If FoundationEssentials is available use it

[adam-fowler]

Use FoundationEssentials if it is available

Motivation:

Reduce size of any application using swift-metrics on linux

[Lukasa]

As noted in other repos, this is a semver major change, so we’ll need to work out how to stage this in.

[adam-fowler]

As noted in other repos, this is a semver major change, so we’ll need to work out how to stage this in.

Is this a breaking change?

It was a breaking change in swift-nio-transport-services because the import NIOFoundationCompat was removed, but removing an import Foundation doesn't make Foundation unavailable as an import in packages dependent on swift-metrics.

[Lukasa]

Yes, it's a breaking change. Reproducing my comment from swift-crypto:

It doesn’t make the full API available, but it makes some things available. An example of one such problem is available at https://forums.swift.org/t/pitch-fixing-member-import-visibility/71432.

The Swift team consider this a bug, and I agree, but our semver policies take the language as it is, not as it is intended to be. So this will require us to burn a major.

[adam-fowler]

pfff

[t089]

Yeah :( I wonder if we could get away with it, though. Today, most apps (knowingly or unknowingly) anyhow somehow import Foundation somewhere, so unless you are actively trying to avoid it, you are anyhow already importing it. And if you are actively trying to avoid it, you do not use API from it, so this change will not be source breaking.

[Lukasa]

And if you are actively trying to avoid, you do not use API from it, so this change will not be source breaking.

There's a certain "if a tree falls in the forest and no-one is around to hear it" quality to this argument. It's true that whether a change is source-breaking is a function of whether anyone is actually using the API. The wrinkle is that it's very hard for us to validate that they aren't. My bias is strongly against taking that risk, usually.

In the case of metrics in particular, there are good reasons to think that we may want to modernise the API. We could do these two things at once, put the more modern API in a new module and deprecate the old interface. This will cause a natural community migration away from the deprecated API and toward a new one, while avoiding a source break.