choice of fixed point type

choice of fixed point type

[tmoran]

Why does
    type duration is delta 1.0E-13 range 0.0 .. 0.000_1;
    f : float;
    i : integer := 10_000;
begin
    f := float(0.000_0001/i);
    ada.text_io.put_line(float'image(f));
    f := float(duration'(0.000_0001)/i);
    ada.text_io.put_line(float'image(f));
    f := 0.000_0001/float(i);
    ada.text_io.put_line(float'image(f));
print
 0.00000E+00
 9.94760E-12
 1.00000E-11

   Is <universal real> / <integer> always treated as a call on
function "/"(Left : Standard.Duration; Right : Integer)
return Standard.Duration;  ?

Re: choice of fixed point type

[tmoran]

I asked:
>   Is <universal real> / <integer> always treated as a call on
>function "/"(Left : Standard.Duration; Right : Integer)
>return Standard.Duration;  ?
  It turns out that was merely a problem with a single compiler.
Three out of four compilers said the division was ambiguous.
  But, in the absence of any other fixed point types, Ada accepts
    Long_Float_Functions.Cos(Long_Float(Ada.Numerics.Pi/integer(i)))
and calculates pi/i as type Duration, then converts that to
a Long_Float and then takes the cosine.  That typically means
the argument is not nearly as accurate as the person using
Long_Float would expect, and of course neither is the resulting cos.

Re: choice of fixed point type

[Richard D Riehle]

In article <9vTv4.603$zz5.39941@news.pacbell.net>,
	tmoran@bix.com wrote:

>I asked:
>>   Is <universal real> / <integer> always treated as a call on
>>function "/"(Left : Standard.Duration; Right : Integer)
>>return Standard.Duration;  ?
>  It turns out that was merely a problem with a single compiler.
>Three out of four compilers said the division was ambiguous.

Keeping us in suspense, Tom?  Which three gave you the good information?  Which
one got it wrong?   Eager minds want to know.  :-)

Richard Riehle
 

Re: choice of fixed point type

[tmoran]

>Keeping us in suspense, Tom?  Which three gave you the good information?  Which
>one got it wrong?   Eager minds want to know.  :-)
  Given:
with Ada.Text_Io;

procedure test is

  type duration is delta 1.0 range 0.0 .. 60.0;
  f : float;
  i : integer := 3;

begin
  f := float(16.0/i);
  ada.text_io.put_line(float'image(f));

end test;

  Gnat 3.12p NT gave no error message and used, apparently,
Standard.Duration

  ObjectAda 7.1.2.205 said
--------------------Target: Win32 (Intel) Debug--------------------
test.adb: Error: line 10 col 18 LRM:8.6(31), Expression is ambiguous when interpreted as an expression of any, Returning first interpretation

test.adb: Error: line 10 col 8 The expression is ambiguous, Continuing

Front end of ..\examples\test.adb failed with 2 errors.
Tool execution failed.

    "David W. Glessner" <dwglessn@collins.rockwell.com> tells me:
Sure 'nuff, Rational Apex 3.0.0b doesn't compile your program.

08:55:32 >>> Line 7: F := Float (0.000_0001 / I);
08:55:32 *** 0.000_0001 / I is ambiguous
08:55:32 *** "/" could be (in TEST_FIXED_DURATION)
Test_Fixed_Duration.Duration."/"
08:55:32 *** "/" could be (in STANDARD) "/"

  Janus 3.1.1e NT said:
In File F:\test.ada at line 10
--------------
    9:  begin
   10:    f := float(16.0/i);
-------------------------^
*ERROR* Unable to resolve expression (6.4.6) [RM 5.2(4)]

Re: choice of fixed point type

[Robert Dewar]

In article <S_iw4.2837$im1.54677@news.pacbell.net>,
  tmoran@bix.com wrote:

>   Gnat 3.12p NT gave no error message and used, apparently,
> Standard.Duration

As one would expect, since this type is most certainly
visible. However, if a compiler has other fixed point types
visible (certainly permissible), then it would be ambiguous.
The code in question is really non-portable at best, wrong
at worst, but not illegal if there is only one fixed-point
type visible. At least that's my reading.



Sent via Deja.com http://www.deja.com/
Before you buy.

Re: choice of fixed point type

[David W. Glessner]

In article <89v3nq$rs7$1@nnrp1.deja.com>,
  Robert Dewar <robert_dewar@my-deja.com> wrote:
> In article <S_iw4.2837$im1.54677@news.pacbell.net>,
>   tmoran@bix.com wrote:
>
> >   Gnat 3.12p NT gave no error message and used, apparently,
> > Standard.Duration
>
> As one would expect, since this type is most certainly
> visible. However, if a compiler has other fixed point types
> visible (certainly permissible), then it would be ambiguous.
> The code in question is really non-portable at best, wrong
> at worst, but not illegal if there is only one fixed-point
> type visible. At least that's my reading.

But doesn't Tom's example have two fixed-point types visible?
Namely, Standard.Duration, and Test.Duration.

(I thought GNAT 3.12p compiled, without warning, the example
even when "type Duration is delta ..." was replaced with
"type T is delta ...".  I don't have access to 3.12p right
now, so I can't verify that.)

--
David


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: choice of fixed point type

[tmoran]

>(I thought GNAT 3.12p compiled, without warning, the example
>even when "type Duration is delta ..." was replaced with
>"type T is delta ...".  I don't have access to 3.12p right
>now, so I can't verify that.)
  Using "type D is delta ..." also fails to generate any warning.

>> The code in question is really non-portable at best, wrong
>> at worst, but not illegal if there is only one fixed-point
>> type visible. At least that's my reading.
  The code in question arose from testing around an "Ada
surprise".  An execution error had been traced to where a
programmer had written something like
  Long_Float(N * Ada.Numerics.pi)   instead of
  Long_Float(N) * Ada.Numerics.pi
resulting in the use of (relatively imprecise) type Duration
for the multiply instead of Long_Float.  Standard.Duration
was of course (silently) visible, and there was no other
fixed point type in this code, so the compiler did what the
programmer said, but not what he wanted.