I work on a database project with many tables which are scattered in 6 schema. So long I integrated all tabled I needed into one dataset. But as I need to touch more and more tables this dataset is hard to manage and I wonder whether it is practicable to implement many datasets, one dataset for each single schema. Is there something obvious which integrates multiple datasets into one class, such that I can pass transaction around and keep foreign key relations?
Whenever I use a dataset I make it specific to the application's needs rather than trying to make it a representation of the database - it will rapidly get really out of hand if you try and map your entire database into it..
If it gets too big even as a dataset that is specific to the application, I then break it down further into sub parts that use a dataset for operations/a set of operations. In a windows app form sense this can even be a dataset per form for some forms, or a dataset per group of similar forms for other areas of the application