Explanation of using FOR XML PATH('') to Concatenate Rows in SQL

I was recently shown a cool way to concatenate an unlimited number of rows into a single cell in SQL, but was left wondering how it worked under the hood. The trick involves using FOR XML PATH('') to build multiple values and then cast them back into a single string. Starting from a simple query, and getting increasingly more complex, let’s look at how this actually works.

I’ll do everything on SQLFiddle. You can follow along by starting your own fiddle or clicking the results link for each step. Let’s start off by creating a basic table called Person with two columns and three rows of data:

Create A Table

Start by making a table:

CREATE TABLE Person 
(
  FirstName varchar(20), 
  LastName varchar(20)
);
INSERT INTO Person
  (FirstName, LastName)
VALUES
  ('Kyle', 'Mit'),
  ('Bob', 'Builder'),
  ('Jimi', 'Hendrix');

Plain SQL Query

This is just a run of the mill SQL query returning 3 rows of data:

SELECT FirstName, LastName
FROM Person 

Results:

FirstName LastName
Kyle Mit
Bob Builder
Jimi Hendrix

SQL Query as XML Grouped by ‘Person’

Here, the output of the Query is formatted into XML with each row represented as a Root node named ‘Person’ and each column as a child element:

SELECT FirstName, LastName
FROM Person 
FOR XML PATH('Person')

Results:

<Person>
    <FirstName>Kyle</FirstName>
    <LastName>Mit</LastName>
</Person>
<Person>
    <FirstName>Bob</FirstName>
    <LastName>Builder</LastName>
</Person>
<Person>
    <FirstName>Jimi</FirstName>
    <LastName>Hendrix</LastName>
</Person>

XML Query With Custom Named Child Nodes

The name of each child node under person is determined by the final column name (this is important later). When columns are selected, those are used by default, however you can manually give any column a specific name.
Note: The name of the child nodes has changed based on our select statement:

SELECT FirstName AS First, LastName AS Last
FROM Person 
FOR XML PATH('Person')

Results:

<Person>
    <First>Kyle</First>
    <Last>Mit</Last>
</Person>
<Person>
    <First>Bob</First>
    <Last>Builder</Last>
</Person>
<Person>
    <First>Jimi</First>
    <Last>Hendrix</Last>
</Person>

XML Query with comma and NAMED child node

In SQL, when you perform any kind of aggregation or selection function to a column, it no longer applies a default name. We’d like to concatenate our fields with a comma, so we’ll add a comma to the select. As an intermediary step, we’ll explicitly specify a name just to show that the comma itself isn’t modifying the result set:

SELECT ',' + LastName AS Last
FROM Person 
FOR XML PATH('Person')

Results:

<Person>
    <Last>,Mit</Last>
</Person>
<Person>
    <Last>,Builder</Last>
</Person>
<Person>
    <Last>,Hendrix</Last>
</Person>

XML Query with comma and UNNAMED child node

By removing the explicit namimg, SQL isn’t able to guess the column name. Consequently, the data is just stuffed into the Person Element.
From the FOR XML PATH documentation:

Any column without a name will be inlined. For example, computed columns or nested scalar queries that do not specify column alias will generate columns without any name.

NOTE: This is a very important step towards concatenation. We need to somehow get rid of the xml markup around our data and we’ve dropped an entire node. In the next step we’ll find out how to get rid of the second one.

SELECT ',' + LastName
FROM Person 
FOR XML PATH('Person')

Results:

<Person>,Mit</Person>
<Person>,Builder</Person>
<Person>,Hendrix</Person>

XML Query with no root node

The command we’ve been using all along is FOR XML which has four different modes:

  1. RAW
  2. AUTO
  3. EXPLICIT
  4. PATH

In this case, we’re using PATH, which will wrap the data elements in a parent element named for the table from which it came. Optionally, you can add a string parameter to Path to override the root element name. The last trick:

If you specify a zero-length string, the wrapping element is not produced.

By manually specifying the path as an empty string, all the data elements are shown right next to each other.
Note the file name: this is still returning XML (just poorly formatted XML)

SELECT ',' + LastName
FROM Person 
FOR XML PATH('')

Results:

XML_F52E2B61-18A1-11D1-B105-00805F49916B
,Mit,Builder,Hendrix

Get result of the XML sub query

By selecting the result of the entire query, we transform the XML into a value:

SELECT (
        SELECT ',' + LastName
        FROM Person 
        FOR XML PATH('')
)

Results:

COLUMN_0
,Mit,Builder,Hendrix

Remove the First Comma

To remove the leading comma, we’ll use STUFF(character_expression, start, length, replaceWith_expression). The following query will start at position 1 and replace the 1st character with '':

SELECT STUFF((
              SELECT ',' + LastName
              FROM Person 
              FOR XML PATH('')
), 1, 1, '')

Results:

COLUMN_0
Mit,Builder,Hendrix

Cast into VARCHAR for type safety

Finally, we’ll take the whole query and make sure it’s of type VARCHAR. Also, for good measure, we’ll give the returned column a name:

SELECT CAST(STUFF((
              SELECT ',' + LastName
              FROM Person 
              FOR XML PATH('')
), 1, 1, '') AS VARCHAR(MAX)) AS LastNames

Results:

LastNames
Mit,Builder,Hendrix

And that’s how to go from constructing XML to a relatively simple concatenation. There are other ways to do this and other customizations you can add to this query, but this should help you understand a little bit more of the fundamentals underneath it all.

6 comments:

  1. Thank you very much! Very helpful breakdown

    ReplyDelete
  2. Awesome and very useful explanation. Thank you

    ReplyDelete
  3. Detailed article with nice presentation. Thank you.

    ReplyDelete
  4. Thank you So much for your expalanation...:)

    ReplyDelete
  5. Your annotated bibliography is far more than just a simple list of references listed out to detail the author, the title and so on. The annotated part means that you either have to provide a summary https://essaysrescue.com/edubirdie-review/ of the document in question or you have to assess the work in relation to your specific topic and research.

    ReplyDelete
  6. Women are also in business, so it is important for them to have a competent resume, for this I advise this site https://www.insightssuccess.com/resumes-tips-for-women-in-business/, because there you can learn about Tips for writing a resume for women in business

    ReplyDelete