I've used a binary load once in a production application. I basically have two applications that are clones of each other, but one of them has had all of the revenue data stripped out. The stripped application is available to a wide group of users. The one with the revenue data to only a small number of users. This probably could have been handled with some other security model, but at the time we weren't even using access point, and we wanted to restrict the users with Windows permissions, not even letting users into the directory with the revenue data.
A much better use of binary load is what Oleg pointed out, where one group is responsible for script, and another for the display objects. The group writing script could just create QVWs without any user objects. You could have an Orders QVW with all of the tables related to customer orders, for instance. You could have multiple user applications created from this. Each would binary load the data, avoiding any duplication of script, and keeping the responsibilities separate. And then each application could have the display built however was appropriate for that application. Even if the duties are all in the same group, or even the same person, it would still be beneficial to binary load if you have several applications with the same data model.
We do use QVD files everywhere. We have a lot of exceptions, but the general rule is that applications load only from QVDs.
We haven't used include files. Very little script is shared in common between applications, so I'm not sure there would be much benefit for us personally using include files. I do think I could benefit from a calendar include and a text mapping include at some point, though. Those are getting more complicated yet more similar between applications all the time. At some point I need to bite the bullet and switch to a more integrated approach.
We haven't used variables for expressions very often, and where we have, they have not been shared between applications, only used within the same application. I've run into bugs with variables holding expressions that make me nervous about using them more widely. But so many people are using them happily that I think I just got really unlucky in how I first started using them.
Oleg Troyansky wrote:Once various pieces of logic are heavily shared between developers, departments, applications, etc... - at some point, you can't make any changes without calling a summit of all involved parties.
We've been lucky in this regard. First, we've been lucky that most of our shared resources (mostly the QVD data warehouse) were somehow built correctly the first time, so we've had little need for any disruptive changes. When we do need changes, our department is in charge of almost the full process end to end, from the business systems that create the data in the databases, to building the QVD data warehouse from those databases and other sources, to user application script, user application design, and setting up the Publisher tasks. While I have no official authority (I'm just a cubicle grunt), I've taken on the role of benign dictator for much of this. So when I want to make a change, I make a change, often without even consulting anyone, and I just make sure I check and as necessary fix anything that could be affected. So far no complaints or conflicts of any significance that I'm remembering. *crosses fingers*
That probably wouldn't be typical of the QlikView group(s) in most companies, though. There ARE aspects of what we do that DO suffer from this sort of paralysis, with a 5 minute change taking 3 weeks. They just aren't on the QlikView side of the house.
