[seam-issues] [JBoss JIRA] Updated: (SEAMREMOTING-4) Optimize Seam Remoting XML Marshalling

[ https://jira.jboss.org/browse/SEAMREMOTING-4?page=com.atlassian.jira.plug... ] Kevin Galligan updated SEAMREMOTING-4: -------------------------------------- Attachment: Final_seam_stuff_2.patch First shot at reducing XML size. Changes are as follows: - Most tag names are significantly shorter. That includes types and structural tags (bean, map, number, element, member, etc). Where possible, they are 1 or 2 letters. - Members of beans are listed in alphabetical order instead of using the 'name' attribute. On the client side, the metadata is used to inflate the data objects. This significantly cuts down on data size for larger object graphs, but introduces a major potential issue discussed below. - The tag names for the simple types are shortened, and I defined them in one place in the code which is dynamically generated in the Remote servlet. This was to simply to make that clearer and easier to change in the short term. Once the structure is set, they could be put directly into the JS again. - Some remote.js changes. The type handling/registration code between DTO types and components had some duplication, so I rolled that into a single object, but kept the original method names in case somebody had redefined them. That could be trimmed in the future. Potential issue. If the interface js file is cached, and the object model changes, the positional member resolution will fail badly. Still need to sort that out. The file size savings are significant, however, so we should figure that out. We did some light testing locally and our data size was cut down to about 34% of the original. We could squeeze out a bit more by cutting down the tag names at the top (envelope, header, conversationId, etc). Its relatively small, though. The other place to cut would be the bean type names, but that would have required some real changes to how the API worked that I didn't want to get into. I also changed the interface file generation. This actually saves quite a bit. Our major interface file is cut back to 20% of the original size, from about 75k to 15k. It uses the metadata to build pretty much everything, including declaring the fields and getter/setter methods. In the current interface file there is a minor bug. Sort of a bug, anyway. It works, but its incorrect. The getter/setter methods are defined on the prototype, but the are defined each time the constructor is called. It works, but the prototype methods really only need to be defined once. Running that each time is just adding some proc overhead. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira