Working with one of my clients (who is hiring BTW), we decided to use Docker as deployment platform. Since many Linux systems now use Python for many utilities, it''s advisable to install your own Python next to the system one and use it.
Installing CPython from source requires some system packages, libraries, headers
and some knowledge. The much easier path it to use MiniConda (from the wonderful people at Continuum). Not only the Python installation is super simple,
but also the conda package
manger will get you a lot of packages pre-compiled so you don''t have to install
gcc and header files for C extensions. And if you can''t find the package you
need with conda, pip is also available.
Here''s a little project to demonstrate how to do this. The application is an image server with has two entry points /edge for edge detection
and /resize for image resizing. We''ll be using scikit-image and Pillow for image manipulation and Flask as web server. All of them can be conda installed.
Here''s the Dockerfile for the project. Build with docker build -tag imgsrv,
Run with docker run -p 8080:8080 imgsrv (see Makefile).
- ► 2016 (17)
- ▼ 2015 (18)
- ► 2014 (24)
- ► 2013 (35)
- ► 2012 (22)
- ► 2011 (29)
- ► 2010 (17)
- ► 2009 (38)
- ► 2008 (45)
- ► 2007 (26)