Adds ability to capture all the db queries at once

[kingbuzzman]

When managing multiple databases—over six in our case—we don’t focus on tracking queries for each one individually (e.g., db1 vs. db2 vs. db3). Instead, our priority is the overall number of queries executed. If that total seems off, we then dive deeper to identify the specific issue.

Old approach:

@pytest.mark.django_db(databases=["default", "log", "backoffice", "cart"])
def test_models(django_assert_num_queries):
    with django_assert_num_queries(3), django_assert_num_queries(2, using="log"), django_assert_num_queries(3, using="backoffice"), django_assert_num_queries(1, using="cart"):
        Model.objects.create(...)
        ...

Since we use a django database router these model's data store can change location due to size at any moment. When this happens, all our tests will break.

New approach:

@pytest.mark.django_db(databases=["default", "log", "backoffice", "cart"])
def test_models(django_assert_num_queries_all_connections):
    with django_assert_num_queries_all_connections(9):
        Model.objects.create(...)
        ...

[kingbuzzman]

@bluetech let me know if there is something you want me to change

[bluetech]

This seems sensible to me.

One thing, pytest-django tries to mostly follow Django upstream, and just uses django.test underneath as you know. So ideally, this feature suggestion would go through Django upstream.

Would you consider sending your suggestion to the Django developer list (https://forum.djangoproject.com/c/internals/5), or open a Django ticket to discuss it?

[adamchainz]

Yes, this seems like a good feature to add upstream first. I would propose it extending assertNumQueries to allow using to accept both an iterable of database names, like using={"default", "other"}, and some special value for all databases, probably using="__all__". There is precedent for __all__ in TestCase.databases.

Adding another fixture with “all connections” part in the name (django_assert_num_queries_all_connections seems unnecessary to me.