Personally I'm not married to the current grammar, but we do need a plausible alternative, and we need it soon.

An alternative syntax needs to have:

- an obvious terminator (a comma is not that, especially for assert)

- somewhere to do the preamble (pre:, post r:, assert: )

- not be current valid syntax.

Thinking out loud here, but could braces work?

auto f(auto const x, int const y) noexcept -> void

    requires integral<decltype(x)>

    pre {

         x > 0;

         y > x;

    }

    pre(axiom) {

        {x % 2 == 0};

    }

    post(audit, new) [r=return, x, y] {  // seems like lambda-like captures fit in

        {r % x == 0};

        {r % y == 0};

    }

{

/* function body */

   assert {

      { x > 0 };

   };

}

That looks pretty much like the grammar for the "requires" expression except with runtime values. Hm.....

From reading the discussion, this seems to solve some real concerns:

• Per Ville's observation, it offers a solution for the concerns about non-const parameters.
• It avoids the concern about the syntax looking like an attribute while not technically being one.

Additional benefits (in my opinion)

• The syntax is reminiscent of function-try-blocks (though, since they are so rarely used, that probably has little appeal for non-enthusiasts).
• It borrows from Concepts in a meaningful way. One could argue that
    pre { x > 0; }
  is the run-time equivalent of:
    requires { requires x > 0; }
  Perhaps it would be worthwhile to extend the requires clause directly to encapsulate Contract declarations; that would avoid the need for context sensitive pre and post keywords.
  
• Extension points are clear and won't require new potentially context sensitive keywords to introduce new contract categories like axiom and audit.
  

What are the cons?

• This is new syntax relative to what has been discussed for Contracts in the past; prior education is lost (this may be a good thing if what gets standardized deviates significantly from semantics associated in prior efforts).
• Others?

With regard to function-try-blocks, I would expect that try would have to follow any contract declarations.

void x::f(int x)
  pre { x > 0; }
try : member(x) {}
catch (...) {}

Tom.

On Wed, Sep 22, 2021 at 5:48 PM Aaron Ballman via SG21 <sg21@lists.isocpp.org> wrote:

On Wed, Sep 22, 2021 at 11:06 AM Jens Maurer via SG21
<sg21@lists.isocpp.org> wrote:
>
> On 22/09/2021 12.19, Andrzej Krzemienski via Liaison wrote:
> >
> >
> > śr., 22 wrz 2021 o 09:42 Ville Voutilainen via Liaison <liaison@lists.isocpp.org <mailto:liaison@lists.isocpp.org>> napisał(a):
> >
> >     On Wed, 22 Sept 2021 at 00:02, Jens Maurer via Liaison
> >     <liaison@lists.isocpp.org <mailto:liaison@lists.isocpp.org>> wrote:
> >     > (Personally, I think contracts should be a first-level
> >     > language feature that should not be hidden inside an
> >     > attribute-looking syntax atrocity.  At least in C++,
> >     > the space where they are does allow for context-sensitive
> >     > keywords without much hassle; cf. override and final.)
> >
> >     Well, yeah.. if we were to entertain a contract declaration preceding
> >     the decl-specifier,
> >     so that it could do forward-lookup for the parameters, then the case
> >     for an attribute-like
> >     syntax would be relatively strong. If the contract declaration has to
> >     appear where context-sensitive
> >     keywords appear, then why not use a context-sensitive keyword?
> >
> >
> > The current contracts proposal (P2388R2 <http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2021/p2388r2.html>) as well as C++20 contracts offer(ed) three new annotations:
> >  * preconditions and postconditions that appear in function declarations,
>
> ... in a place where we have good experience with context-sensitive
> keywords.
>
> >  * an assertion that appears in the function body.
> >
> > This third kind is in the position of a regular statement inside a block scope, so a context-sensitive keyword will not work; unless we were to drop these assertions or treat them in a different, irregular way.
>
> Well, we could do
>
>   pre: conditional-expression
>   post identifier: conditional-expression
>   assert: conditional-expression  // in a block
>
> and the only potential conflict would be with a user-defined label "assert".
>
> (The colon should be a pretty good disambiguator for pre and post;
> if that's not enough, we can require a logical-or-expression and/or
> add a comma to separate contracts in declarations.)
>
> > Also a natural future extension of the currently proposed contracts is class invariants, which would also appear as a declaration in class-scope, so no room for context-sensitive keywords.
>
> With the colon, I'm not seeing a serious problem.

This sort of syntax would alleviate my personal concerns with the
proposed syntax.

In terms of whether this would be palatable to WG14, it's a bit less
clear. C has no context sensitive keywords and I have no memory of
WG14 discussions to add any, so I don't have much experience to draw
from. However, I think context sensitive keywords are a reasonable
idea worth seriously exploring as it seems at least plausible (we have
plenty of implementation and usage experience with override and final
in C++ that we can point to as prior art that's mildly C related, but
if anyone knows of a context-sensitive keyword in a C implementation,
that would strengthen the case in WG14 greatly).

~Aaron

>
> Jens
>
> _______________________________________________
> SG21 mailing list
> SG21@lists.isocpp.org
> Subscription: https://lists.isocpp.org/mailman/listinfo.cgi/sg21
> Link to this post: http://lists.isocpp.org/sg21/2021/09/1212.php
_______________________________________________
SG21 mailing list
SG21@lists.isocpp.org
Subscription: https://lists.isocpp.org/mailman/listinfo.cgi/sg21
Link to this post: http://lists.isocpp.org/sg21/2021/09/1216.php

_______________________________________________
SG21 mailing list
SG21@lists.isocpp.org
Subscription: https://lists.isocpp.org/mailman/listinfo.cgi/sg21
Link to this post: http://lists.isocpp.org/sg21/2021/09/1234.php