Hints to build with Win10 / VS2019 / Python 3.7

[incensesmoke]

For a Win10/VS2019 build with a Python3.7 wrapper and SWIG4.0.1 I had to make the following changes in the current master BTK-core:

• Support building for Python 3 #26 Support building for Python 3
• Visual Studio 2017 warning about std::tr1 #23 Avoid tr1 with VS2017: 2b7a9a8
• Moved the *.tpp includes from the .h to the .cpp. #16 Moved the *.tpp includes from the .h to the .cpp.
• Fix the problem with using BTK in other libraries (like OpenSim) on Windows with MSVC. #12 Fix the problem with using BTK in other libraries (like OpenSim) on Windows with MSVC.
• Adapt ...\BTKCore-master\Utilities\eigen3\Eigen\src\Core\VectorwiseOp.h on line 585 to include the following:
  #ifdef EIGEN_VECTOROP_PLUGIN
#include EIGEN_VECTOROP_PLUGIN
#endif

Perhaps this can save somebody else some trouble. As a n00b this took me a long time to figure out ;) Now it passes all tests and I'm happy!

Edit: spelling

[jimmyDunne]

May want to try https://github.com/pyomeca/ezc3d

[incensesmoke]

Thanks for the link! I'll look into it

[melund]

@incensesmoke Could you push the changes you did to your fork of BTK?

I would really love to get conda-forge to build conda packages of BTK for Python 3.

[melund]

I have created a conda package for BTK which supports Python 3. It can be installed with:

conda install -c conda-forge btk

It currently has support for Windows. If anyone would help getting Linux/mac working that would great. I don't think it will be hard, I just haven't had the opportunity to try other compilers.

The build recipe is here:
https://github.com/conda-forge/btk-feedstock

I hope this will serve for anyone who needs BTK for older projects but can't use Python 2.7. I will not recommend it for new projects since BTK development has stopped.

Thanks a lot to:
@allardhoeve, @klshrinidhi, @chrisdembia, @aroy-hc
for all your fixes and commits which help get BTK Python 3 support.

also thanks to:
@incensesmoke, @CameronJGrant, @aaa34169, @jjnurminen, @demotu
Your posts and comments different places on Github helped me a lot to get the build working.

[aaa34169]

Hi Morten Sound good. Just one question : have you run all native implemented unit tests ? regards Fabien Le jeu. 9 janv. 2020 à 14:53, Morten Enemark Lund <notifications@github.com> a écrit :

…

I have create a conda package for BTK which supports Python 3. It can be installed with: conda install -c conda-forge btk It currently has support for Windows. If anyone would help getting Linux/mac working that would great. I don't think it will be hard, I just haven't had the opportunity to try other compilers. The build recipe is here: https://github.com/conda-forge/btk-feedstock I hope this will serve for anyone who needs BTK for older projects but can't use Python 2.7. I will not recommend it for new projects since BTK development has stopped. Thanks a lot to: @allardhoeve <https://github.com/allardhoeve>, @klshrinidhi <https://github.com/klshrinidhi>, @chrisdembia <https://github.com/chrisdembia>, @aroy-hc <https://github.com/aroy-hc> for all your fixes and commits which help get BTK Python 3 support. also thanks to: @incensesmoke <https://github.com/incensesmoke>, @CameronJGrant <https://github.com/CameronJGrant>, @aaa34169 <https://github.com/aaa34169>, @jjnurminen <https://github.com/jjnurminen>, @demotu <https://github.com/demotu> Your posts and comments different places on Github helped me a lot to get the build working. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#28?email_source=notifications&email_token=AAXUMBLTMSIQWAZ5LCJQJQTQ44T6PA5CNFSM4JSNGRI2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIQL3NY#issuecomment-572571063>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAXUMBLITOWVTZRSYLAJQGTQ44T6PANCNFSM4JSNGRIQ> .

-- Fabien

---------------------------------------------------------------------------------------------- Fabien Leboeuf Ingénieur "analyste du mouvement" du Pole 10, CHU Nantes, France Chercheur associé de l'Université de Salford, Manchester, Royaume uni Laboraratoire d'analyse du mouvement 85 rue saint Jacques 44093 Nantes, FRANCE ------------------------------------------------------------------------------------------------

[melund]

Good question and the answer is yes. The conda build recipe is setup to do that as part of the build. The build can only succeed if all the tests pass.

I had to adapt the test runner a little bit to handle the setup on the CI server used by conda forge. You can see it here:
https://github.com/conda-forge/btk-feedstock/blob/master/recipe/run_test.py

[aaa34169]

well done. i have not seen you used circle-ci. Envoyé de mon iPhone

…

Le 9 janv. 2020 à 19:36, Morten Enemark Lund ***@***.***> a écrit : Good question and the answer is yes. The conda build recipe is setup to do that as part of the build. The build can only succeed if all the tests pass. I had to adapt the test runner a little bit to handle the setup on the CI server used by conda forge. You can see it here: https://github.com/conda-forge/btk-feedstock/blob/master/recipe/run_test.py — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[incensesmoke]

Thanks @melund for doing the work I gave up on! Is it a lot of effort to also make a PyPi package? Perhaps to merge/augment @aaa34169's contribution there?

[melund]

Is it a lot of effort to also make a PyPi package?

I actually have no idea. I guess it is possible these days using wheels. But I think packages like this is exactly what conda is for. Distributing binaries with the python and real tracking of dependencies.

[jjnurminen]

@melund Thanks, this is great! Installs and works without a hitch on Win10 64-bit/Python 3.7.

[jjnurminen]

@melund Now that you established the build procedure, would it be possible to also build a 2.7 version of btk into conda-forge with small effort? I know this may sound silly, but I still need to maintain 2.7/3.7 versions of my package for some time. It would be great to be able to get the same btk conda-forge dependency regardless of whether Python 2.7 or 3.7 is used.