they are expensive to git fetch and git clone and GitHub would rather we didn't do that
2. they are slow to git fetch and git clone and this provides a bad experience to end users
3. we auto-run brew update in many cases and the (by far) slowest part of this is git fetch on homebrew-core and homebrew-cask

[MikeMcQuaid]

Looking great! A few comments.

[BrewTestBot]

Review period ended.

[MikeMcQuaid]

Some suggested/optional tweaks (that may not be needed if I've misunderstood). Merge when you're happy! Great work as usual.

[Rylan12]

My latest push adds tests for the BottleAPI class and changes the prioritization so that installing from a JSON file is preferred even if homebrew/core is tapped (as long as HOMEBREW_JSON_CORE is set).

To-do items for future PRs:

• Don't have brew update update homebrew/core if the env variable is set and it's auto-run
• Have brew info show the latest version info instead of whatever is tapped locally
• Don't have brew install, brew reinstall, and brew upgrade download bottles for every dependency unless the dependency is already installed (and at the latest version)
• Create command to un-dev-cmd yourself