Trouble installing scikit-bio on Windows

Question

When attempting to install the scikit-bio toolkit via pip on Windows XP using Python 2.78 and Visual C++ 2008 Express Edition, the process is interrupted with the following message issued by VC:

cl : Command line error D8021 : invalid numeric argument '/Wno-error=declaration
-after-statement' 

Concerning this error, Microsoft Developer Network website just says:

invalid numeric argument 'number'

A number greater than 65,534 was specified as a numeric argument.

I did not (yet) attempted to install scikit-bio under Linux (Ubuntu 12.04 Precise), but I am under the impression that it will work fine (as everything goes as regards Linux).

Has anyone ever succeeded installing scikit-bio under Windows (XP, 7, 8)? Any hints?

Thanks in advance!

Flathead

Answer 1

I've installed Anaconda's Python (3.5) on a Windows 10 machine. As you should know, there is no official version of scikit-bio for this platform, but you can install it with this workaround:

1. Download the latest source from the Scikit-bio Github repository.
2. Using a tool like 7-zip, unzip it at your home directory.

3. Open a command line client and change to the source's directory

   cd %HOMEPATH%\scikit-bio-master
   

4. Using Notepad++ edit the file setup.py

   "C:\Program Files (x86)\Notepad++\notepad++" setup.py
   

5. Look for the line where the ssw_extra_compile_args variable is defined and change it. You can comment the previous version and redefine the variable in a new line:

   #ssw_extra_compile_args = ['-Wno-error=declaration-after-statement']
   ssw_extra_compile_args = []
   

6. Save the changes, close the editor and run the installation with this command:

   python setup.py install
   

7. Hopefully you won't receive any error messages. Open an Anaconda's Python session (using the command python) and test if Scikit-Bio was installed rightly using print(skbio.art).

By the way, I've previously installed Visual Studio 2015 Community Edition with C++ SDK features enabled, in order to meet the compiler requirements for the package (Scikit-Bio) installation.

Closeknit

Glad you have a work-around! FYI, scikit-bio will officially support Windows in 0.5.0, which is scheduled for July 2016 (timed with SciPy 2016). – Imbrue

Just to say, you inspired me a similar solution: I failed to install skbio on python 3.6/anaconda 4.4 with windows 64; there are no conda package and pip install failed. So from sources, I modified setup.py so that if sys.platform == 'win32' : ssw_extra_compile_args = [] reads 'win64' (I actually added an "or sys.platform == 'win64'", but this has no importance). It worked like a charm. Thanks/ – Ashby

Answer 2

At this time, scikit-bio doesn't officially support windows. We'd definitely be interested in hearing about people's efforts to install, test, and use scikit-bio on Windows, but we don't have developers with expertise in this area.

This is related to biocore/skbio#941.

Imbrue