Group data when shredding xml to multiple tables in SQL server

Question

This follows on from this question:
http://stackoverflow.com/questions/624206/modify-xml-in-sql-server-to-add-a-root-node/

In addition to the problem posed and now fixed in that question, I am also splitting the existing table into 2, where the 2nd table is a sub-table of the first, with a foreign key back to the first.

So currently

table1(ID, col1, col2, col3....coln)

Is becoming

table1(ID, col1, col2, col3)

and

table2(PK, FK_table1, col4, col5, ...coln)

At the moment I haven't removed anything from the existing table, just created the second, and I have discovered the OUTPUT construct so that I can insert into both tables at once, including putting the PK of the 1st table into the 2nd as FK in the 1 insert - so something like;

INSERT INTO table1 col1, col2,...etc...

OUTPUT inserted.ID, col1, col2, ....
INTO table2(FK_table1, col1, col2, .... )

Select  col.value('node1[1]', 'int') col1,
 col.value('node2[1]', 'varchar(50)') col2,
....etc......
FROM @xml.nodes('//Items/Item') doc(col)

This successfully inserts into both tables at once BUT relies on the fact I haven't removed the columns from table1 yet, *AND results in a 1:1 relationship between the tables, when what I need is for table1 to just have rows for distinct col1, col2, col3 , with table2 holding the more detailed info

I'm wondering if I am going about this the right way? Am I going to be able to do this in 1 query or am I going to have to break it down? I'm thinking maybe I could shred the xml first into table1, then shred again separately into table2, joining on table1 to get the FK for the insert? Anyone done this sort of thing before?

Answer 1

Yes, I'm done similar and used both suggestions depending on what other work I need to do

Options:

1. Shred the XML once into a tabe variable SELECT DISTINCT from that into table1, SELECT into table2

2. Shred the XML twice.

I'd be inclined to do the expensive operation (shredding XML) first in one go.