• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 683
  • Last Modified:

How to update with a group count from another table?

I have two tables - PERMIT_SUMMARY and WELL.DWIGHT_SCOUT. I need to update the column PRMT1W in the former table with the count(PRMTDATE) from the latter table that grouped by a substring of UWI with records for the past 7 days. The common field in these two tables are SCID and substr(UWI,1,5). Heres what I have so far. Without where scid in& the result never came back.

update permit_summary b
set prmt1w =
(select count(prmtdate) from well.dwight_scout
where prmtdate > to_char(sysdate - 7, 'yyyymmdd')
and substr(uwi,1,5) = b.scid
group by substr(uwi, 1, 5))
where scid in &
;

Thanks!
0
eduliu
Asked:
eduliu
  • 5
  • 5
  • 2
1 Solution
 
Daniel WilsonCommented:
OK, let's see about derived table syntax ...

update permit_summary b
Inner Join (Select count(prmtdate) as countDate, 
  substr(uwi, 1, 5) as FK
  from well.dwight_scout
  Group by substr(uwi, 1, 5)) A
on b.SCID = A.FK
Set prmt1w = A.countDate

Open in new window

0
 
Daniel WilsonCommented:
Hold on ... I saw this in the MySQL zone ... but now I see it's also in PL/SQL.

What DB are you using?
0
 
eduliuAuthor Commented:
Oracle. Thanks!
0
VIDEO: THE CONCERTO CLOUD FOR HEALTHCARE

Modern healthcare requires a modern cloud. View this brief video to understand how the Concerto Cloud for Healthcare can help your organization.

 
Daniel WilsonCommented:
So ... does the code I posted work or error out?  Basic SQL syntax is the same between the DB's ... but this has pushed a bit beyond basic :-)
0
 
eduliuAuthor Commented:
I'm getting the following error:
ORA-00971: missing SET keyword
0
 
Daniel WilsonCommented:
Seeing something similar here ... http://forums.databasejournal.com/showthread.php?t=42650

Your syntax looks right (with the possible exception of the WHERE scid IN & on which I can't comment).

You say it never came back ... How many rows are in each table?  That's an inefficient join that uses the substring ... but if the tables are relatively small, you should get by with it.
0
 
eduliuAuthor Commented:
permit_summary has 3,219 rows with scid column indexed. well.dwight_scout has 515,836 rows with uwi indexed. In the following example, if I specify specific scid, the result came back in seconds. But if I keep as it is, nothing came back after 10min. My questions are two:
1. Is there a way to speed up this, to limit full table scan, etc?
2. Will PL/SQL work better here? Can you provide some hint?

Thanks!

update  permit_summary b
set prmt1w =
(select count(*) from well.dwight_scout
where prmtdate > to_char(sysdate - 7, 'yyyymmdd')
and substr(uwi,1,5) = b.scid
group by substr(uwi, 1, 5))
where b.scid in (select distinct scid from permit_summary);
0
 
Daniel WilsonCommented:
I'm in over my head on the Oracle thing.  I thought it was MySQL when I jumped in.  Sorry.

Please click the Request Attention link and ask that the MySQL zone be removed and the Oracle zone be added.

That should get some Oracle / PL SQL gurus looking at it.
0
 
GGuzdziolCommented:
Probably function-based index would help
create index x_dwight_scout_01 on well.dwight_scout (substr(uwi,1,5))
update  permit_summary b
set prmt1w =
(select count(*) from well.dwight_scout
where prmtdate > to_char(sysdate - 7, 'yyyymmdd') -- why to store date as a string?
and substr(uwi,1,5) = b.scid -- this makes index on uwi unusable
group by substr(uwi, 1, 5)) -- this is not necessary
 
Maybe you should do that more like
 
update  permit_summary b
set prmt1w =
(select count(*) from well.dwight_scout
where prmtdate > sysdate - 7 -- make prmtdate DATE datatype and additionally you can index it 
and substr(uwi,1,5) = b.scid -- for this function-based index on substr(uwi, 1, 5) would work
)

Open in new window

0
 
eduliuAuthor Commented:
The creating index on substr(uwi,1,5) helped the most. Using my original sql, the result came back in 6 sec. A couple of questions from your advice:
- group by substr(uwi, 1, 5) is necessary, as the count has to be rolled up and added together based on substr(uwi,1,5)? Otherwise, I'm getting a single count.
- prmtdate was created by someone before me. They want to capture a format like 20081009. Is there an alternative?

Thanks!
0
 
GGuzdziolCommented:
Again, you do not need group by. If it was needed, your query would fail with ,,single row subquery returns more then one row'' ;-) Strictly: you are filtering on certain substr(uwi, 1, 5) - grouping by on the same cannot change anything, since there already is only one! That's because subquery is logically (and probably also physically) evaluated for each row of the table being updated. This will do no harm, but makes your code less readable.

Alternative is putting it explicitly as DATE using TO_DATE conversion while inserting. Another option - if you *have to* insert is as string and can do nothing about it - could be setting NLS parameter on session level to use YYYYMMDD as date format (NLS_DATE_FORMAT - or something similiar - I do not remember by heart, and do not have time to check right now).
0
 
eduliuAuthor Commented:
Thanks!
0

Featured Post

Get quick recovery of individual SharePoint items

Free tool – Veeam Explorer for Microsoft SharePoint, enables fast, easy restores of SharePoint sites, documents, libraries and lists — all with no agents to manage and no additional licenses to buy.

  • 5
  • 5
  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now