From chemistry-request |-at-| ccl.net Thu Dec 12 21:32:20 1991
Date: Thu, 12 Dec 1991 18:17:56 PST
From: "W. Todd Wipke" <wipke-: at :-SECS.UCSC.EDU>
To: chemistry: at :ccl.net
Subject: Large (100,000 lines of code) programs                          
Status: R

John Rupley asked to hear from someone involved in large fortran programs.
I have been involved in several, e.g., our SECS synthesis planning program
that involved more than 60 person-years of programming, MACCS, REACCS,
are other examples.  The SECS program has its own virtual memory system,
dynamically allocated arrays, lists, sets, a compiler and interpreter,
i.e., many complex data structures for AI applications.  Now we are 
writing much code in C, mainly because of ease of interface with X-windows.
I have found that the fortran code is more readable and much easier to
debug.  C provides automatically many features we had to program in
Fortran, but we built into our Fortran implementations checks to catch
wild pointers.  Once we made an interface layer in C to catch wild pointers,
debugging became easy, and the code looked more like our Fortran code
with function references rather than direct pointer access.  Readability
is a key for long term projects.  I worked on TACOS-II in 1965 and the
program is still being used today, it simulated the Patriot missles we
saw in action not long ago.  Thats pretty good longevity and transportability.
C is like a double-edged razor blade, it takes skill to avoid hurting
yourself!