rbarzic / dita-ng Goto Github PK

Implement a conversion process to generate DTD-syntax vocabulary modules and 
document type shells that conform to the DITA DTD coding requirements.

Define  run-time parameter to define the static parts of the .ent and .mod 
public IDs. Can't assume OASIS SGML-style public IDs. Need to accommodate URNs 
like those used in DITA for Publishers.

Class attribute declaration lists should be formatted like this:

<!ATTLIST sl        %global-atts;  class CDATA "- topic/sl "         >

The code as of rev 78 is generating only parameter entity declarations. It 
needs to also generate element and attribute list declarations for each element 
type. 

For example, given this RNG:


  <define name="mathmlref.content">
    <empty/>
  </define>

  <define name="mathmlref.attributes">
    <ref name="xref.attributes"/>
  </define>

  <define name="mathmlref.element">
    <element name="mathmlref">

      <ref name="mathmlref.attributes"/>
      <ref name="mathmlref.content"/>
    </element>
  </define>

The DTD result should be one parameter entity for mathmlref.content, one for 
mathmlref.attributes, but then an ELEMENT and ATTLIST declaration pair that 
references those entities, e.g.:

<!ENTITY % mathmlref.content 
"..."
>
<!ENTITY % mathmlref.attributes
 "..."
>

<!ELEMENT mathmlref
  %mathmlref.content;
>
<!ATTLIST mathmlref
  %mathmlref.attributes;
>

There is never a parameter entity for the element type itself.

The naming convention on the <define> elements is consistent, so that any 
define with the name "*.element" should never result in a parameter entity. 
Rather, it should result in the ELEMENT/ATTLIST declaration pair as shown. The 
ELEMENT and ATTLIST declarations should only ever have the single parameter 
entity references, so the only information that is needed is the element type 
name: the names of the parameter entities will always follow the patter 
"{typename}.content" and "{type name}.attributes", so there is no need to 
process of the contents of any <define> with a name matching "*.content" unless 
it is to get the element type name from the contained <element> element.

Need XML markup for RNG annotations to provide the details required by both the 
DITA coding requirements as well as the TC's conventions for comment content in 
TC-defined modules and document type shells.

As of revision 91 this declaration is showing up in the .mod file:

<!ENTITY % definitions 
  PUBLIC "-//OASIS//ENTITIES DITA XML Domain - RNC//EN" 
         "xmlDomainMod.ent" 
>
%definitions;


It should only be in the shell DTD.

Need to get the module header comments from the input RNG file. We can't assume 
that the input is an OASIS-defined module.

The .ent files for domain modules only include the "integration" entities, 
which are the entities that extend the base types. In the RNG these are defines 
where the name starts with the name of the domain and a trailing dash, e.g. 
"hi-d-":


  <define name="hi-d-ph">
    <choice>
      <ref name="b.element"/>
      ...
    </choice>
  </define>

You could get the module name from the domains-atts-value declaration: it 
should always be the second token within the parenthesized value:

  <define name="domains-atts-value" combine="choice">
    <value>(topic hi-d)</value>
  </define>

The <defines> that reference the "integration" defines would go in the shell 
DTD, not the module itself, because in DTD syntax they are specific to each 
shell. So those entities would never get generated in the .ent or .mod file, 
but only in the DTD shell.