Search for a command to run...
Copy TOM objects from one model to another in Power BI / Fabric
Principal Program Manager, Microsoft Fabric CAT helping users and organizations build scalable, insightful, secure solutions. Blogs, opinions are my own and do not represent my employer.
Last week Microsoft released an open-source text embedding model called Harrier in three sizes- 270M, 0.6B and 27B. I have been testing it in my RAG pipeline and so far it has crushed all my metrics.
For Power BI Copilot and Data agents with semantic models, you must use Prep Data for AI configuration to ground the responses in the context added in Prep for AI. In this blog, I will show you how yo
At FabCon Atlanta last week, the updated notebook Copilot for data engineering and data science was announced. It brings agentic capabilities to the Copilot and is much more intelligent and Fabric-awa
Using Fabric data agent Python SDK
Use workspace monitoring to monitor MCP server use
The new version of Semantic Link Labs is out (v 0.9.11), and as always, Michael Kovalsky has added many new features that make working with Semantic Models in Fabric items much easier. One new method introduced is copy_object(), which, as the name suggests, makes copying semantic model objects from one semantic model to another a breeze. Previously, you could do this using TOM, but now it has its own function, so you can use the boilerplate function. You can also use Tabular Editor (either manually copy/paste or using C#). I have done many migrations and have always used a C# script I developed. In this case, we use Python. You can copy paste between Import < - > Direct Lake models as well (as long as the objects/properties are supported).
The method is easy to use:
#%pip install semantic_link_labs --q
import sempy.labs as labs
with labs.tom.connect_semantic_model(dataset="SourceDataset", workspace="SourceWorkspace") as tom:
# to copy Sales table from source dataset to target dataset
table = tom.model.Tables["Sales"]
tom.copy_object(
object=table,
target_dataset="TargetDataset",
target_workspace="TargetWorkspace"
)
For demo purposes, I created two models model_1 and model_2 . model_1 has relationships, calculated columns, calculated tables and measures. model_2 has the two import tables without any relationships and calc tables/calumns and measures. I want to copy from model_1 to model_2. I will define the functions:
from sempy_labs.tom import connect_semantic_model
import sempy.fabric as fabric
#if workspaces are None, current workspace is used
def copy_relationships(
source_dataset: str,
target_dataset: str,
source_workspace: str = None,
target_workspace: str = None
):
"""
Copy all relationships from the source semantic model to the target semantic model.
"""
with connect_semantic_model(dataset=source_dataset, workspace=source_workspace) as src_tom:
for rel in src_tom.model.Relationships:
print(f"Copying relationship: {rel.Name}")
src_tom.copy_object(
object=rel,
target_dataset=target_dataset,
target_workspace=target_workspace
)
print("All relationships have been copied.")
def copy_calculated_columns(
source_dataset: str,
target_dataset: str,
source_workspace: str = None,
target_workspace: str = None
):
"""
Copy all calculated columns from the source semantic model to the target semantic model.
Uses sempy.fabric.list_columns to reliably identify calculated columns.
"""
calc_cols_df = fabric.list_columns(source_dataset, workspace=source_workspace)[['Table Name', 'Column Name', 'Type']]
calculated_columns = set(
tuple(x) for x in calc_cols_df.query('Type == "Calculated"')[['Table Name', 'Column Name']].values
)
with connect_semantic_model(dataset=source_dataset, workspace=source_workspace) as src_tom:
for table in src_tom.model.Tables:
for column in table.Columns:
if (table.Name, column.Name) in calculated_columns:
print(f"Copying calculated column: {column.Name} from table: {table.Name}")
src_tom.copy_object(
object=column,
target_dataset=target_dataset,
target_workspace=target_workspace
)
print("All calculated columns have been copied.")
def copy_calculated_tables(
source_dataset: str,
target_dataset: str,
source_workspace: str = None,
target_workspace: str = None
):
"""
Copy all calculated tables from the source semantic model to the target semantic model.
"""
calculated_tables = set(
fabric.list_tables(source_dataset, workspace=source_workspace)
.query('Type == "Calculated Table"')['Name']
)
with connect_semantic_model(dataset=source_dataset, workspace=source_workspace) as src_tom:
for table in src_tom.model.Tables:
if table.Name in calculated_tables:
print(f"Copying calculated table: {table.Name}")
src_tom.copy_object(
object=table,
target_dataset=target_dataset,
target_workspace=target_workspace
)
print("All calculated tables have been copied.")
from sempy_labs.tom import connect_semantic_model
def copy_all_measures(
source_dataset: str,
target_dataset: str,
source_workspace: str = None,
target_workspace: str = None
):
"""
Copy all measures from every table in the source semantic model to the target semantic model.
"""
with connect_semantic_model(dataset=source_dataset, workspace=source_workspace) as src_tom:
for table in src_tom.model.Tables:
for measure in table.Measures:
print(f"Copying measure '{measure.Name}' from table '{table.Name}'...")
try:
src_tom.copy_object(
object=measure,
target_dataset=target_dataset,
target_workspace=target_workspace)
except:
print(f"Error with {measure.Name}, check again")
continue
print("All measures have been copied.")
copy_relationships(
source_dataset="model_1",
target_dataset="model_2",
source_workspace="a79cbb27-3cc-d64bf25ca405",
target_workspace="a79cbb27-3bf64bf25ca405"
)
copy_calculated_tables(
source_dataset="model_1",
target_dataset="model_2",
source_workspace="a79cbb27-3cc-d64bf25ca405",
target_workspace="a79cbb27-3bf64bf25ca405"
)
copy_calculated_columns(
source_dataset="model_1",
target_dataset="model_2",
source_workspace="a79cbb27-3cc-d64bf25ca405",
target_workspace="a79cbb27-3bf64bf25ca405"
)
copy_all_measures(
source_dataset="model_1",
target_dataset="model_2",
source_workspace="a79cbb27-3cc-d64bf25ca405",
target_workspace="a79cbb27-3bf64bf25ca405"
)
When you copy tables, make sure to update the connection and refresh the target semantic model to reflect the changes and ensure all objects are copied correctly. When you copy objects, all its properties are also copied (e.g. formatting, annotations etc.) You can also copy any TOM object, such as partitions, hierarchies, perspectives, and more.
You need XMLA Write enabled in the tenant and capacity settings. You also need Contributor+ access (or build permissions) to both the semantic models. This works in any Premium/Fabric workspace.
Download notebook from here.