I agree it would be nice to be able to share the common C++ library and then have separate bindings for other languages, although based on the current pip packages which come out of bazel (currently contain all the C++ binaries and related header files etc), it would require a lot of patching of the bazel builds to allow all the bindings to link with the C++ library. If anyone knows enough about bazel to properly set this up, without significant modification to the source, I'm happy to go with your approach, otherwise we will see what the trusted users prefer. On Mon, Apr 25, 2016 at 4:15 PM, Pieter ROBYNS wrote:

I believe there's an issue to consider here. Tensorflow is not a Python module. It is a C++ library that happens to provide Python bindings. Bindings for other scripting languages (e.g. Go) could be introduced in the future. There is a C API and C++ API as well, though these are - presently - not as easy to use [1].

I'm wondering whether it would make more sense to have two packages: "tensorflow" and "tensorflow-git", where both python2 and python3 bindings are installed by default. A similar approach is taken in the "gnuradio" package for example: it also provides a C++ library and the python2 bindings are implicitly installed.

Perhaps a TU can help decide which approach should be taken. If they consider the python2-*, python-* approach to be cleaner, I will remove "tensorflow-git" from the AUR.

[1]: https://www.tensorflow.org/versions/r0.8/get_started/basic_usage.html

2016-04-25 7:41 GMT+02:00 :

...

qazokm [1] filed a request to merge tensorflow-git [2] into python2-tensorflow-git [3]:

Ignore previous deletion request. Both packages are near identical just newer one supports CUDA. Can they be merged to prevent confusing over naming

[1] https://aur.archlinux.org/account/qazokm/ [2] https://aur.archlinux.org/pkgbase/tensorflow-git/ [3] https://aur.archlinux.org/pkgbase/python2-tensorflow-git/