C

Numerical Recipes in C++: The Art of Scientific Computing

Format: Hardcover
Author: William H. Press
ReleaseDate: February, 2002
Publisher: Cambridge University Press
Rating:

Reader unfriendliness frozen in time
And in staunchly defiance of all the incredible progress in Computing Science. This book surprises to stay frozen in time through all its editions. No matter what the title says about the computer language used for the implementation of the numerical algorithms in the book, it is, in fact, the old unfriendly FORTRAN code from the first edition. Hastily made to comply with different language compilers.

Back then, it was a cheap substitute for a numerical library and it served its purpose pretty well. If you'd need a simple cut & paste solution of a numerical problem and if you want to roughly know how the result is obtained, the book is still OK. (The inclusion of limited number support functions (actually: "subroutines" . . . Oh, God!) is quite straightforward).

But if you'd need to make some adjustments to the code you are in for a lot of frustration and significant expenditures of time. I've had a lot of the both in the past with the wavelet chapter. Recently, I made the mistake to attempt a translation of the simplex algorithm from the book into a Java Script. Huge mistake! The way the code is written is mind boggling. I'm not sure if an obfuscator could render any program less readable than this.

Even the fact that most of the content of the book can be found on the web doesn't reduce the cost of using it.

A must-have book
Spendid exposition, unbeateable code. This is the kind of book you MUST have if you have to manage numbers and have to compete with numerical problems. A little more expensive, maybe if more would purchase the book.

Not as much as a step forward as the C version
It has been very useful. I have dipped into the C version of NR since it came out almost ten years ago. Having recently got the C++ version I may have had unrealistically high expectations. It is a good book, but I don't judge that it is that much better from the C version. All modern C++ compilers accept C programs so you can still get the formidable benefits of NR without moving to this version. I particularly disliked the size of the header file in the C++ version. It has to be included in all code (or else write your own individual header files) and since it runs to many thousands of characters, surely it was obvious that no one would spend time entering it? A bad decision for the reader and user, but one that definitely encourages the purchase of the downloadable code package.