The simplest way to get more performance on select on xmltype column?

Question

We have a table with xmltype column:
CREATE TABLE "TK"."PLAN_2" ("UNID" NUMBER NOT
NULL,
"XMLSTRING" "TK"."XMLTYPE",
CONSTRAINT "SYS_C001142583" PRIMARY KEY("UNID")
USING INDEX
TABLESPACE "TX"
STORAGE ( INITIAL 64K NEXT 0K MINEXTENTS 1 MAXEXTENTS
2147483645 PCTINCREASE 0) PCTFREE 10 INITRANS 2 MAXTRANS 255)
The data in column xmlstring looks like this:
<Agent>
<OP End="2520000" Id="001_8232_0_91" Origin="Autos" Quantity="1.0 ea" Start="2520000">
<OSP Key="ADM_ID" String="001_8232_0" />
<OSP Key="AddAutoNr" String="0" />
<OSP Key="CURRENT_STATE" String="Tages-Vorgabe" />
<OLP Key="ChUser">
<OIP Key="(T_SP) test2" Long="1218125408541" />
<OIP Key="(T_SP) admin" Long="1219071385346" />
<OIP Key="(T_SP) test2" Long="1237540381906" />
</OLP>
<OSP Key="Debicode" String="9903" />
<OSP Key="Resource" String="SBN" />
<OSP Key="TrafficDay" String="91" />
<OSP Key="AutoNrSchema" String="001" />
<OSP Key="AutoNumber" String="8232" />
<OPE End="0" Id="Prod" Origin="Autos|Prod" Start="0">
<OProd Duration="0" End="0" Id="Produce" OriginDirect="Autos|Prod|Produce" Quantity="1.0 ea" Start="0" Type="Produce">
<OSP Key="RESelection.Parameter" String="Resource" />
<Resource Id="Autos" />
<Alloc From="OP|001_8232_0_91|Prod|Produce" Id="" Quantity="1.0 ea" To="CO|001_8232_0_91|001_8232_0_91" Type="PRODUCEREQ_TO_CONSUMEREQ" />
</OProd>
</OPE>
<OPE End="2520000" Id="Step_0" Origin="Autos|Step" Start="0">
<OQP Key="Delay" Overridden="true" Quantity="0.0 s" />
<OQP Key="Distance" Quantity="44.84 km" />
<OQP Key="Duration" Quantity="2520.0 s" />
<OQP Key="Fahrt" Quantity="2520.0 s" />
<OSP Key="From" String="BS" />
<OSP Key="I1" String="0" />
<OSP Key="PLACE" String="1" />
<OQP Key="StartDelay" Overridden="true" Quantity="0.0 s" />
<OQP Key="TWait" Overridden="true" Quantity="0.0 s" />
<OSP Key="To" String="OL" />
<OUse Duration="2520000" End="2520000" Id="LStep_0" Name="L" OriginDirect="Autos|Step|LStep" Quantity="1.0 ea" Start="0" Type="Step">
<OSP Key="I2" String="0" />
<OIP Key="LPOSITION" Long="1" Overridden="true" />
<OSP Key="PLACE" String="1" />
<OQP Key="Wait" Overridden="true" Quantity="0.0 s" />
<Resource Id="vs_RABe525_13" />
</OUse>
<OUse Duration="480000" End="0" Id="LPre_0" Name="L" OriginDirect="Autos|Step|LPre" Quantity="1.0 ea" Start="-480000" Type="Pre">
<OQP Key="Fahrbereit" Overridden="true" Quantity="2 min" />
<OSP Key="From" Overridden="true" String="BS" />
<OSP Key="I2" String="0" />
<OSP Key="PLACE" String="1" />
<OSP Key="StepIndex" String="0" />
<OSP Key="To" Overridden="true" String="BS" />
<OQP Key="Vorbereitung" Overridden="true" Quantity="6 min" />
<Resource Id="vs_RABe525_13" />
</OUse>
<OUse Duration="600000" End="3120000" Id="LPost_0" Name="L" OriginDirect="Autos|Step|LPost" Quantity="1.0 ea" Start="2520000" Type="Post">
<OQP Key="Abstellzeit" Overridden="true" Quantity="4 min" />
<OSP Key="From" Overridden="true" String="OL" />
<OSP Key="I2" String="0" />
<OQP Key="Nachbereitung" Overridden="true" Quantity="6 min" />
<OSP Key="PLACE" String="1" />
<OSP Key="StepIndex" String="0" />
<OSP Key="To" Overridden="true" String="OL" />
<Resource Id="vs_RABe525_13" />
</OUse>
<OUse Duration="2520000" End="2520000" Id="FStep_0" Name="F" OriginDirect="Autos|Step|FStep" Quantity="1.0 ea" Start="0" Type="Step">
<OSP Key="I2" String="0" />
<OSP Key="PLACE" String="1" />
