https://pantsbuild.org/ logo
Join Slack
Powered by
is it normal for the `pantsv2` `test` goal to run ...
# general
c
is it normal for the 
pantsv2
test
goal to run on 
__init__.py
files in the test target ? If so, has anyone encountered situations where it will fail , even though they are empty files ? Tried running with debug logs but nothing meaningful is shown other than the exit code being 5
h
I don't see anything in 
python_tests
target that suggests that. Here's the default for sources and you can see init files shouldn't be showing up.
Do you need the init files around if they're empty?
Not sure it'd be the recommended way, but you could pants_ignore them
Also, exit code 5 is what you'd expect. There's nothing in the module so no tests are run. Pants treats all non-zero exit codes as a process failure and raises indiscriminately. Ref
I had raised some similar odd test behavior here
c
the 
__init__.py
are there to help with the import of the source code. The codebase comes from an older state where it was stored as 
src
and 
tests
style and without 
_init_.py
, the tests will have a hard time importing the code frm 
src
h
Maybe you don't need the inits anymore now that pants helps locate things?
c
I have seen example repos structure the code so that test and source files are in the same directory ? Is that the recommended approach ?
w
which Pants version is this? anything after 
2.x
should skip those files and not run them
h
My interpretation is that pants does not intend to have any recommended structure and aims to be flexible to user needs.
c
2.9.0
w
what is the 
BUILD
content in the relevant directory?
c
BUILD
Copy code
dtl_version = {}
with open("products/datalogue/dtl-python-sdk/datalogue/version.py") as fp:
    exec(fp.read(), dtl_version)

python_sources(
    name="datalogue",
    sources=["datalogue/**/*.py"],
    dependencies=['3rdparty/datalogue/python:requests',
                  '3rdparty/datalogue/python:python-dateutil',
                  '3rdparty/datalogue/python:validators',
                  '3rdparty/datalogue/python:pytest',
                  '3rdparty/datalogue/python:numpy',
                  '3rdparty/datalogue/python:pyyaml',
                  '3rdparty/datalogue/python:pyarrow',
                  '3rdparty/datalogue/python:pandas',
                  '3rdparty/datalogue/python:pbkdf2'
                  ]
)

python_distribution(
    provides=setup_py(
        name="datalogue",
        version=dtl_version["__version__"],
        author="Nicolas Joseph",
        author_email="<mailto:nic@datalogue.io|nic@datalogue.io>",
        license="""
        Copyright 2021 Datalogue, Inc.

        This Datalogue SDK is licensed solely pursuant to the terms of the Master Software License executed between you as Licensee and Datalogue, Inc.

        All rights reserved.
        """,
        description="SDK to interact with the datalogue platform",
        long_description="",
        long_description_content_type="text/markdown",
        url="<https://github.com/datalogue/platform>",
        classifiers=[
            "Programming Language :: Python :: 3",
            'Programming Language :: Python :: 3.6',
            'Programming Language :: Python :: 3.7',
            'Programming Language :: Python :: 3.8',
            'Programming Language :: Python :: 3.9',
            "Operating System :: OS Independent"
        ],
        python_requires=">=3.6",
        setup_requires=['pytest-runner'],
        tests_require=['pytest>=3.6.3', 'pytest-cov>=2.6.0']
    )
)

python_sources(
    name="tests-utils",
    sources=["tests/util/*",
             "tests/kit/*"]
)

python_tests(
    name='unit-tests',
    sources=["tests/unit/**"],
    dependencies=[":datalogue",
                  ":tests-utils",
                  ":test_resources"
                  ],
)

python_tests(
    name='integration-tests',
    sources=["tests/integration/**", "tests/unit/**"],
    dependencies=[":datalogue",
                  ":tests-utils",
                  ":test_resources"
                  ],
)

resources(
    name='test_resources',
    sources=['products/datalogue/dtl-python-sdk/tests/resources/**/*.json',
             'products/datalogue/dtl-python-sdk/tests/resources/**/*.csv',
             'products/datalogue/dtl-python-sdk/tests/resources/**/*.yaml'
             ],
)
and command is
Copy code
./pantsv2 test products/datalogue/dtl-python-sdk:unit-tests -ldebug
w
@creamy-airplane-38079: so, by default, 
python_tests
does not glob 
__init__.py
files, but by globbing 
tests/integration/**", "tests/unit/**"
, you are including them here
➕ 2
it defaults to 
"test_*.py", "*_test.py", "tests.py"
c
so I remove the 
sources
field
it should run on defaults ?
or just set that to the directory ?
w
@creamy-airplane-38079: the default is not recursive… targets default to owning files in their directory. so you will still need to set it with this layout
h
That glob pattern is only that directory. You'd need 
tests/unit/**, tests/unit/**/!__init__.py
(I think)
w
so in this case, you’d need something like:
Copy code
"tests/integration/**/test_*.py", "tests/unit/**/test_*.py", ...
…etc, for the relevant patterns
c
gotcha
thanks
w
it’s a larger conversation to have, but in general: rather than recursive globbing like this, we recommend 111, where targets live in BUILD files in the same directory as their files
ah, yea: or what @high-yak-85899 said: you can do a full glob, and then exclude the 
__init__.py
files.
c
got it , thanks for the feedback
w
re: 111: if you were to remove that BUILD file entirely, and then run 
./pants tailor $dir
for that directory, it would generate a very different layout, which might be a bit more maintainable
👀 1
but adjusting the globs should also work.
c
any docs or paragraphs that describe the pros of the 111 approach ? i see the docs mentioning to keep the build files with the code they reference to but can't remember the benefits
nevermind I found the doc I was looking for
👍 1
Open in Slack
PreviousNext
Powered by