rob@kaa.eng.ohio-state.edu (Rob Carriere) (09/09/88)
In article <632@muffin.cme-durer.ARPA> libes@cme-durer.arpa (Don Libes) writes: > [ Numerical Recipes in C is FORTRAN translated by people who don't > quite know what they are talking about; evidence from a review in > Micro/Systems Journal that quotes matter from the introductory > chapter. > ] > Isn't there a better book for numerical programming in C? The FORTRAN origins of the NRC code are at points quite clear, there are several routines that could easily have been coded base 0, and were coded base 1, and so on. The introductory chapter contains several remarks about C that to the experienced C programmer, and even to me, sound rather inane. All true. However, I do not know of any book, in *any* language, that contains the quality and quantity of numerical material that NRC has. You should take into account that good numerical analysis is more than 80% math, so even if the programming job were botched (and it isn't), the book would be worth its money. Further, apart from the base-0/base-1 array issue that has been beaten to death in this group already, the actual C code (as opposed to their philosophy about it) is good. The reason for this is simply that while numerical code may have very intricate analysis behind it, the actual code tends to be rather simple -- a couple of for loops and a handful of if's is typical. Finally, the problem with the base-1 arrays can simply be solved by a minor change to the vector(), dvector(), ivector(), matrix(), dmatrix() and imatrix() code, with a corresponding change in the free_<vector, etc> routines. I posted the change for vector() a while ago. In summary, anyone who claims the book to be of little or no value is not doing it justice. Rob Carriere