coverage does not account for numba decorators

[pbacon42]

In fact the problem is related to Sonarqube (SQ) which in a way or another relies on reports produced by the coverage library. I noticed that functions decorated with numba decorators (jit, njit) are ignored when running SQ. Those functions are well tested but they are not taken into account when estimating the coverage.

I read on the coverage documentation there is the --timid flag but it has no effect (besides making the scan slower).

I am running Python3.6.3 and numba version is 0.43.1. The structure of my code is as follows:

integrate_tools.py:
`from numba import njit

@njit
def integrate(f, a, b):
# do numerical integration here...
`

test_integration_tools.py:
`import integrate_tools as it

def test_integrate():
# call integrate here...
`

Please let me know if I forgot any information

[nedbat]

Thanks. Could you make a small complete example that I can run? With all of the code, all of the dependencies listed, etc? I don't know how to flesh out the example you've given.

Also, I doubt the problem lies with Sonarqube, but just to be sure, can you look at reports generated by only coverage.py to see that the information you want is missing?

[wbadart]

Here's a pretty minimal example based on some work I'm doing that bumped up against this issue:

# foo.py
import numba

@numba.jit
def foo(a, b):
    return a + b  # <-- reported missing

# test_foo.py
from foo import foo

def test_foo():
    a = 1
    b = 2
    assert foo(a, b) == 3

$ coverage run -m pytest -q test_foo.py
.                                                     [100%]
1 passed in 0.69s

$ coverage report -m
Name          Stmts   Miss  Cover   Missing
-------------------------------------------
foo.py            4      1    75%   6
test_foo.py       5      0   100%
-------------------------------------------
TOTAL             9      1    89%

I created a conda env to run this example with

conda create -n numbatest numba coverage pytest

Here's that environment exported, with all the versions pinned down:

name: numbatest
channels:
  - defaults
dependencies:
  - _libgcc_mutex=0.1=main
  - attrs=19.3.0=py_0
  - blas=1.0=mkl
  - ca-certificates=2020.1.1=0
  - certifi=2020.4.5.1=py38_0
  - coverage=5.0=py38h7b6447c_0
  - intel-openmp=2020.0=166
  - ld_impl_linux-64=2.33.1=h53a641e_7
  - libedit=3.1.20181209=hc058e9b_0
  - libffi=3.2.1=hd88cf55_4
  - libgcc-ng=9.1.0=hdf63c60_0
  - libgfortran-ng=7.3.0=hdf63c60_0
  - libstdcxx-ng=9.1.0=hdf63c60_0
  - llvmlite=0.32.0=py38hd408876_0
  - mkl=2020.0=166
  - mkl-service=2.3.0=py38he904b0f_0
  - mkl_fft=1.0.15=py38ha843d7b_0
  - mkl_random=1.1.0=py38h962f231_0
  - more-itertools=8.2.0=py_0
  - ncurses=6.2=he6710b0_0
  - numba=0.49.0=py38h0573a6f_0
  - numpy=1.18.1=py38h4f9e942_0
  - numpy-base=1.18.1=py38hde5b4d6_1
  - openssl=1.1.1g=h7b6447c_0
  - packaging=20.3=py_0
  - pip=20.0.2=py38_1
  - pluggy=0.13.1=py38_0
  - py=1.8.1=py_0
  - pyparsing=2.4.6=py_0
  - pytest=5.4.1=py38_0
  - python=3.8.2=hcf32534_0
  - readline=8.0=h7b6447c_0
  - setuptools=46.1.3=py38_0
  - six=1.14.0=py38_0
  - sqlite=3.31.1=h62c20be_1
  - tbb=2020.0=hfd86e86_0
  - tk=8.6.8=hbc83047_0
  - wcwidth=0.1.9=py_0
  - wheel=0.34.2=py38_0
  - xz=5.2.5=h7b6447c_0
  - zlib=1.2.11=h7b6447c_3
prefix: /home/wbadar/workspace/.miniconda3/envs/numbatest

And here's the above as a gist:
https://gist.github.com/wbadart/25a110ff1ef5b8160c6a6c7a0449e266

[vitorsr]

@wbadart, I suggest testing the pure Python counterpart alongside the Numba-compiled function. It is available as an attribute of the Dispatcher object. (See 1.)

Considering your example, that would be foo.py_func.

Note that this is a bypass to obtain the coverage metrics. Both the original issue and your MWE still apply.

1. http://numba.pydata.org/numba-doc/latest/reference/jit-compilation.html?highlight=py_func#dispatcher-objects

[vitorsr]

test_foo.py

    assert foo(a, b) == 3
+   assert foo.py_func(a, b) == 3

$ coverage run -m pytest -q test_foo.py
.                                                    [100%]
1 passed in 1.12s

$ coverage report -m
Name          Stmts   Miss  Cover   Missing
-------------------------------------------
foo.py            3      0   100%
test_foo.py       6      0   100%
-------------------------------------------
TOTAL             9      0   100%

[pbacon42]

Thanks and sorry for missing your replies. I left this stuff aside for a while. Will try your solutions.

[peterhys]

Hi I think I have the issue with numba.vectorize, which compiles the function to a ufunc, but it does not have the py_func attribute.
coverage 5.1
python 3.6.9
Numba 0.50.0

# distance.py

import numba as nb


@nb.vectorize(
    [nb.float64(nb.float64, nb.float64, nb.float64)],
    nopython=True,
    target="parallel",
)
def distance(dx, dy, dz):

    return dx ** 2 + dy ** 2 + dz ** 2

# test_dis.py

from distance import distance


def test_distanec():
    a = 1
    b = 1
    c = 1
    assert distance(a, b, c) == 3

coverage run -m pytest -q test_dis.py     
.                                                                        [100%]
1 passed in 1.49s

$ coverage report -m
Name          Stmts   Miss  Cover   Missing
-------------------------------------------
distance.py       3      1    67%   11
test_dis.py       6      0   100%
-------------------------------------------
TOTAL             9      1    89%

Please let me know if there's a workaround for this thank you.