Fix Typescript type inference for returning() to better support wildcard ("*") calls

[CodyEakins]

Environment

Knex version: 0.19.3
Database + version: Postgres 11
OS: Windows 10, WSL

@lorefnon - Hopefully you can address this... Or at least tell me if I am wrong.

Bug

When using .returning() in a Query Builder chain, it's possible to pass in a wildcard ("*") to indicate that you want all of a table's columns returned with the result.

For example, when using Typescript Generics:

interface User {
  id: number;
  email: string;
  password: string;
}

const queryResult = await knex<User>("user").insert({
  email: "name@example.com",
  password: "passwordHash"
}).returning("*");

One expects queryResult's type to be User[] here, but it is actually Partial<User>[].

This result makes sense when calling returning() with a column identifier (e.g. - returning("id")), but not when it is called with a wildcard. When called with a wildcard, the result set is an array of fully qualified User objects, not partial ones.

---

Note:

I know you can pass the User type to returning() more directly, like so:

.returning<User>("*")

And get the expected result... But this bug is more about the type inference from the main Query Builder to returning().

---

Suggested Fix:

Maybe update the definition to something like:

returning<TResult2 = SafePartial<TRecord>[] | TRecord[]>(
  column: string | string[]
): QueryBuilder<TRecord, TResult2>;

[KristjanTammekivi]

Mildly OT but maybe enable a zero-parameter .returning() which automatically means return all columns as a possible fix for this

[lorefnon]

#3446 adds overrides that special case '*' so that the returned value will be the complete record.

Maybe update the definition to something like:

returning<TResult2 = SafePartial[] | TRecord[]>(
column: string | string[]
): QueryBuilder<TRecord, TResult2>;

This is unlikely to get you what you want because T extends Partial<T> and hence T | Partial<T> is effectively Partial<T>.

[lorefnon]

I also noticed that travis no longer run lint:types. Was this an intentional change ? cc. @kibertoad

[kibertoad]

@lorefnon Huh, I wonder how did that happen. Opened #3450 to address that.

[CodyEakins]

Thanks for the fast and thorough response to this.

[kibertoad]

Released in 0.19.5