Support aliases.

[MikeMcQuaid]

Also some other general cleanup and refactoring and ensuring that tests have 100% coverage.

CC @xu-cheng

[xu-cheng]

I wonder do we need to cache this? Considering, it will only be run once for all of brew bundle subcommands. Beside by removing the cache, we can remove formulae_info_reset! at the same time.

If you want to keep it, you want to change @@formulae_info to @formulae_info

[MikeMcQuaid]

It's a class-level cache so it can be reused across instances.

[xu-cheng]

Actually I mean it make more sense to cache the result after it is sorted based on dependency order.

In the core, we prefer to use singleton object variable over class variable. Although both can be used to share data among objects.

Since all of these class are only used once. I wonder should we turn all of them to singleton class? This may help to simplify the code

[MikeMcQuaid]

We're using class variables elsewhere so I thought it was better to just be consistent here. Caching after sorted makes sense though 👍

[xu-cheng]

You need to update Cleanup::formulae_to_uninstall to support alias.

Another problem is that only using name to check formula is not enough. We have to handle the case when full name is used in Brewfile like foo/bar/baz or foo/bar/baz-alias but current installation is from another tap or core.

[MikeMcQuaid]

You need to update Cleanup::formulae_to_uninstall to support alias.

👍

Another problem is that only using name to check formula is not enough. We have to handle the case when full name is used in Brewfile like foo/bar/baz or foo/bar/baz-alias but current installation is from another tap or core.

Will test that case.

[MikeMcQuaid]

It's 💚. @xu-cheng let me know what you think.

[xu-cheng]

You need to update Cleanup::formulae_to_uninstall to support alias.

You seem to miss that 😉

I don't think we need to handle that case for now unless it's specifically requested.

We still need to somehow support old name. Otherwise, check and cleanup will be broken to users whenever we rename formulae.

[MikeMcQuaid]

You seem to miss that

Whoops, yep, I'll get it.

We still need to somehow support old name. Otherwise, check and cleanup will be broken to users whenever we rename formulae.

That's fine. I don't want to expose old_name as a public API. In those cases users will need to rename. What do you propose we do.