S
At 06:34 PM 1/12/2003 -0800, Craig R. McClanahan wrote:
As we've discussed a couple of times, the last major functionality change we had discussed for Struts 1.1 was to migrate to dependence on commons-resources, rather than the proprietary message resource facilities inside og.apache.struts.util. As you might recall, Michael Schacter took a first crack at factoring out the Struts resources classes out to create this commons package, which is currently in the sandbox.I've recently gone through it, and did a major refactoring of commons-resources, to the point where I'm now ready to propose that we modify Struts to depend on it. I'd like the other committers to evaluate the current state of commons-resources, and my proposed integration plan below, to see what they think of this idea. The nightly build of commons-resources.jar included in recent Struts nightly builds is the code that I'm proposing. You can see the Javadocs for this code at: http://jakarta.apache.org/commons/resources/api/ and get the sources via either CVS (from jakarta-commons-sandbox) or nightly snapshots: http://jakarta.apache.org/builds/jakarta-commons/nightly/commons-resources/ In terms of Struts integration, I propose: (1) Most Struts classes declare a static MessageResources instance for the messages unique to that Struts package. For example, org.apache.struts.taglib.bean.CookieTag has this: protected static MessageResources messages = MessageResources.getMessageResources ("org.apache.struts.taglib.bean.LocalStrings"); This would be migrated to the new Messages class from commons-resources: protected static Messages messages = Messages.getMessages("org.apache.struts.taglib.bean"); The calls to actually retrieve message strings are compatible with the existing code, as well as the properties files used to acquire the message strings, so no other changes should be required. (2) Convert o.a.s.u.MessageResources (and its friends) to wrappers around equivalent functionality from commons-resources (much like GenericDataSource now wraps commons-dbcp), and deprecate them. This protects existing apps that are customizing these APIs, but allows us to remove the o.a.s.u classes in a future version. (3) Modify the <message-resources> initialization element to allow the selection and configuration of an appropriate ResourcesFactory from commons-resources, wrapped by a Messages instance. This is primarily a change in the interpretation of the "factory" attribute, and should not affect anyone that uses the current default. (4) Modify all internal uses (including in tag libraries) of org.apache.struts.util.MessageResources to use org.apache.commons.resources.Messages instead. This will be transparent to users that use the standard implementations, but will require folks who have subclassed the MessageResources classes to migrate their code as well. What do you think? Should we go ahead and do this migration? Is the commons-resources package as it stands now as complete and functional as it needs to be (obviously, it'll need to be promoted to a standard Commons package and released so we can rely on it, which will require a couple of volunteers willing to help me maintain it). Should we do the entire migration outlined above, or maybe only part of it? Thoughts, please. Craig -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
