Fixed point attributes

Fixed point attributes

[brian.b.mcguinness]

I do not understand why Ada fixed point types have the Round attribute
but not Floor or Truncation.  Adding the later two would make fixed
point types more useful, and reduce the need for type conversions.

--- Brian

Re: Fixed point attributes

[Randy Brukardt]

<brian.b.mcguinness@lmco.com> wrote in message
news:1110998197.165590.63800@z14g2000cwz.googlegroups.com...
> I do not understand why Ada fixed point types have the Round attribute
> but not Floor or Truncation.  Adding the later two would make fixed
> point types more useful, and reduce the need for type conversions.

Ada fixed point types don't have "Round"; that's only available for decimal
types (intended primarily to be used for money). I don't know what the
thinking was for omitting Truncation for decimal types; perhaps Bob knows.

In general, you can't have attributes like that for fixed point types,
because there is no requirement that the a fixed point type can represent
integers. For instance, if you have:
    type Weird_Fixed delta 0.77 range 0 .. 100.0;
    for Weird_Fixed'Small use 0.77;

the only integers that can be represented exactly are 0 and 77. So what
would Round or Truncation mean for that sort of type? What is the result of
Weird_Fixed'Truncation (20.5)? The closest value smaller than 20 is 19.25.
That doesn't seem to be a sensible result to me.

This of course is a case where Ada's generality gets in the way. It would be
nice to have such attributes, but it would mean outlawing arbitrary small
values. I don't think smalls like this are used much, but it would be
incompatible, and there would be no real workaround if such smalls were
outlawed. (You'd have to switch to floating point, or write your own
pseudo-fixed package.) I think that would be too much of an incompatibility.

(Note that only allowing the attributes on some fixed point types would be a
generic contract model violation, so that isn't a viable solution, either.)

                       Randy.

Re: Fixed point attributes

[brian.b.mcguinness]

I see your point, but according to
http://www.grammatech.com/rm95html-1.0/rm9x-03-05-10.html,

---
  The following attributes are defined for every fixed point subtype S:

  ...
  S'Round
  S'Round denotes a function with the following specification:
  (13)
           function S'Round(X : universal_real)
             return S'Base

  (14)
  The function returns the value obtained by rounding X (away from 0,
if X is midway between two values of the type of S).
---
So I was thinking that if they could define Round they could define the
other two as well.

--- Brian

Re: Fixed point attributes

[Randy Brukardt]

<brian.b.mcguinness@lmco.com> wrote in message
news:1111066266.099494.162830@f14g2000cwb.googlegroups.com...
> I see your point, but according to
> http://www.grammatech.com/rm95html-1.0/rm9x-03-05-10.html,
>
> ---
>   The following attributes are defined for every fixed point subtype S:

Nope; it's only defined for decimal types. You missed 3.5.10(6):

The following additional attributes are defined for every decimal fixed
point subtype S:

And Round follows that.

BTW, a better looking version of the RM (which also includes the Corrigendum
updates) can be found at:

    http://www.adaic.com/standards/ada95.html

You can also find the working draft of Ada 2005 there (with some searching).

                   Randy.




> So I was thinking that if they could define Round they could define the
> other two as well.

Makes sense, but you misread the Standard.