Rather than mapping a lexical symbol in a context-free production rule to an EString (if its not recognized as an enumeration or predefined data type), it should be mapped to a user-defined data type with Instance Type Name = java.lang.String. Users then have the option to implement their data type in Java or to use the String.

An even better solution would be to actually generate an implementation of the custom data type (i.e. translate the lexical syntax productions into a Java class that implements the data type).

Submitted by Oskar van Rest on 25 October 2012 at 10:40