In the Semantic sources window, each source system row unfolds into the tables it reaches and the transformations behind them, so you can see where a system's data lands before you edit, truncate or delete anything. Creating and managing the sources and systems themselves is covered on Semantic sources, source systems and logical models.
The first time you expand a row, the list loads; while it does, the body shows Loading tables linked to followed by the system's name. If the fetch fails, the body reads "Linked tables did not load. Collapse and expand the row to retry." A system that nothing points at yet reads "No tables use this system yet."
Expanding is scoped to the source you are viewing: open another source on the left and its systems start collapsed.
Each table row carries the table's kind icon, its name, and a summary of its transformations:
Under the table row, the transformations that read from this system are listed, each with its status badge; the Warehouse workspace tour explains what Pending and Skipped mean.
Click a table to open its window, or a transformation to open it. These are the same windows the tree opens; Open object editors lists the other ways in.
A table reaches the list in either of two ways, and one of them is enough:
A transformation always belongs to exactly one table and one source system, so the rows nested under a table are exactly the loads of that table that come from this system.
Transformations on tables you cannot see are counted, not dropped. The list then ends with a note such as "3 transformations on tables not shown here." so it never quietly disagrees with the Transformations count on the system row.
The window can open already pointed at one system, for example from a chat message that links to it. It selects the system's source, highlights the system row, and scrolls the systems list so that the row sits in the middle. Choosing a system from the window's search focuses it the same way.
The scroll runs twice: right away, and again after the window's opening animation, because a position measured mid-animation lands in the wrong place. Only the list moves; the window itself does not jump.
Opening the window again while it is already open brings that window forward and points it at the system; it never opens a second copy.
A table appears when it names the system itself or when at least one of its transformations reads from the system. The second case is why a table that never lists the system can still show up under it.
The table has five transformations in all. Two of them read from this system and are listed under the table; the other three read from other systems and do not appear in this list.
They act on tables you cannot see, so those tables are not listed. The note at the end counts them, for example "3 transformations on tables not shown here.", so the list stays consistent with the Transformations count on the system row.
A to Z by name, ignoring capital letters and reading numbers naturally, so a table numbered 2 comes before one numbered 10.
The table opens in its own window and the transformation opens in the transformation window; both are the windows the tree opens as well. See Open object editors for the other ways to open an object.
The window opened pointed at that system, for example from a chat message that links to it. It selects the system's source and scrolls the list until the row is in the middle, so a row far down the list is not lost off screen. The same happens when you choose the system from the window's own search.
The window could not fetch the system's tables. Collapse the row and expand it again to retry, as the row itself suggests.
Not across sources. Expanding belongs to the source you are viewing, and another source's systems start collapsed.