Re: generic queue packages

[mfeldman@seas.gwu.edu (Michael Feldman)]

In article <1993Feb26.031650.3447@leeweyr.sccsi.com> bill@leeweyr.sccsi.com (Bill Lee) writes:
>In article <1993Feb24.212641.8326@evb.com> jgg@evb.com (John Goodsen) writes:
>>wallace@cookie.enet.dec.com (Richard Wallace) writes:
>>
>>>Remember that a variant record will have the maximum size elements used when
>>>allocating the memory for the elements in your queue.  
>>>
>>
>>Not true.  It is implementation dependent.  An implmentation
>>*may* choose to use run-time logic and allocate the proper
>>size based upon the variant.
>>
>>Doesn't DEC Ada do just this ???
>
>Yes.
>
So does much-maligned Meridian.

Folks, what we have here is a classical time-space tradeoff. You can't
have it both ways. Either allocate the space statically, for the
maximum size, or do it dynamically and re-allocate every time the
size changes. (Or allocate in blocks, which is a compromise between
the two.)

This is almost an FAQ. The most intelligent way to deal with variant
records is to make sure that the discriminant has a reasonable type.

TYPE Stupid (Length: Positive := 8) IS RECORD
  String_Part: String(1..Length);
END RECORD;

is a recipe for big trouble. Yet I see it a lot, and not just in old
textbooks. Surely the writer didn't REALLY mean that the string
might possibly be Positive'Last characters long, right? Basically
the writer was too lazy to define a subtype that revealed the realistic
maximum length - 64 maybe, 255, but surely not Positive'Last. That's
32767 even if Integer is 16 bits.

SUBTYPE Realistic IS Integer RANGE 1..64; -- for example
TYPE Smart (Length: Realistic := 8) IS RECORD
  String_Part: String(1..Length);
END RECORD;

will waste at most 64 characters' worth, whether the compiler chooses
static or dynamic allocation. That's a whole lot better than the
roughly 2 gigabytes you'd try to allocate for a record of the first
type, if the compiler used static allocation and Integer was 32 bits.
(TeleSoft and Verdix do this, for example.) So it's (more) portable.

Variant records are a GREAT way to test intelligent design decisions.
Don't blame bad ones on the language or the compiler, friends.

Mike Feldman
------------------------------------------------------------------------
Michael B. Feldman
co-chair, SIGAda Education Committee

Professor, Dept. of Electrical Engineering and Computer Science
School of Engineering and Applied Science
The George Washington University
Washington, DC 20052 USA
(202) 994-5253 (voice)
(202) 994-5296 (fax)
mfeldman@seas.gwu.edu (Internet)

"The most important thing is to be sincere, 
and once you've learned how to fake that, you've got it made." 
-- old show-business adage
------------------------------------------------------------------------