[Date Index][Thread Index]
[Date Prev][Date Next][Thread Prev][Thread Next]

Behaviour configuration of slice (was: Re: Multi Language Support Question)

Denis Barbier wrote:
> > -p: like -o, but write output file only, if all *p*ositive (+) slices
> >     in the slice term exist (to substract something, that doesn't
> >     exist, doesn't hurt :-)
> [...]
> But how to deal with
>      A-(B+C)
> and
>      A-B-C
> if C is undefined?
> I do not find a good definition for `positive' slice.

Hmmm, I thought of every slice, which will add text to the overall
result, so in the above cases both will write the file, because C is
substracted in both cases. 

But I see that this definition is not very precise and may leed to

I got this idea when working with slices for meta files. I wanted
something, that is robust against SLICE-META_* if META_* doesn't
match, because I don't need meta files in all cases.

Due to my and Karl-Heinz Marbaise's examples were of the kind
ALL-FOO_*, I forgot, that SLICE+FOO_* could also happen. So my idea of
"positive" is just one of two cases.

Well, although there may be cases where expicit mentioned slices do
not exist, the main problem in our example seem wildcard slices, that
do not match. So we perhaps should focus on them for solving that
problem, instead of my initial approach with 'positive slices'...

Hmmm, now I managed to confuse myself.

OK, we better start at the beginning: What do we need?

Defining the behaviour of slice if
  a) a mentioned slice doesn't exist,
  b) a mentioned slice wildcard doesn't match and
  c) the whole output file just contains (or would contain) only
     white spaces or the empty string

(BTW: Should case (a) include case (b)?)

I each case, we have at least two possible behaviours:
  1) write the file nevertheless, setting all undefined slices to the
     empty string.
  2) Don't write the file at all.

In the first two cases (and perhaps also in third case), we have a
third possibility:
  3) Exit with error message.

To get the most powerful solution, we have to assign the behaviours
1-3 to the cases a-c, so single on/off flags won't work. :-(

P.S.: Hope, the rest of the list doesn't mind my brain storming on
this case... :-)

		Regards, Axel
Axel Beckert - abe@cs.uni-sb.de - http://abe.home.pages.de/
Student of Computer Science, University of Saarland (Germany)
Artificial Intelligence Laboratory (AI Lab), Prof. Dr. W. Wahlster;
WWW-/FTP-Administrator IBFI Schloß Dagstuhl; Students Representative CS
Website META Language (WML)                www.engelschall.com/sw/wml/
Official Support Mailing List                   sw-wml@engelschall.com
Automated List Manager                       majordomo@engelschall.com