Re: Attribute markings - a question

It may be just early in the morning and the coffee hasn't yet kicked in, but I don't see the problem.


I checked in my implementation, making the grammar unambiguous in the process:



 S : @able, baker, @charlie.
 able: string.
 baker: string.
 charlie: string.
 string: ["abc"]*, ".".


Input:
 aaa.bbb.ccc.


Result:
 <S able='aaa.' charlie='ccc.'>
    <baker>
       <string>bbb.</string>
    </baker>
 </S>


Which was what I was expecting.


So assuming I'm not missing something obvious, I suspect that you need to reread the serialisation section of the spec:


"
A nonterminal attribute is serialised by outputting the name of the node as an attribute, and serialising all non-hidden terminal descendants of the node (regardless of marking of intermediate nonterminals), in order, as the value of the attribute.
"
which I think covers what you are asking for.


The other side of this coin is:


"
A nonterminal element is serialised by outputting the name of the node as an XML tag, serialising all exposed attribute descendants, and then serialising all non-attribute children in order. An attribute is exposed if it is an attribute child, or an exposed attribute of a hidden element child (note this is recursive).
"


Steven

On Wednesday 13 October 2021 04:19:52 (+02:00), C. M. Sperberg-McQueen wrote:

> Consider the grammar
>
> S : @able, baker, @charlie.
> able: string.
> baker: string.
> charlie: string.
> string: ~[]*.
>
> Is this grammar OK? (Yes, it’s hopelessly ambiguous, but that’s beside the point.)
>
> If we ignored the annotations, a raw parse tree for this grammar might look like this:
>
> <S>
> <able mark=“@"><string>aaa</string></able>
> <baker><string>aaa</string></able>
> <charlie mark=“@"><string>ccc</string></able>
> </S>
>
> Note that ‘string’ is implicitly marked serializable (^).
>
> When a nonterminal marked to be serialized as an element appears as a child of a nonterminal marked to be serialized as an attribute (as ’string’ here appears as a child of @able and @charlie), is the rule
>
> - Raise an error because the grammar cannot be serialized that way?
>
> - Omit the content of ’string’ from the value of @able and @charlie by analogy with what happens when we calculate the text node children of an element?
>
> - Ignore the marking on ’string’ on the grounds that we have already been told that @able is an attribute. Since elements cannot appear within attributes, the implicid ^ marking on ’string’ is ignored.
>
> The grammar for ixml offers two examples that seem relevant: in a raw parse tree, @name will dominate nodes labeled namestart and namefollower, which are explicitly marked non-serializable (-). @dstring and @sstring similarly dominate nodes labeled dchar and schar, which are implicitly marked ^. The attributes @from and @to directly dominate nodes labeled ‘character’ (marked -) and indirectly dominate nodes labeled ‘dchar’ and ’schar’ (implicitly ^).
>
> In the spirit of making things as simple as possible for the grammar authors, I suppose the right rule is “when constructing the value of an attribute, treat nonterminals marked ^ and - the same: recur through them” (the last possibility mentioned above).
>
> I apologize if this has been discussed before - I have the guilty sensation that it has been, and that I did not retain the answer.
>
> Michael
>
>
>
>