Re: isetl developments and updates

[Jan Hlavacek <lahvak@math.ohio-state.edu>]

On Tue, Jul 13, 1999 at 12:03:19AM -0700, jim@mercury.laney.edu wrote:
> OK, in other news, I have tried my build of isetl, having become familiar
> a bit with the language. EVERYTHING I tried works fine! The only problem
> is the interface, which just looks wrong. Works good, looks bad.
> 
> I get the feeling that isetl might be able to draw pictures; I am
> currently considering that a wild unsubstantiated rumor :) So, isetl
> as I have built it is text-only.

DOS, Mac, and windoze versions do have graphics, Unix version never did.
I had some serious trouble with windoze graphics last year, when I tried
it, but Ed Dobinsky claims that it must be a problem with my windows
instalation.  It's possible, my windoze are all f@#$ed up.  I deleted
lot of stuff which I probably should have kept.  But anything else I
tried works.  Graphics is one of the things that will have to be done.
It shouldn't be that hard, graphics in isetl is very simple, from what I
saw in textbooks, you have to draw coordinate axis and points on the
plane.  The idea is that you don't connect the dots when drawing graphs.
The program should not pretend to do more than it really does, if you
know what I mean.  

> The original problem was it kept segfaulting in the function cbreak(),
> a libcurses command that changes how keyboard input works. I added
> code to initialize curses and code to undo the effect of initializing
> curses for when the user wants out of isetl.

Yes, I remember that when I was compiling isetl on solaris, I had to fix
this, too.  That was about 1.5 years ago.

> I was also very interested in making isetl rock solid, and having the 
> compiler help me do that. I noticed that improvement in this area was
> possible when I noticed that all the function definitions did not tell
> the compiler the types of arguments, so I altered a lot of them so that
> they would. Doing so would ensure that argument data is lined up properly
> so that the function would look in the exact right places on the stack
> for it.

That's great.  I think that it also improves the readability of the
source.

> Third, I noticed the parser was not getting built correctly because the
> data type for items of the syntax stack defaulted to the wrong type.
> The parser is built using a special tool which takes a language grammar
> and other information and automatically generates a parser that can
> understand and break down passages written in that language. I saw to
> it that the parser can always be built from the grammar description,
> that "make clean" would get rid of the parser that was generated, and
> that the syntax stack items would be the right data type.

Yes, I recall having some problems with the parser, but don't remember
exactly what it was.  

Anyway, I tried to compile your version on solaris with gcc, worked
fine, now I have two isetls, both seem to be working the same.  I didn't
do any extensive testing, but the test I did worked fine on both.  Your
version is probably more stable, you did much more work on it than I.

> YES, let's contact them, but remember, while the interpreter IS
> working, it's not ready for much more than an evaluation.

My reason for contacting them was mostly to figure out the differences
and improvements from isetl 3.0 (DOS, Mac, Unix) to ISETLW.  I think
they did some improvements in the interpreter, but I don't know exactly
what.

> Ya know, I don't see any activity in "Calculus, Computers, and
> Cooperative Learning" (aka C4L) for the past 2 years, nor do I see
> anything current they're doing with isetl. They offer it, but that's
> about it. Is anyone aware of any activity in C4L or anyone else using
> setl in education?
> 
> I expected to find MUCH more material, but as far as I can see from
> the web pages, either C4L and isetl interest has gone to -273.1 deg C
> (i.e., absolute zero) or else the thought of their updating their web
> pages is like getting their teeth pulled slowly without novacaine... 
> To me, it was disheartening to see what amounted to a total lack of 
> interest. I hope I'm wrong about this... isetl looks promising.
> 
> The last time they did a teacher seminar was in 1997, as far as I can tell.

They did CLUME seminar last summer (actually two of them).  But tht's
not about isetl, that's only cooperative learning.  Ed Dubinsky was here
(at OSU) last spring and had a talk about his teaching methods, and
mentioned isetl several times.  Aparently, he uses it a lot. 

There is several textbooks available which use isetl.  Just from my
memory, there is Calculus with ISETL (or some such), Learning Discrete
Mathematics with ISETL, one other discrete math book, I don't remember
the title, Learning Abstract Algebra with ISETL, and at least one other,
I don't remember what it was about, maybe precalc?

> So let's ask them for updated information on the C4L project, and suggest
> they update their web pages. And let's let them know that isetl is close
> to being ready for linux.

That sounds like a good idea.

> 
> > > Platforms I tested my version on: debian slink i386, debian potato
> > > i386, sparc, dec alpha. All perform similarly, but some show more
> > > compiler warnings than others; I'll clean those up soon. 
> > 
> > As I said, I don't have a linux with a compiler right now, but I could
> > give it a try on solaris.  I'll let you know how it works.
> 
> I'm not sure my stuff will compile on solaris... On the other hand,
> I'm also not sure the curses stuff won't be better either. If that's
> the case, it means the curses libraries are somewhat different. As far
> as I can see, curses support just wasn't completed.  

Se above.  It compiles with gcc, and the curses support doesn't work any
better on solaris.  That part of code will definitely need some more
work.  I think that two main parts that need improvement are the
interface and graphics.  As for the interface, the DOS version uses some
weird line editor (the only part of isetl that's not free software,
AFAIK).  I don't know if unix version uses it, I don't think so.  Maybe
that could be replaced with gnu readline?  We would have to look at the
textbooks, if they make any reference to the editor, and how do they use
its features (mostly history editing, as far as I can tell).  Gnu
readline is fairly flexible, we may be able to do something with it.

> In any case, this test build is definitely for linux; I may have done
> some things in the code to break portability with other platforms. I
> plan to run thru it again, this time encapsulating my changes into
> #ifdef linux .... #endif directives.

I wouldn't use `linux', unless you do something really linux specific.
Maybe use `gcc' instead?

-- 
Jan Hlav\'{a}\v{c}ek
lahvak@math.ohio-state.edu  (Blind Carbon Copies will bounce)
www: http://www.math.ohio-state.edu/~lahvak/