As per this ticket Pylons/pyramid#1625.

Would an approach like this sound reasonable?

https://code.rhodecode.com/rhodecode-appenlight/files/58cb9fcf39dff182fd74dead41df90a3e091ff16/backend/src/appenlight/scripts/migratedb.py#L37

This is what I used in the past and could port the code into cookiecutter.
I would like some feedback on this.

We have to run tests on Windows for Python 2.7, but the current version of pytest has a bug that causes an infinite recursion. There is an open issue at pytest-dev/pytest#3506. When pytest fixes that issue, then we can change the minimum version of pytest to that. The issue is open as of pytest 3.7.3, released 2018-08-26.

related Pylons/pyramid#3220 (comment)

See Pylons/pyramid#3318

referencing via this post: https://groups.google.com/forum/#!topic/pylons-discuss/_MJflNUcjdg

the standard ways to implement sqlalchemy for years has been to add a finished_callback (https://docs.pylonsproject.org/projects/pyramid-cookbook/en/latest/database/sqlalchemy.html) or use a tween to close the dbSession.

maybe i missed something, but this cookiecutter doesn't seem to do either or import anything that does.

Hi! Thanks for the work in this cookiecutter :)

I was following along but found an issue,

https://docs.pylonsproject.org/projects/pyramid/en/latest/narr/testing.html

Says, in the bottom, that for functional tests we should use

class FunctionalTests(unittest.TestCase):
    def setUp(self):
        from myproject import main
        app = main({})
        from webtest import TestApp
        self.testapp = TestApp(app)

    def test_root(self):
        res = self.testapp.get('/', status=200)
        self.assertTrue(b'Pyramid' in res.body)

If we try to apply this on this cookiecutter we get an error from the main({}) part with a SQLAlchemy KeyError on url, following the unit test I was able to augment the cookie cutter with

class BaseFunctionalTest(unittest.TestCase):
    def setUp(self):
        self.config = testing.setUp(settings={
            'sqlalchemy.url': 'sqlite:///:memory:'
        })
        self.config.include('bookshare.models')
        settings = self.config.get_settings()

        from bookshare import main
        app = main(settings)

        from webtest import TestApp
        self.testapp = TestApp(app)

    def tearDown(self):
        testing.tearDown()
        transaction.abort()

However I still got the same error, after some google-ing and Stackoverflow I saw that it should be

app = main(global_config={}, **settings)

Of course that I could've checked better the docstring of the functions and so on but still, for newcomers, not that obvious :)

Thanks!

Would you be so kind as to provide a working uWSGI entry point for this cookiecutter?

There already exist tutorials on setting up uWSGI to serve REALLY BASIC pyramid applications. But their methods don't work with this cookiecutter, because there is not just one single python file that has an "application" defined.

How to Recreate Issue

Say I create a new application named "umbrella":

python3 -m cookiecutter gh:Pylons/pyramid-cookiecutter-alchemy

After following the instructions output from the cookiecutter, running pserve works fine:

env/bin/pserve production.ini  # this works fine!

But say now I want to run it from uWSGI like this:

uwsgi --chmod-socket=020 --enable-threads --plugin=python3 -s /tmp/umbrella.sock --manage-script-name --mount /=wsgi:application

Based on the "--mount /=wsgi:application" part of the invocation, uWSGI expects to have an entry point (python file) by the name of "wsgi.py" that defines a variable named "application". Would you be so kind as to provide such a wsgi.py file that will work?

My best effort is the following:

# wsgi.py
from umbrella import main
import pyramid


global_config = {'__file__': '/home/ubuntu/umbrella/production.ini',
                 'here': '/home/ubuntu/umbrella'}

settings = {'pyramid.reload_templates': 'false',
            'pyramid.debug_authorization': 'false',
            'pyramid.debug_notfound': 'false',
            'pyramid.debug_routematch': 'false', i
            'pyramid.default_locale_name': 'en',
            'sqlalchemy.url': 'sqlite:////home/ubuntu/umbrella/umbrella.sqlite',
            'retry.attempts': '3'}

application = main(global_config, settings=settings)

if __name__ == '__main__':
    # uWSGI won't call this section, but it gives us a way to test the code above
    # using env/bin/python wsgi.py
    from waitress import serve
    serve(application, host='0.0.0.0', port=6542)

But whether it's invoked via the uwsgi incantation shows above or via a simple env/bin/python wsgi.py, it gives the same error:

$ env/bin/python wsgi.py
Traceback (most recent call last):
  File "wsgi.py", line 15, in <module>
    application = main(global_config, settings=settings)
  File "/home/ubuntu/umbrella/umbrella/__init__.py", line 12, in main
    config.include('.models')
  File "/home/ubuntu/umbrella/env/lib/python3.6/site-packages/pyramid/config/__init__.py", line 839, in include
    c(configurator)
  File "/home/ubuntu/umbrella/umbrella/models/__init__.py", line 69, in includeme
    session_factory = get_session_factory(get_engine(settings))
  File "/home/ubuntu/umbrella/umbrella/models/__init__.py", line 17, in get_engine
    return engine_from_config(settings, prefix)
  File "/home/ubuntu/umbrella/env/lib/python3.6/site-packages/sqlalchemy/engine/__init__.py", line 464, in engine_from_config
    url = options.pop('url')
KeyError: 'url'

Not saying this is necessarily a bug with the cookiecutter, but hoping you can offer
some advice anyway.

I am a bit doubtful about the need to have the [alembic] section in all ini files. If I am not mistaken, the [alembic]script_location must be same in all config files for migrations to work. Then I think putting it into the config file is sort of confusing and can cause mistakes. What about adding something like this into alembic's env.py?

if not config.get_main_option("script_location"): config.set_main_option("script_location", "{{ cookiecutter.repo_name }}/alembic")

Thus, the settings in the ini files would continue working but they would be optional.

If we agree on this, I will prepare a pull request...

refs: Pylons/pyramid#3069