"end of declaration"

"end of declaration"

[Mehdi Saada]

can someone explain in simple terms what are the rules for types' "end of declaration" ?

I define a type in the private part of the package, and in the body I clone it (type t_proba2 is new t_proba) to benefit from its predefined operations... since it's a fixed point type I need to, or bam ! recursion... also I want the accuracy etc to stay the same so I would do the same for floating point types too.

   type T_Proba is delta 0.00001 range 0.0 .. 1.0;
   for T_Proba'Small use 0.00001;

package body P_Proba2 is
      type t_proba2 is new T_Proba2;
-- type t_proba2 can't be use before the end of its declaration
   package Point_Fixe_io is new Ada.Text_Io.Fixed_IO(T_Proba);

so I rather defined a different type with the same specifications. ugly.
no nicer way to access predefined operations for these cases ?

Re: "end of declaration"

[AdaMagica]

0012...@gmail.com schrieb am Mittwoch, 10. Februar 2021 um 19:40:00 UTC+1:
> package body P_Proba2 is 
> type t_proba2 is new T_Proba2; 
> -- type t_proba2 can't be use before the end of its declaration

I don't unDerstand. DeclAration ends herE.
Why do you use a derived type?

Re: "end of declaration"

[Egil H H]

On Wednesday, February 10, 2021 at 7:40:00 PM UTC+1, 0012...@gmail.com wrote:
> 
> I define a type in the private part of the package, and in the body I clone it (type t_proba2 is new t_proba) to benefit from its predefined operations... 
What? The predefined operations are already visible in the body.

> since it's a fixed point type I need to, or bam ! recursion...
huh? 

> package body P_Proba2 is 
> type t_proba2 is new T_Proba2; 
> -- type t_proba2 can't be use before the end of its declaration 
This is because of a typo... Try
    type t_proba2 is new T_Proba; 
instead. Except that you don't need this derived type, so just delete the line.

> package Point_Fixe_io is new Ada.Text_Io.Fixed_IO(T_Proba); 
...and here you're using the original type anyway, so...

> no nicer way to access predefined operations for these cases ?
Just call them, they are visible...

-- 
~egilhh

Re: "end of declaration"

[Mehdi Saada]

@øþĸłµ¤££%¤µ]ßĐÐß»¢ĐŊГ¢Ŋ@ÐĐ@ßĐÆ#{#[~#{#Þ]!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

I hope it will stop happening...
otherwise, operations for t_proba are redefined in the visible (and body) part so recursion will occur.
But I saw what I should have done:
type t_proba is private;
private
type Nombre is ... with the representation clause
type t_proba is new nombre;

I'm ashamed I didn't come up with that before...

Re: "end of declaration"

[Shark8]

On Wednesday, February 10, 2021 at 11:40:00 AM UTC-7, 0012... wrote:
> can someone explain in simple terms what are the rules for types' "end of declaration" ? 
> 
> I define a type in the private part of the package, and in the body I clone it (type t_proba2 is new t_proba) to benefit from its predefined operations... since it's a fixed point type I need to, or bam ! recursion... also I want the accuracy etc to stay the same so I would do the same for floating point types too. 
> 
> type T_Proba is delta 0.00001 range 0.0 .. 1.0; 
> for T_Proba'Small use 0.00001; 
> 
> package body P_Proba2 is 
> type t_proba2 is new T_Proba; 
> -- type t_proba2 can't be use before the end of its declaration 
> package Point_Fixe_io is new Ada.Text_Io.Fixed_IO(T_Proba); 
> 
> so I rather defined a different type with the same specifications. ugly. 
> no nicer way to access predefined operations for these cases ?
Use more meaningful names if you can.
You had "Type K is new K", which is obvious nonsense.
Considering the context, I think something like
> type Probability is delta 0.00001 range 0.0 .. 1.0; 
> for Probability'Small use 0.00001;
and
> type Internal_Probability is new Probability;
would be much better readability-/maintainability-wise

Re: "end of declaration"

[Shark8]

> otherwise, operations for t_proba are redefined in the visible (and body) part so recursion will occur. 
What?
No, "Type J is new K" gives you a completely new type.
To use a function that requires a K-value with a J-value you have to explicitly convert; given "Steve : J" and "-"(Right: K) return K, you would have to use -(K(Steve)) to call the unary-minus for K.

Re: "end of declaration"

[Niklas Holsti]

On 2021-02-10 23:08, Shark8 wrote:
>> otherwise, operations for t_proba are redefined in the visible
>> (and  body) part so recursion will occur.
> What?
> No, "Type J is new K" gives you a completely new type.

Yes, but J inherits all primitive operations of K. And you can 
type-convert between J and K with no loss of information.

> To use a function that requires a K-value with a J-value you have to
> explicitly convert; given "Steve : J" and "-"(Right: K) return K, you
> would have to use -(K(Steve)) to call the unary-minus for K.
If the "-" function is a primitive operation of K, there is an identical 
inherited operation for J. If the "-" with a K-parameter is not 
primitive for K, it is not inherited by J, and a J-value j must be 
converted to a K-value before that "-" is called: - K(j).

You can override the inherited "-" with a new operation, specific to J. 
Within that operation, calling "-" on a J-value j would make it 
recursive. The K-operation can be called with - K(j) and that will not 
be a recursive call.