Add support for TypeScript class constructor in TypeScript mode

[briandipalma]

Describe the issue

From the TypeScript documentation: https://www.typescriptlang.org/docs/handbook/jsdoc-supported-types.html
We can use class constructors in params:

/**
 * A parameter can be a class constructor, using Google Closure syntax.
 *
 * @param {{new(...args: any[]): object}} C - The class to register
 */
function registerClass(C) {}

But this fails to parse when using the ESLint JSDoc plugin.

[simonseyock]

Thanks for the report! I will look into it.

[brettz9]

The construct signature with new keyword can apparently be used with arrow style too:

function greet(ctor: new () => Base) {
  const instance = new ctor();
  instance.printName();
}

https://www.typescriptlang.org/docs/handbook/2/classes.html#abstract-construct-signatures

FWIW, this is a high priority for us for a PR with the purpose of generating a declaration file for espree after some jsdoc manipulations (hopefully eventually leading to the same for ESLint).

[brettz9]

@simonseyock : any chance of being able to take a look at this in the near future?

[simonseyock]

@brettz9 I will look into this in the next days.

[jsdoc-type-pratt-parser-chores]

🎉 This issue has been resolved in version 3.1.0 🎉

The release is available on:

• GitHub release
• npm package (@latest dist-tag)

Your semantic-release bot 📦🚀

[brettz9]

I'm not seeing this item from the OP work in the tester: {new(...args: any[]): object}

[brettz9]

Am reopening as the specific case of an object with the new keyword indicating a constructor isn't yet resolved.

[brettz9]

@simonseyock : Any chance you might have some time to take a look?

[simonseyock]

the type expression would be new(...args: any[]): object not {new(...args: any[]): object}. The former does work in the tester, the latter does not make sense for me.

[simonseyock]

ok I might not have understood that correctly. I will see what I can do.

[simonseyock]

This might be complicated as this has to do with the whole problematic of needing a different parsing context inside object literals. This is in work, but will need some time.

[brettz9]

Much appreciated, thanks!

This feature should be particularly handy since without defining and pointing to an actual class, it is the only way to indicate the pattern of a constructor (and define the class' static properties between the same curly brackets). In cases like the PR I'm working with, one may have a factory which produces a class rather than having any publicly exported class that can be pointed to.

[brettz9]

@simonseyock : I know you said it'd need some time, but just wondering if you had any status update to share? Thanks!

[simonseyock]

I looked into this again today, but I did not find a solution that worked properly. The problem with the contexts is solvable, but no I hit some problems with the types of the results. The problem is that I use one result type for jsdoc and typescript syntax but it is at some points so different that it does not really make sense anymore. I am thinking about splitting the result types and to make the Parser generic. But I don't know if that works properly.
Sorry for being so inconsistent, but theses changes need a lot of work and it is hard for me to find that much time in short succession.

[brettz9]

No worries! Appreciate all the investigations and effort...