Re: Grammar analysis

Point taken.

I note, for example, that Relax NG (a) defines explicit terms with the
required meanings, and (b) calls one of them 'notAllowed' rather than
(for example) 'empty-set'.

Michael

Graydon <graydonish@gmail.com> writes:

> On Sun, Aug 27, 2023 at 09:43:40AM -0600, C. M. Sperberg-McQueen scripsit:
>> Graydon <graydonish@gmail.com> writes:
>> > ∅ (U+2205 EMPTY SET) might be another character possibility.
>> 
>> Please no.  In some discussions of formal languages (at least), ∅ is
>> used to denote the empty language -- that is, the language that has no
>> sentences.  It is not the same as the language consisting of the empty
>> string '', which is often denoted ε (U+0385 GREEK SMALL LETTER EPSILON),
>> as in Norm's proposal.
>
> I was unaware!
>
> That would indeed make things confusing; suggestion withdrawn.
>
> [snip]
>> For me, part of the appeal of ixml is its relative simplicity.  Adding
>> keywords or symbols that do not change the expressive power of the
>> language is a slippery slope.  I won't say never, but I'm skeptical.
>
> I think "as simple as possible, but no simpler" is a supportable principle. 
>
> Having meaningful implicit nothingness is simple in a "language rules"
> sense but not simple in a "using the language" sense.  It creates a
> category of syntax errors which can't be reported by machine parsing
> because there's nothing mechanically wrong with the syntax even when
> there is a significant error in meaning.
>
> A rule for "No meaningful implicit nothingness" (which is how I
> understand Norm's proposal) strikes me as a large gain in
> utility. Such a rule need not include a standard representation of
> nothingness, fundamentally it just has to require the parser to throw
> an error.
>
> I would prefer a standard representation of nothingness, but I need the error.
>
> I use ixml in production to test that generated text conforms to
> agreed patterns of natural language.  I don't think the second time I
> do this the grammar will be more effort than the code that generates
> the text, but it isn't going to become trivial, either. "Making ixml
> easier to use" might not be a sensible goal—grammars are not on the
> list of things easy to use—but I think introducing this particular
> error condition is a large gain for small cost.
>
> -- Graydon


-- 
C. M. Sperberg-McQueen
Black Mesa Technologies LLC
http://blackmesatech.com

Received on Sunday, 27 August 2023 18:22:31 UTC