Collation Behavior Change in lakehouse to warehouse joins.Is COLLATE DATABASE_DEFAULT Still Required

[FabricUser_5037]

Hello,

I have observed a change in behavior regarding collation when querying Lakehouse tables from a Warehouse.

Earlier :
When joining columns from a Lakehouse table with a Warehouse table, I had to explicitly specify COLLATE DATABASE_DEFAULT to avoid collation mismatch errors. For example:

SELECT k.[address] COLLATE DATABASE_DEFAULT
FROM Lakehouse.dbo.employee k
JOIN dbo.master m ON k.[name] COLLATE DATABASE_DEFAULT = m.[fullname]
 
Recently, I noticed that in similar queries involving joins between lakehouse and warehouse tables, I am no longer receiving collation related errors,even without explicitly using COLLATE DATABASE_DEFAULT.
Has there been a recent update or fix that addresses collation mismatches between Lakehouse and Warehouse tables?
Any official clarification or documentation on this change would be greatly appreciated.

[deborshi_nag · Accepted solution]

Hi @FabricUser_5037 

Looks like the last Collation update came in Aug 2025 - might your case be related to this?

Streamlining Data Management with Workspace Collation Settings in Microsoft Fabric Warehouse and SQL...

Hope this helps - please appreciate by leaving a Kudos or marking it as a Solution! 

[svenchio · Accepted solution]

Hi @FabricUser_5037  as indicated by @deborshi_nag   it seems the change you observed is due to August 2025 introduction of  workspace-level collation settings for Fabric Data Warehouse, and tied all warehouses and all SQL analytics endpoints to that workspace collation, furthgermore, new warehouses and all SQL analytics endpoints are configured based on the workspace’s Data Warehouse default collation.  

Because a Lakehouse SQL analytics endpoint is just another SQL analytics endpoint backed by OneLake,  that means that any  lakehouse-to-warehouse joins created after that change, the collations match out-of-the-box 😁🤞 and your old pattern now work even without the COLLATE DATABASE_DEFAULT clauses as you observed. 

In summary and to answer the questions: 

#1. Has there been a recent update or fix that addresses collation mismatches between Lakehouse and Warehouse tables? Yes, from Microsoft Fabric Updates Blog article published August 5 🔗 2025 https://blog.fabric.microsoft.com/en-us/blog/streamlining-data-management-with-collation-settings-in... 

#2. is COLLATE DATABASE_DEFAULT still required?  Not required anymore if both sides of the join now share the same collation, but still useful / required in mixed-collation scenario

Hope this further clarifies your question, great if you mark this as the answer to  your question, a kudos would be nice to ... best of lucks mate, cheers! 

[FabricUser_5037]

Thanks @svenchio for the detailed explanation

[deborshi_nag · Accepted solution]

Hi @FabricUser_5037 

Looks like the last Collation update came in Aug 2025 - might your case be related to this?

Streamlining Data Management with Workspace Collation Settings in Microsoft Fabric Warehouse and SQL...

Hope this helps - please appreciate by leaving a Kudos or marking it as a Solution! 

[FabricUser_5037]

Thanks @deborshi_nag for sharing the blog