This is a very good question. The beauty of Xcalar is in the fact that you do not have to. Xcalar manages the dataflow graph for you. We believe as a user, you simply explore the data using Xcalar Design GUI and keep discovering relationships. You have an ability to perform countless number of operations, change your mind at times, and not only answer your questions, but also know how to ask new questions, that you haven't even begun to think about, as you uncover aspects of your data.
As your mind carves out the algorithm for data modeling and exploration, behind the scene, Xcalar Design creates a rich and accurate dataflow graph that represents the current state of your algorithm at any point. This is a visual representation of your algorithm. You do have the ability to click on tables and operations in this graph, and explore a lot of interesting information. You can peek at the schemas, even revert to an older table, and start all over again. You can decide to add table back to worksheet from the dataflow graph, and this makes it easy for you to analyze your dataflow and all it's intermediate tables. The dataflow graph tracks the data lineage of every single field in your table. You can see the path of each field back to the originating table.
Hope this gives you a bit more perspective on how datraflows work in Xcalar.