FOR XML causing high cardinality in update statement.

Question

Wasn't really sure how to word this, but here it goes. I have a simple update statement that uses a sub-query to concatenate a column from a record set using the most recently recommended fashion of FOR XML. The row estimations through the XML reader show
as 42 million rows, but the actual rows are 18k'ish.
update d
set d.DraftDocumentReadyDate = stuff(isnull((select ', ' + convert(varchar, wa.WorkAudit_Date, 110)
from livelink.WAuditTrail wa inner join livelink.KUAF ku1
on wa.WorkAudit_PerformerID_Name = ku1.Name
where d.VolumeID = wa.WorkAudit_WorkID
and wa.WorkAudit_Task_Title = 'Draft Document'
and wa.WorkAudit_Status = 21
for xml path (''), type).value('.','nvarchar(4000)'), ''), 1, 2, '')
from #ltbl_DataDump d
Adding image of cardinality:
John M. Couch

Answer 1

I would change this query a bit (since I like to simplify):
;with cte as (Select workAudit_workID, stuff(isnull((select ', ' + case when ku1.FirstName IS NULL and ku1.LastName IS NULL and ku1.Name IS NULL then 'None'
when ku1.FirstName IS NULL and ku1.LastName IS NULL then ku1.Name
else ku1.FirstName + ' ' + ku1.LastName
end
from livelink.WAuditTrail wa inner join livelink.KUAF ku1
on wa.WorkAudit_PerformerID_Name = ku1.Name
where wa.WorkAudit_WorkID = wa1.workAudit_workID
and wa.WorkAudit_Task_Title = 'Draft Document'
and wa.WorkAudit_Status = 24
for xml path (''), type).value('.','nvarchar(4000)'), ''), 1, 2, '') as DocName,
STUFF((SELECT ', ' + convert(varchar(20), wa.WorkAudit_Date, 110)
from livelink.WAuditTrail wa
where wa.WorkAudit_WorkID = wa1.workAudit_workID
and wa.WorkAudit_Task_Title = 'Draft Document'
and wa.WorkAudit_Status = 24
for xml path ('')),1,2,'') as Dates
FROM livelink.WAuditTrail wa1
where wa1.WorkAudit_Task_Title = 'Draft Document'
and wa1.WorkAudit_Status = 24
GROUP BY wa1.workAudit_WorkId)
MERGE #ltbl_DataDump d
USING cte ON d.WolumeID = cte.workAudit_Workid
WHEN MATCHED
THEN UPDATE
SET
DraftDocumentStartedPerformer = cte.DocName,
d.DraftDocumentStartedDate = cte.Dates
For every expert, there is an equal and opposite expert. - Becker's Law
My blog
My TechNet articles