Power BI - using UTF8_LCASE column returns error Unable to understand the type for column

[wgrusznis]

What happened?

Using ADBC driver in Power BI returns error for COLLATE UTF8_LCASE columns
hard to say exactly where is it coming from - PBI wraps drivers in powerQuery layer, almost sure the driver itself is the csharp one

An error occurred in the ‘’ query. Expression.Error: Unable to understand the type for column
Details:
c2

This is PowerQuery that I used for testing and replicating this bug.

= Value.NativeQuery(
Databricks.Catalogs(
"<DATABRICKS_HOST>",
"<SQL_WAREHOUSE_PATH>",
[
Catalog = "<CATALOG_NAME>",
Database = null,
QueryTags = null,
EnableAutomaticProxyDiscovery = "enabled",
Implementation = "2.0"
]
){[Name = "<CATALOG_NAME>", Kind = "Database"]}[Data],
"select#(lf)
collation('Test') as t1#(lf)
,collation('Test' COLLATE UTF8_LCASE) as t2#(lf)
,'Test' as c1#(lf)
,'Test' COLLATE UTF8_LCASE as c2",
null,
[EnableFolding = true]
)

Stack Trace

No response

How can we reproduce the bug?

PowerQuery Implementation = "2.0" switches to ADBC driver, leaving it as null returns to ODBC mode (where UTF8_LCASE works correctly)

= Value.NativeQuery(
Databricks.Catalogs(
"<DATABRICKS_HOST>",
"<SQL_WAREHOUSE_PATH>",
[
Catalog = "<CATALOG_NAME>",
Database = null,
QueryTags = null,
EnableAutomaticProxyDiscovery = "enabled",
Implementation = "2.0"
]
){[Name = "<CATALOG_NAME>", Kind = "Database"]}[Data],
"select#(lf)
collation('Test') as t1#(lf)
,collation('Test' COLLATE UTF8_LCASE) as t2#(lf)
,'Test' as c1#(lf)
,'Test' COLLATE UTF8_LCASE as c2",
null,
[EnableFolding = true]
)

Environment/Setup

windows, power bi Version: 2.153.1206.0 64-bit (April 2026) , databricks via SQL warehouse (v 2026.10)

[lidavidm]

I believe this was answered here, does this match your problem? #4245

[wgrusznis]

Hi David, thank you for prompt answer. It looks very much related with my issue.
although @hb0313 error on fabric is much clearer in powerbi desktop is less explicit

I build the csharp driver and replaced it in my local power bi installation and the error is still there.
I'll try to extract the metadata queries power bi produces to see if there is anything unusual there.

[wgrusznis]

alright I made some digging in Power BI connector (M query/PowerQuery wrapper)
first finding: although I replaced csharp ADBC driver with my build still what it outputs for UTF8_LCASE is ' STRING COLLATE UTF8_LCASE' -> but here might be catch, PowerBI likes to revert some of the files when it noticed they were altered.

but then I looked in the PBI databricks connector. in DatabricksAdbc.pqm there is already code that handles type mappings
originalType -> adjustedType

i added one line in there

                    else if Text.StartsWith(connectorType, "STRING") then "STRING"

build new version and this actually solves it.
So I think actually the Power Bi wrapper needs to be patched, and the ADBC driver works as expected.

[davidhcoe]

To be clear, are you using the driver from https://github.com/adbc-drivers/databricks/? What version of Power BI do you have? cc @eric-wang-1990

[wgrusznis]

I'm using latest as for today power bi desktop - Version: 2.153.1206.0 64-bit (April 2026)
it ships with the csharp adbc driver
for testing i build https://github.com/adbc-drivers/databricks/ csharp version and replaced it in powerbi folder
if I understand correctly eric's change STRING COLLATE [anything...] should still be exposed as STRING
https://github.com/adbc-drivers/databricks/pull/285/commits

but for UTF8_LCASE i still saw -> STRING COLLATE UTF8_LCASE, but don't get me wrong, it might be that what I tried to override with my cshar build was not in correct placa/might be re-writen by powerbi if checskum were not what expected etc.

but in my opinion actually the powerbi connector wrapper is missing the line to just treat STRING [add anything here] as STRING.
right now it sees STRING COLLATE UTF8_LCASE and says "I have no idea what to do with that" (query. Expression.Error: Unable to understand the type for column)

[eric-wang-1990]

@davidhcoe Can you check if that commit lands on Mar 3rd made into Apr release? adbc-drivers/databricks#285 I guess it did not.
@wgrusznis Can you try build the ADBC driver and copy all the content under \databricks\csharp\artifacts\AdbcDrivers.Databricks\Debug\net472 to C:\Program Files\Microsoft Power BI Desktop\bin\Adbc Drivers\Databricks?

Note you need to copy the net472 folder, and Debug/Release based on which you build.

[eric-wang-1990]

@wgrusznis Actually I just reproed with latest driver, it seems this only occurs for NativeQuery, if you are dealing with a table the error does not exist.
Yeah I think we need to patch the connector in this case since for NativeQuery pbiConnector use different way to get metadata, I will put a fix

[wgrusznis]

awesome, thank you @eric-wang-1990 for looking into it.