Recently I got a copy of NumPy Cookbook to read and review from Packt Publishing and I must say that I was positively surprised. Focus is of course NumPy, but the book touches SciPy too.
The book is laid out nicely and is generally easy to read and digest. I really loved how instead of doing examples as Python programs or even interpreter commands, they chose to use IPython, which is like regular Python shell, but in steroids (as my colleague eloquently put it). IPython makes experimenting and sharing the experiments with others fun and easy.
My only experience with NumPy before was from time when I was writing a simple ray tracer with Python. I knew that the library had lot to offer besides simple things I was doing, but did not really have good way to dig in into it. This book has over 70 different ways of using NumPy, SciPy, PIL among other libraries that can be used to analyze and manipulate data. It also briefly touches subject of quality assurance, which of course is very close to my heart.
Focus is all the time in showing how to do things with brief examples. This suited very well for me, since I’m more about learning by doing than learning by reading type of person. While the book is relatively thin (around 200 pages + index), it has quite lot in it. For seasoned NumPy / SciPy user it probably does not offer that much new, but for a person not familiar with the libraries it offers a fast way getting started.