1 comments about Numerical Algorithms With Fortran.

1. An excellent advanced book on numerical methods. The authors discuss a variety of methods, and furnish example code written in Fortran. A good indication of the serious nature of the book comes in seeing that right at the second chapter, it discusses solving nonlinear systems. More elementary books often neglect this, or put give more emphasis on handling linear systems, since these are amenable to matrix techniques.
   
   The choice of Fortran for coding underscores that language's continued dominance of numerical work in science and engineering. Though if you are skilled in another language, you may still want to consider this book. The descriptions of the algorithms are explicit enough that you should be able to code from those in your language.

No comments about Optimization Techniques With Fortran.

3 comments about C++ Scientific Programming : Computational Recipes at a Higher Level.

1. This review is based on a cursory first impression of the book C++ Scientific Programming. When I ordered this book I was excited with anticipation of discovering high quality, well-documented code that I could use with little modification in various projects. Unfortunately, as I briefly scanned the first three chapters of the book, I found one particular error immediately that appears consistently throughout the book. The error involves de-allocation of memory allocated with the new[] array operator. Dr. Berryhill consistently uses the plain delete operator instead of the delete[] operator. That is an error resulting in potentially huge memory leaks. Furthermore, it is a very basic error which in my opinion should be caught by the author not the reader five minutes after opening the package in which the book arrived. This error immediately stifled my excitement and I became more skeptical. Further complaints include the lack of templated classes; every numerical class uses a double as the underlying fundamental type. Also, the include files in the source code follow this general format:
   #include "C:/jcb32/book/source/xxx.h"
   I want the code to compile in whatever directory I choose to install it in. I don't necessarily want to replicate the directory structure shown above on my hard drive in which the root directory is obviously named after the author.

   To the book's credit, however, many appealing subjects are covered, and perhaps with the code provided and the book's discussion of the algorithms, a reader could implement his/her own library without the egregious errors.

   My suggestion would be to first look at libraries such as MTL and Blitz++ which truly use modern C++ programming techniques.




2. The previous reviewer is absolutely correct about the code flaws in the book. Rather unfortunate that these exist. But if you step back and take a broader view, then the book might appear in a more flattering light. Berryhill shows how important computational ideas in linear algebra can be accomplished in C++. Of the 14 chapters, most concern linear algebra, while the last chapter discusses Fourier transforms, like FFTs. The title of the book is overly broad, therefore.
   
   Within linear algebra, the treatment is suitable for an undergraduate, who already knows the basics of C++, and has had some exposure to matrices. You could perhaps use this book as an introduction to the latter, but that would not be recommended for most readers.
   
   As for the flaws in the code samples, most samples are in fact ok. Just treat them with a grain of salt. You could use this to your advantage by going over them carefully for any errors, and thus improve your C++ abilities. Which is the point of the book, after all. Make the most of it.



3. Very clear and exhaustive. A lot of examples. A good mathematical background is obviously required. Every programmer involved in projects on a domain requiring mathematical optimizations should have it on its desk.

2 comments about Classical Fortran.

1. I am unsure how much new this book really offers over Fortran texts of the last decade. It goes over the standard syntax in an adequate fashion. But no better or worse than most such other texts.
   
   The author has chosen to describe Fortran 77, which may be still the most widely used variant of Fortran. But perhaps more discussion about the newer versions that came out in the 1990s might make this book more relevant to some potential readers.
   
   The style of the examples is totally procedural, as befits that version of Fortran. To some extent, this may not matter, given the length of the examples. And it lets students concentrate on coding the numerical aspects of their solutions. But it can leave them unaware of the scaling difficulties when such programs grow. Which is an acknowledged problem with Fortran libraries, given the length of time [decades] that some of these libraries have been maintained and added to.



2. While this book doesn't include many of the newer additions to the Fortran language, it does cover all of the basics in enough detail that a beginning scientific or engineering programmer can easily write their own programs using it. However, the real value of this book is that it is written in a very simple and easy to follow (and sometimes even entertaining) style. It includes only the necessary details, but doesn't assume that the reader is familiar with any of the idiosyncracies of the Fortran language. Every topic includes a fundamental description of how the feature works, instructions and examples to demonstrate its use, as well as examples of how not to use it. The index is very thorough, making this a great reference book as well.
   
   I highly recommend this book to anyone that is interested in learning to write scientific or engineering applications in Fortran. The chapters on coding style and "what not to do" alone are in my opinion worth the cost of the book.

2 comments about Atlas for Computing Mathematical Functions: An Illustrated Guidebook for Practitioners With Programs in Fortran 90 and Mathematica.

