Re: Some thoughts about Sather

[Norbert Nemec]

Hi Kipton, nice to have yo with us! 

On Sat, Dec 02, 2000 at 02:37:26AM -0500, Kipton M Barros wrote:
> The tutorial on the GNU web site has been pretty invaluable to me for
> figuring out what's going on, but...  it's a little tough to grok the
> System Libraries.  The class browser seems like a really good idea, but
> I've been having some trouble with it.  Is there a way to have it browse
> *all* system classes in "SATHER_HOME"?

In general, you can only browse through a program. Simplest way to get the 
effect you ask for is to write a dummy program that references every class in 
the library. (all-classes.sa is supposed to do that, but it may be outdated.)

> (Idea: make this the default when run w/o arguments).

The browser could need much work besides that. Unfortunately, there is much 
more fundamental work to do at the moment.

> I'd really like to see more "overview" library documentation, maybe in
> HTML.  For example, I have to use a balanced binary tree; does this
> currently exist in the sather library?  If not, are you taking
> contributions?  How? 

There is a completely new library growing to completeness at the moment. You 
might want to look at the 1.3-beta series if you are interested in that. Keith 
Hopper is just writing documentation for it as well.

> Really useful for the newbie would be some additional discussion in the
> tutorial about actually using the package.  Like, the basics of sacomp,
> and module files.  Maybe options like "-output_C -pretty" should be on by
> default.

Again, the compiler implementation is about just as buggy as the browser. 
Before any minute is put into polishing the surface of the existing compiler, 
it is much more urgent to have a new compiler written from scratch. Work is 
being done here as well.

> Nitpick: has anyone considered adding support for multiline comments
> (* ... *) in the emacs Sather mode?  What's involved?

Mainly: knowing elisp. If yo ant to look into it, we would be very happy to 
merge in your changes. (Unlike the old library and the current compiler, I do 
not see any reason why anybody should do rewrite those files from scratch - 
therefore, work being done here might be worth the time spent on it.)

> This one's a little radical:  I'm wondering about language support for
> tuples. Was this ever discussed? I'm imagining dynamic packing/unpacking. 
> It would be neat if you could write...

Nice idea. Personally I really like it - anyway, the Sather language definition 
is seen as stable at the moment. There have been tons of great ideas how the 
language itself could be improved, anyhow: once you start doing changes like 
this, it is hard to find a point to stop, and in the end you have a new 
language. There are a few points that really are needed (like 
modules/namespaces to avoid and handle name clashes between classes) but beyond 
those, I do not think it would do any good to try to "improve" the language.

> Another thought -- has anyone considered extending closures to be closer
> to lambda expressions?  Ie, allowing nested function calls.  For
> implementation Sather could define an anonymous function at compile time. 
> Doesn't it do this already when some of the parameters are bound?

Again, this would mean a major change of language concepts. Neat idea, but I 
bet it would have far deeper consequences than anyone could oversee at the 
moment.

Ciao,
Nobbi

-- 
-- ______________________________________________________
-- JESUS CHRIST IS LORD!
--          To Him, even that machine here has to obey...
--
-- _________________________________Norbert "Nobbi" Nemec
-- Hindenburgstr. 44  ...  D-91054 Erlangen  ...  Germany
-- eMail: <address@hidden>   Tel: +49-(0)-9131-204180