Hi PeterGoertz 

Always change the data type of columns before making sure that data agent is working. The problem here is python doesn't support particular data type and parquet with _delta_log need a specific data type ( STRING etc..). Also, Azure SQL using different data type ( for ex. VARCHAR in Azure SQL ) 

so, make sure that you should to have correct data type and use python data agent or Dataflow Gen II. Either or 

Hi PeterGoertz ,

Thanks for your input. I did validate the column types already even reduced the target to a single column/row with standard types but the issue still persists. The error message itself

Failed to execute 'ariaNotify' on 'Document'. The provided value 'none' is not a valid enum value of type AriaNotifyPriority

Points to a Fabric Dataflows Gen2 front-end bug introduced with the latest release. The mashup script runs fine, but the web client fails when it tries to announce a status update with "priority" "none", which isn’t allowed in the FluentUI control.

So while checking types is always a good practice (especially for parquet vs. SQL/Python agent compatibility), in this case the root cause is the UI regression.

Go ahead and raise a support ticket with Microsoft and include the session ID so the product team can trace it. To raise a support ticket with Microsoft so the issue is officially logged and can also help influence future improvements:
Create a Microsoft Support Ticket.

Thanks,
Akhil.

I am getting the same issue as detailed below.Could you help  please!

---------- Message ----------
Failed to execute 'ariaNotify' on 'Document': Failed to read the 'priority' property from 'AriaNotificationOptions': The provided value 'none' is not a valid enum value of type AriaNotifyPriority.

---------- Stack ----------
React.useCallback(()@webpack://node_modules/@fluentui/react-aria/lib/AriaLiveAnnouncer/useAriaNotifyAnnounce.js:21:23
announce()@webpack://@powerquery/monolith-ui/ClientShared/Scripts/Components/Dialog/DialogRenderer.tsx:33:45

---------- Session ID ----------
7a064a6a-bdb6-4cf5-b342-f2be20ffb917

---------- Mashup script ----------
[StagingDefinition = [Kind = "FastCopy"]]
section Section1;
shared SalesOrderDetail = let
Source = Parquet.Document(AzureStorage.DataLakeContents("https://datalake0610.dfs.core.windows.net/data/parquet/SalesOrderDetail.parquet")),
#"Removed other columns" = Table.SelectColumns(Source, {"OrderQty", "ProductID", "SalesOrderID", "SalesOrderDetailID"})
in
#"Removed other columns";
shared SalesOrderHeader = let
Source = Parquet.Document(AzureStorage.DataLakeContents("https://datalake0610.dfs.core.windows.net/data/parquet/SalesOrderHeader.parquet")),
#"Removed columns" = Table.RemoveColumns(Source, {"Status", "OnlineOrderFlag", "SalesOrderNumber"}),
#"Changed column type" = Table.TransformColumnTypes(#"Removed columns", {{"OrderDate", type date}}),
#"Removed columns 1" = Table.RemoveColumns(#"Changed column type", {"TotalDue"}),
#"Added custom" = Table.AddColumn(#"Removed columns 1", "TotalAmt", each [SubTotal]+[TaxAmt]+[Freight]),
#"Changed column type 1" = Table.TransformColumnTypes(#"Added custom", {{"TotalAmt", type number}})
in
#"Changed column type 1";
[DataDestinations = {[Definition = [Kind = "Reference", QueryName = "FactSales_DataDestination", IsNewTarget = false], Settings = [Kind = "Manual", AllowCreation = false, ColumnSettings = [Mappings = {[SourceColumnName = "SalesOrderID", DestinationColumnName = "SalesOrderID"], [SourceColumnName = "OrderDate", DestinationColumnName = "OrderDate"], [SourceColumnName = "CustomerID", DestinationColumnName = "CustomerID"], [SourceColumnName = "SalesPersonID", DestinationColumnName = "SalesPersonID"], [SourceColumnName = "TerritoryID", DestinationColumnName = "TerritoryID"], [SourceColumnName = "SubTotal", DestinationColumnName = "SubTotal"], [SourceColumnName = "TaxAmt", DestinationColumnName = "TaxAmt"], [SourceColumnName = "Freight", DestinationColumnName = "Freight"], [SourceColumnName = "TotalAmt", DestinationColumnName = "TotalAmt"], [SourceColumnName = "SalesOrderDetailID", DestinationColumnName = "SalesOrderDetailID"], [SourceColumnName = "OrderQty", DestinationColumnName = "OrderQty"], [SourceColumnName = "ProductID", DestinationColumnName = "ProductID"]}], DynamicSchema = false, UpdateMethod = [Kind = "Replace"], TypeSettings = [Kind = "Table"]]]}]
shared FactSales = let
Source = Table.NestedJoin(SalesOrderHeader, {"SalesOrderID"}, SalesOrderDetail, {"SalesOrderID"}, "SalesOrderDetail", JoinKind.LeftOuter),
#"Expanded SalesOrderDetail" = Table.ExpandTableColumn(Source, "SalesOrderDetail", {"OrderQty", "ProductID", "SalesOrderDetailID"}, {"OrderQty", "ProductID", "SalesOrderDetailID"})
in
#"Expanded SalesOrderDetail";
shared FactSales_DataDestination = let
Pattern = Fabric.Warehouse([HierarchicalNavigation = null, CreateNavigationProperties = false]),
Navigation_1 = Pattern{[workspaceId = "41518f36-d390-4d15-81f6-f17c3efa77d3"]}[Data],
Navigation_2 = Navigation_1{[warehouseId = "6a38ab53-f2e6-47b5-b1d7-d0b239f9e9db"]}[Data],
TableNavigation = Navigation_2{[Schema = "dbo", Item = "FactSales"]}[Data]
in
TableNavigation;

