Re: RE: Multiple parms to procedures

["David Morris" <dmorris@xxxxxxxxxxxxx>]

Walden,

There are restrictions on how you can use a CONST parameter.  You cannot 
retrieve the address is one of them.  Say I want to pass that constant to 
another routine and I don't know, or care what the type is.  The strict type 
checking prevents me from passing it directly.  So I have to pass the address 
which I can't retrieve.  We use a lot of black-box functions that use CONST 
wherever possible.  I would like to have the benefit of a CONST without the 
restrictions.  I can simulate a protected memory space using authority, so why 
can't the compiler do it?  The compiler has access to a lot lower level 
functions and could do this more efficiently.

I agree that compile time checking is best.  Saves time.  The compiler could 
still return an error with a low severity if a line of code could potentially 
change a CONST variable.  Similar to the way a CALLP to a procedure that 
returns a value that I don't need to use.

I also agree the compiler should use the most optimal method possible.  For 
example the compiler allows logical record blocking if certain conditions are 
met.  You even get feedback from the compiler telling you when blocking can be 
used.  You can use that feedback to tune your application if it is important to 
you.  I would want the same type of feedback for constants if the 
implementation affected performance.

Another alternative would be to create a new keyword.  I don't care for CONST 
anyway.  Something like PROTECTED.

Thanks,

David Morris

>>> Walden Leverich <walden@techsoftinc.com> 07/22 8:07 AM >>>
OK, I'll bite, why do you want the CONST parms protected "as if they had
a debug WATCH ensuring they are not changed rather than having the
compiler evaluate source statements"? 

It sounds to me like you are saying you want the checking done at
runtime, not at compile time. Why? Compile time checking will find all
errors when I have the code available to edit, runtime will only find it
when the program blows up (hopefully in testing). 

Also, compiler based checking also allows some optimizations that may
not be possible if the compiler can't assume the const-ness of the
parameters.

-Walden


+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---