Algebra and unsafe about algebra

[WaDelma]

Other choice is to have same kind of thing as the Approx* variants, but that creates lots of bloat.

However it has the benefit of being more flexible: If the unsafe just needs the one algebraic property to be safe, it can be encoded.

[milibopp]

I really don't know… This does not actually provide any type safety, it's more like waving a red flag: “Hey, you might potentially mess up memory management by using this!”

Rust is missing dependent types for this to be actually useful. And in the current situation I'm not sure whether I would really use this for encoding semantics that are not inherently memory-unsafe. After all, would you really rely on such properties of an abstraction when implementing unsafe code? This isn't going to make it any safer.

[WaDelma]

Well the whole unsafe keyword is basically just waving red flag too so I don't see big difference with this vs it.

I do agree that dependent types would make it more useful, but still it's better than nothing?

And yeah I don't really have good use case for this yet.

[milibopp]

Sure, but you are mostly asserting the memory-safety of whatever you are doing inside. I find it somewhat confusing to use the keyword to convey other semantics, that can't be strictly enforced at compile-time. I guess, it's a trade-off?

But yeah, a use case would certainly be a nice example to drive this discussion. Comparing what it looks like and how the old behaviour might surprise a user, where adding an unsafe trait like that would help.

[brendanzab]

Hey @WaDelma, would you be interested in taking over this library? I don't currently have a great deal of time to spend on it, and you seem interested in pushing it forwards - which is great!

[WaDelma]

Well I would be interested if I had stronger knowledge on algebra... I really would like this to be thing that can be used, but I don't know if I can make it work.