And then once that data model has been loaded, all the tables are there and available for you to manipulate or dump to QVDs.
But this isn't generally going to make it faster to generate QVDs, because you had to load the data into the original QVW in some way. You can't really get around loading the data from your database somewhere, somehow.
I do binary load then write out QVDs in two cases. I had requests to combine data from two different QVWs into a third, greatly simplified view. The QVWs in question did some data transformation. They did not merely load straight from QVDs. Probably the right thing to do would have been to build a new application that did the data transformations and stored the results as QVDs, then loaded those QVDs into both the original and new applications. Under time pressure and also hoping the new applications would go away soon, I simply did a binary load of the original applications and stored the data I wanted to QVDs to load into the new application. This was only faster to create QVDs in the sense of "saved development time" and not "saved computer time". It's not something I would generally recommend. Sloppy programming. Just something I'm confessing to having done.
