Lesson 05 of 7 · The model in Fabric
Same model. New source.
Good news: your measures don't need to learn anything new. Only the plumbing changes.
How much of the model changes?
The kit's model is a folder of files: tables, columns, relationships, measures and roles. It's moving from Import to .
How much of the TMDL do you expect to change?
One table, before and after
Toggle between the two versions of the Supply table. The measure and the column don't move; the partition does.
// Supply.tmdl, Import (today)
table Supply
measure 'Total supply (kt)' = SUM ( Supply[Kilotonnes] )
column Kilotonnes
dataType: decimal
partition Supply = m
mode: import
source =
let
Source = gold_Supply
in
Source
A quick way to tell the flavours apart: Direct Lake on OneLake uses the AzureStorage.DataLake connector; Direct Lake on SQL uses a SQL connector.
Stays, changes, or moves upstream?
The measure Total supply (kt)
The relationships between Supply, Commodity, Country, Year and Scenario
The report opening on Base case in the scenario slicer
The row-level security roles
Each table's partition
The Power Query steps and the file-path parameter
The Criticality band calculated column
A user-defined hierarchy on the Year table
Tools and checks
An open-source notebook library, Semantic Link Labs, can migrate an Import model to Direct Lake. What won't it bring across?
After the switch, the kit's validator fails every partition check. What do you do?
Three things to keep
- 1The semantic model barely moves: measures, relationships and roles read the same.
- 2What changes is each table's partition, which now names a Delta table through one shared expression.
- 3Power Query steps and move upstream. Update the spec before the files.