Spyder install questions.

[Aaron]

Hello everyone,

Our mechanical engineering department at our university is currently
working to replace MATLAB with a Python Suite and I've recently
discovered Spyder along the way. In order for this project to go
forward, we need to have install procedures prepared for Windows,
Ubuntu, and Mac OS X. The situation with Windows is quite simple with
the install of python(x,y), but this masks a lot of details that need
to be addressed when doing an Ubuntu install (have not tried Mac
yet). I was hoping that someone in this group could please tell me
the bare minimums that need to be installed for Spyder to operate as a
suitable MATLAB replacement. For example, using the Synaptic Package
Manager in Ubuntu 11.10, I was able to complete a Spyder install.
Afterwards, I opened the program and began using commands listed in
the numpy tutorial. The question is, have all dependencies been taken
care of, allowing me to begin running Spyder after installing via
Synaptic Package Manager? Or do I need to follow a more detailed
procedure? In case you need to know, I have moderate experience in
Ubuntu and I am a decent MATLAB user trying to break through to the
Python world. I don't have a learned Python community in my
department and any help that you could give would be great.

[todd rme]

Unlike MatLab, python isn't really monolothic, which means you can
have as many or as few capabilities as you want. For a Matlab
replacement, absolutely essential modules would be:

numpy - this provides a matlab-like multidimensional array and some
other matlab-like capabilities like basic signal processing.
matplotlib - this provides matlab-like 2D plotting (and experimental
3D plotting). You can't use this without numpy.

To make a more complete matlab replacement, these would be very useful:
scipy - provides a lot of advanced scientific and engineering
functions and tools, built on top of numpy. Numpy provides the
equivalent of the core matlab installation, while scipy provides the
equivalent of many of the more specialized matlab toolboxes
sympy - provides symbolic manipulation, like maple or mathematica
pylint - a source code style and error checker, useful if you use
mlint in matlab
mpi4py - if you are using supercomputer, clusters, or other obscenely
parallel systems or use mpi, this provides a good python interface for
them.
rpy - provides a python interface for the R statistics language.

There are other modules in these categories (for instance there are a
bunch of plotting packages), and many more specialized modules.

If we knew more about what you wanted to do, it might make it easier.

-Todd

[David Verelst]

As mentioned by Todd, installing the different python packages on Linux shouldn't be hard. The important ones (numpy, matplotlib, scipy, etc) are all available in the repositories of all the major distributions. And dependencies are taken care of by the package manager so you don't need to worry about that (at least not for the mainstream packages we are considering here). It might be a little more work if you want to have a complete clone of the Python(x,y) environment you have on Windows. I haven't checked up on Python(x,y) for a while so I don't know exactly what's in there; it will be more than just numpy, scipy, matplotlib and spyder. But than again, it will strongly depend on what you exactly want to do with your Python environment.

Although a bit comprehensive, I can surely recommend following literature:

Python Scientific Lecture Notes
http://scipy-lectures.github.com/

Numpy Guide for Matlab users
http://www.scipy.org/NumPy_for_Matlab_Users

The Python g-node also organises a yearly summer school on scientific programming with Python. All the lecture presentations are available on the website.
Advanced Scientific Programming in Python (St Andrews 2011 Course)
https://python.g-node.org/wiki/schedule

MIT open course ware
6.189 A Gentle Introduction to Programming Using Python
http://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-189-a-gentle-introduction-to-programming-using-python-january-iap-2008/

How to Think Like a Computer Scientist: Learning with Python 2nd Edition
http://openbookproject.net/thinkCSpy/

There are also some books on scientific Python programming published by Springer:
http://www.springerlink.com/content/978-3-642-02474-0
http://www.springerlink.com/content/978-3-540-73915-9

And this list goes on...google, duckduckgo, and others are your friends ;-)

Hope this helps you anyway forward with you Python adoption!

Regards,
David

[Ghislain Vaillant]

If you're using Ubuntu, the python-scitools packages (look for it in
the software manager or type in "sudo apt-get install python-
scitools*" in a command prompt) already contain most of the packages
recommanded in earlier posts. Not sure if Ipython will be pulled from
spyderlib or python-scitools dependancies, though it's definitely a
good addition for quick testing of small code snippets, like in the
David's post references.

I see more and more Academics adopting Python for scientific computing
and Spyder as their prefered IDE. I did it and never looked back.