用于XML自动问题的SQL Server

Question

Hi I am trying to get my db to output XML and I am succeeding somewhat.

When I do a standard select with various joins ect and all works fine, I add the following line:

for xml auto, root('MyRoot')

This generated the required XML

However, when I use a cast or convert in the select one of my tags get combined into its parent and I dont have any understanding why or how to fix it. Everything else is perfect.

Below is the code which outputs the XML as I want it

select  Tbl1.id, 
        'xyz   ' as [randCol], 
        Tbl2.id,
        Tbl2.name, 
        aDate as [date] --(aDate is the date field, date here is not data type, but a colum name for output) 
from dbo.table1 as Tbl1 
inner join dbo.table3 as Tbl3 
on Tbl1.id = Tbl3.table1id 
inner join table2 as Tbl2 
on Tbl3.table2id = Tbl2.id 
where Tbl1.id = 1 

for xml auto, root('MyRoot')

So the resulting XML is like this:

<MyRoot>
  <Tbl1 id="1" randCol="xyz   ">
    <Tbl2 id="10001" name="John">
      <Tbl3 date="2011-10-19T22:59:00" />
    </Tbl2>
    <Tbl2 id="10002" name="Brian">
      <Tbl3 date="2011-10-19T22:59:00" />
    </Tbl2>
    <Tbl2 id="10003" name="Jimmy">
      <Tbl3 date="2011-05-19T23:00:00" />
    </Tbl2>
  </Tbl1>
</MyRoot>

The problem is that when I replace the "aDate as date" line with the following line to format the date my XML gets messed up.

CONVERT(date,aDate) as [date]

This causes the following output:

<MyRoot>
  <Tbl1 id="1" randCol="xyz   ">
    <Tbl2 id="10001" name="dave" date="2010-11-17" />
    <Tbl2 id="10002" name="harry" date="2010-11-16" />
    <Tbl2 id="10003" name="lenny" date="2010-06-15" />
  </Tbl1>
</MyRoot>

What has caused this change and what do I do to get the date formatted the way I want it and keep it on a separate line like in the first example.

EDIT2: Below is an ERD of the db (I removed the explanation of the erd and just put an erd. also with some sample data: Following is the dataset I want in my xml.

id    randCol   id        name      date
1     xyz       10001     John      2011-10-19
1     xyz       10002     Brian     2011-10-19
1     xyz       10003     Jimmy     2011-05-19

a full dataset run with select * will show the following: (for the purposes of explaining the joins).

id  table1id   table2id   aDate                 id      name
1   1          10001      2011-10-19 22:59:00   10001   John
1   1          10002      2011-10-19 22:59:00   10002   Brian
1   1          10003      2011-05-19 23:00:00   10003   Jimmy

What I am trying to acheive is XML output in the following format.

<MyRoot>
  <Tbl1 id="1" randCol="xyz   ">
    <Tbl2 id="10001" name="John">
      <Tbl3 date="2011-10-19" />
    </Tbl2>
    <Tbl2 id="10002" name="Brian">
      <Tbl3 date="2011-10-19" />
    </Tbl2>
    <Tbl2 id="10003" name="Jimmy">
      <Tbl3 date="2011-05-19" />
    </Tbl2>
  </Tbl1>
</MyRoot>

However I need to acheive this without using Explisit, so need to know if I should be using Raw, Auto or Path, and with what combination of prameters.

Answer 1

I think the problem is that, with the conversion (as with any reasonably complex function), the resulting column no longer "belongs" to Tbl3 (imagine a function taking columns from multiple tables - which table would it "belong" to?). As Using AUTO Mode says:

This does not provide much control over the shape of the XML generated from a query result. ... Using EXPLICIT Mode and Using PATH Mode provide more control

and,

When a column in the SELECT clause cannot be associated with any of the tables identified in the FROM clause, as in the case of an aggregate column or computed column, the column is added in the XML document in the deepest nesting level in place when it is encountered in the list. If such a column appears as the first column in the SELECT clause, the column is added to the top element.

So I'd suggest switching to using PATH . In simplifying your query, I think you've oversimplified your joins, since they're all now on the same id value, but the following generates the right shape of query:

declare @table1 table (
    id int not null
)
declare @table2 table (
    id int not null,
    name varchar(10) not null
)
declare @table3 table (
    table1id int not null,
    table2id int not null,
    aDate datetime not null
)

insert into @table1 (id) select 1
insert into @table2 (id,name) select 10001,'John' union all select 10002,'Brian' union all select 10003,'Jimmy'
insert into @table3 (table1id,table2id,aDate)
select 1,10001,'2011-10-19T22:59:00' union all
select 1,10002,'2011-10-19T22:59:00' union all
select 1,10003,'2011-05-19T23:00:00'
select  Tbl1.id as [@id],
        'xyz   ' as [@randCol],
        (select 
          Tbl2.id as [@id],
          Tbl2.name as [@name],
          CONVERT(date,aDate) as [Tbl3/@date] --(aDate is the date field, date here is not data type, but a colum name for output)
        from
          @table3 as Tbl3 
             inner join
          @table2 as Tbl2 
             on
                Tbl3.table2id = Tbl2.id 
        where
          Tbl1.id = Tbl3.table1id 
        for xml path('Tbl2'), type)
from @table1 as Tbl1 
where Tbl1.id = 1 

for xml path('Tbl1'), root('MyRoot')

Result:

<MyRoot>
  <Tbl1 id="1" randCol="xyz   ">
    <Tbl2 id="10001" name="John">
      <Tbl3 date="2011-10-19" />
    </Tbl2>
    <Tbl2 id="10002" name="Brian">
      <Tbl3 date="2011-10-19" />
    </Tbl2>
    <Tbl2 id="10003" name="Jimmy">
      <Tbl3 date="2011-05-19" />
    </Tbl2>
  </Tbl1>
</MyRoot>

Answer 2

I think you have to use XML Path in this case. Try something like this:

select  Tbl1.id,
        'xyz   ' as [randCol],
        (SELECT Tbl2.id,
                Tbl2.name,
                (convert(date,aDate,102)) as [date]
                FROM (dbo.table1 as Tbl1 inner join table3 as Tbl3 on Tbl1.id = Tbl3.id inner join table2 as Tbl2 on Tbl3.id = Tbl2.id where Tbl1.id = 1)
                FOR XML PATH('Table2'), TYPE
        ) 
from table1
FOR XML path('Table1'), root('myroot'), ELEMENTS;

If it doesnt work, can you post your tables' structures so its easier to test?