Using autoDtd
The autoDtd program creates a DTD file from an XML file example. The program also generates an "attributed tree" and a summary of the actual data in the file.
Listing Four (available electronically) shows the DTD autoDtd generates from the XML file in Listing One. It starts out by defining XML entities to represent numerical data. Most XML parsers will treat these entities as synonyms for strings (CDATA). However, autoDtd and xmlToSql know to map them to numerical types in C and SQL, respectively. The DTD continues with ELEMENT and ATTLIST definitions. The elements correspond to the tags in the XML file. The parenthesized lists after the element name describe the legal child tags. The child tags followed by "?" are optional. "*" indicates that any number of children of this type can follow. "+" indicates there must be at least one child of this type. The ATTLIST describes the attributes within a tag. Note that the x attribute is of type %float; and the y attribute is of type %int;. This is because of the value 3.0 for the x attribute in one of the points in the example, while there are no decimal points in the y values.
Listing Five (available electronically) shows the "attributed tree" output of autoDtd. This is a denser, more readable representation of the same data structures. There is a line for each type of tag in the XML file. Children are indented in lines following their parent. Attributes, if any, are on the same line following the tag name. The data type, if any, is indicated by a letter before the tag or attribute name, with "$" representing strings, "#" representing integers, and "%" representing floating-point numbers. Tag names may be followed by the characters ?, *, or + with the same meaning they have in DTD files.
The stats output of autoDtd (Listing Six, available electronically) summarizes the contents of the XML file. It is also used by the xmlToSql program to decide which fields to index. The stats output begins with a comment that describes the format.