105

I have a table that contains a Xml column:

SELECT * 
FROM Sqm

enter image description here

A sample of the xml data of a row would be:

<Sqm version="1.2">
  <Metrics>
    <Metric id="TransactionCleanupThread.RecordUsedTransactionShift" type="timer" unit="µs" count="1" sum="21490"   average="21490"   minValue="73701"    maxValue="73701"                               >73701</Metric>
    <Metric id="TransactionCleanupThread.RefundOldTrans"             type="timer" unit="µs" count="1" sum="184487"  average="184487"  minValue="632704"   maxValue="632704"                              >632704</Metric>
    <Metric id="Database.CreateConnection_SaveContextUserGUID"       type="timer" unit="µs" count="2" sum="7562"    average="3781"    minValue="12928"    maxValue="13006"    standardDeviation="16"     >12967</Metric>
    <Metric id="Global.CurrentUser"                                  type="timer" unit="µs" count="6" sum="4022464" average="670411"  minValue="15"       maxValue="13794345" standardDeviation="1642047">2299194</Metric>
    <Metric id="Global.CurrentUser_FetchIdentityFromDatabase"        type="timer" unit="µs" count="1" sum="4010057" average="4010057" minValue="13752614" maxValue="13752614"                            >13752614</Metric>
  </Metrics>
</Sqm>

In the case of this data, I would want:

SqmId  id                                                   type   unit  count  sum      minValue  maxValue  standardDeviation  Value
=====  ===================================================  =====  ====  =====  ======   ========  ========  =================  ======
1      TransactionCleanupThread.RecordUsedTransactionShift  timer  µs    1      21490    73701     73701     NULL               73701
1      TransactionCleanupThread.RefundOldTrans              timer  µs    1      184487   632704    632704    NULL               632704
1      Database.CreateConnection_SaveContextUserGUID        timer  µs    2      7562     12928     13006     16                 12967
1      Global.CurrentUser                                   timer  µs    6      4022464  15        13794345  1642047            2299194
1      Global.CurrentUser_FetchIdentityFromDatabase         timer  µs    1      4010057  13752614  13752614  NULL               13752614
2      ...

In the end I'll actually be performing SUM(), MIN(), MAX() aggregation. But for now I'm just trying to query an xml column.

In pseudo-code, I would try something like:

SELECT
    SqmId,
    Data.query('/Sqm/Metrics/Metric/@id') AS id,
    Data.query('/Sqm/Metrics/Metric/@type') AS type,
    Data.query('/Sqm/Metrics/Metric/@unit') AS unit,
    Data.query('/Sqm/Metrics/Metric/@sum') AS sum,
    Data.query('/Sqm/Metrics/Metric/@count') AS count,
    Data.query('/Sqm/Metrics/Metric/@minValue') AS minValue,
    Data.query('/Sqm/Metrics/Metric/@maxValue') AS maxValue,
    Data.query('/Sqm/Metrics/Metric/@standardDeviation') AS standardDeviation,
    Data.query('/Sqm/Metrics/Metric') AS value
FROM Sqm

But that SQL query doesn't work:

Msg 2396, Level 16, State 1, Line 2
XQuery [Sqm.data.query()]: Attribute may not appear outside of an element

I've hunted, and it's amazing how poorly documented, or exampled, Xml querying is. Most resources rather than querying a table, query a variable; which I'm not doing. Most resources only use xml querying for filtering and selection, rather than reading values. Most resources read hard-coded child nodes (by index), rather than actual values.

Related resources that I read

Update: .value rather than .query

I tried randomly using .value, in place of .query:

SELECT
    Sqm.SqmId,
    Data.value('/Sqm/Metrics/Metric/@id', 'varchar(max)') AS id,
    Data.value('/Sqm/Metrics/Metric/@type', 'varchar(max)') AS type,
    Data.value('/Sqm/Metrics/Metric/@unit', 'varchar(max)') AS unit,
    Data.value('/Sqm/Metrics/Metric/@sum', 'varchar(max)') AS sum,
    Data.value('/Sqm/Metrics/Metric/@count', 'varchar(max)') AS count,
    Data.value('/Sqm/Metrics/Metric/@minValue', 'varchar(max)') AS minValue,
    Data.value('/Sqm/Metrics/Metric/@maxValue', 'varchar(max)') AS maxValue,
    Data.value('/Sqm/Metrics/Metric/@standardDeviation', 'varchar(max)') AS standardDeviation,
    Data.value('/Sqm/Metrics/Metric', 'varchar(max)') AS value
FROM Sqm

But that also doesn't work:

Msg 2389, Level 16, State 1, Line 3 XQuery [Sqm.data.value()]:
'value()' requires a singleton (or empty sequence), found operand of type 'xdt:untypedAtomic *'

s.dragos
  • 582
  • 2
  • 8
  • 24
Ian Boyd
  • 246,734
  • 253
  • 869
  • 1,219

