The Diagram tab of a semantic model draws its BI tables as cards, with a line for each relationship between them. Use it to arrange the tables, find one on a large model, refresh the data behind the diagram, and open the lineage view that maps a model to its reports.
Open the diagram
Open the semantic model and click Diagram in its tabs.
If the model has no BI tables yet, the canvas shows No BI tables found and the note No BI tables in this semantic model. Add tables from the model first; see Semantic models and BI tables. While no semantic model is selected, the panel reads "Select a semantic model to view its diagram."
You need edit access to the model and the Edit DataPrism Semantic Models permission to change anything here. Without them the diagram is read-only: you can still search, zoom and read it. Actions that change the model appear only for a saved semantic model. See Who can do what in DataPrism.
On a desktop, right-click the canvas for New relationship and Add BI table, or right-click a table's card to start a relationship from that table. See Relationships and the core table for creating and editing the lines. On a phone or in a narrow window, the editor does not open and shows "The BI diagram editor requires a larger screen. Open this page on desktop to edit relationships."
The cards on the canvas
Each card is headed by the table's name and a type icon; the core table's card is highlighted.
A collapsed card lists 10 fields. Columns used in relationships and key columns come first, and the card grows to show every one of them, so none is hidden; the rest follow in alphabetical order. Show more lists every field; Show less collapses the card again.
Click a card to select the table. Click a field row to select that column. Double-click a card to open the table's popup.
Hold Ctrl or Cmd to select several tables at once.
Drag a card to move it. Drag its edge to change its width; the width you set is kept.
Arrange the tables
Use Auto Layout in the toolbar to arrange the whole diagram. Under Layouts, pick a preset:
Preset
What it does
Star
Arranges each connected group around its hub table.
Compact
The default. The tightest arrangement.
Horizontal
Lays the diagram out from left to right.
Spacious
Like Compact, with more room between tables.
Under Unconnected, choose where tables with no relationship wait: under the connected tables (Below) or beside them (Side).
A reorder without a chosen preset first asks you to confirm: Reorder Diagram with "Are you sure you want to reorder the diagram?". Choose Reorder to continue, or Cancel to stop.
When an arrangement is applied, a Layout Applied message says which preset was used.
The first time you open a diagram with nothing saved, DataPrism arranges the tables for you (when the model has relationships). After that, the arrangement you leave is kept.
The first time you switch to a datasource, the tables are adjusted for it once.
Undo and redo
Undo and Redo sit in the toolbar; each is greyed out when there is nothing to undo or redo.
Undo covers table, relationship and dependency changes: adding, removing and editing them. Undoing a removed table brings its relationships and dependencies back with it.
Layout changes are not undo steps. Moving or resizing a card is saved on its own, and the buttons do not bring an old arrangement back.
Where your arrangement is kept
Your arrangement is saved with the semantic model, so it comes back the next time the diagram opens. When nothing is saved, the diagram computes a fresh arrangement.
Saving is automatic: a moment after you finish dragging a card, and right away when a layout is applied. Only finished moves are saved.
A copy in your browser makes loading faster. It is used only while it is less than a day old; what is saved on the model wins.
Your view preferences, such as the datasource you picked and your layout choice, are remembered in this browser for 30 days.
Tables and columns that change in the model appear on the canvas without a reload: when tables join or leave the model, or a column is renamed or hidden from reports, the canvas updates itself.
Find a table or column
Type in Search... in the toolbar.
The dropdown groups results under Tables and Columns, shows up to 10 results with tables first, and highlights the matching text. Column results also name their table.
Search matches names by text, ignoring case, as you type.
Use ArrowDown and ArrowUp to move through the list, Enter to select, Escape to close.
Selecting a result brings the table or column into view and highlights it for two seconds.
If nothing matches, the list reads "No tables or columns found matching" followed by your search text.
Zoom, pan and the minimap
Zoom runs from 10% to 500%; use the zoom controls in the toolbar.
Scrolling pans the canvas, so you can move around a large model without leaving the table you are on.
Show all fits the whole diagram in view.
Show Minimap adds a small overview of the whole diagram; Hide Minimap removes it. It is off by default.
