So what you're trying to do is generate a sample document from the schema?

There have been many tools that do this. The problem is, there's 
generally a wide range of possible documents that match a given schema 
and no clear way to decide which choices to make. You either need to 
pick randomly and accept that your synthesized document may not in fact 
be similar to any "real" document (due to semantic constraints not 
expressed in the schema), or provide additional semantic info from some 
other source.

There are a number of tools out there which will do this to a greater or 
lesser extent, for the purposes of generating "churn" test suites. It 
sounds like you ought to be investigating those before spending too much 
time attempting to solve the same problem.


On the other hand, if you're just saying that you want to pretty-print 
XML, and (unlike most attempts in that direction, which risk breaking 
the document semantics when they insert the whitespace) you want to try 
to use the schema to recognize then it's safe (ie, when you're dealing 
with "element content")... I'm sure that too has been addressed many 
times, but at least it's something that has a clearer definition.

On the other other hand, if you just want to pretty-print *with* the 
risk that indentation will change the document's semantics, you don't 
need the schema at all -- just track element nesting.