DQ is just a java-based way for defining an HQL statement for eventual processing by Hibernate. It doesn't really offer much beyond what is supported by HQL.

Whatever HQL supports, DQ will support.

Again, you're thinking about this all wrong. If your table had 5 million rows, any kind of function like this is going to be a table scan action to pull out the data followed by X amount of time to process the value on each row. Regardless how small X is, when multiplied by 5 million you're looking at a huge performance impact on your database and a slow performing query.

Even if you respond with "Well I will only have Y rows, not 5 million" you must still agree that X will basically be a constant and that constant will be multiplied by Y for the performance impact.

Then consider how many times the query might be called, so you have another multiplier based on how often people are using the portlet which needs the data.

Save yourself some time and headache now. Flatten your data model so you can get to the week using a simple indexed query. You're going to save yourself and your DBA a lot of time and problems.