Filtering xml records from CLOB database column

Question

Hi,
I have a xml data stored in a CLOB field in database.
I need to retrieve the contents that match my filter criteria.
Say my xml is like below,
<ROOT>
<EMP>
<EMP_NO>1</EMP_NO>
<SAL>100</SAL>
</EMP>
<EMP>
<EMP_NO>2</EMP_NO>
<SAL>200</SAL>
</EMP>
<EMP>
<EMP_NO>3</EMP_NO>
<SAL>300</SAL>
</EMP>
</ROOT>
I want to fetch the entire xml tree but filtered based on the SAL field, WHERE SAL>100.
So my resulting xml should be
<ROOT>
<EMP>
<EMP_NO>2</EMP_NO>
<SAL>200</SAL>
</EMP>
<EMP>
<EMP_NO>3</EMP_NO>
<SAL>300</SAL>
</EMP>
</ROOT>
I will be having 'n' number of columns, and I cant specify everything in COLUMNS clause.
I want to select all nodes, but filter based on few fields.
I tried using it as below, but I feel it took some time for it to display results.
select x.*
from emp t
, xmltable(
'/ROOT'
passing xmltype(t.doc)
columns VSAL number PATH './EMP/SAL'
, OTHER xmltype PATH '.'
) x
where x.VSAL>100
Please suggest me a solution that doesnt consume much time.
I think the problem with my query is OTHER xmltype PATH '.'
Here 'xmltype' datatype is expensive I think, because after I included that, the query took some time.
I cant hardcode the SAL value of 100 as its dynamic. So I am unable to use this method of filtering.
select x.*
from emp t
, xmltable(
'/ROOT[.//EMP/SAL>100]'
passing xmltype(t.doc)
) x
Thanks!
Edited by: Chit on Mar 24, 2012 10:12 AM

Answer 1

Hi,
I want to fetch the entire xml tree but filtered based on the SAL field, WHERE SAL>100.Then you have to rebuild the document keeping only the elements satisfying the condition (and yes, you can bind XQuery variables too) :
SQL> select xmlserialize(document x.doc as clob indent)
  2  from emp t
  3     , xmltable(
  4         '<ROOT>
  5          {
  6            for $i in /ROOT/EMP
  7            where $i/SAL > $minsal
  8            return $i
  9          }
10          </ROOT>'
11          passing xmlparse(document t.doc)
12                , 100 as "minsal"
13          columns doc xmltype path '.'
14       ) x
15  ;
XMLSERIALIZE(DOCUMENTX.DOCASCL
<ROOT>
  <EMP>
    <EMP_NO>2</EMP_NO>
    <SAL>200</SAL>
  </EMP>
  <EMP>
    <EMP_NO>3</EMP_NO>
    <SAL>300</SAL>
  </EMP>
</ROOT>
Or you can just delete the nodes you don't want :
SQL> var minsal number
SQL> exec :minsal := 100
PL/SQL procedure successfully completed
SQL> select xmlserialize(document
  2           deletexml( xmlparse(document t.doc)
  3                    , '/ROOT/EMP[SAL<='||:minsal||']' )
  4           as clob indent
  5         )
  6  from emp t
  7  ;
XMLSERIALIZE(DOCUMENTDELETEXML
<ROOT>
  <EMP>
    <EMP_NO>2</EMP_NO>
    <SAL>200</SAL>
  </EMP>
  <EMP>
    <EMP_NO>3</EMP_NO>
    <SAL>300</SAL>
  </EMP>
</ROOT>
Please suggest me a solution that doesnt consume much time.
I think the problem with my query is OTHER xmltype PATH '.'The real problem is the CLOB text storage.
If you want performance, use binary XMLType or object-relational.