[TS] Fix generation of struct members in object api

[tira-misu]

If a struct is used in a struct TS generates wrong code.
Example:

struct A {
testBool: bool;
}

struct B {
teststruct: A;
}

Furthermore:

• null value for type bool was wrong ("false" instead of "0")
• fixed reserved word argument -> arguments
• field accessor function must be unescaped

[CasperN]

Hello, I'm not the TS maintainer but I am working on standardizing our code generation and would like you opinion on #7142. I'm curious why "field accessor function must be unescaped" does TS/JS first resolve fields before trying to find keywords?

I noticed you only changed idl_gen_ts. I think you should make a test schema that demonstrates your changes. I think adding ts generation to keyword_test.fbs should do it.

Also, pull again - I managed to break master when adding a printf and I fixed that.

[tira-misu]

About reserved word as functions in TS. Yes - its a bit strange, but indeed reserved words are allowed as function/method names. In other languages that's not the case. So TS seams to be a special snowflake here. But in my opinion its more convenient for a user of TS to not escape these method names.

I will add a test case to the fbs file and try to run the test.

[CasperN]

@krojew can we get a review?

[krojew]

This seems quite odd - you first allow a to be undefined (?), but then you assert the value is present (!). Looks contradictory. The same pattern repeats for other fields in this PR.

[tira-misu]

Of cause this looks a bit strange. We could simplify this line from (this.a === null ? 0 : this.a?.id!) to (this.a?.id!). It is equivalent and should compile.

[krojew]

This would still be wrong - the result of the ?. operator can be undefined, so you can't later assert it's not by using !.

[tira-misu]

What would you suggest?
Would this solution be ok? :

• struct of struct: (this.a === null ? 0 : this.a.id!)
• struct of struct of struct: (this.a?.b === null ? 0 : this.a.b.id!) <- i think this will not compile

[krojew]

Why do you need ?. in the first place? There is already a null check present.

[krojew]

Ok, I see what the problem might be for nested structs, but the proposed code doesn't solve it. For example, this.a?.a === null will never be true if this.a is null, thus defeating the check. Instead, something like this should work: this.a?.a ?? 0.

[tira-misu]

You are right. The solution is:

• struct of struct: this.a?.id ?? <nullvalue> (wrong: this.a?.id && <nullvalue>)
• struct of struct of struct: this.a?.b?.id ?? <nullvalue> (wrong: this.a?.b?.id && <nullvalue>=

is false for type bool and 0 for all other simple data types.
I will try it out. If it works i will alter this pull request. I will also write a test case for this.

[krojew]

No, this is not what I proposed - using && instead of ?? will not work.

[tira-misu]

Of cause - a typo at my side. Thanks

[dbaileychess]

Are we all good to go on this?

[tira-misu]

From my point of view yes.

[dbaileychess]

Thanks as always!