RE: Preventing Nested Table Full Access

From: Mark W. Farnham <mwf_at_rsiz.com>
Date: Mon, 1 Apr 2013 12:17:48 -0400
Message-ID: <035301ce2ef4$748208e0$5d861aa0$_at_rsiz.com>



Hmm. One strategy for the sort of thing you have described is to use a hash function on your input parameters, add a non-uniquely indexed hash_value column to the table, and then you'll only have to paw through the rows that match the hash of your array arguments. Pawing through the nested table arrays is probably the slow part.  

Oh well, it is April 1.  

mwf  

From: Anupam Pandey [mailto:my.oralce_at_gmail.com] Sent: Monday, April 01, 2013 12:04 PM
To: Mark W. Farnham
Cc: oracle Freelists; anupam pandey
Subject: Re: Preventing Nested Table Full Access  

Thanks Mark and Andrew for looking into it .  

We have a application from where we get these list of arrays . Based on the array values we formulate a query and return it back to the application . Logic to create the query is very complicated and takes around 15 to 20 sec for execution . So we thought of building a table

where we will store this array values and the queury which was returned to the application .  

So first time when application calls my procedure I will find out the query through process and will store the array elements and the corresponding query in that table .Next time when we get the same call we can return the query for same from the table itself rather than going through entire process.  

But this is now becoming bottleneck as the query for selecting the "query" is doing FTS and also taking time .    

Is there anyway to index the nested table or any other way I can try out for speeding up the selection process ..  

Thanks,

Anupam      

On Mon, Apr 1, 2013 at 9:10 PM, Mark W. Farnham <mwf_at_rsiz.com> wrote:

well, you have no indexes on any of your predicate columns, so it indeed will have to do a full table scan.

Your query looks vaguely like you might want to be querying a bitmap index, but you seem to be storing hard wired predicate values in nested tables, and you're trying to look up a candidate set of clob payloads.

I am at a loss to discern what you are actually trying to accomplish. Perhaps a little more exposition about what you're trying to accomplish would be helpful.

-----Original Message-----

From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Anupam Pandey
Sent: Monday, April 01, 2013 11:23 AM
To: oracle Freelists
Cc: anupam pandey
Subject: Preventing Nested Table Full Access

Hi ,

       I am using a table as a cache for an oracle object type collection.The query to build the object collection is time consuming .So idea is to build the cache object once and store the elements in cache table and when next time its needed then return from cache table itself rather than going through entire process .
But the problem is Oracle does a Full Table Scan always when I select from this cache table . So select time increases as number of rows in table increases .Following is the script for what I am using ..

CREATE OR REPLACE TYPE VALUE_arr as table of varchar2(30);

CREATE TABLE OBJECT_CACHE
(

  CACHE_ID           NUMBER,
  DARRAY          VALUE_ARR,
  MARRAY       VALUE_ARR,
  FARRAY   VALUE_ARR,
  AARRAY       VALUE_ARR,

  SARRAY VALUE_ARR,
  QUERY CLOB,
  CONSTRAINT key_pk

       PRIMARY KEY(cache_id)
)
NESTED TABLE DARRAY STORE AS DARRAY_TBL , NESTED TABLE MARRAY STORE AS MARRAY_tbl , NESTED TABLE FARRAY STORE AS FARRAY_TBL , NESTED TABLE AARRAY STORE AS AARRAY_TBL , NESTED TABLE SARRAY STORE AS SARRAY_TBL ; Select query which I am using is following ..

select query
from OBJECT_CACHE
where darray = value_arr( 'a','b','c','d') and marray =value_Arr('a','b')
and farray = value_arr('a','b');

As of now this table has 1000 rows and select statement is taking nearly 11 secs for fetching the query .

Is there a way to avoid the FTS or is there any other way to work out this solution ?

Thanks,
Anupam

--

http://www.freelists.org/webpage/oracle-l  

--

http://www.freelists.org/webpage/oracle-l Received on Mon Apr 01 2013 - 18:17:48 CEST

Original text of this message