an entity with signature NONE. 
Author Message
 an entity with signature NONE.

When an entity [ except functions with side-effects ] is defined with
signature NONE, it seems that its value becomes known at compile time
(Void) and that access to the entity can be optimized.

I assume the implementation only contains one object of type NONE and
that this is the object returned by the Void feature of class
GENERAL. Any creation instructions of type NONE should attach this
object to the entity. [ This is AFAIK implementation-defined - but
this seems like a sane thing to do - having only one instance of

Since any reference is initially assigned the 'Void' object, an entity
of type NONE cannot be attached any other object than this object,
even if it is left uninitialized.

Since we know the type of the entity at compile-time, we know the
value of the entity _at compile time_ [ this knowledge is already used
by compilers to optimize `if my_entity = Void then..' - type of
statements ]. Knowing information at compile-time means that we don't
have to use storage as run-time.

So what?

This knowledge can be used to save resources where entities normally
need to be stored. The most important case is object instances,
but parameters and the Result entity are also interesting uses.

 * Redefining the signature of an attribute to NONE means that the
   implementation can remove the attribute from the memory image for
   the object, thus reducing the memory footprint. This means that the
   size of an object doesn't necessarily have to increase as we
   descend a class hierarchy.

 * Redefining a parameter to NONE means the implementation don't have to
   include it in the call sequence.

 * Redefining the result type of a function to NONE means the
   implementation doesn't have to return a result [ the call-sequence
   effectively becomes as if it was a procedure ], and if the function is
   without side-effects, the whole call can be removed and no code
   need to be generated for the function [ this has the same debugging
   consequences as inlining though - setting a breakpoint in an
   inlined feature can be difficult ].


Sat, 13 Nov 1999 03:00:00 GMT  
 an entity with signature NONE.

I think duplicating the methods of the spammers is a bad idea:

People who want to talk to you just get frustrated, and maybe your
variations of wrong EMail addresses get into some patch databases,
changelogs, bug-reports, or whatever. Up to now I either (spammers
read this!) delete messages with forged from-addresses or well-known
spam-addresses, or I take the time to complain if the address seems to
be valid. I envision the day, when all the from addresses will be
validated and thus valid.

Up to now I can resist changing my address to "...uni-regensburg.despam"

> Newsgroups: comp.lang.eiffel
> Date: 27 May 1997 16:59:20 +0200
> Organization: Guardian Networks AS
> Lines: 46
> NNTP-Posting-Host:
> X-Newsreader: Gnus v5.4.37/XEmacs 19.15


Tue, 16 Nov 1999 03:00:00 GMT  
 [ 3 post ] 

 Relevant Pages 

1. None not None in Excel COM?

2. PythonCOM: com_error: (-2147418113, 'Catastrophic failure', None, None)

3. Proposal: min(None, x) and max(None, x) return x

4. SRNC 2.001: Dynamic Signatures

5. Signature Revealing Naming Convention (for ST)

6. Why session beans and entity beans?

7. SRNC 2.001: Dynamic Signatures

8. Electronic Signatures

9. Help needed with entity classes...

10. Interface Signatures -- an alternative to Interfaces?

11. Naming Oracle entities within Clarion

12. Scanning and Signature Capturing


Powered by phpBB® Forum Software