SQL & PL/SQL :: Rewriting Without Analytic Functions
Nov 12, 2011
How can I rewrite this without the analytic functions?
SELECT employee_ID, first_name, salary,
RANK() OVER(ORDER BY salary desc) toprank_desc,
RANK() OVER(ORDER BY salary ASC) toprank_asc
FROM employees
ORDER BY first_name
SQL> ed Wrote file afiedt.buf 1 select * 2 from (select deptno, ename, sal 3 ,dense_rank() over (partition by deptno order by sal desc) rank 4 from emp)
[code]...
why is that i just added ename on the ORDER BY part of the DENSE_RANK and then SQL> ed Wrote file afiedt.buf 1 select * 2 from (select deptno, ename, sal 3 ,dense_rank() over (partition by deptno order by sal desc, ename) ran
[code]...
ADAMS and WARD we're removed from the result, why is it? did it rank it as UNIQUE per sal and ename?
I have a table (events) with this structure: customer_id, event_id, ... For each customer_id there can be several rows in the table. I need to run a query of the format: select customer_id, expensive_function(customer_id),... from events.
The expensive_function to be applied to customer_id in the query is really expensive (a Java class calculating a check sum) and the events table has billions of rows.
Rows in events table have same customer_id for a few rows, then continue with a different customer_id ang again coming back to the first, etc.
I was thinking that it should be a way to trigger calculation of expensive_function only when customer_id changes, in order to reduce the number of calls. Only my knowledge about SQL is not going that far and I cannot use PL/SQL or any other procedural language, need to stick to standard SQL (or Oracle version of it).
I am building a reporting table using the count analytic function in order to count up several different attributes in one statement.What I find is that this method quickly eats up my TEMP space. This is 10gR2. I have attempted to use MANUAL workarea policy with as large ofsort_area_size as possible (2G) but that does not seem to have any effect on performance or TEMP usage. The RAW table is about 12G with 75 million rows. I am not that concerned about execution time, but rather TEMP usage.
--INSERT into <object>... select distinct file_sid,filename,control_numb,processing_date,file_class, vendor_id,vendor_desc, c_status_id,c_status_desc,
[code]...
I am not seeing any increase in onepass or multipass executions on the PGA during execution of this statement using...
SELECT CASE WHEN low_optimal_size < 1024*1024 THEN to_char(low_optimal_size/1024,'999999') || 'kb <= PGA < ' || (high_optimal_size+1)/1024|| 'kb' ELSE to_char(low_optimal_size/1024/1024,'999999') ||
[code]...
I'd like to get a better explaination of how analytics use the instance resources and TEMP space. For example if I add a count with a different window (such as the last two columns commented in the above query) I blow out my temp space (70G). Is the critcal factor the use of distinct? or multiple windows? or something else?
I need to return an ordered list of documents. The documents may belong to a set id (optional) and if so, are either a "master" or a "duplicate" type. For each set there can be only one master but many duplicates. My goal is to group all the sets together such that each master is proceeded by its duplicates.
There's also a documents table containing the documentid and among other things a page_count. In the following example I want to sort the documents first by page count but preserving the master/dupe grouping. Any documents which don't belong to a set or are just a duplicate without a master i want at the end of my set but also ordered by page count.
Here's an example set that I would want to order by:
As you can see I have a little bit of everything here. Docs 2001 and 2002 are the typical set of 1 master and its duplicate. Docs 2010, 2011, and 2012 is the same just a set of 3. Doc 2004 is a master but without any duplicates. Docs 2003 and 2014 are duplicates without a master (these docs have a master in the table but that doc isn't in the set i need to order by). Docs 2008 and 2009 do not belong to a set and as such do not have a master/dupe type.
The result i'm looking to achieve will be ordered as follows:
As I said above I first want to get the groupings of master/dupes and order ascending on the masters page count. For each duplicate of a master I then want to order the duplicates by page count. After I finished ordering all the master/dupe groups I then want to move on to the rest of the documents which will contain documents that don't belong to a set along with documents which are duplicates but have no master in my set. However, documents which are masters but without duplicates should have been ordered along with the other master/dupes groupings.
With this all in mind I have just been completely overwhelmed as to where to even start. Am I using analytic functions? Hierarchical stuff?
I have a question regarding analytic functions. I've been working with some functions, but I can't achieve the one which gives me the pretend result. I know to resolve this without using a function, with a internal select, but I think the analytical function is faster and proper.
I can' use sequence in the group by function and if I get equivalent analytic for above group by even then I can't write row_number as the order by gives detail record
I don't want to wrap this select inside other select
"representant" acct_id per group (about 300 rows total)acct_repres as( select distinct acct_id, origin_id, acct_parm_id from ( select a.* , source_id , dense_rank() over (partition by source_id origin_id order by acct_nbr nulls first, acct_id) as odr from account a join account_parm ap on (a.parm_id = ap.acct_parm_id) ) where odr = 1)select col1 , col2 , ( select accct_id from acct_repres ar where ar.acct_parm_id = t2.acct_parm_id) col3 , ( select count(1) from acct_repres) col4from some_table t1join other_table t2 on (....)
And here it comes.
The "acct_repres" subquery returns more than 300 rows when executed separately. But when used in CTE sometimes (depending on execution plan) it seems to have only one row - the value in the column col4 is "1",while value for col3 is NULL for most of the cases. It looks like the the dense_rank function and the condition "where odr =1" are evaluated at the very end.
When I use MATERIALIZE hint the result was the same. But when I put the result of account_repres into dedicated table and use that table instead of CTE the output is correct.
