Experimental Physics and
Industrial Control System

Ben Franksen <[email protected]> · Thu, 9 Sep 2010 19:02:17 +0200

On Donnerstag, 9. September 2010, Ralph Lange wrote:
> On 08.09.2010 18:57 Benjamin Franksen wrote:
> > On Donnerstag, 9. September 2010, Andrew Johnson wrote:
> >> I'll be on vacation until next Wednesday now.
> >>
> >> I don't like the keyword "scope", can you find some other word
> >> instead?  It has too much of a "computer-science theory"
> >> connotation. Even "block" would be better, but that's not very
> >> descriptive either.
> >
> > I'll try to come up with something that doesn't scare people off,
> > is reasonably short, and still conveys the meaning. Any suggestions
> > anyone?
>
> What about "group"?
>  From all the stupid ideas that I had, this sounded most applicable
> and comprehensive even to non-geek people.

It's a possibility. Others I have thought of are "var" and --now my 
favourite-- "let", as it has been used in many programming languages to 
mean "local definitions follow". (It is also familiar from mathematical 
texts as in "let x be defined as...", though I guess it's just geeks 
like me who view this as a helpful mnemonic.)

> BTW: I assume that every call to msi will implicitly be a group. 

Yes, in the sense that definitions given on the command line are visible 
throughout the whole substitution file.

> Will 
> - in addition to that - also each substitutions file have its own
> implicit group? (Making it work much like C file scope.)

I have not planned this, it seems this would re-introduce globals 
through the back door. See below.

> The crucial issue is how concatenating subst files and running msi on
> the result will differ from running msi on the same set of subst
> files. (That was a Diamond use case, iirc.)

Yes, which is the reason I would rather not allow that. The problem is 
that you can have defaults when referencing a macro, e.g. "$(X=abc)". 
This means it makes a difference whether the variable is defined or 
not. This is really something for which there is no equivalent in 
languages like C. I think you can do it in Perl with something like

  my x = defined(x) ? x : abc;

though I would have to try it out to be sure ;-)

One thing I would like to add instead is to treat every file block as if 
it contained an implicit let-block, i.e. allow definitions to come 
before any substitution instances (or the 'pattern' keyword). These 
would scope over all the substitution instances. I think it would be 
highly useful to be able to factor out common definitons for each 
instance of a certain template without having to enclose the whole 
thing in an extra let-block.

BTW, I am working on a formalization of the semantics and the abstract 
syntax by using Haskell as a specification language. Doing so promptly 
uncovered a few bugs in the spec. Yay!

Cheers
Ben
-- 
"Never confuse what is natural with what is habitual."
                                                 -- Mahatma Gandhi

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System