fishers_exact_test
fishers_exact_test copied to clipboard
brentp
Build issue on Windows
I'm getting an error trying to install this on Windows. At first I tried this using conda, and after it failed there, I tried using just a barebones python2.7 install.
It seems that the issue stems from the fact that the Visual C++ for Python compiler that Microsoft released does not include an lgamma in its math.h
This appears to be the only issue. If I just replace the call to lgamma (in cfisher.c) with '0', then the package builds just fine.
See output from building a wheel below.
C:\Users\David\fishers_exact_test>C:\Python27\Scripts\pip.exe wheel --wheel-dir=wheel .
Processing c:\users\david\fishers_exact_test
Building wheels for collected packages: fisher
Running setup.py bdist_wheel for fisher ... error
Complete output from command c:\python27\python.exe -u -c "import setuptools, tokenize;__file__='c:\\users\\david\\appdata\\local\\temp\\pip-cljy6z-build\\setup.py';exec(compile(getattr(tokenize, 'open', open)(__file__).read().replace('\r\n', '\n'), __file__, 'exec'))" bdist_wheel -d c:\users\david\appdata\local\temp\tmpih00pjpip-wheel-:
running bdist_wheel
running build
running build_py
running build_ext
building 'fisher.cfisher' extension
creating build\temp.win-amd64-2.7
creating build\temp.win-amd64-2.7\Release
creating build\temp.win-amd64-2.7\Release\src
C:\Users\David\AppData\Local\Programs\Common\Microsoft\Visual C++ for Python\9.0\VC\Bin\amd64\cl.exe /c /nologo /Ox /MD /W3 /GS- /DNDEBUG -Ic:\python27\lib\site-packages\numpy\core\include -Ic:\python27\include -Ic:\python27\PC /Tcsrc/cfisher.c /Fobuild\temp.win-amd64-2.7\Release\src/cfisher.obj -O3
cl : Command line warning D9002 : ignoring unknown option '-O3'
cfisher.c
c:\python27\lib\site-packages\numpy\core\include\numpy\npy_1_7_deprecated_api.h(12) : Warning Msg: Using deprecated NumPy API, disable it by #defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
src/cfisher.c(2106) : warning C4013: 'lgamma' undefined; assuming extern returning int
src/cfisher.c(3442) : warning C4244: '=' : conversion from 'npy_intp' to 'int', possible loss of data
C:\Users\David\AppData\Local\Programs\Common\Microsoft\Visual C++ for Python\9.0\VC\Bin\amd64\link.exe /DLL /nologo /INCREMENTAL:NO /LIBPATH:c:\python27\libs /LIBPATH:c:\python27\PCbuild\amd64 /LIBPATH:c:\python27\PC\VS9.0\amd64 /EXPORT:initcfisher build\temp.win-amd64-2.7\Release\src/cfisher.obj /OUT:build\lib.win-amd64-2.7\fisher\cfisher.pyd /IMPLIB:build\temp.win-amd64-2.7\Release\src\cfisher.lib /MANIFESTFILE:build\temp.win-amd64-2.7\Release\src\cfisher.pyd.manifest
cfisher.obj : warning LNK4197: export 'initcfisher' specified multiple times; using first specification
Creating library build\temp.win-amd64-2.7\Release\src\cfisher.lib and object build\temp.win-amd64-2.7\Release\src\cfisher.exp
cfisher.obj : error LNK2019: unresolved external symbol lgamma referenced in function __pyx_f_7cfisher_lnfactorial
build\lib.win-amd64-2.7\fisher\cfisher.pyd : fatal error LNK1120: 1 unresolved externals
error: command 'C:\\Users\\David\\AppData\\Local\\Programs\\Common\\Microsoft\\Visual C++ for Python\\9.0\\VC\\Bin\\amd64\\link.exe' failed with exit status 1120
----------------------------------------
Failed building wheel for fisher
Running setup.py clean for fisher
Failed to build fisher
ERROR: Failed to build one or more wheels
maybe we need:
extra_compile_args = ['-O3', '-std=c99']
in setup.py (line 10)
can you try that and let me know if it works. I don't have access to a windows machine for testing.
Didn't help. The real issue seems to be that there just isn't any lgamma function in math.h for the Windows C compiler. To verify, I opened the file myself on my system and yep, definitely no lgamma.
See here for a compatibility table: http://www.johndcook.com/blog/math_h/
It seems to be the only function that's causing an issue, though. It all builds if I just remove lgamma (though of course, it's non-functional then).
It looks like there used to be an "lngamma" function, defined in the fischer source, that took care of this. I imagine the best solution would be to sense if lgamma exists and if not, use a fallback, though I don't know if there is an easy way to set that up in Cython.
On Wed, May 11, 2016 at 12:54 PM, Brent Pedersen - Bioinformatics < [email protected]> wrote:
maybe we need:
extra_compile_args = ['-O3', '-std=c99']
in setup.py (line 10)
can you try that and let me know if it works. I don't have access to a windows machine for testing.
— You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub https://github.com/brentp/fishers_exact_test/issues/16#issuecomment-218570752
I thought npy_math.h has lgamma but it doesn't, so the best way would be to simply resurrect the hand-rolled version.
Yes, looks like scipy has it - http://docs.scipy.org/doc/scipy/reference/generated/scipy.special.gammaln.html#scipy.special.gammaln - though not sure how hard it would be to borrow the implementation. Or if it really is worth the effort - performance-wise.
