Generate passing live tests when schema and table names are equal

[rmoorman]

Generate passing live tests when schema and table names are equal

When the live generator is used with a schema name that is the same as
the table name, the generated tests should not fail.

This change makes sure that the generated tests correctly target the
collection item elements by using schema.collection instead of
schema.table.

The following generator calling variations should therefore work as
intended:

• same names, default pk name: mix phx.gen.live Books Series series name:string
• different names, default pk name: mix phx.gen.live Books Author authors name:string
• same names, custom pk name: mix phx.gen.live Books Covers covers --primary-key covers_id front:text back:text
• different names, custom pk name: mix phx.gen.live Books Reviewer reviewers --primary-key reviewer_id name:string

For context, before this change, calling the generator with the same
word for schema name and table name (e.g. mix phx.gen.live Books Series series name:string) resulted in failing tests with errors like:

  1) test Index updates series in listing (ExampleWeb.SeriesLiveTest)
     test/example_web/live/series_live_test.exs:52
     ** (ArgumentError) selector "#series_collection-c9129c62-f6b4-4749-854d-6d56f6a3b4b2 a" did not return any element within:

     <div id="phx-GGSpIya1xsOZigmM" data-phx-main="" data-phx-session="SFMyNTY.g2gDaAJhBnQAAAAIdwJpZG0AAAAUcGh4LUdHU3BJeWExeHNPWmlnbU13B3Nlc3Npb250AAAAAHcKcGFyZW50X3BpZHcDbmlsdwZ...
  ...

Reason being, that the stream used to generate the listing was named
after the schema.collection which was in this case suffixed with
_collection while the tests used the schema.plural value.

For reference see:

• https://github.com/phoenixframework/phoenix/blob/v1.8.1/lib/mix/phoenix/schema.ex#L102
• https://github.com/phoenixframework/phoenix/blob/v1.8.1/priv/templates/phx.gen.live/index.ex#L54

[SteffenDE]

I guess it would make our lives easier if we changed the stream_configure to also use schema.collection instead?

[rmoorman]

That is indeed way cleaner! I amended the PR to reflect the suggestion @SteffenDE .

On an unrelated note (not wanting to derail this little thing), I usually setup phoenix to use binary ids by default, but when I just tested this again, I did the first run without the --binary-id flag for the installer and it appears that the generated test code (the fixtures in particular) didn't work for the cases I set a custom primary key through --primary_key .... Is this some known issue?

Errors like

 12) test edit post renders form for editing chosen post (ExampleWeb.PostControllerTest)
     test/example_web/controllers/post_controller_test.exs:44
     ** (Postgrex.Error) ERROR 23502 (not_null_violation) null value in column "post_id" of relation "posts" violates not-null constraint

         table: posts
         column: post_id

     Failing row contains (null, some title, 2025-09-24 14:59:37, 2025-09-24 14:59:37).