Benedikt Franke

> If you i.e. introduce new field deep down within the type tree and forget to add @shared, whole subscription will switch to private mode, without you even notice, which...

There are some things omitted from this schema. In an actual application, `socialSecurityNumber` would only be accessible to authenticated users, and users would only be able to query the field...

After thinking this through for a bit, I do get your point about marking entire subscriptions with `@shared`. We would have to put a big disclaimer on it, explaining that...

This functionality would have to be implemented here: https://github.com/nuwave/lighthouse/blob/0d8f0fb8a3dd2791dda4c772f3db4b53ee072fb3/src/Execution/Arguments/ArgumentSet.php#L239. The syntax should probably be the same as in array validation: `activities.upsert.*.user_id`. What is supposed to happen when the client does...

In order for this merge request to move forward please address each of my comments.

@yondifon I think your approach should in fact work, but it has some problems. `NodeList::merge()` does not eliminate duplicates. Use `ASTHelper::mergeUniqueNodeLists()` instead. I would like to see a test case...

Looks much better, definitely on the right track. > I think we've resolved all the comments. There are still 3 open comments which are not sufficiently addressed.

> They are outdated No. I have looked at them. They are still relevant. Please address them.

@SirLamer good analysis. I think we will have to dig a bit deeper into the inner workings of `graphql-php` to either grab the unparsed values - or turn off parsing...

Right, the current behaviour is somewhat nonsensical. To properly support scopes on polymorphic relations, the definition has to take the type of the related model into account. I got a...