Upsert unique constraint column names are not normalized

[danielrearden]

The column names passed in through the uniqueConstraintColumnNames parameter should be normalized just like the named value bindings are. In other words, if I submit { emailAddress: '[email protected]' } as the value bindings, I should be able to submit ['emailAddress'] as the unique constraint column. Right now, I have to normalize the name myself (i.e. ['email_address']).

[gajus]

The logic for handling normalization of keys is because ESLint may enforce id-match rule.

The same reason does not apply for unique constraints.

In some thread it was even discussed that we should allow to override this using explicit sql.identifier, e.g.

update(
  connection,
  'user',
  {
    [sql.identifier(['given_name'])]: 'foo'
  }
);

What's the logic for needing the unique constraint to be normalized?

[danielrearden]

Beyond just consistency, it would also allow for some additional type safety when entering the unique column names:

function upsert<T extends NamedValueBindingsType> (
  connection: DatabaseConnectionType,
  tableName: string,
  namedValueBindings: T,
  inputUniqueConstraintColumnNames: ReadonlyArray<keyof T> | null = null,
  inputConfiguration: UpsertConfigurationType | null = null,
): Promise<unknown>

[gajus]

Interesting. I like this. Will patch it up.