Migration to Direct Lake 03-24-2025 13:45 PM - last edited 03-24-2025 14:54 PM

[mikova]

Download this notebook from: semantic-link-labs/notebooks/Migration to Direct Lake.ipynb at main · microsoft/semantic-link-labs ·...

Watch the video below to see a walkthrough of the Direct Lake Migration process

Install the latest .whl package

Check here to see the latest version.

%pip install semantic-link-labs

Import the library and set initial parameters

import sempy_labs as labs
from sempy_labs import migration, directlake
import sempy_labs.report as rep

dataset_name = '' #Enter the import/DQ semantic model name
workspace_name = None #Enter the workspace of the import/DQ semantic model. It set to none it will use the current workspace.
new_dataset_name = '' #Enter the new Direct Lake semantic model name
new_dataset_workspace_name = None #Enter the workspace where the Direct Lake model will be created. If set to None it will use the current workspace.
lakehouse_name = None #Enter the lakehouse to be used for the Direct Lake model. If set to None it will use the lakehouse attached to the notebook.
lakehouse_workspace_name = None #Enter the lakehouse workspace. If set to None it will use the new_dataset_workspace_name.

Create the Power Query Template file

This encapsulates all of the semantic model's Power Query logic into a single file.

migration.create_pqt_file(dataset = dataset_name, workspace = workspace_name)

Import the Power Query Template to Dataflows Gen2

• Open the OneLake file explorer and sync your files (right click -> Sync from OneLake)

• Navigate to your lakehouse. From this window, create a new Dataflows Gen2 and import the Power Query Template file from OneLake (OneLake -> Workspace -> Lakehouse -> Files...), and publish the Dataflows Gen2.

Important! Make sure to create the Dataflows Gen2 from within the lakehouse window. That will ensure that all the tables automatically map to that lakehouse as the destination. Otherwise, you will have to manually map each table to its destination individually.

Create the Direct Lake model based on the import/DQ semantic model

Calculated columns are not migrated to the Direct Lake model as they are not supported in Direct Lake mode.

import time
labs.create_blank_semantic_model(dataset = new_dataset_name, workspace = new_dataset_workspace_name, overwrite=False)

migration.migrate_calc_tables_to_lakehouse(
    dataset=dataset_name,
    new_dataset=new_dataset_name,
    workspace=workspace_name,
    new_dataset_workspace=new_dataset_workspace_name,
    lakehouse=lakehouse_name,
    lakehouse_workspace=lakehouse_workspace_name
)
migration.migrate_tables_columns_to_semantic_model(
    dataset=dataset_name,
    new_dataset=new_dataset_name,
    workspace=workspace_name,
    new_dataset_workspace=new_dataset_workspace_name,
    lakehouse=lakehouse_name,
    lakehouse_workspace=lakehouse_workspace_name
)
migration.migrate_calc_tables_to_semantic_model(
    dataset=dataset_name,
    new_dataset=new_dataset_name,
    workspace=workspace_name,
    new_dataset_workspace=new_dataset_workspace_name,
    lakehouse=lakehouse_name,
    lakehouse_workspace=lakehouse_workspace_name
)
migration.migrate_model_objects_to_semantic_model(
    dataset=dataset_name,
    new_dataset=new_dataset_name,
    workspace=workspace_name,
    new_dataset_workspace=new_dataset_workspace_name
)
migration.migrate_field_parameters(
    dataset=dataset_name,
    new_dataset=new_dataset_name,
    workspace=workspace_name,
    new_dataset_workspace=new_dataset_workspace_name
)
time.sleep(2)
labs.refresh_semantic_model(dataset=new_dataset_name, workspace=new_dataset_workspace_name)
migration.refresh_calc_tables(dataset=new_dataset_name, workspace=new_dataset_workspace_name)
labs.refresh_semantic_model(dataset=new_dataset_name, workspace=new_dataset_workspace_name)

Show migrated/unmigrated objects

migration.migration_validation(
    dataset=dataset_name,
    new_dataset=new_dataset_name, 
    workspace=workspace_name, 
    new_dataset_workspace=new_dataset_workspace_name
)

Rebind all reports using the old semantic model to the new Direct Lake semantic model

rep.report_rebind_all(
    dataset=dataset_name,
    dataset_workspace=workspace_name,
    new_dataset=new_dataset_name,
    new_dataset_workpace=new_dataset_workspace_name,
    report_workspace=None
)

Rebind reports one-by-one (optional)

report_name = '' # Enter report name which you want to rebind to the new Direct Lake model

rep.report_rebind(
    report=report_name,
    dataset=new_dataset_name,
    report_workspace=workspace_name,
    dataset_workspace=new_dataset_workspace_name)

Show unsupported objects

dfT, dfC, dfR = directlake.show_unsupported_direct_lake_objects(dataset = dataset_name, workspace = workspace_name)

print('Calculated Tables are not supported...')
display(dfT)
print("Learn more about Direct Lake limitations here: https://learn.microsoft.com/power-bi/enterprise/directlake-overview#known-issues-and-limitations")
print('Calculated columns are not supported. Columns of binary data type are not supported.')
display(dfC)
print('Columns used for relationship must be of the same data type.')
display(dfR)

Schema check between semantic model tables/columns and lakehouse tables/columns

This will list any tables/columns which are in the new semantic model but do not exist in the lakehouse

directlake.direct_lake_schema_compare(dataset=new_dataset_name, workspace=new_dataset_workspace_name)

Show calculated tables which have been migrated to the Direct Lake semantic model as regular tables

directlake.list_direct_lake_model_calc_tables(dataset=new_dataset_name, workspace=new_dataset_workspace_name)

https%3A%2F%2Fgithub.com%2Fmicrosoft%2Fsemantic-link-labs%2Fblob%2Fmain%2Fnotebooks%2FMigration%2520to%2520Direct%2520Lake.ipynb

[parry2k]

Thanks for sharing @mikova !