An Ada Typing problem

An Ada Typing problem

[Marc.Graham]

I would like to determine if Ada can be made to handle NULL values in an 
elegant, efficient way. The problem arises in interfacing a database
language (namely SQL) and Ada.

In brief, I would like to define a type which properly includes all values
and operations of the predefined (package STANDARD, appendix F) type
INTEGER and also includes one new value: NULL. The rules of arithmetic
remain unchanged for values in INTEGER. The rule extending to NULL is: any 
operation involving the NULL value returns the NULL value.

I want the solution to have the following properties: any expression valid
for objects of type INTEGER is valid for objects of the new type. The
solution must be ***very efficient at run time*** and reasonable at compile
time. Run time efficiency includes: evaluation of arithmetic operations;
conversion costs.

For the conversion problem: Assume the DBMS delivers (and expects) values in 
the form of pairs <x,y> where y is an indicator and takes on some special 
value when the value represented is null; otherwise, the value represented
is the value of x.  ***Make no assumptions about x and y (e.g. length in
bits, encoding scheme, etc)*****  Can a solution requiring ****zero****
runtime conversion costs be produced?

Mail your solutions to me (marc@sei.cmu.edu) or, more interestingly, post
them here for comments.

An Ada Typing problem

[Marc.Graham]

Date: 2 Sep 1987 12:38:45-EDT

I would like to determine if Ada can be made to handle NULL values in an 
elegant, efficient way. The problem arises in interfacing a database
language (namely SQL) and Ada.

In brief, I would like to define a type which properly includes all values
and operations of the predefined (package STANDARD, appendix F) type
INTEGER and also includes one new value: NULL. The rules of arithmetic
remain unchanged for values in INTEGER. The rule extending to NULL is: any 
operation involving the NULL value returns the NULL value.

I want the solution to have the following properties: any expression valid
for objects of type INTEGER is valid for objects of the new type. The
solution must be ***very efficient at run time*** and reasonable at compile
time. Run time efficiency includes: evaluation of arithmetic operations;
conversion costs.

For the conversion problem: Assume the DBMS delivers (and expects) values in 
the form of pairs <x,y> where y is an indicator and takes on some special 
value when the value represented is null; otherwise, the value represented
is the value of x.  ***Make no assumptions about x and y (e.g. length in
bits, encoding scheme, etc)*****  Can a solution requiring ****zero****
runtime conversion costs be produced?

Mail your solutions to me (marc@sei.cmu.edu) or, more interestingly, post
them here for comments.