1. According to the author, the practicality of books on functions is often limited by the graphical content, and that books of tables like Abramowitz & Stegun "serve a very limited purpose" today (primarily to check function evaluating software). This "Atlas" tries to fill this void, and it does a respectable job. However, detailed function exploration and visualization is left to Mathematica users, the primary audience for this book. Without Mathematica 2.2 or later, and/or a Fortran 90 complier, a great deal in this title cannot be fully appreciated. Of course, the reader can often explore function behavior by plotting an equation using common spreadsheet software, or a graphing calculator, without using F90 or Mathematica. But much of this high-priced textbook consists of small blocks of F90 code and everlasting references to Mathematica notebook calls. Part II contains ~200 pages of Mathematica notebook descriptions and ~100 pages of F90 driver programs, not to mention the F90 code documented throughout the first 600 pages.

   Most the "Atlas" graphics are small (~2") gray-scale screenshots of Mathematica plots. The quality of some graphics leaves something to be desired though (Figure 7.2.2 p. 117 for example). Many figures are obviously grainy - vertical lines and text characters often appear as broken line segments, not unlike a tilted faxed image. These gray-scale images are fairly bland; I expected at least a little color and only the highest quality graphics for a book calling itself an "atlas", especially for the asking price.

   "Atlas" is no substitute for the timeless books of tables and equations such as the (inexpensive) A&S (ISBN 0486612724) or the CRC Standard Math Tables. In computing the error function (erf), for example, Dr. Thompson defines erf in terms of a function call of the gamma function, while A&S provides many, many more alternatives suitable for machine solution. The discussions here, while more generous than A&S, are often not quite as insightful as Numerical Recipes, which the author often references. Instead, pictorial surveys primarily forego a lot of the detailed explanation of the underlying function theory. A few of the functions, such as the Voight distribution, are hard to find in the classic references, but the reader will find very few new topics here. "Atlas" is a well packaged presentation but not quite the insightful, general purpose book for which I had hoped.

   The contents are almost identical to an earlier C version by the same name (ISBN 0471002607). The availability in C, F90 and Mathematica is commendable, although it seems that the F90 version may now be out-of-print having been listed at a price for almost two hundred dollars for several years. Programmers of the older Fortran 77 standard will find the level of F90 programming reasonably suited for translation back to the older standard - or even C itself, if necessary. Therefore, Mathematica users in particular will find the used but now heavily discounted F90 copies the much greater bargain.




2. Don't expect an in-depth coverage of the theory behind the calculations. Said that, the book is a "concise" presentation of the programs used to calculate special functions present on the CD. For every program, the author shows the underlying equations and the bibliographic sources, plus nice discussions on accuracy/performance when you have the option of several methods.
   
   I would say the book is more useful to the engineer/physicist having to do certain calculations, than to the mathematician/student wanting to learn the ins and outs of special functions theory.
   
   The only dissapointment is a rather poor produced CD: the names of the programs on the CD are in 8.3 format, and I think (not sure though) that not all the code of the driver programs shows up on the CD- there are names of program snippets that are not on the CD- this may be an unnecessary hasle for someone trying to follow the examples in the book to the letter.

No comments about Fortran.

1 comments about Fortran 77: Language and Style : A Structured Guide to Using Fortran 77.

1. I needed to learn Fortran quickly and this book helped me to do it in 5 hours flat. Just type in all the worked out examples and you are on your way. The book is also a good reference.

1 comments about Fortran 77: With Numerical Methods for Engineers and Scientists/Book and Disk.

1. I used this book during my masters degree and I must say that it really helped me in building my own programs from scratch or modifying those of others'. The person who reads this book need not have previous knowledge of any programming language. I enjoyed it very much and took with great enthusiasm the well-selected examples, showing how ti apply Fortran for specific problems. A real achievement this book.

No comments about The Fortran cookbook.

2 comments about Programming in F.

1. I found this book quite by accident. After looking at the F programming language, I found that while F is quite general, there is a class of programming problems for which there is no F-ing way to solve. I am hopful that the future CORBA or COM IDL compilers will start recognizing the "-fthis" command-line switch and start allowing me to F with COM (or is that COM with F)? DB



2. F is Imagine1 Inc.'s new language (or rather a dialect), consisting of the actual 90 / 95 parts of Fortran 90 / 95, available for Windows / Macintosh / Unix / Linux. In other words, it doesn't have the junk (go to, strange If's) of old Fortran 66 / 77 but includes pointers and structures a la C. The F language is easier to use than C (and Java), and deserves to be the Turbo Pascal of the 2000's for statisticians / behavioral scientists / engineers / scientists, in terms of it being a great, general language (although it could do with graphics / interface / commercial numeric libraries). This book is not the sort of thing that can be read in an afternoon, but it is both very comprehensive and very clear, if you're willing to keep at it. Recommended!