Parses UNQ data to a DTD's specification, and returns an appropriate
structure of parsed data. The structure consists of label/value
pairs like standard UNQ data, but the values are different.
The value of each depends on the type that the DTD claims the
field will have.
A DTD 'int', 'float' or 'string' will have the corresponding
value type in DGD. A DTD 'phrase' will have a DGD phrase object
as its value. A DTD 'unq' entry will simply pass that piece
of preparsed UNQ through to its value slot. The DTD 'unq' type
is useful for passing data through to be parsed later.
A DTD type followed by a modifier like '?', '*' or '<3..7>'
will match some number of entries of that type. The values will
be returned as an array, even if only one (or none at all)
are parsed. Each element of the array will be a value like
the DTD type would generate if the field were by itself. For
instance, type 'int<4>' would yield an array of four integers.
A DTD structure such as '{int, string*, unq}' will have an array
of tag/value pairs as its value. Within that array of tag/value
pairs, the tags are field names, and the values are as described
throughout this section. For instance, if '~nums{float+}' is a field
of that structure, the tag will be the string 'nums' and the
value might be an array of seven DGD floating-point numbers.
A named field in a sequence, such as '{int, nums, unq}', will
appear as a tag/value pair as well.
Parse_to_dtd clears the parse error stack every time it gets
called. Don't rely on that stack to persist across calls.
