Conference turris::languages

    It depends on who you ask (I believe C has different meanings),
    but from the VAX Pascal perspective,
    
    o	casting is an explicit request to re-interpret a set of
    	bits (which are the representation of a particular value
    	of a particular data-type) as being a representation from
    	a different data-type, e.g.  You put 3.14159 into a float
    	and then view it as a character string (and get nonsense);
    
    o	coercion is the implicit conversion of one data-type to 
    	another in such a manner that the "meaning" is maintained,
    	e.g. you put 3 into an integer and then multiply it by
    	2.1 (a float), so the the 3 is coerced into 3.0.  Note
    	that the bit pattern representing 3.0 can be quite different
    	from that which represents 3.
    
    		-John Bishop

Re .0:

I agree with your definitions, for both C, and in general.


Re .1:

I daresay the definition of "casting" you attribute to the VAX Pascal
perspective is commonly known as "representational viewing".  Also, no decent
context comes to my mind where it has that definition you give.  (Ask me what
indecent context in which it arose).  Got any good references?
				/AHM/THX

    _Programming_in_VAX_PASCAL_, section 7.2.6, p 7-12:
    
    "...the actual representation of the object being cast is never
    altered by the type cast operator."
    
    There's an example of casting a float as a record to look at the
    mantissa, etc.
    
    			-John Bishop

Re .3:

Thanks, that's as horrifically authoritative as one could want.  Pray tell, is
this feature (or nomenclature) part of any ISO Pascal standard, or adopted
from another vendor's Pascal, or is it a Digital extension?
				/AHM/THX

The VAX Pascal concept of a "cast" came from an early definition of
Ada.  There are "casts" and there are "conversions".  Conversions are
what programmers usually do; casts are what magicians usually do.  In
a conversion, the interpretation of the value never changes although
the representation may change.  In a cast, the representation never
changes although the interpretation of the value may change.  (Note
that sometimes programmers are also magicians when they need to cast a
spell over a particularly recalcitrant piece of data).

The VAX Pascal use of "cast" is in conflict with the Algol-68 use of
the term.  In Algol-68, a cast always had the semantics of a
conversion of some sort.  In C, the term cast is not used uniformly.
In some cases, a cast in C will change the representation without
changing the value while in other cases a cast in C will change the
value without changing the representation.  (On the other hand, C is a
language for magicians whose spells don't always work).

    See also BLISS' MAP declaration.  Just a "cast" with a scoped syntax.
    
    				-John

    Also, Apollo's Pascal has real "casts" instead of "coercions".
    
    				-John

Re .5:

>... in other cases a cast in C will change the value without changing the
>representation.

Offhand, the only such cases I can recall all involve overflow, and are thus
*not* defined by K&R, or ANSI C.
				/AHM/THX