4 Answers4

133

Actually you're close to your goal, you just need to use nodes() method to split your rows and then get values:

select
    s.SqmId,
    m.c.value('@id', 'varchar(max)') as id,
    m.c.value('@type', 'varchar(max)') as type,
    m.c.value('@unit', 'varchar(max)') as unit,
    m.c.value('@sum', 'varchar(max)') as [sum],
    m.c.value('@count', 'varchar(max)') as [count],
    m.c.value('@minValue', 'varchar(max)') as minValue,
    m.c.value('@maxValue', 'varchar(max)') as maxValue,
    m.c.value('.', 'nvarchar(max)') as Value,
    m.c.value('(text())[1]', 'nvarchar(max)') as Value2
from sqm as s
    outer apply s.data.nodes('Sqm/Metrics/Metric') as m(c)

sql fiddle demo

Roman Pekar
  • 107,110
  • 28
  • 195
  • 197
  • 2
    How do i get the *"value"* of the node itself? There seems to be no way to `select m.*` to see the secret, magical, intermediate table it constructed. What's the syntax to query the value of an element? e.g. the value of `8675309` is *"8675309"* – Ian Boyd Oct 03 '13 at 17:34
  • 2
    @IanBoyd sorry, missed that, see updated. You can use '.' or text if there could be nested elements – Roman Pekar Oct 03 '13 at 17:42
  • 4
    What do the aliases `s`, `m`, and `c` represent in this query? – Ian R. O'Brien Nov 10 '16 at 15:23
  • 6
    @IanR.O'Brien `m` is resultset returned by `nodes()` function, `s` is `sqm` table itself, `c` is column with data type xml in the resultset returned by `nodes()` function – Roman Pekar Nov 10 '16 at 16:32
  • This is perfect – Sarath Mohandas Jan 28 '21 at 09:29
15

I've been trying to do something very similar but not using the nodes. However, my xml structure is a little different.

You have it like this:

<Metrics>
    <Metric id="TransactionCleanupThread.RefundOldTrans" type="timer" ...>

If it were like this instead:

<Metrics>
    <Metric>
        <id>TransactionCleanupThread.RefundOldTrans</id>
        <type>timer</type>
        .
        .
        .

Then you could simply use this SQL statement.

SELECT
    Sqm.SqmId,
    Data.value('(/Sqm/Metrics/Metric/id)[1]', 'varchar(max)') as id,
    Data.value('(/Sqm/Metrics/Metric/type)[1]', 'varchar(max)') AS type,
    Data.value('(/Sqm/Metrics/Metric/unit)[1]', 'varchar(max)') AS unit,
    Data.value('(/Sqm/Metrics/Metric/sum)[1]', 'varchar(max)') AS sum,
    Data.value('(/Sqm/Metrics/Metric/count)[1]', 'varchar(max)') AS count,
    Data.value('(/Sqm/Metrics/Metric/minValue)[1]', 'varchar(max)') AS minValue,
    Data.value('(/Sqm/Metrics/Metric/maxValue)[1]', 'varchar(max)') AS maxValue,
    Data.value('(/Sqm/Metrics/Metric/stdDeviation)[1]', 'varchar(max)') AS stdDeviation,
FROM Sqm

To me this is much less confusing than using the outer apply or cross apply.

I hope this helps someone else looking for a simpler solution!

Michael Bray
  • 14,998
  • 7
  • 42
  • 68
Ryan Dorendorf
  • 151
  • 1
  • 4
10

use value instead of query (must specify index of node to return in the XQuery as well as passing the sql data type to return as the second parameter):

select
    xt.Id
    , x.m.value( '@id[1]', 'varchar(max)' ) MetricId
from
    XmlTest xt
    cross apply xt.XmlData.nodes( '/Sqm/Metrics/Metric' ) x(m)
Moho
  • 15,457
  • 1
  • 30
  • 31
7

I don't understand why some people are suggesting using cross apply or outer apply to convert the xml into a table of values. For me, that just brought back way too much data.

Here's my example of how you'd create an xml object, then turn it into a table.

(I've added spaces in my xml string, just to make it easier to read.)

DECLARE @str nvarchar(2000)

SET @str = ''
SET @str = @str + '<users>'
SET @str = @str + '  <user>'
SET @str = @str + '     <firstName>Mike</firstName>'
SET @str = @str + '     <lastName>Gledhill</lastName>'
SET @str = @str + '     <age>31</age>'
SET @str = @str + '  </user>'
SET @str = @str + '  <user>'
SET @str = @str + '     <firstName>Mark</firstName>'
SET @str = @str + '     <lastName>Stevens</lastName>'
SET @str = @str + '     <age>42</age>'
SET @str = @str + '  </user>'
SET @str = @str + '  <user>'
SET @str = @str + '     <firstName>Sarah</firstName>'
SET @str = @str + '     <lastName>Brown</lastName>'
SET @str = @str + '     <age>23</age>'
SET @str = @str + '  </user>'
SET @str = @str + '</users>'

DECLARE @xml xml
SELECT @xml = CAST(CAST(@str AS VARBINARY(MAX)) AS XML) 

--  Iterate through each of the "users\user" records in our XML
SELECT 
    x.Rec.query('./firstName').value('.', 'nvarchar(2000)') AS 'FirstName',
    x.Rec.query('./lastName').value('.', 'nvarchar(2000)') AS 'LastName',
    x.Rec.query('./age').value('.', 'int') AS 'Age'
FROM @xml.nodes('/users/user') as x(Rec)

And here's the output:

enter image description here

Mike Gledhill
  • 27,846
  • 7
  • 149
  • 159