Todo

  • Make compiler easier to use. Try an Ant CompilerAdapter approach
  • Using two classloader in impl..is this necessary
  • Property setting for attributes. Named and positional parameters for init of attribute.
  • Do not serialize the attribute in the .class file. Instead store the text that will be used at runtime to create the attribute.
  • Do we really need an Attribute interface. No...
  • Remove dependency on using javadoc, use of other javadoc like engines QDoc etc. More control, faster etc.
  • Create a taglet so that attribute can appear in javadoc generated documentation.
  • Eclipse integration. Look into how aspectj is supported inside eclipse
  • Write better tutorial/sample docs.
  • Add more control over inheritance. Similar to 'AttributeUsage' in .Net
  • 'AttributeUsage' AttributeTargets: specify only for methods, classes etc.
  • 'AttributeUsage' Can attributes be inserted multiple times.
  • Finish up support for Javadoc tags that are not @attribute. Specify which javadoc @tags should be processed.
  • Clean/think about Attributes interface that is used to retrieve attributes.
  • Write better tests for method parameter matching.
  • Special handling for ctor in the API and implementions. The name of the constructor method in BCEL is (init)
  • Change Attribute.java to return List of attributes instead of object arrays?
  • Resolve symbolic names out of attribute classes as in "@TransactionAttribute(TransactionalAttribute.TRANSACTION_REQUIRED)"
  • Attributes for parameters? Attributes for jar files?
  • Check interoperablity between bytecode manipulation libraries. interoperablity between bytecode manipulation libraries at compile time and runtime. Attributes are stored as byte arrays, make sure no bcel/cfparse specific stuff is stored in the classfile.
  • Consider creating base Attribute class for use in bcel implemention instead of using their Unknown class.
  • Expose at part of the configuration the name under which the custom property is stored in the byte code. This will make it easier to migrate to what JSR 175 uses.
  • Rename BCELParseAttributeExtractor to BCELAttributeExtractor.
  • Clean up exception handling in Attribute.getAttrExtractor
  • More clean up of coding conventions using Checkstyle.

Applications

There are many application in which annonated source code will add significat ease of use benefits to existing applications. Examples include Web services, much like what was done in BEA WebLogic Workshop and also examples taken from .Net, such as NUnit - the xUnit port for .Net. .Net applications are interesting to watch closely since they already incorporate attributes as part of the language. Here is a small incomplete list of things off the top of my head that

  • JUnitAttrib - like NUnit, see nunit.org
  • Special support for XDoclet/J2EE tags?
  • Support for some opensource containers to extract deployment information in order to create j2ee components 'just in time' like jsp compilation.