Refresh from the database
Refresh from Database in the toolbar reloads the latest tables and relationships for the diagram. While it runs, the control is disabled and its icon spins.
Local copies are cleared first, then the refresh steps run side by side, so one failing step does not stop the others.
Progress reads Refreshing diagram with "Fetching latest data from the database..."; on success it reads Diagram refreshed with "Successfully loaded latest data from the database."
A failed refresh reports Refresh failed. A single failed step reports Refresh step failed. If the local copy cannot be cleared, Failed to clear local diagram cache appears and the refresh continues.
Read the lineage of a model
The lineage view is a read-only map of a semantic model: semantic models in a left column, their reports in a right column, with a line from each report to its model.
Each node is labelled with its kind: Semantic model or Report. The current model carries an Open badge, and hovering a node highlights its connections.
Click a node to open it.
Fit to view in the bottom-right corner frames the whole map, and the map re-frames itself when it changes.
Zoom runs from 10% to 200%, and the nodes cannot be moved: it is a map, not an editor.
Remove a table from the model
Select the table's card and click Remove from semantic model on its toolbar. The toolbar shows on a desktop, for a saved semantic model.
The dialog Remove from semantic model? reads Remove "{name}" from this semantic model? This permanently deletes the BI table with its columns, relationships and report filters. For the core table it reads "{name}" is the core table. Removing it clears the core-table mark and permanently deletes the BI table with its columns, relationships and report filters.
Confirm with Remove, or with Delete anyway when reports use the table.
Removing deletes the BI table itself, the same as Delete in the model's BI tables tree; to point a table at another source instead, use Replace table (Replace a table's source).
Notices and limits
While the diagram loads, badges read Loading diagram... and Auto-arranging tables...; the canvas overlay reads Loading canvas.
Dropping a table that is already on the canvas shows Table already exists in diagram and brings the existing card into view.
With All Datasources selected, the table list is empty; pick a single datasource to see its tables.
If the diagram cannot load its data, it reports Failed to load diagram.
Questions, answered
How do I open a model's diagram?
Open the semantic model and click Diagram in its tabs. The tab draws the model's BI tables and their relationships. If the model has no BI tables yet, the canvas says No BI tables found; add tables from the model first. See Semantic models and BI tables.
Why did my tables move on their own?
Two moments arrange the diagram for you: the first time it opens with nothing saved (when the model has relationships), and the first time you switch to a datasource. After that, your arrangement is kept. Auto Layout re-arranges only when you ask it to.
Where is my arrangement kept?
On the semantic model, so it returns when the diagram opens again. A copy in your browser speeds up loading and is used only while it is less than a day old. View preferences, such as the layout choice and the datasource you picked, are kept in this browser for 30 days.
Can I undo a layout change?
No. Undo covers tables, relationships and dependencies. Moving and resizing cards is saved on its own and is not part of the undo steps.
What does Refresh from Database do?
It reloads the latest tables and relationships from the database. Your saved arrangement stays. The steps run side by side, so one failed step does not stop the rest, and each failure gets its own message.
Why did the search highlight disappear?
The highlight clears itself after two seconds, so it does not hide the diagram while you work. Search again to bring the table or column back into view.
Why is the table list empty when I choose All Datasources?
With All Datasources selected, the diagram shows no tables. Pick one datasource to see its tables.
What does the lineage view show?
A read-only map: semantic models in the left column, their reports in the right column. Click a node to open it; Fit to view frames the whole map; the nodes cannot be moved.
Why do I not see Add BI table or New relationship?
They appear only for a saved semantic model, and only when you can edit it: you need edit access to the model and the Edit DataPrism Semantic Models permission. Without them you can still search, zoom and read the diagram. See Who can do what in DataPrism.
Does Remove from semantic model keep the BI table?
No. It deletes the BI table with its columns, relationships and the report filters that target it, as the dialog says. Measures stay in the model. When reports use the table, the dialog lists them before you confirm with Delete anyway.
Why does a table show only a few columns?
A card starts at 10 fields to stay readable, with key columns and columns used in relationships first. It grows past 10 when there are more of those, so none of them is hidden. Show more lists every field; Show less collapses it again.