We have a little bit of a catch22 going on with this one. There is a
function to build up the view model, but the view models on this
project have a ton of data that they need. Typically I would follow a
single call, compose the VM and return to controller for consumption.
This project calls down into a service for calls such as for
MediaItems, ContentBlocks, etc. Some of the aggregate roots have
multiple value objects needed for a single screen. Because many of
these calls can be in parallel we have made it so. So you may get:
Parrallel.Invoke(
() => _mediaService.GetStuff(),
() => _contentService.GetStuff()
);
//Compose the ViewModel here and return
Alright, thanks for the input guys, might have to rethink some of this
and figure out which negatives are the ones we can live with while
still meeting the reqs.