Switch from codespan-reporting to a custom crate based on codespan

[Stupremee]

This PR replaces codespan-reporting with annotate-snippets.
To actually use annotate-snippets in a nice manner and to make the switch easier,
a new module, rslint_errors, is introduced to provide an API similair to codespan-reporting
but build on top of annotate-snippets.

Resolves #29

[RDambrosio016]

A diagnostic struct with a lifetime is going to cause a myriad of issues, i pretty much rewrote half of the old linter because of it, so i would rather not do that

[RDambrosio016]

oops

[Stupremee]

What is the sense of the Applicability enum in your code in #29?

[RDambrosio016]

Its from rustc_errors https://doc.rust-lang.org/nightly/nightly-rustc/rustc_errors/enum.Applicability.html

[Stupremee]

Do we really need the Applicability enum? I would not add it to RSLint because it doesn't seem to affect the error reporting

[RDambrosio016]

It does not affect error reporting but it does affect the LSP, since it needs to know what suggestions it can offer which dont have placeholders

[Stupremee]

Should we implement a similar suggestion API like rustc_errors? So you can "fix" the error by substituting some strings that are provided by the suggestion? I think this can be very nice in a linter. The parser can then try to fix many issues, that are obviously typos, by providing a suggestion, which will get automatically applied by auto-fix.

[RDambrosio016]

Yes the LSP can provide autofix options based on the suggestions from the diagnostic, although im not sure how we are going to do this. Id like to have a fixer for rules, but then the parser cant use it. I think we will keep the fixer for rules but have other suggestions to.

[RDambrosio016]

Eventually we will have to separate the AST and syntax node stuff into its own module, that way rslint_errors can provide a span for ast nodes and syntax nodes. That should allow us to make the fixer a standalone crate too, which means we can offer the fixer interface through rslint_errors for suggestions

[Stupremee]

So, I added all the basic structs and representations that now just need methods and logic to report.
I'm currently not sure if we should stay with SmolStr everywhere, or if we should use String.
I have no idea how much of a speed boost / memory saving SmolStr would give us.

[RDambrosio016]

SmolStr wouldnt do much, most messages are > 22 chars

[Stupremee]

I think rslint_errors is ready to replace codespan now, except for suggestions, I have to fix some lifetime issues first, and I have to provide an immutable Diagnostic (or public methods to read the data) so that the LSP and other stuff can actually apply the suggestions. But other than that it should be ready.

[RDambrosio016]

I dont really see why the diagnostic's data should not be public

[Stupremee]

Yea that works too

[Stupremee]

annotate-snippets uses lifetimes everywhere. That's sooo annoying. It blocks everything I want to do 😢

[RDambrosio016]

In what places does it cause issues?

[Stupremee]

I have a method convert which will convert a Diagnostic to a Snippet, but all the strings used in Snippet have to be a &str and thus a lifetime is required, so i can't create new strings in the convert method and use them in a Snippet. But I think a found a fix.

[RDambrosio016]

The changes seem to work well, some of the rendering is wonky but we can fix that later, i think we can merge it now

[RDambrosio016]

I take that back, it seems we are suffering from the same issue as denoland/deno_lint#391 which is a little odd

[Stupremee]

The issues are super weird

Edit: I know the cause for both of them, I will fix them in our local copy.

[RDambrosio016]

Are you sure? it still seems to be a bit broke 🤔

error[SyntaxError]: expected `'('` but instead found `length`
  --> tests\main.js:20:7
   |
...
19 |
20 |   for length;
   |       ^^^^^^ unexpected
   |
error[SyntaxError]: Expected an expression, but found none
  --> tests\main.js:26:5
   |
...
25 |   // when jumpSize = √array.length.
26 |   const jumpSize = Math.floor(Math.sqrt(arraySize));
   |     ^^^^^ Expected an expression here
   |

[RDambrosio016]

Switching from annotate-snippets to codespan has solved all of the issues, essentially the rendering is the same as using codespan on its own but we have a cool new unified builder as well as suggestions, and with all of the nice codespan rendering 🎉