I'd like to see the update routines only update the fields that have been modified. As it is, I need to ensure that I have either gotten the record from the database, update the data, and then update the database; or that I have gotten all fields from the client application. Right now, we're taking the approach that the client apps always send all attributes to the server so that we don't need to retrieve the record in order to save it, but this also means that our network traffic is larger than it needs to be. (Obviously, this also implies that the server is sending all available attributes to the clients for any data request so that the client has the data to return...)
As a cheap (i.e. minimal lines of template code modified) work-around for those cases where the client may not have all the available data, I've modified the cst files so that the update stored procs only write values that are not provided as NULL. (If the data isn't provided to the objects, the code passes a NULL value to the stored proc - the only real change to the code side template is that you drop the requirement that only null-able fields will send NULL to the stored proc is not present). This has a pretty obvious limitation though, in that if you want to set a column to NULL, you can't. For my current project, this is OK, but I don't see this as a long term solution. If there isn't any plan to do this in the forsee-able future, I'll probably dive into the cst's and do a better solution.