Re: Oberon and Wirthian languages

[Simon Clubley <clubley@remove_me.eisner.decus.org-Earth.UFP>]

On 2014-04-24, Randy Brukardt <randy@rrsoftware.com> wrote:
>> On 14-04-24 21:51 , Simon Clubley wrote:
> ...
>>> It will be interesting to see if Randy and company have any comments on
>>> this syntax.
>
> Not really, other than that someone needs to come up with syntax that has 
> the right "feeling". None of the proposals I've seen have that yet, but keep 
> trying.
>

Ideas on what you think is wrong with the current suggestions will be
useful.

Do you have some general approach in mind which you would prefer ?

For example, do you think it should be done as part of an assignment
statement as Niklas is suggesting, or something more Pascal like as
I was thinking of ?

I would not want to see something fully Pascal like, with random
statements within the with, but something directed towards updating
the variable only. I also now think the atomic on the statement may
be redundant as it can be made a part of the variable definition using
the atomic pragma.

My revised version looks like this:

	update uart0.config1 using
		.flag1	:= 0;
		.enable	:= 1;
		.txdis	:= 0;
	end update;

I think the leading dot is required to positively identify something as
a member of config1 (in this case) and avoid conflicts between variable
definitions.

However, I'm also tempted to ban the use of config1 component references
on the RHS of the assignment. This avoids issues, if the component is
modified within the update statement, around whether the component value
used on the RHS is the value on entry to the update statement or the
modified value.

There would also have to be wording along the lines of "An atomic variable
shall be read exactly once and written exactly once by the update statement.
The compiler must reject the update statement if this cannot be achieved."

Although my current preference is for something like the above, as an
alternative, would you be open to a new assignment operator which was
used for "updates" instead of "becomes" semantics ?

Something along the lines of:

	uart0.config1 {updates_operator} (flag1 => 0, enable => 1, txdis = 0);

or

	uart0.config1'Update := (flag1 => 0, enable => 1, txdis = 0);

although I'm not sure what characters would be used for the updates_operator.

Once again, the atomic update would be enforced by the atomic pragma.

Simon.

-- 
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world