Sibling rivalry

[Jon Bosak <bosak@xxxxxxxxxxxxxxxxxxxxxxx>]

I'm formatting the latest version of the XML spec and have hit a wall
in handling the productions.

In the old version of the spec, a production had a left hand side and
one or more right hand sides, with each right hand side optionally
followed by a comment, a validity check, or a well-formedness check.

   <!element prod (lhs,(rhs,(com | wfc | vc)?)+)>

This was handled by the following code:

   (element PROD
     (make table
	   (make table-column
		 width: 2pi)
	   (make table-column
		 width: 7pi)
	   (make table-column
		 width: 1.5pi)
	   (make table-column
		 width: 16.5pi)
	   (make table-column
		 width: 9pi)
	   (process-children-trim)))

   (element LHS
     (sosofo-append
      (make table-cell
	    column-number: 1
	    font-weight: 'bold
	    (make paragraph
		  use: para-style
		  start-indent: 0pt
		  font-weight: 'bold
		  (literal
		   (string-append
		    "["
		    (number->string (element-number (ancestor "PROD")))
		    "]"))))
      (make table-cell
	    column-number: 2
	    (make paragraph
		  use: monopara-style
		  font-weight: 'bold
		  (process-children-trim)))
      (make table-cell
	    column-number: 3
	    (make paragraph
		  start-indent: 0pt
		  (literal ":=")))))

   (element RHS
     (make table-cell
	   starts-row?: (not (first-sibling?))
	   column-number: 4
	   n-columns-spanned: (if (absolute-last-sibling?) 2 1)
	   (make paragraph
		 use: monopara-style
		 (process-children-trim))))

   (element COM
     (make table-cell
	   column-number: 5
	   (make paragraph
		 use: para-style
		 font-posture: 'italic
		 start-indent: 0pt
		 (sosofo-append
		  (literal "/* ")
		  (process-children-trim)
		  (literal " */")))))

   (define ($vc-wfc$ vc-or-wfc)
     (make table-cell
	   column-number: 5
	   (make paragraph
		 font-family-name: %title-font-family%
		 font-size: 9pt
		 start-indent: 1pi
		 first-line-start-indent: -1pi
		 (sosofo-append
		  (literal
		   (string-append "[" vc-or-wfc ": "))
		  (process-children-trim)
		  (literal "]")))))

   (element VC ($vc-wfc$ "VC"))
   (element WFC ($vc-wfc$ "WFC"))

Now, however, the editors are allowing multiple comments and/or
validity checks and/or well-formedness checks to follow each RHS:

   <!element prod (lhs,(rhs,(com | wfc | vc)*)+)>

For example:

   <prod><lhs id='NT-Attribute'>Attribute</lhs>
   <rhs><nt def='NT-Name'>Name</nt> <nt def='NT-Eq'>Eq</nt> 
   <nt def='NT-AttValue'>AttValue</nt></rhs>
   <vc name='Attribute Value Type'>Attribute Value Type</vc>
   <wfc name='No External Entities References'>No External Entity 
   References</wfc></prod> 

What I want to do here is to create a new cell only if the COM, VC, or
WFC is the first one following the close of an RHS, and to create just
a paragraph if it's not.  But I don't see any function in the core
query language that will let me do this.

The alternative is to create a new container for the optional stuff
following the RHS, let's call it PINFO:

   <!element prod (lhs,(rhs,(pinfo)*)+)>
   <!element pinfo (com | wfc | vc)+ >

Suddenly life for the stylesheet designer becomes much easier:

   (element PINFO
     (make table-cell
	   column-number: 5
	   (process-children)))

   (element COM
     (make paragraph
	   use: para-style
	   font-posture: 'italic
	   start-indent: 0pt
	   (sosofo-append
	    (literal "/* ")
	    (process-children-trim)
	    (literal " */"))))

   (define ($vc-wfc$ vc-or-wfc)
     (make paragraph
	   font-family-name: %title-font-family%
	   font-size: 9pt
	   start-indent: 1pi
	   first-line-start-indent: -1pi
	   (sosofo-append
	    (literal
	     (string-append "[" vc-or-wfc ": "))
	    (process-children-trim)
	    (literal "]"))))

   (element VC ($vc-wfc$ "VC"))
   (element WFC ($vc-wfc$ "WFC"))

Unfortunately, life for the draft editors suddenly becomes much more
difficult, and it appears that a quick perl hack isn't going to solve
the problem.  (The eds seem to have commented out arbitrary spans
within the productions...)

Does anyone see a stylesheet solution to this problem, or am I going
to have to make the editors unhappy?

(If you see a solution, mail me directly as well as responding to the
list; I subscribe by daily digest, and I'd like to find out sooner
rather than later.)

Jon