Page 27 - Compact Numerical Methods For Computers
P. 27

A starting point                         17
                        Section 1.8 notes some other details of algorithm expression which relate to the
                      ease of use of the codes.



                                            1.7. GENERAL NOTATION

                       I have not attempted to avoid re-use of symbols within this work since this would
                       have required an over-large set of symbols. In fact, I have used greek letters as
                      little as possible to save my typists’ and typesetters’ effort. However, within
                       chapters and within a subject area the symbols should be consistent. There follow
                       some brief general points on notation.
                      (i) Absolute value is denoted by vertical bars about a quantity, | |.
                      (ii) The norm of a quantity is denoted by double vertical bars, || ||. The form of
                      this must be found, where necessary, from the context.
                      (iii) A closed interval [u, v] comprises all points x such that u <  x < v. An open
                      interval (u, v) comprises all points x such that u < x < v.
                      (iv) The exponents of decimal numbers will be expressed using the symbol E as in
                                                         -5
                                               1·234 * 10  = 1·234E-5
                      and
                                                       2
                                               6·78 * 10  = 678 = 6·78E2.
                      This notation has already appeared in §1.2.



                                      1.8. SOFTWARE ENGINEERING ISSUES
                       The development of microcomputer software for users who are not trained in
                       computer science or related subjects has given rise to user interfaces which are much
                       more sophisticated and easy to use than were common when the first edition
                       appeared. Mathematical software has not escaped such developments, but source
                       code collections have generally required the user to consolidate program units, add
                       driver and user routines, and compile and run the programs. In my experience, the
                       lack of convenience implied by this requirement is a major obstacle to users learning
                       about software published in source code form. In our nonlinear estimation software
                       (Nash and Walker-Smith 1987), we were careful to include batch command files to
                       allow for easy consolidation and execution of programs. This philosophy is con-
                       tinued here, albeit adapted to Turbo Pascal.
                           1. All driver programs include code (from the fragment in file startup.pas) to
                           allow the user to specify a file from which the program control input and the
                           problem data are to be input. We refer to this as a ‘control input file’. It has a
                           name stored in the global string variable infname, and is referred to by the
                           global text variable infile. Algorithms which need input get it by read or readln
                           statements from infile. The input file can be ‘con’, the console keyboard.

                       WARNING: errors in input control files may cause source code files to be destroyed. I
                       believe this is a ‘bug’ in Turbo Pascal 3.01a, the version used to develop the codes.
   22   23   24   25   26   27   28   29   30   31   32