Ada reference manual rule for doubled parentheses around if_expression?

Ada reference manual rule for doubled parentheses around if_expression?

[Stephen Leake]

A question recently came up on the Emacs ada-mode list; my Emacs
ada-mode parser rejects this if_expression in an association list:

   Tmp : Boolean := Id (if True then True else True);

but accepts this:

   Tmp : Boolean := Id ((if True then True else True));

I remember seeing some rule in the Ada reference manual that says the
doubled parens are not needed here. But now I can't find the rule, and I
don't want to edit my grammar without a reference.

Anyone know where that rule is? or what to search for?

-- 
-- Stephe

Re: Ada reference manual rule for doubled parentheses around if_expression?

[Bob Duff]

Stephen Leake <stephen_leake@stephe-leake.org> writes:

> A question recently came up on the Emacs ada-mode list; my Emacs
> ada-mode parser rejects this if_expression in an association list:
>
>    Tmp : Boolean := Id (if True then True else True);
>
> but accepts this:
>
>    Tmp : Boolean := Id ((if True then True else True));
>
> I remember seeing some rule in the Ada reference manual that says the
> doubled parens are not needed here. But now I can't find the rule, and I
> don't want to edit my grammar without a reference.
>
> Anyone know where that rule is? or what to search for?

4.5.7(7).  This is a syntax rule, despite not being written in BNF.
I suggest you read the AARM annotations.

- Bob

Re: Ada reference manual rule for doubled parentheses around if_expression?

[Simon Wright]

Bob Duff <bobduff@theworld.com> writes:

> Stephen Leake <stephen_leake@stephe-leake.org> writes:
>
>> A question recently came up on the Emacs ada-mode list; my Emacs
>> ada-mode parser rejects this if_expression in an association list:
>>
>>    Tmp : Boolean := Id (if True then True else True);
>>
>> but accepts this:
>>
>>    Tmp : Boolean := Id ((if True then True else True));
>>
>> I remember seeing some rule in the Ada reference manual that says the
>> doubled parens are not needed here. But now I can't find the rule,
>> and I don't want to edit my grammar without a reference.
>>
>> Anyone know where that rule is? or what to search for?
>
> 4.5.7(7).  This is a syntax rule, despite not being written in BNF.  I
> suggest you read the AARM annotations.

The question to which Stephe refers also shows

   function F2 (V : Integer) return Boolean
   is (True and then True)
      or else True;

which GNAT accepts but probably shouldn't? the parens in ARM 6.8(2)[1]
wrap the whole expression.

[1] http://www.ada-auth.org/standards/12rm/html/RM-6-8.html#p2

Re: Ada reference manual rule for doubled parentheses around if_expression?

[Stephen Leake]

Bob Duff <bobduff@theworld.com> writes:

> Stephen Leake <stephen_leake@stephe-leake.org> writes:
>
>> I remember seeing some rule in the Ada reference manual that says the
>> doubled parens are not needed here. But now I can't find the rule, and I
>> don't want to edit my grammar without a reference.
>>
>> Anyone know where that rule is? or what to search for?
>
> 4.5.7(7).  This is a syntax rule, despite not being written in BNF.
> I suggest you read the AARM annotations.

Thanks.

There's a similar rule for quantified expressions in 4.5.8(4).

GNAT also accepts this:

   function Expression_Function_1 (V : Integer) return Boolean
     is (V = V and then True)
          or else True;

I think that's a bug, because there are no parens surrounding the
entire expression. 

-- 
-- Stephe

Re: Ada reference manual rule for doubled parentheses around if_expression?

[Bob Duff]

Stephen Leake <stephen_leake@stephe-leake.org> writes:

> Bob Duff <bobduff@theworld.com> writes:
>
>> Stephen Leake <stephen_leake@stephe-leake.org> writes:
>>
>>> I remember seeing some rule in the Ada reference manual that says the
>>> doubled parens are not needed here. But now I can't find the rule, and I
>>> don't want to edit my grammar without a reference.
>>>
>>> Anyone know where that rule is? or what to search for?
>>
>> 4.5.7(7).  This is a syntax rule, despite not being written in BNF.
>> I suggest you read the AARM annotations.
>
> Thanks.

You're welcome.

> There's a similar rule for quantified expressions in 4.5.8(4).

Right.

> GNAT also accepts this:
>
>    function Expression_Function_1 (V : Integer) return Boolean
>      is (V = V and then True)
>           or else True;
>
> I think that's a bug, because there are no parens surrounding the
> entire expression. 

I agree that's a bug.  It doesn't seem to have anything to do with the
special rules about conditional/quantified expressions.  Or does it?

If you report this to report@adacore.com, it will get fixed.

- Bob

Re: Ada reference manual rule for doubled parentheses around if_expression?

[Randy Brukardt]

"Bob Duff" <bobduff@theworld.com> wrote in message 
news:87mvybktvt.fsf@theworld.com...
> Stephen Leake <stephen_leake@stephe-leake.org> writes:
>
...
>> GNAT also accepts this:
>>
>>    function Expression_Function_1 (V : Integer) return Boolean
>>      is (V = V and then True)
>>           or else True;
>>
>> I think that's a bug, because there are no parens surrounding the
>> entire expression.
>
> I agree that's a bug.  It doesn't seem to have anything to do with the
> special rules about conditional/quantified expressions.  Or does it?
>
> If you report this to report@adacore.com, it will get fixed.

Looks like a bug to me, too. I remembered that we had changed something 
about these, but what it was was to accept an aggregate as well as a 
parenthesized expression:

    function Expression_Function_1 (V : Integer) return Rec is (A => 1, B => 
V);

rather than

    function Expression_Function_1 (V : Integer) return Rec is ((A => 1, B 
=> V));

This was another case of double paren elimination. But of course we don't 
have any rules that eliminate all parens!

                                  Randy.

Re: Ada reference manual rule for doubled parentheses around if_expression?

[EGarrulo]

On Friday, July 31, 2015 at 7:00:13 PM UTC+2, Stephen Leake wrote:
> A question recently came up on the Emacs ada-mode list; my Emacs
> ada-mode parser rejects this if_expression in an association list:
> 
>    Tmp : Boolean := Id (if True then True else True);
> 
> but accepts this:
> 
>    Tmp : Boolean := Id ((if True then True else True));
> 
> I remember seeing some rule in the Ada reference manual that says the
> doubled parens are not needed here. But now I can't find the rule, and I
> don't want to edit my grammar without a reference.
> 
> Anyone know where that rule is? or what to search for?

The Rationale says it in the "1.3.2 Expressions" chapter.