akvo / akvo-mis / #765

89%

main: 89%

Pull #280

coveralls-python

[#269] Record tenant ownership on the definition-root tables (#279)

* [#269] Add tenant ownership to the six definition-root tables

Levels, Administration, Forms, Organisation, Entity and
AdministrationAttribute are the only tables with no FK path to a
tenant; everything else derives through existing relations. Each gains
a nullable PROTECT tenant FK, so existing rows and the test seeders
stay valid.

Uniqueness moves per-tenant where it was global or implicit: level
numbers (previously convention only), one root administration per
tenant, and organisation names. NULL tenants are exempt because
Postgres treats NULLs as distinct, which keeps seeded test data
unconstrained.

* [#269] Backfill existing rows to a default tenant

A data migration creates a tenant with subdomain "default" and stamps
it onto every pre-existing row of the seven tenant-owned tables.
Existing single-tenant deployments therefore migrate with no behavior
change: their data simply becomes the default tenant's.

The migration is idempotent (filtered update), reversible (FKs nulled,
tenant removed), and uses _base_manager so soft-deleted rows are
stamped as well.

* [#269] Create a per-tenant hierarchy on every registration

With Levels and Administration carrying tenant ownership, the
get-or-create sharing of the first tenant's hierarchy is retired:
each registration creates its own unnamed level 0 and its own root
unit inside the same transaction. Simpler code than the conditional
version, and hierarchies are disjoint from birth.

The registration tests now count tenants through a helper that
excludes the "default" tenant. The backfill migration runs when Django
builds the test database, so that row is present in every test — the
bare Tenant.objects.count() assertions written in the previous
iteration would otherwise be off by one.

* [#269] Resolve each tenant superadmin to their own root administration

Superuser profile scoping substituted the first root it found, which
was c... (continued)

5707 of 6609 branches covered (86.35%)

Branch coverage included in aggregate %.

10789 of 12011 relevant lines covered (89.83%)

0.9 hits per line