This directory consists of pairs of test cases (an XSD file) and expected
output (a JSON file). For every pair, xsd.Parse is called on a document
containing the fragments. The resulting type set is compared with the json
file using the following rules:

* Every member in the top-level JSON object is the name of a type (xmlns
  omitted).
* The value of each member is a partial JSON encoding of the xsd.Type.
* Each type in the top-level JSON *must* be in the xsd.Schema.
* Types may be omitted from the top-level JSON.
* If a field is present in the JSON value for a type member, it must match the
  field in the parsed xsd.Type.
* If a field is not present in the JSON value, it is not checked on the
  xsd.Type.
