@odow This looks like it's on the right track. https://github.com/julia-vscode/CSTParser.jl/blob/master/src/spec.jl is the best reference for node types. The only new types introduced in the formatter are PLeaf types.

You shouldn't be editing the CST from CSTParser, it's just used to create the PTree. We probably also want to add an additional check besides just val[end] == '.'. I'm not sure if this occurs in other cases besides floats without a trailing zero.

This should work:

        val = x.val
        if x.kind === Tokens.FLOAT && val[end] == '.'
            val *= '0'
        end
        return PTree(x, loc[1], loc[1], val)

Token types are found in https://github.com/JuliaLang/Tokenize.jl/blob/master/src/token_kinds.jl

BTW Does the trailing zero come up at all with complex numbers or would this already handle that?

LGTM. Let me know if it's ready to merge in or if you're still making changes.

Thanks for you help. I was missing the fact that the x.val argument to the PTree didn't have to be x.val.

I believe Tokens.FLOAT is the only case where there can be a trailing .. I've added tests with complex number examples.

What's the policy on running format("src") and format("test") on a PR? Because there seemed to be a lot of indentation changes, but I didn't include them because they seemed unrelated.

No problem. I need to add more documentation to this thing so it's kind of my fault 🙃 .

Generally I do format("src") and format("test/runtests.jl").

I guess one hack would be to add

format(joinpath(dirname(@__DIR__, "src")))
format(@__FILE__)

to the bottom of test/runtests.jl so that anyone running the tests also auto-formats things.

that or GH support can get back to me so I can get this to work https://github.com/domluna/JuliaFormatter-action 😄

Anyway I'm gonna merge this in, thanks again!