[sv-ec] Mantis 2715, ID#21: is string a singular or aggregate type?

Here's an attempt to summarize what the ballot draft LRM 
says about whether strings are aggregate or singular.

Clause 7.1 clearly includes strings in its list of 
aggregate types, but it's out on a limb - everywhere else,
aggregates are specified to be unpacked structs, arrays and
unions; everything else is singular.  Although strings look
somewhat like unpacked arrays of bytes, there is nothing 
that requires them to be so.

So let's see if it would be OK to make strings singular.
There are several places where a singular type is mandatory.
Taking each of these in turn...

1) The target of a $cast must be of a singular type.
~~~~
I can't see why one would ever need to $cast(str, something);
on the other hand, it's not inherently unreasonable.  This
doesn't sway the choice one way or the other.

2) Event expressions must evaluate to a singular value.
~~~~
Two out of three simulators accept @(str) and deal with
it in the obvious way; a third rejects it as non-singular.
I'm guessing that users would like to be able to do this.

3) The target of a force or a procedural assign must be singular.
~~~~
Although this isn't nice, I can easily imagine why users might
wish to force a string - perhaps to modify the flow of a 
testbench.  But I can see that there are implementation 
difficulties with forcing a dynamic-size variable, and 
no tool I tried supports it.  If we eventually decide that 
strings are singular, there will need to be a special 
restriction prohibiting force/assign for strings.

4) The left operand of "inside" must be singular.
~~~~
The LRM has examples of (str inside {"str1", "str2"}) but
tool support for it is patchy.  It's clearly something that
one would like to be able to do.

5) Subprogram arguments, and module ports, must be singular 
  if they are to have a default value.
~~~~
Even in the tool that treats strings as aggregates, string
arguments can be defaulted.  It's unimaginable to forbid it.

6) %p descends through an aggregate until it finds a singular.
~~~~
The LRM clearly states in 21.2.1.7 that a string is singular 
in this respect.  The only tool I could find that supports %p
exactly follows the LRM here, even though in most respects it
treats strings as being aggregates.


In summary, then: on balance it is clear to me that strings 
should be singular, and should be removed from clause 7;
it would be easy to put the text of 7.8 into 6.16, and to 
add a restriction outlawing force and procedural assign
to string targets.

Comments?
-- 
Jonathan Bromley
Consultant

Doulos - Developing Design Know-how
VHDL * Verilog * SystemVerilog * SystemC * PSL * 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                        http://www.doulos.com

--------------------------------------------------------------------------------
Doulos Ltd is registered in England and Wales with company no. 3723454
Its registered office is 4 Brackley Close, Bournemouth International 
Airport,
        Christchurch, BH23 6SE, UK. 

This message may contain personal views which are not the views of
Doulos, unless specifically stated.


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