These are chat archives for spyder-ide/public

15th
Sep 2018
hofingermarkus
@hofingermarkus
Sep 15 2018 09:48
Sounds like cool new features in spyder! :)
One question about that spyder-kernels package.
Is it possible to set other environment variables - like a CUDA_HOME path?
For example, tensorflows binaries require different CUDA & CUDNN versions depending on the tf version.
And another question. Does this mean the other environment can now deviate completely from spyder requirenments. E.g. having spyder runnging with pyqt5 and the other environment with pyqt4? Or are there some basic dependencies for the spyder-kernels package that circumvent that?
Olaf
@randomgambit
Sep 15 2018 12:27
Hi There, I saw there is a new version of Spyder coming in 2019. IS there a stable beta version already? Can we try it? Thanks!!
Carlos Cordoba
@ccordoba12
Sep 15 2018 13:20

@hofingermarkus, my answers:

Is it possible to set other environment variables - like a CUDA_HOME path?

For your environment variables to be passed to the kernel, you need to start it in a terminal with the command python -m spyder_kernels.console

Does this mean the other environment can now deviate completely from spyder requirenments

Yes, any environment can have the packages you want on it.

Or are there some basic dependencies for the spyder-kernels package that circumvent that?

The basic dependencies of spyder-kernels are installed when you install it with conda/pip install spyder-kernels

@randomgambit, Spyder 4 will come in the first months of 2019. You can install its first beta following the instructions mentioned here: https://github.com/spyder-ide/spyder/releases/tag/v4.0.0b1
Olaf
@randomgambit
Sep 15 2018 13:27
@ccordoba12 thanks! is it safe to use or its going to mess up my dataframes? :)
Carlos Cordoba
@ccordoba12
Sep 15 2018 13:29
Beta1 is quite stable, current master is a bit unstable but we'll stabilize it before the next beta
CAM Gerlach
@CAM-Gerlach
Sep 15 2018 13:32
@hofingermarkus Specifically, the basic two are ipykernel and cloudpickle, and the total list of subdependencies are backcall, colorama, decorator, ipython, ipython_genutils, jedi, jupyter_client, jupyter_core, libsodium, parso, pickleshare, prompt_toolkit, pygments, python-dateutil, pyzmq, simplegeneric, six, tornado, traitlets, wcwidth, and zeromq, with only three total packages (ipykernel, pyzmq and jupyter-client) having specific >= version requirements. So yes, you can have Qt4 (though really, it is time to upgrade since it or PyQt haven't been supported for many years and will be dropped from Anaconda sooner or later) or any other packages you want, so long as spyder-kernels is installed and something doesn't require > the versions specified (4.8.2, 17 and 5.2.3) for the three specific mentioned (which is rather unlikely).
Olaf
@randomgambit
Sep 15 2018 13:48
@ccordoba12 thanks carlos, but want I wanted to know if whether the beta can crash (and thats fine) or it can actually corrupt stuff in memory (more worrysome). Being a GUI I would not expect the latter but who knows?
Carlos Cordoba
@ccordoba12
Sep 15 2018 13:51
@randomgambit, there are no dangerous changes in our beta1, so it's safe to use it.
Olaf
@randomgambit
Sep 15 2018 13:52
@ccordoba12 thanks!
CAM Gerlach
@CAM-Gerlach
Sep 15 2018 13:55
@randomgambit In general, Spyder won't ever touch any of your data in memory unless you specifically open it in a Variable Explorer dialog, modify it, and hit "Save and Close"; furthermore, its just in memory, so your script can always be run to regenerate it (at least, if it is properly reproducible as it should be).