tables with same name and different schemas breaks migration generator logic

[sezaru]

Describe the bug
In my system, I have two schemas, public and raw. And inside both schemas, I have a table called records.

They have the same name, but they are very different tables

Seems like the resource_snapshots store both tables json file in the same resources directory and that seems to create some conflicts.

For example, I first added the raw.records table and created a migration for it, then, I created the public.records table and its migration.

Now I'm adding an index to the raw.records table, but when I run mix ash_postgres.generate_migrations, Ash will try to recreate the full raw.records table:

create table(:records, primary_key: false, prefix: "raw") do
...
end

To Reproduce
Create two resources (A and B) with the same table name but with distinct schemas.
First generate a migration for the resource A, then to B.
Now try to create an index for the A resource and try to generate a migration for that change.
AshPostgres will add the A resource table creation to the migration file.

Expected behavior
AshPostgres should not be confused by resources that share the same name but distinct schemas.

** Runtime

• Elixir version 1.16.0
• Erlang version 26.2.1
• OS Fedora Silverblue 39
• Ash version 2.17
• any related extension versions AshPostgres main

[zachdaniel]

The solution here is to include the prefix in the file name when there is a statically configured prefix on the resource. I will address this early next week, but if anyone wants to take a stab at it before then they are more than welcome. The fix will be in the file name determination logic in migration_generator.ex.