How to install bokeh for Python3

Question

I installed bokeh via pip and here is the information of the versions that are installed

pooja@X1-Carbon-6:~$ python3 --version
Python 3.5.2
pooja@X1-Carbon-6:~$ python --version
Python 2.7.12
pooja@X1-Carbon-6:~$ bokeh --version
0.13.0

for python2, it works fine and could import bokeh

lkhr@X1-Carbon-6:~notebooks$ python
Python 2.7.12 (default, Dec  4 2017, 14:50:18) 
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import bokeh
>>> 

However, it complains when I use python3

olkhr@X1-Carbon-6:~$ python3
Python 3.5.2 (default, Nov 23 2017, 16:37:01) 
[GCC 5.4.0 20160609] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import bokeh
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: No module named 'bokeh'
>>> 

I want to use bokeh in my Python3 Jupyter-Notebook and having problem cos of this, if any suggestions are available, please let me know.

Many thanks,

Solution 1

It will sound trivial but you need to install both (jupyter notebook and bokeh) under the same environment (virtual or not).

If you installed jupyter notebook using a snippet from the jupyter's website (pip3 install jupyter) then you have it installed in a non-virtual environment and from what I've understood you are trying to import bokeh which is installed in a virtual one.

There are two solutions:

1. You run everything under non-virtual environment (which is probably not the best/cleanest choice):

   • install the notebook using pip3 install jupyter
   • install bokeh using pip3 install bokeh
   • start the notebook using jupyter notebook (without activating virtual environment)

2. You run everything under virtual environment:

   • activate your virtual environment
   • install the notebook using python -m pip install jupyter
   • install bokeh using python -m pip install bokeh
   • start the notebook using jupyter notebook

Solution 2

I had the same problem here. Apparently, there are some problems with pip installation. I solved my problem with re-installing bokeh with conda.

Solution 3

You can have both versions of pip together (pip2, pip3).

Also pip can be linked to whether pip2/python2 or pip3/python3.

For me, pip is linked to the Python 3.x, pip2 to the Python 2.7 and pip3 to the Python 3.x:

$ pip -V
pip 9.0.3 from /usr/local/lib/python3.6/dist-packages/pip-9.0.3-py3.6.egg (python 3.6)

$ pip2 -V
pip 10.0.1 from /usr/local/lib/python2.7/dist-packages/pip (python 2.7)

$ pip3 -V
pip 9.0.3 from /usr/local/lib/python3.6/dist-packages/pip-9.0.3-py3.6.egg (python 3.6)

---

Python versions:

$ python -V
Python 2.7.12

$ python3 -V
Python 3.6.6

---

In this case, you should install bokeh for both versions of Python:

pip3 install bokeh
pip2 install bokeh

---

Test:

$ python
Python 2.7.12 (default, Dec  4 2017, 14:50:18) 
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import bokeh
>>> 

$ python3
Python 3.6.6 (default, Jun 28 2018, 04:42:43) 
[GCC 5.4.0 20160609] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import bokeh
>>> 

---

[NOTE]:

Refer to the following links to install both versions of pip:

• Link_1
• Link_2

Comments

[Violet]

I could install bokeh on virtual python3 environment, could do 'bokeh import' in python3, but still fail to import it in python3 jupyter-notebook, when opened in python3 virtual environment. (disappointed)

[Benyamin Jafari - aGn]

I posted an answer, hope help you up

[Violet]

Thanks for sharing your experiences. I am personally not a big fan of conda cos of my past experiences. So would like to have things working without conda. Lets see, if I succeed or would have to give Conda a try again. thanks again!

[Ashkan Farhadi]

@Pooja If you succeed, please let us know how. Thank you.

[bigreddot]

@Pooja I'd encourage you to look at conda again. All conda environments are completely self-contained and don't touch the system or other Python at all, which is a great feature for safety and reproducibility. Additionally many of the packages shipped with conda are more optimized (e.g. NumPy ships with Intel MKL linked)