What you would like to accomplish: Introduce catalog-level “join groups” in Fabric/SQL to enable multiple columns across different tables (e.g., fact and dimension tables) to share a consistent dictionary encoding for join keys. This would ensure that identical string or UUID values are assigned the same internal integer code, making joins and group-bys on these keys significantly faster and more memory-efficient. DuckDB Feature Request I made: Shared / Common Dictionary Encoding for Join Keys includes more details on the motivation and potential approaches. BigQuery Feature Request I made https://issuetracker.google.com/issues/442600668  One important detail to highlight is that typical dictionary encoding is local to a column or file, so even if two tables both use dictionary-encoded columns (e.g., customer_id as a string), the integer codes won't match between them. This forces the engine to fall back to comparing the original string values during joins, completely negating the performance benefits of encoding. That’s the core issue this proposal aims to address. How this might work: Allow explicit declaration of “join groups” at the dataset or project level, so that specific columns across tables share the same dictionary domain. Example of a possible SQL API: -- Define a join group to align dictionary codes across tables CREATE JOIN GROUP customer_id_group ( fact_customer_daily.customer_id, fact_customer_orders_daily.customer_id ); -- Optionally rebuild or refresh the dictionary ALTER JOIN GROUP customer_id_group REBUILD; Internally, Fabric/SQL would persist and reuse a common dictionary for the grouped columns, ensuring that equivalent string/UUID values map to identical encoded integer IDs across tables and partitions. If applicable, reasons why alternative solutions are not sufficient: Manual mapping tables: Today, teams often maintain surrogate key lookup tables to replace string/UUID identifiers with integer IDs. This adds complexity, operational overhead, and data engineering effort. Per-table dictionary encoding (as in Parquet/ORC): While these formats support dictionaries, they are scoped to row groups or files, and don’t provide cross-table consistency. Other information (workarounds you have tried, documentation consulted, etc): Current practice in many warehouses (includingFabric/SQL) is to maintain mapping tables for surrogate integer keys derived from string or UUID identifiers. While effective, this approach requires extra ETL complexity and maintenance. Implementing catalog-level shared dictionaries inFabric/SQL could provide operational simplicity and significant performance improvements for users with large, join-heavy analytic workloads. That said, I also want to acknowledge that it's entirely possible systems like Fabric/SQL may already be implementing something along these lines internally, and as an external user, I wouldn’t necessarily have visibility into whether or how that’s happening. ... View more

Currently complext type (json, struct) are supported in Onelake endpoint but not in SQL endpoint which prevents these important payload data types to be explored or used in Power BI. In big data world such fields are very common as many data generated through IoT are partially parsed in individual columns and raw json payload is left for future use or as on need basis. Before parquet - we use to store them as MapType which make them easy to further use. Adding this feature will do both - parsing those field easy for end user and also migration of existing data possible. ... View more

I recently came across that scalar functions are now available in sql end points for warehouses and lakehouses however you are unable to call them within CTEs and other sql objects , please make this feature available  ... View more

When connecting to SQL Database from dataflow, Power BI reads/load "datetimeoffset "  as DATE/TIME/TIMEZONE, but in order to set incremental refresh we need just DATE/TIME. If change it to DATE/TIME on SQL Databse, searching is changing from "seek" to "scan" and searching time is increased.  So we would like Power BI load "datetimeoffset " as DATE/TIME ... View more

In Fabric Data Warehouse I can select multiple queries and delete them in a single operation (hold down CTRL or SHIFT and select the queries I want to delete, then right click and click delete).   However, in Fabric SQL Database it doesn't seem to be possible to select and delete multiple queries this way.   Please make it possible, thanks! ... View more

Hi,   I am looking for the connector to connect database using pyspark notebook in the same Workspace, and unfortunatlly I have not get the right soluction. could you please help me on how can we connect?   Thanks in Advance, Venkat ... View more

We are having issues inititiating the mirrored tables for large tables. There is very little info on the net about removing the timeout. It would be nice if that is not a restriction for seeding new large tables ... View more

Hi, I think it is very importante that the Fabric SQL database workload could support fully Power BI, also with Direct Lake mode, like it happens for a lakehouse or a warehouse. Fabric SQL database should be fully integrated in the Fabric platform. Thanks ... View more

I suggest including the field parameters functionality in Azure Analysis Services,same as what already exists in power bi ... View more

Overtime queries ran on a sql end point in a lake house build up in the my queries folder. They have to be individually deleted which can take a lot of time. There should be a way to delete all or highlight an delete multiple queries  ... View more

When conectint to a Fabric SQLDB in SSMS  its is very difficult as we need to also enter the datbase name  (this is not so for any other SQL Edition)   We request that the default database for users be set the the Fabric SQLDB, so that they do not need to specify the database name in connections   It makes no sense to make master the defautl database if no one is allowed connect to it. ... View more

As an experienced SQL developer, the ability to quickly covert columns to lists greatly improves workflows.   I have developed functions that convert columns to lists that are of much use to me. Further, my column-to-list functions can be set to do a column-to-bracketed list or a column-to-quoted list. That functionality is a game changer.    Simple idea, but I guanrentee that SQL developers will appreciate it. Most do not develop their own column-to-list functions. Instead they often copy and paste lists in excel or 3rd party apps to convert.  ... View more

I would suggest that any data scientists who use Power BI do outlier detection directly from a data connection in Power BI inside a pivot table. You can easily program the Interquartile Range (IQR) statistical outlier rule into the creation of your pivot table from a data connection and have the results visualized inside your final pivot table. It works great and is faster than trying to pull data using Power BI onto a woksheet and then doing analysis from there. I created this video so that you can see how the rule actually works to analyze and visualize a simple column of data and then how a dashboard can be directly created from it. ... View more

Currently, if you start a manual refresh for a dataset (Azure SQL in my case), then start a refresh schedule for that dataset before the manual refresh finishes, the refresh schedule will be wiped when the manual refresh finishes (I have reproduced this many times).   This is either a bug, or this feature should be changed. The current behavior of wiping scheduled refreshes makes no sense. ... View more

We want to create Power BI data connection in our on premise gateway server to point to the secondary node of SQL server instead of the primary. But currently there is no way we can put a parameter ApplicationIntent to ReadOnly for same. This feature should be available ... View more

Currently the server name for SQLDB (and DW) is a double GUID of {tenant-Id-workspace-id.}database.fabric.microsoft.com   This causes a lot of issues as 1. This is impossible to remember 2. We ahve no way to distinguish test from PRD   We request user frirnedly DNS Names like  Prodata-Sales-DEV.database.fabric.microsoft.com   This would help clearly label an worksapce as DEV or PRD and avoid confusion and make the platform easier to work with., ... View more

Currently when using SQLDB if we enter the server name / dns into SSMS or another tool it will fail as the connection defaults to master and the yser has no rights. We then have to manually add database, which is a GUID iso impossible to remember!   Please allow the user(s) public rightsr in the master so that we can connect in SSMS and other tools and then select the database from pull down.   This is how every other edition of SQL works! ... View more

When we connect to the DB we have to enter the Database Name  whgich is a GUID. Please make this database name equal to the same name as trhe SQL End Point and database name in Fabric. This imporves usability, and allows for the same 3 part TSQL to run on SQL DB v SQL End Point. At the moment we have different TSQL Statements if we query SQLDB v SQL End Point which is not optimum   ... View more