Run IPython Notebook on Docker

Let's run IPython Notebook with Python 3.4 in Docker. There are some libraries that are not fully compatible with Python 3.4, but the migration is progressing gradually, so Python 3.4 is a good choice for future writing. You can check the compatibility of the library at Python 3 Wall of Superpowers.

Why use Docker:

• It can be a hassle when using Python 3.4 Installation can be difficult when using * * numpy * or * scipy * (especially on Windows)
• When the amount of calculation increases, it is convenient to be able to change the instance type with EC2

I think this is the reason why IPython Notebook is used for data analysis.

• Here is a mechanism to reduce fraudulent papers. Data material created with iPython Notebook. --Qiita
• Use IPython Notebook as an ad hoc analysis environment
• Example displayed by nbviewer
• The latter half will be the topic of C expansion, so the content for those who use it to some extent

Public image

Docker images for Python and IPython are registered on Docker Hub or published on Github. I feel that there is no definitive edition yet, so it would be nice to read and compare some Dockerfiles.

It's basically built on Ubuntu, but the C library that depends on which Python module you use is different. Roughly speaking, there are those that assume numerical operations such as * numpy * and those that assume database connections such as PostgreSQL. In addition, the version of Python in the first place is also different between 2 series and 3 series.

Self-made image and usage

This time, for my own study, I made a Docker image by hand and registered it in Docker Hub in order to use the latest version of each library. If you create it with Automated Build linked with Github, the image build will proceed automatically, which is convenient.

• skitazaki/python34-ipython

Launching the container with the image will launch the IPython Notebook on port 8888. You can specify the port number to forward from the host machine with the -p option. (Here, the port number is changed so that you can see how to specify it.)

$ docker run -d -p 8080:8888 skitazaki/python34-ipython

If you want to share files with the host machine, mount / notebook.

$ docker run -d -p 8080:8888 -v $PWD:/notebook skitazaki/python34-ipython

You can use the IPython Notebook by accessing port 8080 with a browser.

You can create a new notebook by clicking "New Notebook". You can see that you can enable * pylab * (* it seems that only * matplotlib * is better behaved in the future) and use * pandas * to draw the graph.

You can expose port 80 when booting on EC2, or you can use it for SSH tunneling. Let's set the security group and disclosure range according to the target data.

After editing the notebook, you can output it in HTML format, and you can download the Python source code and the IPython Notebook source code from the screen. This is often sufficient for checking data summaries and drawing graphs, such as in a pivot table. IPython Notebook allows you to add explanations in Markdown format, so I think it's more convenient to manage than just source code.

If you want to use a simple interactive shell, launch the container as follows.

$ docker run -it skitazaki/python34-ipython ipython

If you want to install additional Python libraries, specify / bin / bash in the command and install withpip3 install {LIBNAME}. After installation, you can start the IPython Notebook server with the following command.

container

root% ipython-notebook-startup.sh /notebook

Other

Useful Python libraries are summarized here. If you look at it once in a while, you may discover something new.

• vinta/awesome-python

When using Docker, it is convenient to set the alias written in the cheat sheet (wsargent / docker-cheat-sheet). Let's add it to .bashrc / .zshrc etc.

.bashrc

alias dl='docker ps -l -q'

I think the following article is easy to understand as an introduction to Docker.

• Let's try! Docker to remember during summer vacation --Knowledge of Sakura
• Let's create a container for Docker from scratch --Knowledge of Sakura