Export data back to Lakehouse

It is possible to export the results of your analysis back to the source tables or to export the entire data set back to the source Lakehouse. The import model used to perform the Lakehouse-to-graph transformation governs which nodes and relationship types are displayed in the configuration tab.

Export is only available for graph datasets created from Fabric and for the nodes and relationships included in the original Lakehouse to graph transformation. There are no limits on adding or changing Neo4j properties of those nodes/relationships.

This feature is not available for Fabric graph datasets created by connecting to existing Aura databases.

Configuration tab

To begin setting up an Export Job for your graph dataset:

  1. Open your Fabric workspace.

  2. Select the desired Neo4j Graph Dataset.

  3. Navigate to Export using the top navigation bar.

  4. In the Configuration view, select which nodes, relationships, and properties to include in your export.

  5. Choose between Update or New Table modes (detailed below).

  6. Save the graph dataset by selecting the save icon in the navigation bar.

  7. Select the green "play" button next to Export to Lakehouse to submit your export job.

History Tab

Once you submit an export job, it will appear in the History tab, right next to the default Configuration tab. Export jobs are placed in your Fabric capacity’s worker queue. Refresh the status of all pending jobs by selecting Refresh. The History view is a summary of Fabric’s monitor page. If you want more details about a job than what is provided here, refer to the Monitor page in the left navigation bar.

Export modes

There are two modes at the bottom of the configuration tab: Update and New table.

Update mode

This mode is only available if your graph dataset’s model has one-to-one mappings from Lakehouse tables to their respective nodes and relationships.

Selected nodes, relationships, and properties are table-joined on the corresponding existing Lakehouse tables when using this mode. Newly added properties in the graph end up as new columns in the Lakehouse table. Newly added nodes or relationships end up as new rows in the Lakehouse table. Updated/changed properties of nodes or relationships are in-place updated in the Lakehouse table. Note that deleted nodes and relationships from the graph are not deleted from the Lakehouse table.

New Table mode

Selected nodes, relationships, and properties are not created in their original Lakehouse tables. Instead, the exported data is written to new tables with the supplied prefixes.

This operation overwrites data, including tables that may already exist with the same name and prefix. Be mindful of the prefixes used to avoid unintended data loss.

If your Lakehouse supports schema, you can target any existing schema in addition to the prefix.