DAX error "Formula is invalid" when updating underlying data, but not changing the formula?

Question

Hello,
first of all: I use the 64-bit versions of Excel 2010 and PowerPivot on Windows Server 2008 R2.
I use this formula to calculate the median of my data:
MINX( FILTER( VALUES( TableName[ColumnName] ),
              CALCULATE( COUNTROWS( TableName ),
                         TableName[ColumnName] <= EARLIER( TableName[ColumnName]
              > COUNTROWS( TableName ) * 0.5 ),
      TableName[ColumnName] )
The data comes from a view on an MS SQL Server and has about 3.5 million rows. With one dataset ("dataset 1"), everything is working out fine and VERY fast ;-). When I change the view on the SQL server to filter for different data ("dataset 2", the result
set still containing about 3.5 million rows) and update the PowerPivot data and then the pivot table, the status bar reads "Executing OLAP query..." and excel.exe utilizes one CPU core at 100% for a long time and its memory usage increases significantly, but
nothing happens. If I interrupt that process by pressing Esc, I get the following error (original German text included):
============================
Fehlermeldung: (Error message:)
============================
Ausnahme von HRESULT: 0x800A03EC (Exception from HRESULT: ...)
Das Median-Feld konnte der PivotTable nicht hinzugefügt werden, weil die Formel ungültig ist. (Could not add the field "Median" to the PivotTable because the formula is invalid.)
============================
Aufrufliste: (Stack trace:)
============================
Server stack trace:
Exception rethrown at [0]:
   bei System.Runtime.Remoting.Proxies.RealProxy.HandleReturnMessage(IMessage reqMsg, IMessage retMsg)
   bei System.Runtime.Remoting.Proxies.RealProxy.PrivateInvoke(MessageData& msgData, Int32 type)
   bei Microsoft.Office.Interop.Excel.PivotTable.AddDataField(Object Field, Object Caption, Object Function)
   bei Microsoft.AnalysisServices.Modeler.FieldList.ExcelInterOpUtil.AddToDataFields(ICalculatedMember calculatedMember, Int32 positionIndex, Boolean isSpecialColumnBasedNamedSetPresent)
   bei Microsoft.AnalysisServices.Modeler.FieldList.ExcelInterOpUtil.AddToDataFields(ICalculatedMember calculatedMember, Int32 positionIndex, Boolean isSpecialColumnBasedNamedSetPresent)
   bei Microsoft.AnalysisServices.Modeler.FieldList.GeminiPivot.SetFieldOrientation(FieldLocation location, IGeminiColumn column, Int32 positionIndex)
   bei Microsoft.AnalysisServices.Modeler.FieldList.GeminiPivot.SetFieldOrientation(FieldLocation location, IGeminiColumn column, Int32 positionIndex)
   bei Microsoft.AnalysisServices.Modeler.FieldList.GeminiPivot.AddField(IGeminiColumn column, Int32 index)
   bei Microsoft.AnalysisServices.Modeler.FieldList.FieldListControl.fieldsTreeView_AfterCheck(Object sender, TreeViewEventArgs e)
============================
I have analyzed the two datasets for differences and found the following ones:
Dataset 2 has also negative values in its ColumnName column, dataset 1 does not. Filtering dataset 2 (by changing the view on the SQL server) so that the column contains only positive values does not help.
A displayed column contains text with square brackets in it in dataset 2. Changing the SQL view to replace them with an empty string (replace(column2, '[', '')) does not help.
I do not know what else to try. Can anybody help me? The two datasets are very large, but if anyone can give a recommendation how to export them in a reasonable size, I can make them available.
Best regards
Michael

Answer 1

Hello Javier,
I use a new measure (button "New Measure" in the toolbar or option in the context menu of the table). Here are the data samples:
Dataset 1:
Probability    RThreshold    SThreshold    vector    value
0    -1    -1    DHTTestApp: Total GET Success Ratio    0.98
0    -1    -1    DHTTestApp: Total GET Success Ratio    1
0    -1    -1    DHTTestApp: Total GET Success Ratio    1
0    -1    -1    DHTTestApp: Total GET Success Ratio    1
0    -1    -1    DHTTestApp: Total GET Success Ratio    1
0    -1    -1    DHTTestApp: Total GET Success Ratio    0.98
0    -1    -1    DHTTestApp: Total GET Success Ratio    1
0    -1    -1    DHTTestApp: Total GET Success Ratio    1
0    -1    -1    DHTTestApp: Total GET Success Ratio    1
0    -1    -1    DHTTestApp: Total GET Success Ratio    1
Dataset 2:
Probability    RThreshold    SThreshold    vector    value
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.011353711790393
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.20522161505768
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.12309191295875
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.26926457661881
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.1911946574326
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.066699727250186
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.32597014925373
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.11977454203852
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.24751410911045
0    -1    -1    [MKTBR] BaseOverlay: All nodes: Own routing trust value    0.076218041485769
In both datasets, probability has a range from 0 to 1, both thresholds from -1 to 1. In dataset 1, value has a range from 0 to 1, in dataset 2, it's -1 to 1.
Thanks for your help!