StrebenMonch:
Thanks for the ideas.
Optimized the SQL is what i have been thinking about.
I attached the SQL here, I could only see to eliminate
the calculation of the cost where I used 'case' to check null value and use sum function again. Please correct me if you think any other ones could be done in the Perl.
After bind_columns, then use loop and array to do the summation?
select
mvmnt_data.region
,store_name
,nvl(i_a.nat_upc,i_b.nat_upc)
,nvl(i_a.long_description,i_b.long_description)
,nvl(i_a.item_size,i_b.item_size)
,nvl(i_a.item_uom,i_b.item_uom)
,vsi.vend_item_num
,salesUnits
,salesDollars
,case_size
,cost
,ps.ps_prod_subteam
from
(select wm.upc,sum(wm.mvmnt_units) salesUnits
,sum(wm.mvmnt_dollars) salesDollars
,wm.store
,wm.pos_dept
,wm.case_size
,sum(round (case when wm.case_size is null or
wm.case_cost=0
then null
else (wm.case_cost/wm.case_size) *
wm.mvmnt_units end, +3)) cost
,s.store_name
,s.region
from stella.mvmnt_wkly wm,vim.store@vim s
where wm.mvmnt_date between to_date
('10/01/2004', 'mm/dd/yyyy')and to_date
('10/03/2004', '/mm/dd/yyyy')
and wm.store in ('10036')
and wm.store=s.ps_bu
--and wm.upc='000000015232'
and wm.pos_dept='0110'
group by wm.case_size
,s.region
,wm.pos_dept,
wm.upc
,wm.store
,s.store_name) mvmnt_data,
vim.item_region@vim ir,
vim.item_master@vim i_a,
vim.item_master@vim i_b,
vim.reg_pos_dept_ref@vim rpdr
,vim.ps_team_ref@vim ps
,vim.vendor_store_item@vim vsi
where mvmnt_data.region=ir.region(+)
and mvmnt_data.upc=substr(ir.upc(+),-12,12)
and substr(ir.nat_upc, 12, 12)=substr(i_a.nat_upc(+),-
12,12)
and mvmnt_data.upc=substr(i_b.nat_upc(+),-12,12)
and mvmnt_data.region=rpdr.region(+)
and mvmnt_data.pos_dept=trim(leading 0 from rpdr.pos_dept
(+))
and rpdr.pos_dept=ps.pos_dept
and rpdr.region=ps.region
and mvmnt_data.store=ps.ps_bu
and mvmnt_data.upc=substr(vsi.upc,-12,12)
and vsi.region=ps.region
and vsi.ps_bu=mvmnt_data.store
and vsi.v_auth_status='P'
Jewang | [reply] |
This is where I might try to break this into 2 selects. (to the more experienced SQL people, feel free to correct me)
Query 1
===========================================================
SELECT ir.region
,ir.upc
,i_b.nat_upc
,rpdr.region
,rpdr.pos_dept
,ps.pos_dept
,ps.ps_bu
,vsi.upc
,vsi.ps_bu
,store_name
,NVL(i_a.nat_upc,i_b.nat_upc)
,NVL(i_a.long_description,i_b.long_description)
,NVL(i_a.item_size,i_b.item_size)
,NVL(i_a.item_uom,i_b.item_uom)
,vsi.vend_item_num
,ps.ps_prod_subteam
FROM vim.item_region@vim ir
,vim.item_master@vim i_a
,vim.item_master@vim i_b
,vim.reg_pos_dept_ref@vim rpdr
,vim.ps_team_ref@vim ps
,vim.vendor_store_item@vim vsi
WHERE substr(ir.nat_upc, 12, 12)=substr(i_a.nat_upc(+),-12,12)
AND rpdr.pos_dept = ps.pos_dept
AND rpdr.region = ps.region
AND vsi.region = ps.region
AND vsi.v_auth_status = 'P'
===========================================================
The first 9 columns in the select are new and will be used in the second select.
Query 2
===========================================================
SELECT mvmnt_data.region
,wm.upc
,sum(wm.mvmnt_units) AS salesUnits
,sum(wm.mvmnt_dollars) AS salesDollars
,wm.store
,wm.pos_dept
,wm.case_size
,SUM(ROUND (CASE WHEN wm.case_size IS NULL OR wm.case_cost=0 THEN NULL
ELSE (wm.case_cost/wm.case_size) * wm.mvmnt_units end, +3)) AS cost
,s.store_name
FROM stella.mvmnt_wkly wm
,vim.store@vim s
WHERE mvmnt_data.region=ir.region(+)
AND mvmnt_data.upc = ?
AND mvmnt_data.upc = ?
AND mvmnt_data.region = ?
AND mvmnt_data.pos_dept = ?
AND mvmnt_data.store = ?
AND mvmnt_data.upc = ?
AND mvmnt_data.store = ?
AND wm.mvmnt_date between to_date('10/01/2004', 'mm/dd/yyyy') and to_date('10/03/2004', '/mm/dd/yyyy')
AND wm.store in ('10036')
AND wm.store = s.ps_bu
AND wm.pos_dept = '0110'
GROUP BY wm.case_size
,s.region
,wm.pos_dept
,wm.upc
,wm.store
,s.store_name
===========================================================
So what I would do is call the first query then while looping through it you can do the proper substr on the data. And also call query 2.
I would try to optimize query 2 as much as possable by creating indexs if need be and looking at an execution plan. For example this is what I use for MySQL
deleteplan.sql
===========================================================
delete from plan_table
where statement_id = 'TEST'
/
commit
/
===========================================================
test.sql
===========================================================
@deleteplan
EXPLAIN PLAN set statement_id = 'TEST' for
SQL STATEMENT GOES HERE
/
@getplan
===========================================================
getplan.sql (there is some other information you can get from the plan_table, but this is all I need most of the time)
===========================================================
select operation, options, object_name, cost, cardinality, TO_CHAR(timestamp, 'HH:MM:SS')
from plan_table
where statement_id = 'IFXTEST'
start with id = 0
connect by prior id=parent_id and prior statement_id = statement_id
/
===========================================================
So I just put my SQL in test.sql and run it to get my execution plan. From there you can see where your problems are.
| [reply] |