Provide flattened.error.reporting(...).for(...).nested.incompatible.types

[DanielRosenwasser]

let x = { a: { b: { c: { d: { e: { f() { return { g: "hello" }; } } } } } } };
let y = { a: { b: { c: { d: { e: { f() { return { g: 12345 }; } } } } } } };
x = y;

Existing Behavior

Type '{ a: { b: { c: { d: { e: { f(): { g: number; }; }; }; }; }; }; }' is not assignable to type '{ a: { b: { c: { d: { e: { f(): { g: string; }; }; }; }; }; }; }'.
  Types of property 'a' are incompatible.
    Type '{ b: { c: { d: { e: { f(): { g: number; }; }; }; }; }; }' is not assignable to type '{ b: { c: { d: { e: { f(): { g: string; }; }; }; }; }; }'.
      Types of property 'b' are incompatible.
        Type '{ c: { d: { e: { f(): { g: number; }; }; }; }; }' is not assignable to type '{ c: { d: { e: { f(): { g: string; }; }; }; }; }'.
          Types of property 'c' are incompatible.
            Type '{ d: { e: { f(): { g: number; }; }; }; }' is not assignable to type '{ d: { e: { f(): { g: string; }; }; }; }'.
              Types of property 'd' are incompatible.
                Type '{ e: { f(): { g: number; }; }; }' is not assignable to type '{ e: { f(): { g: string; }; }; }'.
                  Types of property 'e' are incompatible.
                    Type '{ f(): { g: number; }; }' is not assignable to type '{ f(): { g: string; }; }'.
                      Types of property 'f' are incompatible.
                        Type '() => { g: number; }' is not assignable to type '() => { g: string; }'.
                          Type '{ g: number; }' is not assignable to type '{ g: string; }'.
                            Types of property 'g' are incompatible.
                              Type 'number' is not assignable to type 'string'.

Proposed

Type '{ a: { b: { c: { d: { e: { f(): { g: number; }; }; }; }; }; }; }' is not assignable to type '{ a: { b: { c: { d: { e: { f(): { g: string; }; }; }; }; }; }; }'.
    'a.b.c.d.e.f(...).g' is incompatible between these types because 'number' is not assignable to type 'string'

[orta]

I explored this a bit with Eloy Durán (@alloy) when thinking about whether we could also use bold/italics to also provide emphasis, instead we used { ..., b: { ... c: } } to provide focus while keeping the pyramid of nested errors.

I prefer the version in OP. I wonder if turning it into three lines would make it more readable:

Type '{ a: { b: { c: { d: { e: { f(): { g: number; }; }; }; }; }; }; }' 
is not assignable to type '{ a: { b: { c: { d: { e: { f(): { g: string; }; }; }; }; }; }; }'.

        'a.b.c.d.e.f(...).g' is incompatible between these types because 'number' is not assignable to type 'string'

Then you get to see the full message without scrolling, and can more easily compare the types.

Ideally, if we're in an environment where we can use bold:

Type '{ a: { b: { c: { d: { e: { f(): { **g: number**; }; }; }; }; }; }; }' 
is not assignable to type '{ a: { b: { c: { d: { e: { f(): { **g: string**; }; }; }; }; }; }; }'.

        'a.b.c.d.e.f(...).g' is incompatible between these types because 'number' is not assignable to type 'string'

Which would really help lead someone's eyes to the right spots to make their own decisions.

[dhruvrajvanshi]

I'd like to try this.

[DanielRosenwasser]

Dhruv Rajvanshi (@dhruvrajvanshi) Go for it! It might be a little bit tricky because this is deep in the logic for checkTypeRelatedTo.

To scope the problem space, I'll call the special diagnostic

'a.b.c.d.e.f(...).g' is incompatible between these types because 'number' is not assignable to type 'string'

a member chain diagnostic.

A member chain diagnostic is created by tracking checks between properties and call signatures, and determining if multiple relevant properties/call signatures are encountered consecutively. A member is relevant for these checks if the member is

• a property that is a valid IdentifierNames (i.e. not Symbol.iterator or "hello there")
• a non-generic call signature, where each containing type has only one call signature (so signatures from { (): number; (x: any): string } and () => boolean shouldn't be included)

---

I like that Orta Therox (@orta). Maybe that can be layered on top for the first message, even if we can't do styling on diagnostics yet. I'd open up a separate issue if you're interested!

[DanielRosenwasser]

(Please forgive me for my poorly written spec 😅)

[DanielRosenwasser]

By the way, we'd like to get this in for the beta if possible. Do you think you can get this done by the end of the month Dhruv Rajvanshi (@dhruvrajvanshi)?

[dhruvrajvanshi]

Daniel Rosenwasser (@DanielRosenwasser) In that case, you might want to assign this to someone else. I can't really commit to a deadline right now.

[austincummings]

Daniel Rosenwasser (@DanielRosenwasser) I think I could commit to getting this done by then. Is there someone/somewhere in particular I should ask for guidance on this? Or just here on the issue?