rmc@cca.UUCP (06/15/83)
A reasonable Modular C is described in the ACM SIGPLAN NOTICES, V18
Nbr4 pp 45-54. It uses defines and specific structures to handle
importing and exporting routines. Some salient features from reading
the article (i have not actually used it as yet):
- There is very little overhead in using it as you have basically
changed the syntax for native C.
- It does hide internal algorithms and data structures as advertised.
- It is readable, but not as nice as Modula et al. To handle
exporting and importing routines, you export a structure of pointers
to the routine entry points. Thus to call "fu" one writes (*fu)(),
which i suppose could be objectionable. The extra time for the
indirection is probably not significant.
- ADA overloading of operators and the PASCAL "with" clause are not
part of C and thus code will in general contain more characters
than if it were written in a language that supported "with" and
overloading.
- The defines necessary to use Modular C and examples of stack, queue
and storage managers are included in the article so you don't have
to reinvent anything, just go ahead and start coding.
In general i would suggest SIGPLAN NOTICES to people interested in
modern programming languages, as it contains lots of articles of small,
quick implementations of new ideas suitable for trying on your own.
Quite a number of them use C as the base language - an object oriented
pre-processor for C comes to mind as a prime example.
R Mark Chilenskas
Chilenskas @ CCA-VMS
(i don't read uucp addrs, so i hope
the header is useful)mcg@tekecs.UUCP (06/28/83)
WRT Chuck Rospach's lambasting of SIGPLAN Notices, and "Finite State Diagrams of Western Swing", that particular article was written by the venerable Donald Knuth, and I found it a very entertaining break from the larger volume of dry technical material I must read. Clearly Mr. Rospach did not realize that it was a joke. S. McGeady
cfv@packet.UUCP (06/30/83)
I knew that 'Finite State Diagrams of ...' was written by Knuth. If it was
written as a joke, it passed by me completely, as the tone of the article
was VERY serious. I picked it out of many possible articles I could have
lambasted simply because it was SO far gone from any base of reality.
The whole problem with SIG{Plan,Ops,Sac,etc} is that rather than sending
out information that is potentially relevant or useful (and in some cases
accurate) they are turned into the last chance for the "Publish or Perish"
groups because they are not refereed. I will agree that there are some
VERY good articles in the publications. I will also state that there are
MANY VERY bad, misleading, poorly written, obscure, useless or simply
incorrect articles. At one point in Sigplan something like 30% of the
publication was being used to correct mistakes in previous issues (mostly
in Ada articles that have sinced moved to AdaTec). Errors in Ada are a
special case since it is hard to hit a moving target but at the same time
if someone bothered to work with the authors a little bit a lot of this
could be reduced.
Finally, I should point out that I am 1) not a compiler writer, and 2)
don't want to be a compiler writer. What I AM interested in is language
design and how language design and compilers interact (and sometimes
interfere) with the programming process. For somebody who is writing
compilers for a living perhaps SIGPlan is a more appropriate publication,
but for me most of it is really worthless.
(Trying to make this my last flame, honest!)
--
>From the dungeons of the Warlock:
Chuck Von Rospach
ucbvax!amd70!packet!cfv
(chuqui@mit-mc) <- obsolete!cfv@packet.UUCP (07/02/83)
I read the article in SIGPLAN on modular C. It was definitely modular. If
was definitely C. If you follow the suggestions they make, it will also
definitely be unreadable and unmaintainable code. It is simply another case
of someone trying to force ideas onto a language that wasn't designed for
them, and I thought that the implementation wasn't extremely well though
out or clear.
--
>From the dungeons of the Warlock:
Chuck Von Rospach
ucbvax!amd70!packet!cfv
(chuqui@mit-mc) <- obsolete!cfv@packet.UUCP (07/02/83)
As long as people are suggesting SIGPLAN, let me NOT suggest it. I have
read SIGPLAN since about 1979, and I let me membership (actually
subscription, but they can call it anything they want) lapse this year when
I renewed. 99% of the articles they publish are rather worthless at best.
The one that will always stick in my mind is (a new low) 'Finite State
Diagrams for Western Swing Dancing'. You must remember that SIGPLAN (and
most if not all of the SIG publications) is NOT a refereed publication, and
it will usually publish anything that is sent in, no matter how trivial the
material (see the above title). I will agree that there are occasional
gems, but spending $22 a year and wading through all of the dreck simply
wasn't worth it anymore, even if I can do it on company time and with a
company membership...
--
>From the dungeons of the Warlock:
Chuck Von Rospach
ucbvax!amd70!packet!cfv
(chuqui@mit-mc) <- obsolete!cfv@packet.UUCP (07/03/83)
Evidently when I made my original assertions about SIGPlan and net news I
did not make some of my points as clearly or lucidly as I had hoped. The
mail I received on some of my comments showed me some of the flaws and I
feel that some apologies and clarifications are in order. Serves me right
to flame when I am in a bad mood. I should know better...
First of all, on SIGPlan: I should have pointed out that for someone much
closer to compiler state of the art than I am it would probably be much
more relevant than it is to me. I still believe that the SIG publications
in general should go through some form of referee to minimize the
triviality or incorrectness that shows up in them.
On net news: What I said before stands: My only NEED for net news is for
the information I get for supporting Unix. If I didn't have a Unix
system to support I wouldn't have net news around (probably). However,
there is a LOT of information available on the net beyond Unix, and all
of it is useful in one form or another to one person or another or it
wouldn't be there. I may not PERSONALLY be interested in a topic, but
as long as someone is, then I fully support its availability and use.
My implication that all non-essential (to me) topics should dry up and
leave me alone was completely off-base, uncalled for, and incorrect and
for that I apologize.
Finally, I would like to take time to thank people like Mark and Scott who
took the time to send me very well thought out and intelligent comments
that made me sit down and think about what I said. Its people like these
that make the net truly interesting and WORK. For all of the nameless
people who sent flames even more mindless than mine were ('Sticks and
stones...') thanks for letting me know that there are others who can write
while sleeping...
--
>From the dungeons of the Warlock:
Chuck Von Rospach
ucbvax!amd70!packet!cfv
(chuqui@mit-mc) <- obsolete!