Re: Logic Errors and Ada (Was: A big hairy thread on Ada, Quality, and Drivers)

[martin.dowie@btopenworld.com (Martin Dowie)]

Simon Wright <simon@pushface.org> wrote in message news:<x7v4r3q67v1.fsf@smaug.pushface.org>...
> My project is using 'subtype Radians is Long_Float' or equivalent,
> with constraints occasionally (though personally I don't now think
> that Radians can sensibly be constrained without a lot of work). Of
> course Latitude could be. The trouble is that the constraints won't
> catch that may errors (in UK waters, anyway).

True - but we tend to have whole-world-coverage type requirements.

It wouldn't be unusual to find something like:

   type A_Degree is digits 15;

   type A_Latitude is private;

   type A_Longitude is private;

   function To_Latitude (D : A_Degree)
      return A_Latitude;

   function To_Longitude (D : A_Degree)
      return A_Longitude;

private

   type A_Latitude is 
      new A_Degree range -90.0 .. 90.0;

   type A_Longitude is
      new A_Degree range -180.0 .. A_Degree'Pred (180.0);

in our code. Given that the algorithms are as old as the sun (e.g.
great circle) the number of times you actually need to do calculations
with both lat and long are fairly rare, so the type conversion just
are much of a hassle.


> I've always thought that the pain of introducing full dimensional
> handling (without language support) is unjustifiable, and that if

We only define what we need. As I mentioned at AdaUK, I've tried
greating *all* combinations before and the text editor just exploded!
:-)

> necessary one should rather go for records in this context:
> 
>   type Position is record
>      Lat  : Latitude;
>      Long : Longitude;
>   end record;
> 
> and inspect the low-level support subprograms.

...and a surprisingly similar...

   type A_Coordinate is record
      Lat  : A_Latitude;
      Long : A_Longitude;
   end record;