Imglyb and PyJNIus conda package maintainers needed for Windows and OSX

windows
imglib2
osx
java
python
imglyb
Tags: #<Tag:0x00007fa30b9ccf88> #<Tag:0x00007fa30b9ccd30> #<Tag:0x00007fa30b9ccb00> #<Tag:0x00007fa30b9cc790> #<Tag:0x00007fa30b9cc498> #<Tag:0x00007fa30b9d7f00>

#1

Apparently, the PyJNIus package I provide for conda is broken for Windows, which makes imglib2-imglyb impossible to install through conda on windows. I tried to re-build the PyJNIus package but I observed the same error. I am not familiar with Windows at all and just setting up the environment to be able to be able to run the conda build scripts is prohibitively slow (it took me a couple of hours to get to the point where I was able to encounter the same issue as in the linked thread above), and I am not even at a point where I could try to debug. This means that, in order for imglyb to be available for Windows on conda I do need support of an experienced Windows user to update (and maintain) conda packages whenever issues like this occur. I personally cannot do that and will have to remove Windows support for now until a maintainer is found.

While I am more familiar with OSX (and it seems that I have had more success maintaining imglyb on OSX in the past), my access to OSX computers is even more limited, and help with keeping OSX conda builds up-to-date would be very helpful and appreciated as well.

I do have helper scripts that generate the necessary files for a conda build on github. If we find maintainers, I will add instructions to those build scripts.


Analysis with ImageJ and Visualization in the Jupyter Notebook
Analysis with ImageJ and Visualization in the Jupyter Notebook
#2

I noticed that both ant and maven are now available on anaconda, which will hopefully make the build process easier.


#3

Hi @hanslovsky,

I’d like to volunteer to help getting the stuff running on Windows. I believe building bridges between the ImageJ and python worlds is a major challenge we are faced to with a potentially enormous impact for the image analysis community. However, I’m more an ImageJ and Windows-API guy and rather not very familiar with conda, python and build scripts. Thus, any kind of detailed documentation of this part might be very helpful. Shall we open a gitter room to discuss details?

Cheers,
Robert


#4

That is a great relief and very much appreciated @haesleinhuepf As a starter, it would be great to see if you can find a way to successfully compile and run the tests of PyJNIus (I removed all lines that are relevant only to the conda build process):

You will need (all available on conda):

  • ant
  • cython
  • nose (optional, for testing)

Then, in the PyJNIus root directory:

call ant all
"%PYTHON%" setup.py build_ext --inplace -f
python -c "import jnius"

The last command fails for me with

ImportError: DLL load failed: The specified module could not be found.

I inspected jnius/jnius.cp36-win_amd64.pyd with the dependency walker (if there are better tools than that for Windows, please share) and saw that some of the dlls required by jnius.pyd could not be found.

If you get that to work, you might want to run the test suite. Keep in mind that I had commented out those tests because they had not worked at some point in the past (but not with a dll import error):

:: run tests
cd tests
set CLASSPATH=..\build\test-classes;..\build\classes
set PYTHONPATH=..
nosetests -v

We can use the imglib2-imglyb2/Lobby on gitter


#5

@haesleinhuepf It seems that I might have figured out the issues that I had with PyJNIus. Shoutout to @jakirkham and @jjahanip for their help!
Check-out conda-forge/staged-recipse#6091


#6

I just updated the imglib2-imglyb and imglyb-examples packages on my conda channel hanslovsky. I confirmed on a Windows 10 machine that it works. Please let me know if there are any issues.