(C#) Keyword before tuple return type is highlighted as a method name

[Youssef1313]

Describe the issue

See https://docs.microsoft.com/en-us/dotnet/csharp/discards#tuple-and-object-deconstruction for example.

Which language seems to have the issue?

C#

Sample Code to Reproduce

private static (int, int) GetOriginCoordinates()
{
    return (0, 0);
}

Expected behavior

static should be highlighted as a keyword (i.e. hljs-keyword), not as a function name (i.e. hljs-title)

To avoid merge conflicts, this should wait until #2679 is merged.

[IEvangelist]

I'm assuming that the title mechanism is based on a RegExp that scan for the first occurrence on a line with the opening parenthesis (. This would cause the return tuple types to cause any keyword before to be treated as a hljs-title, not just static.

[joshgoebel]

I'd guess TYPE_IDENT_RE isn't wide enough to know what a tuple type is. PR would be welcome. :)

[joshgoebel]

Can tuples include other advanced types also such as Vector<string>[] or whatever? (I don't know c# types)

[Youssef1313]

@joshgoebel Here are some different variations (some may be rarely used, but still valid):

using System.Collections.Generic;

public class C
{
    private static (int, int) GetOriginCoordinates()
    {
        return (0, 0);
    }

    private static (int X, int Y) GetOriginCoordinates2()
    {
        return (0, 0);
    }
    
    private static Dictionary<string, (int, int)> GetCoordinates()
    {
        return new Dictionary<string, (int, int)>
        {
            { "Origin", (0, 0) },
            { "One_One", (1, 1) }
        };
    }
    
    private static Dictionary<string, (int X, int Y)> GetCoordinates2()
    {
        return new Dictionary<string, (int X, int Y)>
        {
            { "Origin", (0, 0) },
            { "One_One", (1, 1) }
        };
    }
    
    private static (int DivisibleBy, List<int> Numbers) DivisiblesByTwo()
    {
        return (2, new List<int> { 2, 4, 6, 8, 10 });
    }
}

[joshgoebel]

@joshgoebel Can a tuple include a tuple?

[Youssef1313]

@joshgoebel Yes they can be nested to multiple levels.

[joshgoebel]

e8eef17#diff-8b07d3d641e06fd8cbdabdf8115c0ddeR121

Any other leading keywords I'm not thinking of we should add?

[Youssef1313]

internal and protected.

Pinging @IEvangelist to confirm if I'm not missing any others.

Also this change is for C# not Java

[IEvangelist]

Hi @Youssef1313 and @joshgoebel

How would async be treated? Is that something that is separate? It too would be a keyword that is valid before the return type is defined in a method signature in C#. In fact there are lots of modifiers that could appear before the return type:

• abstract
• async
• extern
• override
• unsafe
• virtual

Are these what you're after? See C# keywords, and look at the Modifiers section. Not all of these are available prior to a return type, but a lot are. This is making me think that we should add more C# tests, with more variations of the syntax. We should also expand the C# example in the demo to include more code.

[joshgoebel]

In fact there are lots of modifiers that could appear before the return type:

Why wouldn't we want to add the full list (of the ones available) - so they would be detected as keywords and never as the accidental name of a function?

[joshgoebel]

@IEvangelist Why would "async" be special?

Again, the context here is "preventing keywords from being falsley detected as name of function"... one easy way to do that (given a finite set) is to just exclude them all and tell it they are modifiers...

[IEvangelist]

I guess I was confused by how this all comes together, I'm not really sure how it works internally - I was more or less stating that there are additional keywords that come before return types in C#. All of these keywords are already captured in my previously merged PR, so why is this change necessary?

[joshgoebel]

Because #context.

        v ----- this looks TOO much like a function named `unsafe`
private unsafe (int, int) GetOriginCoordinates()

So we need a safe list of modifier words to prevent them from being wrongly detected as function names. Then the tuple is wrongly/conveniently detected as "params" and just works. The alternative is someone writing an amazingly complicated "type regex" to detect all possible type declarations, and I'm not even sure that's possible.

So just have a safe list of modifiers seems to be a simple solution.

[IEvangelist]

Ah, ok. Thank you for the added context, that is really helpful. I agree. Then the safe list of modifiers would be something like:

• abstract
• async
• extern
• override
• unsafe
• virtual
• private
• protected
• internal
• public
• static

What do you think @Youssef1313?

[Youssef1313]

Ah, ok. Thank you for the added context, that is really helpful. I agree. Then the safe list of modifiers would be something like:

* `abstract`

* `async`

* `extern`

* `override`

* `unsafe`

* `virtual`

* `private`

* `protected`

* `internal`

* `public`

* `static`

What do you think @Youssef1313?

That list looks great and adds many many that dropped of my mind. Probably add new (when used for shadowing base class method), sealed, and partial to this list too.