Same issue.

Solution: AriaNotify Error in Fabric Dataflows

The Error

If you're encountering this error in Fabric Dataflows:

This is caused by how Power Query handles streaming data transformations and triggers UI progress notifications.

Root Cause

The error occurs when performing column transformations on streaming data from API sources (like Dynamics 365 Business Central, Dataverse, REST APIs, etc.).

When you transform columns on streaming data:

1. Data flows incrementally from the API
2. Each chunk triggers UI progress updates
3. The UI's accessibility notification system gets called repeatedly
4. A bug in the FluentUI component tries to set notification priority to "none" (invalid value)
5. The error is thrown and can cause the dataflow to fail

The Solution: Buffer Before Transform

The fix is simple: buffer your data before performing transformations.

❌ Problem Pattern (Causes Error)

let
    Source = Dynamics365BusinessCentral.ApiContentsWithOptions(...),
    Navigation_1 = Source{[Name = "Environment"]}[Data],
    Navigation_2 = Navigation_1{[Name = "Advanced"]}[Data],
    Navigation_3 = Navigation_2{[Name = "api/v2.0"]}[Data],
    Navigation_4 = Navigation_3{[Name = "entities", Signature = "table"]}[Data],

    // ❌ Transforming streaming data - triggers the bug
    #"Changed column type" = Table.TransformColumnTypes(
        Navigation_4,
        {
            {"modifiedAt", type datetime},
            {"createdAt", type datetime},
            {"status", type text}
        }
    ),
    #"Replaced values" = Table.ReplaceValue(
        #"Changed column type",
        "_x0020_",
        " ",
        Replacer.ReplaceText,
        {"status"}
    )
in
    #"Replaced values"

✅ Solution Pattern (Prevents Error)

let
    Source = Dynamics365BusinessCentral.ApiContentsWithOptions(...),
    Navigation_1 = Source{[Name = "Environment"]}[Data],
    Navigation_2 = Navigation_1{[Name = "Advanced"]}[Data],
    Navigation_3 = Navigation_2{[Name = "api/v2.0"]}[Data],
    Navigation_4 = Navigation_3{[Name = "entities", Signature = "table"]}[Data],

    // ✅ Buffer the data FIRST - load completely into memory
    Buffer_Data = Table.Buffer(Navigation_4),