I am creating a "time aware" (DAY, WEEK, MONTH, QUARTER, and YEAR) dimension using Analytic Workspace Manager.
Let me give you some background. I'm coming from a traditional "Oracle Express" OLAP background where all our data is stored in cubes and these are defined, populated and operated on using OLAP DML, there is no SQL or traditional relational tables involved.
I now want to pull data from relational tables into some OLAP cubes and am using Analytic Workspace Manager to do this (maybe this is not the best way?)
Let me explain what I'm trying to achieve. In OLAP worksheet I can type the following DML commands:
DEFINE MY_DAY DIMENSION DAY MAINTAIN MY_DAY ADD TODAY '01JAN2011'
What this will do is create a "day dimension" and will populate it with values for each and every day between 1st Jan 2011 and today. It will be fully "time aware" and thus you can use date functions such as DAYOF to limit the MY_DAY dimension to all the Fridays etc. Similarly if I define a "month dimension" there will be an automatic implicit relationship between these two dimensions, this relationship and time aware cleverness is built into Oracle.
However, a dimension defined using DML commands (and indeed all objects created using DML language) is not visible in Analytic Workspace Manager (as there is no metadata for them?) and for the life of me I cannot work out how to create such a dimension using AWM. If I create a "Time Dimension" then, as far as I can tell, this is not a proper time dimension but merely a text dimension and I, presume, I have to teach it time awareness.
I have no issues creating, and populating cubes from relational tables using Analytic Workspace Manager, the only issue I have is creating a "proper" time aware dimension.
I have attempted to use the analytic function to keep a running total of the count of active calls based on the connect and disconnect times given for each record row.
I have the following query with analytic function but wrong results on the last column COUNT.
1)I am getting the output order by b.sequence_no column . This is a must. 2)COUNT Column :
I don't want the total count based on thor column hence there is no point in grouping by that column. The actual requirement to achieve COUNT is:
2a -If in the next row, if either the THOR and LOC combination changes to a new value, then COUNT=1 (In other words, if it is different from the following row)
2b-If the values of THOR and LOC repeats in the following row, then the count should be the total of all those same value rows until the rows become different. (In this case 2b-WHERE THE ROWS ARE SAME- also I only want to show these same rows only once. This is shown in the "MY REQUIRED OUTPUT) .
My present query: select r.name REGION , p.name PT, do.name DELOFF, ro.name ROUTE,
[code]...
My incorrect output[PART OF DATA]:Quote: REGIONPT DELOFF ROUTE THOR LOC SEQ COUNT NAASNAAS MAYNOOTHMAYNOOTHR010 DUBLINRD CEL 1 1 NAASNAAS MAYNOOTHMAYNOOTHR010 NEWTOWNRD CEL 2 1
[code]...
My required output[PART OF DATA]-:Quote: REGIONPT DELOFF ROUTE THOR LOC COUNT NAASNAAS MAYNOOTHMAYNOOTHR010 DUBLINRD CEL 1 NAASNAAS MAYNOOTHMAYNOOTHR010 NEWTOWNRD CEL 1 NAASNAAS MAYNOOTHMAYNOOTHR010 PRIMHILL CEL 1
[code]...
NOTE :Count as 1 is correctly coming.But where there is same rows and I want to take the total count on them, I am not getting.
Oracle 11g databaseidval1val2100a110b120c200a220b WITH input AS (SELECT 1 id
[Code].....
input; Output:idval1val2assigned_number100a0110b0120c2200a0220b1 The dense numbering sequence should be assigned to each row based on id and val1 column. For a given Id, the numbering only starts after val1 > 1 till then the assigned_number will be zero.
I would like to use character functions (LPAD, INSTR, SUBSTR, etc) to accomplish what i feel should be rather easy. I would like to take the following character strings:
Trying to understand how Oracle DATE works and how it differs to TO_DATE in the below manner...
SELECT DATE '2012-10-25' FROM DUAL; SELECT TO_DATE('2012-10-25','YYYY-MM-DD') FROM DUAL;
I've had a look through the Oracle Docs but can't find much on this. Basically I want to know how safe the first option is and if it does the same as the second. Also, where does DATE get the Format String from (NLS setting?). Can it be employed to compare a literal with a date that has a TIME portion set to something other than 12:00:00?
Again i getting confused with conversion function especially Explicit data type conversions. some cases oracle server automatically converts the data to the required type. This is called IMPLICIT CONVERSION. Explicit conversions are done by using the conversion functions.
Is there any way to apply the restriction on analytical functions, just like WHERE and HAVING .AS we know that we can apply the restriction on table by using WHERE and grouping functions by using HAVING clause .
For Ex: Departments wise count including all employees record :
SQL> select count(*) over(partition by deptno) dept_Count, ce.* 2 from scott.emp ce 3 order by deptno, job;
DEPT_COUNT EMPNO ENAME JOB MGR HIREDATE SAL COMM DEPTNO ---------- ----- ---------- --------- ----- ----------- --------- --------- ------ 3 7934 MILLER CLERK 7782 1/23/1982 1300.00 10 3 7782 CLARK MANAGER 7839 6/9/1981 2450.00 10 3 7839 KING PRESIDENT 11/17/1981 5000.00 10 5 7788 SCOTT ANALYST 7566 4/19/1987 3000.00 20 [code]....
I have created a package with some functions in it.I want to create a view with from 2 tables but also i wanto to use the functions from some fields of the view. (e.g. if the wview has 10 fields the 5 of them will be from the 2 tables and the other from package functions).