[672] in Coldmud discussion meeting

Re: assignment operator

daemon@ATHENA.MIT.EDU (Wed Mar 8 14:03:33 1995 )

To: The Importance of Being Chuck <cadams@weather.brockport.edu>
Cc: crag@serial.mit.edu (Robert de Forest), coldstuff@MIT.EDU,
        nop@fuji.ccs.neu.edu, eostrom@ccs.neu.edu
In-Reply-To: Your message of "Tue, 07 Mar 95 17:28:40 EST."
             <9503072228.AA09902@lightning> 
Date: Wed, 08 Mar 95 13:55:39 -0500
From: Jay Carlson <nop@ccs.neu.edu>

> Robert de Forest drew these heiroglyphs:

> > I'd like to propose '->' be the assignment operator. It visually represents
> >  what happens to the variable being assigned the new value. It has no chance
> >  of being pronounced, 'equals'. It looks cool...

> it's backwards.  :) 

> though that's been discussed.  What concerns me is that '=' is already
> assignment.  do we really want two different assignment operators?  or
> are we removing the single '=' entirely?  

Perhaps a transition server that parsed both as assignment, but only 
unparsed as <-.

> If so, all I'd do is write
> code in-db that changed them back and forth like JHcore MOOs do with
> comments.  

There are a lot of limitations to what I can do in 
:verbcode_internal_to_external, mostly due to the fact that I only 
have access to the string rep of the code.  In particular, dealing 
with anything that's not a statement can be very hard.  Comments are 
statements.  So is assignment right now. 

The thing that really makes me scream is that we could do far more 
powerful transforms if we had access to the syntax tree that the 
parser has to build ANYWAY.

See my signature.

> which makes the whole issue of what to call it quite
> academic, actually.  but it's not feasable to write code in-db that
> does assignment-as-expression, so let's have it in the server.

Without looking at the problem too hard, I'd say that doing the `<-' 
to '=' rewrite is of middling difficulty.  It's merely a lexical 
transform, so we don't need a full codewalker.

> > What have I left out?

I know what *I've* left out, *my* religious views!

I like <- as assignment operator, with := in a close second.  = is 
RIGHT OUT.

<- used to be in the ASCII character set as a single glyph.  It was 
replaced with _.  There are stil some systems out there that have it 
though; in particular, Smalltalk-80.  It's a pretty big win.

I like the fact that assignment is not an expression. If we get <- as 
an assignment operator, it's not as critical, but with assignment 
looking like equality it eliminates a lot of stupid bugs.

Elsewhere, Jeff Kesselman wrote:

> Well, I'm going to be the dissenting voice here.  +=, -=, etc. all seemed
> very rational to me.  

They seemed very rational to me when I didn't understand that 
compilers aren't stupid.

> The = being an assignment operator ONLY caused me
> some trouble when i forst moved from Pascal to C, and it was minor
> trouble at that.  You learn very quickly to be careful abotu this.

I've been programming in C for ten years now, and = as assignment 
operator has given me trouble the whole time.  I'm tired of being 
really careful about minor syntactic issues when I have big semantic 
problems to deal with.

> The
> (IMO sad) fact of the matter is that fewer and fewer students are even
> taught Pascal before C, so I REALLY think diverging away from C syntax
> just because of some misguided notion that "we can do it better" is a BAD
> idea.  

There are lots of people who come to MUD languages with no experience 
in C.  Anyway, I really think C sucks for many-person collaborative 
projects, especially when the people building the system don't see 
each other face to face, and have no strong commitment to building 
reusable code.

> Currently, the places ColdC differs from C for the most part make
> logical functional sense. (Such as in the for construct, which is a true
> set operator rather then a simple ordinal iterator.)

Oh, cmon, there are a lot of differences from C already, starting 
with dynamic typing, and probably ending with the whole persistent 
object outlook.

Jay Carlson
nop@io.com    nop@ccs.neu.edu

Flat text is just *never* what you want.   ---stephen p spackman