SQL Server 2005 "FOR XML PATH" Grouping

Question

I'm trying to generate a XML document from the SQL Server 2005 database by using "FOR XML" construct.

There are two simple tables in the database with a one-to-many relationship:

1) Magazines

 | Id | Number | Name       |
 ----------------------------
 | 53 | 0001   | Magazine 1 |
 | 54 | 0002   | Magazine 2 |
 | 55 | 0003   | Magazine 3 |

2) Articles

 | Id |   Title   | MagazineId | Size |
 --------------------------------------
 | 1  | Article 1 |        53  | 1205 |
 | 2  | Article 2 |        53  | 817  |
 | 3  | Article 3 |        54  | 1570 |
 | 4  | Article 4 |        54  | 2510 |
 | 5  | Article 5 |        55  | 910  |

Let's assume I have to find all magazines that have an article with size greater than 1000 and to produce the following xml:

<Magazines>
    <Magazine Id="53">
      <Number>0001</Number>
      <Articles>
        <Article Id="1">
           <Title>Article 1</Title>
           <Size>1205</Size>
        </Article>
      </Articles>
    </Magazine>
    <Magazine Id="54">
      <Number>0002</Number>
      <Articles>
        <Article Id="3">
          <Title>Article 3</Title>
          <Size>1570</Size>
        </Article>
        <Article Id="4">
          <Title>Article 4</Title>
          <Size>2510</Size>
        </Article>
      </Articles>
    </Magazine>
</Magazines>

I'm trying to produce such xml by using a "PATH" mode:

SELECT Magazines.Id AS "@Id",
       Magazines.Number AS "Number",
       Articles.Id AS "Articles/Article/@Id",
       Articles.Title AS "Articles/Article/Title",
       Articles.Size AS "Articles/Article/Size"
FROM Magazines INNER JOIN Articles ON Magazines.Id = Articles.MagazineId
WHERE Articles.Size > 1000
FOR XML PATH('Magazine'), ROOT('Magazines'), TYPE

It will produce the following xml:

<Magazines>
  <Magazine Id="53">
    <Number>0001</Number>
    <Articles>
      <Article Id="1">
        <Title>Article 1</Title>
        <Size>1205</Size>
      </Article>
    </Articles>
  </Magazine>
  <Magazine Id="54">
    <Number>0002</Number>
    <Articles>
      <Article Id="3">
        <Title>Article 3</Title>
        <Size>1570</Size>
      </Article>
    </Articles>
  </Magazine>
  <Magazine Id="54">
    <Number>0002</Number>
    <Articles>
      <Article Id="4">
        <Title>Article 4</Title>
        <Size>2510</Size>
      </Article>
    </Articles>
  </Magazine>
</Magazines>

So there are two elements for the Magazine with Id="54" (one for each article) and this is the problem.

I can rewrite the query using a subquery like this:

SELECT M.Id AS "@Id",
       M.Number AS "Number",
    (SELECT Articles.Id AS "@Id",
               Articles.Title AS "Title",
               Articles.Size AS "Size"
  FROM Articles
  WHERE Articles.MagazineId = M.Id
  FOR XML PATH('Article'), ROOT('Articles'), TYPE
    )
FROM Magazines AS M 
FOR XML PATH('Magazine'), ROOT('Magazines'), TYPE

And this produce the following xml:

<Magazines>
  <Magazine Id="53">
    <Number>0001</Number>
    <Articles>
      <Article Id="1">
        <Title>Article 1</Title>
        <Size>1205</Size>
      </Article>
      <Article Id="2">
        <Title>Article 2</Title>
        <Size>817</Size>
      </Article>
    </Articles>
  </Magazine>
  <Magazine Id="54">
    <Number>0002</Number>
    <Articles>
      <Article Id="3">
        <Title>Article 3</Title>
        <Size>1570</Size>
      </Article>
      <Article Id="4">
        <Title>Article 4</Title>
        <Size>2510</Size>
      </Article>
    </Articles>
  </Magazine>
  <Magazine Id="55">
    <Number>0003</Number>
    <Articles>
      <Article Id="5">
        <Title>Article 5</Title>
        <Size>910</Size>
      </Article>
    </Articles>
  </Magazine>
</Magazines>

But by using a subquery I can not filter magazines by articles columns (without complex additional queries).

The "FOR XML AUTO" mode is not suitable, because it is very simple and does not support some "PATH" features (like attributes using @, ROOT, etc..)

So, Is there any possibility in "PATH" mode to group inner table data like in "AUTO" mode?

Thank you!

Answer 1

Well, you can get one step closer by specifying the "size > 1000" inside your subquery:

SELECT M.Id AS "@Id",
       M.Number AS "Number",
    (SELECT Articles.Id AS "@Id",
               Articles.Title AS "Title",
               Articles.Size AS "Size"
     FROM Articles
     WHERE Articles.MagazineId = M.Id
           AND Articles.Size > 1000
     FOR XML PATH('Article'), ROOT('Articles'), TYPE
    )
FROM Magazines AS M 
FOR XML PATH('Magazine'), ROOT('Magazines'), TYPE

What you're missing now is the fact you'll still get magazines that have no articles with a size > 1000. You can eliminate those something like this:

SELECT M.Id AS "@Id",
       M.Number AS "Number",
    (SELECT Articles.Id AS "@Id",
               Articles.Title AS "Title",
               Articles.Size AS "Size"
      FROM Articles
      WHERE Articles.MagazineId = M.Id
            AND Articles.Size > 1000
      FOR XML PATH('Article'), ROOT('Articles'), TYPE
    )
FROM Magazines AS M 
WHERE EXISTS(SELECT * FROM Articles 
             WHERE Articles.MagazineId = M.Id 
               AND Articles.Size > 1000)
FOR XML PATH('Magazine'), ROOT('Magazines'), TYPE

(untested, I don't have a SQL server at hand right now).

Does that work for you? Does it give you the magazines and articles you're looking for?

Marc

Answer 2

Use FOR XML explicit. Is longest code to write but disappear the performance issues provocate by sub queries.