Currently, when data gets loaded into a Direct Lake semantic model, all the rows of the requested columns get loaded into the semantic model.
We want the ability to add a row filter (similar to Filter in Power Query) to tables in a Direct Lake semantic model, to only load filtered rows into the semantic model's memory (e.g. only rows where Year >= 2025).
1 Comment
dbeavon3Memorable Member
This is extremely important for large, multi-year models. The vast majority of users only care about the current and prior year. The majority of our models do not need more than that. There is no point in the delays that we encounter while loading unnecessary data into RAM. For directlake-on-onelake, this could be implemented by selectively excluding deltatable partitions.
