Query with XML Path, Elements XSINIL and other functions - sql

Question

I am looking for guidance in understanding the logic and meaning of this script below. I have tried to break it down but I haven't been successful. I noticed that it merges the table result into one single column with the first part of the query providing the column headers only. I'm unsure though how it does cross apply when there is only one table involved. Where did the "nodes" come from when there is no node column in the table. What does fn:local-name(.)[1] mean as well as the T1.N.Nodes('/*').
Appreciate all the help. Admin here and never had developed a query before.
select top 1 (
select '^^'+T2.N.value('fn:local-name(.)[1]', 'varchar(max)')
from (
select T.*
for xml path(''), type, ELEMENTS XSINIL
) as T1(N)
cross apply T1.N.nodes('/*') as T2(N)
for xml path(''), type, ELEMENTS XSINIL
).value('substring((.)[1], 3)', 'varchar(max)') as systemname
from dbo.tablename as T with (nolock)
UNION ALL
select (
select '^^'+ISNULL(T2.N.value('./text()[1]', 'varchar(max)'),'')
from (
select T.*
for xml path(''), type, ELEMENTS XSINIL
) as T1(N)
cross apply T1.N.nodes('/*') as T2(N)
for xml path(''), type, ELEMENTS XSINIL
).value('substring(./text()[1], 3)', 'varchar(max)')
from dbo.tablename as T with (nolock)
where 1=1

Answer 1

As mentioned, this is a pretty convoluted way of generating a flat-file.
Breaking it down step-by-step:
We start with a single row of values:
select top 1 ...
from dbo.tablename as T with (nolock)
We then make a correlated sub-query as follows:
Take all columns, and turn it into a single XML, keeping null elements and returning it as the xml type
from (
select T.*
for xml path(''), type, ELEMENTS XSINIL
) as T1(N)
Shred it into separate rows, one per node (column), the * selects all nodes. Since , type was used, T1.N is an xml data type, and we can shred it into rows with .nodes
cross apply T1.N.nodes('/*') as T2(N)
From that, select a string ^^ plus the node name, and create another XML. fn:local-name(.)[1] gets you the name of the current node, rather than the inner text value.
Since this is an unnamed column, all the nodes get lumped together. Then use .value('substring((.)[1], 3)' to get that whole string, stripping off the extra ^^. Effectively, this is the same as STRING_AGG
(
select '^^'+T2.N.value('fn:local-name(.)[1]', 'varchar(max)')
for xml path(''), type, ELEMENTS XSINIL
).value('substring((.)[1], 3)', 'varchar(max)') as systemname
Then we union that with the same thing again, this time selecting all rows, and the values rather than the node names
UNION ALL
select (
select '^^'+ISNULL(T2.N.value('./text()[1]', 'varchar(max)'),'')
from (
select T.*
for xml path(''), type, ELEMENTS XSINIL
) as T1(N)
cross apply T1.N.nodes('/*') as T2(N)
for xml path(''), type, ELEMENTS XSINIL
).value('substring(./text()[1], 3)', 'varchar(max)')
from dbo.tablename as T with (nolock)
where 1=1
The results is a big flat-file delimited by ^^
Notes:
I leave you to think about the consequences of nolock, it's generally a very bad idea.
I have no idea why where 1=1 is there, it does nothing.
ELEMENTS XSINIL was not necessary on the outer for xml, because the value is never null.
varchar(max) should be nvarchar(max) if there is any chance of other languages getting in here.
As mentioned, STRING_AGG would have been a better idea on newer versions of SQL Server
OPENJSON can do all of the above in a less convoluted fashion
And to be honest, bcp.exe and other tools are much better at this anyway