Autocomplete at @

[phil294]

cursor after @ does not offer context autocomplete, only this. does.

This cannot be easily solved with a simple replaceAll because it can also mean this without the dot:

a = @

is

var a;
a = this;

[isral]

Can we ignore that case (just type this if you want @)? Probably most @ usage is this. so it's acceptable tradeoff.

[phil294]

ignore that case

This is problematic because it means disallowing the syntax altogether, not just missing type hints. There is only a single one-fits-all transformation step which has no knowledge of the kind of incoming LSP request or cursor position. So replacing all occurences of @ with this. is not okay, it would introduce syntax errors where it is used as plain this without dot.

I just tried another workaround: Replacing @ with this.valueOf() and map cursor position | from @| to this.|valueOf(). This would work, but unfortunately, tsserver does not agree:

//@ts-check
let a = {
    b: 1,
    c() {
        return this.valueOf().b // Syntax error: Property 'b' does not exist on type 'Object'.ts(2339)

  }
}

when it does work - evaluating a.c() results in 1 as expected.

So right now the only solution that I can think of is fixing the extension's architecture but personally, I don't plan on doing that any time soon.

/ come to think about it, maybe we can overwrite typing of .valueOf()? :S

Edit: There was an error in my thinking, tsserver's opinion on .b does not matter, as the emulated cursor is placed before valueOf, not before b. In other words, this .valueOf() approach is possible and I'll push it in a few minutes

[phil294]

ok nevermind what I wrote above, I released v1.1.0 with which everything should work just as expected, even when CS compilation is not possible

If you happen to work with it again @isral, could you please give me feedback on whether it works for you?

[STRd6]

A better replacement may be (this._, this). I haven't thought through all the situations but it seems like it may work to get the dot in the right place for autocompletes while still returning the correct type/value.

[phil294]

A better replacement may be (this._, this). I haven't thought through all the situations but it seems like it may work to get the dot in the right place for autocompletes while still returning the correct type/value.

oh that's clever!