[Avocado-devel] Appropriate test+library layout?

[Amador Pahim]

On Sun, Dec 10, 2017 at 10:32 PM, Paul Graydon <paul.graydon at oracle.com> wrote:
> Hi all,
>
> We're looking at using Avocado to carry out functional testing.  Of
> particular interest is the possibility of providing dimensions for
> testing (need to repeat the same tests but with one or two slight
> differences each time).
>
> I'm running in to some quirks that indicate I'm either approaching
> things wrong, or I need to work around it.  I guess I'm seeking advice
> on good test layout or the best approach to take with avocado.
>
> Here's a toy example that reflects the approach we started with:
>
> $ find .
> .
> ./foo
> ./foo/__init__.py
> ./foo/util.py
> ./tests
> ./tests/__init__.py
> ./tests/test_foo.py
>
>
> $ cat foo/util.py
> def return_true():
>     print "Returning true"
>     return True
>
>
> $ cat tests/test_foo.py
> from foo.util import return_true
> from avocado import Test
>
>
> class TestFooUtil(Test):
>
>     def test_return_true(self):
>         assert return_true()
>
>
> The __init__.py files are blank.
>
> If I run this from ./ it fails because it can't import foo :
>
> $ avocado run .
>
> ...
>
> ImportError: No module named foo.util
>
> ...
>
>
> If I drop sys and print sys.path in to test_foo.py it looks like avocado
> is executing the code from within the tests directory, so imports would
> need to be relative.  I can meddle with the path to make things work, if
> needs be.
>
> If I quickly modify that test_foo.py file and turn it in to a standard
> unittest, I can invoke it from the parent directory and have it import
> correctly:
>
> $ python -m unittest -c tests.test_foo
> Returning true
> .
> ----------------------------------------------------------------------
> Ran 1 test in 0.000s
>
> OK
>
> Or allow it to discover the tests itself:
>
> $ python -m unittest discover
> Returning true
> .
> ----------------------------------------------------------------------
> Ran 1 test in 0.000s
>
> OK
>
>
> Is there a way to approach this using avocado that results in it finding
> all the tests when run from a directory, but executing the python from
> that context, instead of the directory where it found the tests?
> Should I be laying things out in a different fashion?  I don't see
> anything applicable in the documentation, but there's every good chance
> I'm just reading it wrong!
>
> There's lots of tests planned that we're hoping to keep structured
> cleanly under various subdirectories, and a fair bit of utility code
> that is going to be necessary to give us the test / dimensionality we
> really need.
>
> Any advice would be appreciated.

Well, there are many caveats on doing relative imports. We can dig
into the details on how Avocado executes the tests and so on, but to
be objective, my recomendation is that you install your lib and use it
as an absolute import. Example:

$ find .
.
./foo
./foo/__init__.py
./foo/util.py
./tests
./tests/test_foo.py
./setup.py

$ cat setup.py
#!/usr/bin/env python

from setuptools import setup, find_packages


setup(name='avocado-foo',
      description='Avocado Foo Utils',
      version='1.0',
      packages=['foo'],
      include_package_data=True,
      )

$ cat tests/test_foo.py
from foo.util import return_true
from avocado import Test


class TestFooUtil(Test):

    def test_return_true(self):
        assert return_true()


$ python setup.py install --user
running install
running bdist_egg
running egg_info
<..snip..>
Installed /home/apahim/.local/lib/python2.7/site-packages/avocado_foo-1.0-py2.7.egg
Processing dependencies for avocado-foo==1.0
Finished processing dependencies for avocado-foo==1.0

And run the tests from anywhere:



>
> Paul
>
>