Feature Request `PartialEq` for different `Buffer` types

[Daniel-Aaron-Bloom]

Add PartialEq to Literal and all the inner types across different buffers where B1: PartialEq<B2>.

This will enable Literal<String> to be compared to Literal<&str>, which is quite nice.

I'd be happy to send a PR for this if you're interested.

[LukasKalbertodt]

That's a very reasonable request, but unfortunately annoying to implement.

Of course, it's straight forward to hand-roll the PartialEq impls. But that I don't want: apart from adding lots of boilerplate code, it's really easy to forget adjusting those in the future, leading to subtle bugs. One could use a crate that has the derive with more flexibility (like derivative), but that would add a proc-macro as a dependency for this crate, which is also a no-go. I tried a few things to maybe get it working with the built-in derive(PartialEq), but had no luck. I then realized that all fields of all types are always purely derived from raw_input, so it should be sufficient to compare raw_input. Then I noticed that my tests use the PartialEq impl to detect failures, in particular when said assumption doesn't hold. So I would need to switch back to derived PartialEq. Additionally, the assumption might be true now but if something changes about that, it's easy to forget adjusting it.

If you have any good idea, I'd be open to hear it. But assuming there is no awesome way to do this, I probably have to decline to keep maintenance reasonable.

As a workaround, this works: a.into_owned() == b. Of course, that clones the shared buffer.

[Daniel-Aaron-Bloom]

Absolutely fair. A huge appeal of this crate (at least to me) is the extremely limited dependency chain.

Long term this should be doable via declarative macro_rules! attributes (or the derive equivalent when that hopefully becomes a thing). Short term, you can do this with today's macro_rules, it's just somewhat ugly.

As an alternative to into_owned, it should be possible to add an as_ref (can't actually use Borrow or AsRef since they both require &T), which should avoid the shared buffer clone.

[Daniel-Aaron-Bloom]

I've made #20 for this last point.

[LukasKalbertodt]

With 3c329a3, all literal types now have a raw_input method. As I said above: comparing the raw input is equivalent to comparing all struct fields since all others are just purely derived from the raw input. So this is an easy workaround for the limiting PartialEq impl, that does not incur allocation cost.

I might still relax the PartialEq impl in the future (either by using macro_rules! or implement it manually using raw_input for #[cfg(not(test))]). But I will close this issue for now.