[plt-scheme] Re: Programming for non-programmers

Richard Cleis <rcleis at mac.com> writes:

> What about 'Programming for experienced programmers?'  I work with
> eight other programmers (in a science laboratory); none of them have
> any experience with any lightweight languages or functional
> languages (all have computer-science or computer-engineering
> degrees).  I have only been able to convince two people the value of
> scheme or python or anything 'out of the box', but all remain
> entrenched in C++.

We're pretty heavy in C++, but what's appealing about scheme, at least
enough to get us (me) initially looking at it, was not all the cool
features like continuations.  What I liked was (please no rotten
tomatos) the ability to use it as a configuration language, like emacs
uses its version of lisp. 

We like the idea of being able to have config files, command line
options, and the "back door" into our programs to tweak them at
runtime (what we call "guru mode") all processed by the same exact
interpreter.  Traditionally we had a command line parser, config file
parser, and "guru mode" parser that essentially all set the same
variables.  Now we make hooks so that the C++ variables can be changed
by the scripting language, and we can write scripts to drive the C++
program.

Suddenly a whole world of new dynamic behavior is opened up to your
program.  You can inject new behavior by adding a scheme function into
your program's environment, for example, and have it take immediate
effect.  (Just be sure it doesn't have errors... as this is obviously
risky.) 

Anyway, my point is the foot-in-the-door for us to scripting was to
embed the interpreter in our C++ applications, hook it up to most of
the important variables, and use it for simple controls.  From there,
it's surprising how many opportunities appeared where a script was a
better solution.

Another benefit was developing and running unit and regression tests.
They drive our C++ application and are another area where the dynamic
language was invaluable.  Writing tests in C++ is cumbersome and has
extremely slow compile/link cycles.

Dynamic languages require virtually no compile time which makes them
incalculably more valueable for testing compiled applications.

Just some ideas.

-- 
Chris