Expand the json output of mint lint to have more metadata

[bbugh]

I was looking at what it would take to have inline warnings in VSCode, and testing with master mint to use lint --json.

The json that the linter outputs an array of strings with the error messages. In order to effectively use the json output, it would be helpful for it to output an array of objects that include metadata like:

• filename
• line number
• column from
• column to
• basic message (like I was looking for a branch of a case expression but found "Result::Ok" instead.)
• full message

It could be parsed out from the error message, but that would be brittle, and the error message does not include the "end" column number, so any errors in-editor would have to be drawn for the whole line.

[bbugh]

I took a look at the code, and this is my best guess for the changes.

It appears that the parser would have to send more data when the error occurs:

https://github.com/mint-lang/mint/blob/master/src/parser.cr#L64-L67

And the json output would have to be changed, either by overriding the base class (which appears to be Crystal Exception rather than the Mint::Error, so I'm not sure how to do that) or by adding a block:

https://github.com/mint-lang/mint/blob/master/src/commands/lint.cr#L71

But from there I'm not sure how to get that data into Message so that it can be output in the ^^^ lint code.

[Sija]

With #443 it should be relatively easy, having Ast::Node#location.

[gdotdesign]

This would be a good change for the lint command and it's doable but I wonder if it would be better to implement Diagnostics for the language server instead?

[Sija]

@gdotdesign I think we need both, Diagnostics for LS, and a similar feature for the lint command - mostly useful within the CI env.