Package com.sun.msv.verifier.regexp

Class CombinedChildContentExpCreator

java.lang.Object
com.sun.msv.verifier.regexp.CombinedChildContentExpCreator
All Implemented Interfaces:
ExpressionVisitorVoid
public class CombinedChildContentExpCreator extends Object implements ExpressionVisitorVoid
creates "combined child content expression" and gathers "elements of concern" and its "attribute-pruned" content model. Intuitively, "combined child content expression" is a set of content models of "elements of concern", which is appropriately combined to express dependency between elements. "Elements of concern" are ElementExps that are possibly applicable to the next element. These gathered element declarations are then tested against next XML element. "Attribute-pruned" content model is a content model after consuming AttributeTokens and removing unused AttributeExp nodes.

For example, when the current expression is 

    <!-- javadoc escape -->
   <choice>
     <concur>
       <element> ..(A).. </element>
       <group>
         <element> ..(B).. </element>
         ...
       </group>
     </concur>
     <group>
       <element> ..(C).. </element>
       ....
     </group>
   </choice>
then the combined child expression is 

   <choice>
     <concur>
       ..(A').. 
       ..(B').. 
     </concur>
     ..(C').. 
   </choice>
and elements of concern and its attribute-pruned content models are <element> ..(A).. </element> -> ..(A').. <element> ..(B).. </element> -> ..(B').. <element> ..(C).. </element> -> ..(C').. (A'),(B'), and (C') are attribute-pruned content models of (A),(B), and (C) respectively. Note that combined child pattern contains only <choice> and <concur> as its grue (of course, except ..(A').. , ..(B').. , and ..(C').. ). This function object also calculates "continuation", which is the residual expression after eating elements of concern. For example, say the expression is "(A|(B,C))?,D". When EoC is B, then the continuation will be C,D. When EoC is A, then the continuation will be D. When EoC is D, then the continuation will be epsilon. When there are multiple EoC, (say A and B), then the continuation will be meaningless (because it depends on which EoC will be accepted), and thus won't be used. However, the implementator must be aware that it is possible for a binary operator to have EoC on both branch and EoC is still unique. The following expression is an example. (A|B)*,C?,(A|B)* when A is EoC, SequenceExp of (A|B)* and C?,(A|B)* has EoC on both branch.
Author:
Kohsuke KAWAGUCHI