RE: [sv-ec] Mantis 2279 (randomize syntax feedback)

Shalom,

thanks for your quick and detailed response.

> 1. If all three are the same construct and the A.8.2 BNF is changed as
> in the proposal, then I don't see why the two syntax boxes in 
> Clause 17 should be inventing new non-terminals with new names
> and saying that they are not in Annex A. 

Fair enough, but that is not an artifact of 1858; it's been in
the LRM for some time.  We could take this opportunity to fix it.

> It's true that the scope_randomize in Syntax 17-10 is a reduced form
> that omits all the non-relevant parts, but the
> inline_constraint_declaration in Syntax 17-9 seems exactly the same
> except for the prefixed class_variable_identifier. I don't think that
> justifies a new syntax and a new name for the non-terminal.

I think I agree.  It would probably be better to duplicate the
relevant syntax from A.8.2 and G.4 rather than rewriting it.

> 2. The scope_randomize in Syntax 17-10 allows [std::]. The 
> BNF in A.8.2 does not seem to allow it with the prefix,
> only if you prefix it with [package_scope ] as tf_call
> allows you to do, but not via randomize_call.

I had overlooked the fact that package_scope cannot be 
applied to randomize_call.

I suppose that the original writers of the BNF thought that
the description in Annex G is sufficient to cover scope
randomize, and the randomize_call production in A.8.2
was intended only to apply to object.randomize.  But I
don't think this is good enough.  The syntax of std::randomize
cannot be explained-away as a function prototype, because of
its "with{}" syntax.  So I would be far happier to have
Annex G appeal to the randomize_call production in some way.
I know that is somewhat outside the spirit of std:: as a
package; but, until users can write their own subprograms 
that have a with{} construct, it is disingenuous to pretend
that std::randomize() is simply a function prototype that
can be isolated in Annex G.

> 3. The description of std::randomize() appears also in G.4 with yet
> another syntax:
...
> I'd like to see the syntax be consistent throughout.
> Note that scope_randomize in Syntax 17-10 also allows an empty
> variable_identifier_list whereas this syntax does not. 

Clearly it should, because - as you say -

> 17.12 says, "If
> the scope randomize function is called with no arguments, then it
> behaves as a checker and simply returns status." What does 
> "behave as a checker" mean? What status does it return?

Presumably it tests all its with{} constraints and returns
a true/false result?  For consistency with object.randomize
it would make sense to permit a "null" argument here.
 
> The syntax in G.4 also should not end in a semicolon, I think.

OK.  Again it's nothing to do with 1858, but this seems a good moment
to fix such things.
 
> 4. I think the two footnotes in the BNF in the proposal should be
> combined into a single footnote at the end of the production.

If you prefer it, that's fine by me - I thought it was easier to 
describe them if the footnote markers were attached to the bit of 
syntax they affect, but it's easy to change.

I'll upload an alternative proposal in the next few hours.
I'm concerned that we are widening the scope of 2279 and it 
may take longer than anticipated to reach a conclusion.
-- 
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL * Verilog * SystemC * e * Perl * Tcl/Tk * Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, Hampshire, BH24 1AW, UK
Tel: +44 (0)1425 471223                   Email: jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573                           Web: http://www.doulos.com

The contents of this message may contain personal views which 
are not the views of Doulos Ltd., unless specifically stated.

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.