ftp.cs.arizona.edu/ftp.cs.arizona.edu.tar/ftp.cs.arizona.edu/icon/newsgrp/group95b.txt/000065_icon-group-sender _Wed Jun 21 09:56:00 1995.msg

Internet Message Format | 1995-09-18 | 6KB

Received: by cheltenham.cs.arizona.edu; Thu, 22 Jun 1995 08:59:09 MST
Message-Id: <[e-mail]>
Mime-Version: 1.0
Content-Length: 5240
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Date: Wed, 21 Jun 95 09:56 CDT
From: [e-mail]
Subject: Re: ICON and Data Conversion
To: [e-mail]
X-Mailer: SPRY Mail Version: 04.00.06.14
Errors-To: [e-mail]

>A question comes to mind regarding what notations are easy for people to use.

In fact, I think REs are more cryptic and less intuitive than S4 patterns are
(at least, after one gets past "?" and maybe "*").

>From my experience it takes people time to grasp RE, but this time is
spent is CS classes. Snobol patterns also take time to learn, and I think
even more time that RE - but no one really has to learn them in school.
The Icon method is superior in that it offers much more than just string
scanning - it gives you a whole new flow control mechanism, that can be
used for other purposes, and that is integrated in the language.

I agree completely that the Icon string scanning paradigm is more powerful, but
I personally find the Icon paradigm as more complicated than SNOBOL4 for the
more commonly needed situations.  I just think S4 is simpler and easier to use.
(There are also other especially neat and extremely powerful features in SNOBOL4
that Icon simply doesn't have, such as the EVAL() and CODE() functions.)  In
fact, both languages have their uses... I have (and use) both of them, although
find that I use SNOBOL4+ at least an order of magnitude more often than I use
Icon.

Not that I intend that as a criticism of Icon... in general, ever since I've
been a struggling student at the University I've been very impressed with
SNOBOL4, and now with Icon... I sincerely hope to someday meet Ralph Griswold in
person (I've already had the pleasure of meeting his wife Madge on the occasion
of an Icon talk she gave to a PC Users' Group there).  I have a great deal of
professional respect for Professor Griswold, who has clearly contributed SO much
to all of us in the computing profession.

>So we have:
REs - known and semi-powerful (and you can add all kinds of extensions...)

Fine, but the more extensions you add the further it gets from being what people
recognize as "regular expressions".

>S4  - powerful, take time to learn, but seems closer to REs

I don't think that being "closer to REs" is any kind of a virtue.  I think that
SNOBOL4's patterns are fine as they are, a particularly elegant and concise and
intuitive way of expressing how a string is to be parsed.  If there is anything
which I find inconvenient there, it's matching up the corresponding
parentheses... although a good Brief macro (I would suppose) could fix that
problem.

>Icon - Great idea! But takes getting used to the concept.

You could, of course, program Icon in much the way you'd program in primitive
languages like C or Pascal.  But to do so is to lose much of its power and
expressivity.

>Since I think S4 doesn't stand a chance in the future. And I don't want
to make a flame war on this - if you want to say I am wrong, I'd be happy
to learn why.

SNOBOL4 has NEVER really been a mainstream language... it's always been a
language used mostly by the people "in the know", with a dedicated (one might
almost say a fanatical) following.  The wisdom of having written the SIL
implementation (and subsequently the Macro SPITBOL implementation, in turn) in a
machine-independent way, combined with a language design and implementation
which adapts to most any kind of underlying processor and operating system
architecture, has ensured that there have been (and will be) people around who
will be willing to port the language to new machines and operating environments
as they are released.  And now, with the SIL implementation (at least) having
been translated to C, that porting process has become even more simple than it
has been in the past, ensuring the long-term survival of SNOBOL4.

>So, since S4 is a dying language

I don't see it as a "dying language", although I'd agree that I don't expect it
to ever attain the prominence of C++ or something.  Something like Algol (which
USED to be very widely used, particularly in Europe, for example) is much more
curiously a "dying language", probably since it has less of an intrinsic
advantage over C or other competitor languages.  Since SNOBOL4 is _so_ unique
and its features are NOT duplicated in any other language (Icon probably comes
the closest), I don't see any real risk of SNOBOL4 disappearing completely.
Obviously however it IS important that those of us who DO know and use the
language make other newbies aware of it, through things like these newsgroups,
or the links I have on my Web homepage (see the URL below in my signature).

>we can choose between REs which suit people who have to get the job done, and
who know REs from school, from GREP etc. , and Icon's wonderful way of doing
things which is still used only by the elite... :-)

Again, a classic example of "start from an invalid premise, and reach an invalid
conclusion".  :-)

>If you have the time and want to invest it in learning a powerful tool
learn Icon, if you just need a quick hack I am sure Perl, AWK are much
easier to use.

Perl and AWK are LESS powerful and (for most things) LESS easy to use than
SNOBOL4 is.  You've not given any convincing argument of ANY reason why any
thinking person would and should use Perl or AWK in preference to something like
SNOBOL4.

Gordon Peterson
http://www.computek.net/public/gep2/