Question: "Reverse" permission check

[alex-pobeditel-2004]

Hello!

I faced a problem which is trivial at a first glance, but I still didn't find any elegant way of dealing with it.
When creating relations between definitions, is it possible to make "parent" permission depend on "child" permission?

I'll give an example:
I have a site with many sections. Site has admin and moderator roles, where moderator roles are given only to separate sections, and admin role is global.
Here is simplified schema:

  definition site {
    relation site_admin: user

    permission is_site_admin = site_admin
    permission can_view_admin_panel = site_admin
  }

  definition section {
    relation site: site
    relation section_moderator: user

    permission can_view_admin_panel = site->can_view_admin_panel + section_moderator
    permission can_edit_article = site->is_site_admin + section_moderator
  }

From the docs I know that I can inherit permissions from site in the section (to give edit right to admin on any entity, for example). But I don't know if there is a way to inherit permissions backwards?

For example, site has 100 sections and moderator has access to just 1 of them. Now to check if user can enter admin site I'm making 1 check for site + 100 checks (1 per section) for sections, to see if user has can_view_admin_panel permission in at least a one of them.

It seems so temptious to have a way of just checking site->can_view_admin_panel based on "has at least one can_view_admin_panel permission in linked sections"

Is it possible to do this check simple by authzed? Or maybe I designed this schema wrong from the very beginning?
Can you recommend me a way of idiomatic resolution for this kind of problem, please?

[vroldanbet]

Hi 👋🏻 ,

in order to be able to check is_site_admin over a site for a user that happens to be a section_moderator you need to provide a way to traverse the graph from a site, therefore you need a relation section: section under site.

There is also a proposal for a reverse relations that would help making that easier so it's inferred instead of requiring the corresponding writes: #356

[alex-pobeditel-2004]

Hmm, understood!
I'll try doing it in the way you proposed before reverse relations will be added (if they will eventually).
Thank you very much for fast and clear response, I'm closing the issue :)