    // ✅ Now transform the buffered data - no streaming, no UI bug
    #"Changed column type" = Table.TransformColumnTypes(
        Buffer_Data,
        {
            {"modifiedAt", type datetime},
            {"createdAt", type datetime},
            {"status", type text}
        }
    ),
    #"Replaced values" = Table.ReplaceValue(
        #"Changed column type",
        "_x0020_",
        " ",
        Replacer.ReplaceText,
        {"status"}
    )
in
    #"Replaced values"

Why This Works

Streaming vs Buffered Execution

Why You Should Buffer Anyway (Beyond Bug Fix)

Even without this bug, buffering before transformations is a best practice for several reasons:

1. Performance Optimization

• Prevents repeated API calls: Without buffering, each downstream operation can trigger re-evaluation and re-fetch from the API
• Faster transformations: Column operations on in-memory data are much faster than streaming operations
• Reduced API throttling: Single API call instead of potential multiple evaluations

2. Predictable Refresh Behavior

• Consistent data snapshots: All transformations work on the same dataset
• Avoid partial refresh issues: No risk of data changing mid-transformation
• Better error handling: Failures are easier to diagnose

3. Lower API Costs

• Some APIs charge per call or have rate limits
• Buffering ensures you only call the API once per refresh
• Prevents accidental duplicate calls from query folding failures

4. Improved Dataflow Stability

• Reduces memory pressure from streaming operations
• More predictable resource usage
• Fewer timeout issues on large datasets

When to Buffer

Best Practice Pattern

Here's a complete best-practice pattern for API-based dataflows:

let
    // 1. Connect to API
    Source = Dynamics365BusinessCentral.ApiContentsWithOptions(
        "PROD",
        null,
        null,
        [UseReadOnlyReplica = true]
    ),

    // 2. Navigate to your table
    Navigation_1 = Source{[Name = "PROD-ENV"]}[Data],
    Navigation_2 = Navigation_1{[Name = "Advanced"]}[Data],
    Navigation_3 = Navigation_2{[Name = "company/v2.0"]}[Data],
    Navigation_4 = Navigation_3{[Name = "glEntries", Signature = "table"]}[Data],

    // 3. Remove unnecessary columns BEFORE buffering (reduces memory)
    Remove_Columns = Table.RemoveColumns(
        Navigation_4,
        {"dimensionValue", "glAccount"},
        MissingField.Ignore
    ),

    // 4. BUFFER - this is the key step
    Buffer_Data = Table.Buffer(Remove_Columns),

    // 5. Now safely perform all transformations
    Format_Columns = Table.TransformColumnTypes(
        Buffer_Data,
        {
            {"documentType", type text},
            {"postingDate", type datetime},
            {"amount", type number}
        },
        "en-US"  // Use locale for date parsing
    ),

    // 6. Additional transformations (replace, merge, etc.)
    Decode_Text = Table.TransformColumns(
        Format_Columns,
        {
            {"documentType", each Text.Replace(_, "_x0020_", " "), type text},
            {"description", each Text.Replace(_, "_x0020_", " "), type text}
        }
    )
in
    Decode_Text

Additional Tips

Optimize Buffer Placement

1. Remove columns first: Buffer after removing unnecessary columns to reduce memory usage
2. Filter early: If possible, filter data before buffering (but after to refresh window for incremental)
3. Buffer once: Don't buffer multiple times in the same query

Example: Incremental Refresh with Buffer

let
    Source = API_Call,
    Navigation = Navigate_To_Table,

    // Filter to incremental window FIRST (query folding)
    Filtered = Table.SelectRows(
        Navigation,
        each [modifiedAt] >= RangeStart and [modifiedAt] < RangeEnd
    ),

    // NOW buffer the filtered data
    Buffer_Data = Table.Buffer(Filtered),

    // Then transform
    Transformed = Table.TransformColumnTypes(Buffer_Data, ...)
in
    Transformed

Summary

References

• Microsoft Docs: Table.Buffer
• Power Query Performance Tips

Tested and verified with Dynamics 365 Business Central API, Dataverse, and REST API sources in Microsoft Fabric.