[PYTHON] How to easily switch the virtual environment created by Conda on Jupyter

Recently, I've been touching Python, but I've been struggling a little because there are too many new and old methods and various information on the net just to build the environment.

I made it using Conda, which seemed to be the best for building a virtual environment in Python, but when I tried to use it with Jupyter, which is popular recently, I was addicted to it because I could not switch the kernel of the virtual environment well. , I will leave a memorandum about how to do it.

If all goes well, you'll be able to instantly switch between development environments from Jupiter's [Kernel]-> [Change kernel] as shown below, which is quite convenient! !!

Assumption:

Conda and Jupyter installed

** Create an environment where numpy works with conda **

$ conda create -n numpyenv python=3.5 numpy jupyter

It is necessary to install jupyter even for the virtual environment to be created.

** Check the created virtual environment **

$ conda info -e
Using Anaconda Cloud api site https://api.anaconda.org
# conda environments:
#
numpyenv                 /Users/username/anaconda/envs/numpyenv
opencvenv                /Users/username/anaconda/envs/opencvenv
py35                     /Users/username/anaconda/envs/py35
root                  *  /Users/username/anaconda

You can see that an environment called numpyenv has been created.

** Install jupyter_environment_kernels **

A package for switching kernels is provided on Jupyter, so install it.

$ pip install environment_kernels

** Jupyter configuration file generation **

• If you have already created it, you do not need to create it.

$ jupyter notebook --generate-config

Probably ~/.jupyter/jupyter_notebook_config.py I think that the setting file called is generated.

** Addition of settings to automatically load Conda environment **

Add the following to ~ / .jupyter / jupyter_notebook_config.py. ** (Note: may be old) **

c.NotebookApp.kernel_spec_manager_class = 'environment_kernels.EnvironmentKernelSpecManager'
c.EnvironmentKernelSpecManager.env_dirs=['/Users/username/anaconda/envs/']

** (* Addition: 2016/12/09) ** As you mentioned in the comments, in Latest version, it seems that only the following line is all right.

c.EnvironmentKernelSpecManager.conda_env_dirs = [ '/Users/username/anaconda/envs/' ]

For conda_env_dirs, check the path where the virtual environment is generated with conda info -e etc. and set it appropriately.

** Operation check **

$ jupyter notebook

When you start Jupyter, you can see that you can safely select a new kernel called numpyenv.

reference: [Python] Create a virtual environment with Anaconda Cadair/jupyter_environment_kernels Jupyter Notebook and Conda