[plt-scheme] Debugger: How much work to implement?

From: Eric Kidd (eric.kidd at gmail.com)
Date: Tue Oct 12 12:32:46 EDT 2004

On Tue, 12 Oct 2004 08:59:00 -0400, Matthias Felleisen
<matthias at ccs.neu.edu> wrote:
> Eric, if this is the case, we need to study what kinds of scripts
> your users write and what their problems are. Perhaps we can come
> up with even better targeted help. Debuggers are crude tools.

Debuggers *are* crude tools, certainly, and I can see powerful
arguments against using them in a pedagogical environment.  But every
time one of our script authors runs into a problem she or he can't
solve, I get pulled away from low-level coding.  I'd like to reduce
the frequency of interruption.

Our script authors can create non-trivial applications with Macromedia
Director (or Flash).  They code mostly by reading "cookbooks" and
adapting sample code.  A few of them understand abstraction, as a
concept, and can apply it in simple cases.  But in general, even
though the script authors have created actual, useful Director
applications (with attached scripts), they would never think of
enrolling in CS101.

These script authors, most of them full time, are suplemented by
student interns who actually know Scheme fairly well, and have taken a
data structures and algorithms course using PLT Scheme.  We'd use more
senior Scheme people, but so far, we can't hire any.

One of the ways we typically get burnt is that an intern writes some
slightly-too-complicated Scheme, and that our script authors can
*almost*, but not quite, figure out how the Scheme code works.  Right
now, this means that I'm responsible for maintaining every piece of
code ever written by an intern.  This is unfortunate; I'm supposed to
be a tools hacker and not a content developer.

I think that even a fairly rudimentary debugger--one which could stop
the program at a specified break point and examine the state of
lexical variables--would allow our multimedia script authors to
function at least as well in Scheme as they did in Director.  I don't
see any reason why Scheme should be inferior to Lingo or ActionScript
for multimedia development, provided the tools are good enough.


P.S. I just want to be clear--I understand that PLT Scheme is a
pedagogical environment, and that embedding PLT Scheme in production
applications is not your top priority.  I'm OK with this.  We might be
able to support some debugger work; I'll speak with my boss.

Posted on the users mailing list.