(typescript) $class variable is incorrectly highlighted as a keyword

[1aron]

Describe the issue

Which language seems to have the issue?

Are you using highlight or highlightAuto? highlight

...

Sample Code to Reproduce

https://jsfiddle.net/fsLxdrpm/1/

const ElementPrototype = Element.prototype;
const isBool = (v: any) => typeof v === 'boolean';
const $class = (v: any) => typeof v === 'boolean';
const whether = isBool(true);
const test = $class(true);

Expected behavior

1. $class prefix with $ went wrong
2. method with class name went wrong
3. method isn't be highlighted
4. Element isn't be highlighted
5. isBool isn't be highlighted

[joshgoebel]

$class prefix with $ went wrong

This is a limitation of how we process keywords because technically the tiny area between $ and class counts as a word boundary. Is this type of naming common in Typescript?

To fix it we'd have to hardcode a rule to match all $[keyword] instances and ignore them. Or change it to \s vs \b but I bet there are cases that would break such as a keyword following a bracket or something like that, no?

method isn't be highlighted

What method? Also see my comment on isBool.

Element isn't be highlighted

We don't have a completely list of types in our grammar... if you have a suggestion of which ones we should add we could discuss it and see if we can find an agreement - but I'm not sure we should "add them all" because there are a ton of them. For example TypeScript's DOM type definition has 100s of types: https://github.com/Microsoft/TypeScript/blob/master/lib/lib.dom.d.ts

So we'd need some way to decide on the "major" ones...

isBool isn't be highlighted

It wouldn't be, we don't highlight method invocation.

[joshgoebel]

@egor-rogov Actually I wonder if we just extended lexemes to add $ to it (even though it's not used in any keywords)... so now when looking for keywords it matches WHOLE identifiers (perhaps that's the best behavior anyways)... so it would see $class instead of class and go "nope, not a keyword".

See any downsides?

[joshgoebel]

@allejo

[egor-rogov]

@egor-rogov Actually I wonder if we just extended lexemes to add $ to it (even though it's not used in any keywords)... so now when looking for keywords it matches WHOLE identifiers (perhaps that's the best behavior anyways)... so it would see $class instead of class and go "nope, not a keyword".

Yeah, I agree.
My first thought was about Perl (which uses $ for scalars and we have a separate rule for this), but no. In Perl $ has its own meaning, while here it's just a part of identifier (right?).

[joshgoebel]

Yes, in JS/TS $ and _ are just characters like any other. Hence jQuery $ and _ (underscore) libs.

[allejo]

@egor-rogov Actually I wonder if we just extended lexemes to add $ to it (even though it's not used in any keywords)... so now when looking for keywords it matches WHOLE identifiers (perhaps that's the best behavior anyways)... so it would see $class instead of class and go "nope, not a keyword".

I don't see any downsides to this since $ isn't reserved. Just conventions in libraries like jQuery and Angular (they use $$).

[joshgoebel]

Fix will go out in 10.1.