Issues with type annotations when type includes a method, which includes a string

[a-stewart]

Consider the following code:

The type annotations include a callable type which can have string arguments which fail - but then if the string contains either > or ) that ends the type annotation early, resulting in the closing string tag being mistakenly recognized as opening a string.

Note that, I expect that it is not just strings that are not considered here, but other primitives too. I am not familiar with Kotlin though to be fully certain on what is allowed in the language here though.

[fwcd]

I am assuming this issue also occurs while the language server is starting/disabled? In that case, it's probably an issue with the grammar.

[a-stewart]

Hi,

Sorry, yes this is an issue with the grammar, not with the language server.

In particular:

        "type-annotation": {
            "match": "(?<![:?]):\\s*(\\w|\\?|\\s|->|(?<GROUP>[<(]([^<>()]|\\g<GROUP>)+[)>]))+",

The problem is that it does not like it if we include a > or ) in the grammar since that essentially ends where it thinks the type-parameters goes up to. I'm currently playing with the grammar a bit to see if I can find a simple solution to match more generously - but the challenge is trying to find something that should trigger the end of a type annotation.

[fwcd]

Sure no problem, since this repo hosts the canonical version of the grammar, opening the issue here is totally fine! There's even a tiny VSCode extension for tinkering with the grammar here.

[a-stewart]

I have created a PR for a small patch which slightly improves things by not including parameters if there is a quote.

This won't help with the issue fully though, as I think we can also confuse it by introducing parenthesis around numbers.

I think the best solution is likely to involve updating type-annotation to use begin and end - but I am not sure how best to support both generics and parameters in that way.

[themkat]

@a-stewart , did your PR solve the issue completely? Or are there still something missing? Just asking in case we can close this 🙂

[a-stewart]

The PR doesn't solve the issue completely - I think that type annotations which include a string will not be tokenised correctly.