如何加快查询

Question

I am trying to create a table with below sql and it is running for days.

Below is the sql.

create table I83094_Emnmt1 AS
    Select 
     'I83094'                                              AS Audit_Id
     ,rap01.plcy                                           AS plcy
     ,raa02.enddt_t                                        As enddt
     ,RPAD(NVU_GET_TERM_ID(RAP01.PLCY, Raa02.ENDDT_T),1)   AS Term_Id
     ,nvu_get_nxt_proc_enddt(RAP01.PLCY, Raa02.ENDDT_T)    As next_enddt
     ,exception_date                                       AS exception_date
     ,RAP07.MVRDT_t
     ,RAP07.MVRDT_s
     ,rap07a.prefdt_t
     ,RAP07.MRTLST
     ,RAP07.BRTHDT_T                                       AS BirthDate
     ,floor((tt.ja2_effdt_t - RAP07.BRTHDT_T)/365 )        AS Operator_Age
     ,rap07a.perseq
     ,tt.ja2_effdt_t                                       AS TERM_DATE
     ,rap01.j01_PT_LINE_cat_Cd                             AS j01_PT_LINE_cat_Cd
     ,rap01.j01_pt_cdb_part_id                             AS j01_pt_cdb_part_id
     ,Rap01.J01_Pt_State_Cd                                As J01_Pt_State_Cd
     ---
from RAP01
----
Join RAA02
  ON raa02.j46_pt_line_cat_cd  = rap01.j01_pt_line_cat_cd
AND raa02.j46_pt_cdb_part_id   = rap01.j01_pt_cdb_part_id
AND raa02.j46_pt_state_cd      = rap01.j01_pt_state_cd
AND raa02.plcy                 = rap01.plcy
AND raa02.sprodt_t  between '14-OCT-2013' AND '14-OCT-2018'
AND raa02.ahevnt               = '0993'
---
JOIN ewt_prama_term tt
  ON tt.ja2_pt_line_cat_cd     = rap01.j01_pt_line_cat_cd
AND tt.ja2_pt_cdb_part_id      = rap01.j01_pt_cdb_part_id
AND tt.ja2_pt_state_cd         = rap01.j01_pt_state_cd
AND tt.ja2_plcy                = rap01.plcy
and raa02.enddt_t              between tt.ja2_effdt_t and tt.ja2_expdt_t - 1
-----
JOIN rap07a
  ON rap07a.j36_pt_line_cat_cd = rap01.j01_pt_line_cat_cd
AND rap07a.j36_pt_cdb_part_id  = rap01.j01_pt_cdb_part_id 
AND rap07a.j36_pt_state_cd     = rap01.j01_pt_state_cd
AND rap07a.plcy                = rap01.plcy
--and RAP07a.perseq              = rap18.perseq
AND raa02.enddt_t  BETWEEN rap07a.prefdt_t AND (rap07a.dropdt_t  - 1)
----
JOIN RAP07
  ON  RAP07.J37_PT_LINE_CAT_CD =  rap01.j01_pt_line_cat_cd   
AND  RAP07.J37_PT_CDB_PART_ID  =  rap01.j01_pt_cdb_part_id  
AND  RAP07.J37_PT_STATE_CD     =  rap01.j01_pt_state_cd     
AND  rap07.plcy                =  RAP01.PLCY
AND  RAP07.perseq              =  rap07a.perseq
AND RAP07.MRTLST in ('MA','IC','DC','WC','EC','DP')
--AND RAP07.DRVDES_01 IN ('R','P')  
--AND RAP07.MVRDT_T < (select exception_date from I83094_exception)
AND floor((tt.ja2_effdt_t - RAP07.BRTHDT_T)/365) < 50
AND  raa02.enddt_t BETWEEN RAP07.enddt_t AND (RAP07.dropdt_t  - 1)
---
join I83094_exception exp
  ON exp.line_cd               = rap01.line3
AND exp.state_cd               = rap01.state
AND exp.company_cd             = rap01.co3
AND exp.marital_status_cd      = RAP07.MRTLST
---
WHERE Rap01.J01_Pt_Line_Cat_Cd = 'A'
AND Rap01.Line3               IN ('010','019')
AND RAP07.MVRDT_T < exp.exception_date;

field 'J01_Pt_State_Cd' represents the different states, this table contains billions of data, and will take a lot of time to execute. I tried to create table but have to wait for 2 or 3 days. How can I improve the query so that I can perform the query in less time.

Answer 1

The group by works with grouping functions, like sum , max , min and count . When you use some of these functions, you normally want to sum grouping by some attribute, for example sum(sales) group by vendor .

When you use this keyword in a query without grouping functions, it works like a distinct , but you have to include all your fields in the group by.

The error that you are getting is just that, you are not including all your fields in your group by .

Answer 2

I think you are misunderstanding what a group by expression does.

From your comments I assume you are trying to run this on a large amount of data and it is taking a long time. A group by will not help this, group by expressions are used to aggregate data over certain fields, for example if I wanted to find out how long each user has spent logged on to a service I could use a group by on the username.

In order to speed up your query I'd recommend the standard stuff, make sure your tables are indexed properly and make sure all of those joins are necessary.