Small string optimization for Values

[Radvendii]

Motivation

See this tracking issue for more big-picture motivation.

This change in particular stores small strings that have no context directly in the value payload itself. This reduces memory usage by the string length, and also removes a layer of indirection.

Context

• tracking issue for more big-picture motivation.
• @xokdvium already attempted this work, with the conclusion that it doesn't help that much and therefore is not worth a type tag.

---

Add 👍 to pull requests you find important.

The Nix maintainer team uses a GitHub project board to schedule and track reviews.

[Radvendii]

I wanted to open this issue so that i can keep track of the work somewhere, but I discussed with @xokdvium and at the moment this doesn't bring any benefit, so I'll shelve it once CI runs

[Ericson2314]

I wanted to open this issue so that i can keep track of the work somewhere, but I discussed with @xokdvium and at the moment this doesn't bring any benefit, so I'll shelve it once CI runs

OK sounds good. I was going to point out that if --- per the issue --- we're shrinking Value to 4 bytes, then this buys us even less / will stop working.

[Radvendii]

There's some gcc-specific error (i.e. not happening with clang) causing CI to fail. Given that I'm shelving this for now anyways, It's not worth it to me to fix it. If it turns out to be useful in the future I'll fix it then.

[Radvendii]

OK sounds good. I was going to point out that if --- per the issue --- we're shrinking Value to 4 bytes, then this buys us even less / will stop working.

Values payloads will be shrinking to 8 bytes, so it's true we lose a lot of benefit, but it's not as useless as 4 bytes would be.

[Ericson2314]

Oh yeah, the payload. Right. :)