This page explains the ideas behind DataPrism and the terms its screens use. Read it before you create a semantic model, or when a label or a name on a screen is unclear.
A semantic model reads from Lake and Warehouse tables. You add those tables to the model as BI tables, connect them with relationships, and define measures. A report is built on one semantic model. Its pages hold the visuals that read from that model. The new report dialog states it plainly: "Charts pull tables and measures from this model".
Charts run against live data. A saved version of a report keeps the pages and layout, not the numbers, so the figures can differ from when the version was saved. For versions and restoring, see Build a report.
Every semantic model has an owner role and a data mode. You set both when you create the model, and neither can be changed afterwards.
The owner role decides which tables the model's queries run against: the Lake and Warehouse tables that this role can access.
The data mode decides what queries read from PII columns:
| Data mode | What queries read |
|---|---|
| Clear Values | Original values, PII columns included. This is the default. |
| Tokenized PII | Tokenized stand-ins for PII columns. |
You can pick only active roles that you hold as the owner role. Administrators can pick any active role.
The model's header carries the owner role and the data mode as chips; point at them to read Owner role: {name} and Data mode: {label}. A model that uses type columns adds a Type: {name} chip; its tooltip reads Open Type and version columns, and choosing it opens the Type and version columns dialog. If the owner role is later deleted, its name keeps a (deleted) suffix in the model's access details, and the model's queries fail until it is recreated under an active role.
To create a model, see Semantic models and BI tables.
The owner role and the data mode are locked at creation. The model's access details show Locked at creation. To run a model under another owner role, import its model file and choose the new role during the import. See Model files and replacing a table.
Reports and models live in workspaces. There are three kinds that you can reach.
Your access to an item reads Can edit or View only. Edit access is the quiet default; only an item you can open but not change is marked: point at the View only marker to read View only: you can open it but not edit it. What decides this is on Who can do what in DataPrism.
Semantic models cannot be shared directly. To let a team use a model, put it in a role workspace. Reports can be shared with users and roles, and as public links. See Share a report and Workspaces, folders and the home page.
Every model has a core table, the hub of its star schema. Mark one BI table as the core table before you create relationships. Every relationship in the model must involve the core table, and a BI table without a relationship cannot join the star.
The core table carries a Core badge; point at it to read Core table for this semantic model.
Usually the core table is a fact table with measures, such as orders, events or transactions. Dimension tables, such as customers, products or dates, connect to it. For the steps, see Relationships and the core table.
Each column of a BI table has one of these types: Whole Number, Decimal Number, Date, Date Time, Time, Duration, Boolean, Text or Rejected. DataPrism detects the type from the source, and you cannot change it in DataPrism.
A column with the type Rejected has no display format, and you cannot include it in a report.
A measure is a named SQL aggregation, for example sum("amount"). The expression must be an aggregation. Reports built on the model use its measures. For how to write, test and fix them, see Measures.
Row-level security (RLS) limits which rows each viewer of a model gets. Members of the Main group bypass it and see every row.
A model's header and its cards show the switch as a chip: RLS on or RLS off, with Row-level security is active or Row-level security is off on hover.
On a report, badges after the title show the model's settings. An RLS badge means row-level security is on. A PII badge means the model uses Tokenized PII. A model in Clear Values shows no PII badge.
View as runs every chart of a report under another user's row-level security, so you can check what that user sees. See Row-level security and View as.
The screens fall back to a plain name when a real one is missing or cannot be resolved:
See Workspaces, folders and the home page for the tree and feed these labels appear in.
| Term | Meaning |
|---|---|
| Report | Pages of visuals built on one semantic model. |
| Visual | One chart, table, slicer, map or custom visual on a report page. |
| Slicer | A visual that picks the values of one field and filters the other visuals. |
| Field parameter | A list of fields or measures that a slicer lets readers switch between. |
| Filters panel | The report-level filters. Removing a filter there removes it from every page it constrains. |
| Version | A saved state of a report. The newest 100 versions are kept. |
| Public link | A link that anyone can open to view the report without signing in. |
| Subscription | A schedule that emails a report as PPTX, PDF or Excel. |
| Model AI | The AI panel that reads and changes a semantic model, with your approval. |
| BI table | A Lake or Warehouse table added to a semantic model, with its columns and types. |
| Ungrouped | The placeholder that gathers the BI tables that are not part of any semantic model. |
For the screens behind these terms, see Visuals, Filters, slicers and interactions and Email subscriptions.
No. The owner role is locked at creation. To run the model under another role, import its model file and choose the new owner role. BI tables are matched by core table name in the role you pick.
You can pick only active roles that you hold. Administrators can pick any active role. A role that is inactive or deleted is not offered to anyone.
The model may have row-level security turned on, which filters rows for each viewer. Check for the RLS badge after the report title. Members of the Main group always see every row.
Queries read tokenized stand-ins instead of the original values of PII columns. The report shows the PII badge.
A column with the type Rejected has no display format, and you cannot include it in a report.
No. DataPrism detects the type from the source. Change the column in the Lake or Warehouse table that it comes from.
Charts run against live data every time. A saved version keeps the layout, not the numbers, so figures change as the data changes. Restoring an older version brings back its layout, not its numbers.
Ungrouped is the placeholder DataPrism uses when no semantic model is selected: it gathers the BI tables that are not part of any semantic model. It is not a saved model, so model-only parts such as row-level security and the change history are not available for it. Select a semantic model to use its tools.
The item has no name yet, so the screen shows a generic fallback: Untitled report, Untitled semantic model, Untitled BI table and so on. The fallback goes away once the item gets a name.