<OQP Key="Wait" Overridden="true" Quantity="0.0 s" />
<Resource Id="ps_P112" />
</OUse>
<OUse Duration="480000" End="0" Id="FPre_0" Name="F" OriginDirect="Autos|Step|FPre" Quantity="1.0 ea" Start="-480000" Type="Pre">
<OQP Key="Fahrbereit" Overridden="true" Quantity="2 min" />
<OSP Key="From" Overridden="true" String="BS" />
<OSP Key="I2" String="0" />
<OSP Key="PLACE" String="1" />
<OSP Key="StepIndex" String="0" />
<OQP Key="Taxi" Overridden="true" Quantity="0 s" />
<OSP Key="To" Overridden="true" String="BS" />
<OQP Key="Vorbereitung" O verridden="true" Quantity="6 min" />
<OQP Key="Wegzeit" Overridden="true" Quantity="0 s" />
<Resource Id="ps_P112" />
</OUse>
<OUse Duration="600000" End="3120000" Id="FPost_0" Name="F" OriginDirect="Autos|Step|FPost" Quantity="1.0 ea" Start="2520000" Type="Post">
<OQP Key="Abstellzeit" Overridden="true" Quantity="4 min" />
<OSP Key="From" Overridden="true" String="OL" />
<OSP Key="I2" String="0" />
<OQP Key="Nachbereitung" Overridden="true" Quantity="6 min" />
<OSP Key="PLACE" String="1" />
<OSP Key="StepIndex" String="0" />
<OQP Key="Taxi" Overridden="true" Quantity="0 s" />
<OSP Key="To" Overridden="true" String="OL" />
<OQP Key="Wegzeit" Overridden="true" Quantity="0 s" />
<Resource Id="ps_P112" />
</OUse>
</OPE>
</OP>
</Agent>
We try to select:
SELECT UNID, extractValue(value(OUSE_XMLSTRING), '/OUse/Resource/@Id') OUSE_RESOURCE,
extract(XMLSTRING, '/Agent/OP/OSP[@Key = "TrafficDay"]/@String') TT_DAY,
extract(XMLSTRING, '/Agent/OP/@Start') OP_START,
extract(XMLSTRING, '/Agent/OP/@End') OP_END,
extract(XMLSTRING, '/Agent/OP/OSP[@Key = "ADM_ID"]/@String') ADM_ID,
extractValue(value(OUSE_XMLSTRING), '/OUse/@Start') OUSE_START,
extractValue(value(OUSE_XMLSTRING), '/OUse/@End') OUSE_END
FROM tk.plan_2, table(XMLSequence(extract(XMLSTRING, '//Agent/OP//OPE//OUse'))) OUSE_XMLSTRING
where extractValue(value(OUSE_XMLSTRING), '/OUse/Resource/@Id') = 'vs_RABe525_13';
with this explain plan:
PLAN_TABLE_OUTPUT
Plan hash value: 3476314316
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
| 0 | SELECT STATEMENT | | 101 | 22018 | 31 (0)| 00:00:01 |
| 1 | NESTED LOOPS | | 101 | 22018 | 31 (0)| 00:00:01 |
| 2 | TABLE ACCESS FULL | PLAN_2 | 2 | 432 | 2 (0)| 00:00:01 |
|* 3 | COLLECTION ITERATOR PICKLER FETCH| XMLSEQUENCEFROMXMLTYPE | | | | |
Predicate Information (identified by operation id):
3 - filter(EXTRACTVALUE(VALUE(KOKBF$),'/OUse/Resource/@Id')='vs_RABe525_13')
We want to select on the 'ID', which occurs more than once in the xmlstring and the table gonna be big, so we think we need some kind of index on the ID-Identifier. What will be the best way to do this?

Answer 1

Which version of Oracle (4 digits)?
If 10.2 or greater, look into using XMLTable and that is the replacement for the table(XMLSequence(extract()) approach. Also, avoid the use of // unless you really need it. I doubt it will make a difference in your situation but just a good performance approach in general.
We want to select on the 'ID', which occurs more than once in the xmlstringI assume you are going to want the count(...) > 1 from the XML to tell you which you are selecting?
I also have questions about what that XMLType is being stored as but that depends upon what your version is regarding what storage options you have.
Edited by: A_Non on Jun 7, 2010 10:55 AM
Forgot to add, look in the FAQ (under your sign-in name on the upper right) for how to use the tag to better format your posts.