Problems with SELF Join

Question

Post reply

Problems with SELF Join

ghughes

SSCrazy

Points: 2201
More actions
June 28, 2004 at 1:50 pm

#108617

I am trying to do a self join that will take the value of the col called TO_DEPT of the PREVIOUS record and put it in the FROM_DEPT.   I have tried may different queries with frustrating results. Can some on help me out? I need to only update the records from the same O_ID.
What it looks like now
31175, '2003/11/24 14:17','',''
31175, '2003/11/24 14:36','',''
31175, '2003/11/25 13:08','',''
31175, '2003/11/25 13:08','',''
31175, '2003/11/25 13:09','','Pending'
31175, '2003/11/30 23:36','','Quality'
31175, '2003/12/01 15:08','','Technical'
31175, '2003/12/02 10:45','','OM'
31175, '2003/12/03 15:22','',''
31175, '2003/12/03 15:23','',''
31175, '2003/12/03 15:23','',''
31190, '2003/11/25 13:08','',''
31190, '2003/11/25 13:09','','Pending'
31190, '2003/11/30 23:36','','Quality'
31190, '2003/12/01 15:08','','Technical'
31190, '2003/12/02 10:45','','OM'
What I want it to look like
31175, '2003/11/24 14:17','',''
31175, '2003/11/24 14:36','',''
31175, '2003/11/25 13:08','',''
31175, '2003/11/25 13:08','',''
31175, '2003/11/25 13:09','','Pending'
31175, '2003/11/30 23:36','Pending','Quality'
31175, '2003/12/01 15:08','Qualitye','Technical'
31175, '2003/12/02 10:45','Technical','OM'
31175, '2003/12/03 15:22','OM',''
31175, '2003/12/03 15:23','',''
31175, '2003/12/03 15:23','',''
31190, '2003/11/25 13:09','','Pending'
31190, '2003/11/30 23:36','Pending','Quality'
31190, '2003/12/01 15:08','Quality','Technical'
31190, '2003/12/02 10:45','Technical','OM'
31190, '2003/12/03 15:22','OM',''
-----------------------------------------------------------------------
CREATE TABLE [Dummy] (
[O_ID] [decimal](9, 0) NOT NULL ,
[O_TMST] [datetime] NOT NULL ,
[From_Dept] [varchar] (50) COLLATE SQL_Latin1_General_CP1_CI_AS NULL ,
[To_Dept] [varchar] (50) COLLATE SQL_Latin1_General_CP1_CI_AS NULL ,
CONSTRAINT [PK_FSA_ORDER_CHNG_HIST] PRIMARY KEY CLUSTERED
(
  [O_ID],
  [O_TMST]
&nbsp WITH FILLFACTOR = 90 ON [PRIMARY]
) ON [PRIMARY]
GO
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/11/24 14:17','','')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/11/24 14:36','','')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/11/25 13:08','','')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/11/25 13:08','','')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/11/25 13:09','','Pending')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/11/30 23:36','','Quality')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/12/01 15:08','','Technical')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/12/02 10:45','','OM')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/12/03 15:22','','')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31175, '2003/12/03 15:23','','')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31190, '2003/12/03 15:23','','')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31190, '2003/11/25 13:09','','Pending')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31190, '2003/11/30 23:36','','Quality')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31190, '2003/12/01 15:08','','Technical')
INSERT dummy (o_id, o_tmst, from_dept, to_dept) values (31190, '2003/12/02 10:45','','OM')

Viewing 15 posts - 1 through 14 (of 14 total)

You must be logged in to reply to this topic. Login to reply

Steve Jones - SSC Editor SSC Guru Points: 736210 More actions · Answer 1

SELECT a.o_id

, a.o_tmst

, a.from_dept

, a.to_dept

, b.to_dept

from dummy a

inner join dummy b

on a.o_id = b.o_id

where b.o_tmst = ( select max( o_tmst)

from dummy c

where c.o_tmst < a.o_tmst

&nbsp

ghughes SSCrazy Points: 2201 More actions · Answer 2

Steve,

That did not quite work. All the data got pushed down one record.

the records with a timestamp of 13:09 should have data, but they did not, instead the data that should have been on the 13:09 record showed up on the 23:36 record.

You have given me something to work with and that gets me a little closer. Thanks.

Peter Tillotson Hall of Fame Points: 3325 More actions · Answer 3

try:

SELECT a.o_id

, a.o_tmst

, a.from_dept

, isnull(b.to_dept, '')

, a.to_dept

from dummy a

left join dummy b

on a.o_id = b.o_id and b.o_tmst = ( select max( o_tmst)

from dummy c

where c.o_tmst < a.o_tmst and c.o_id = a.o_id)

ORDER BY a.o_id, a.o_tmst

Regards

Peter

rockmoose SSCarpal Tunnel Points: 4601 More actions · Answer 4

Sometimes subselecting can do the trick:

select

outd.O_ID,

outd.O_TMST,

isnull(

( select ind.To_Dept from dummy ind where ind.O_ID = outd.O_ID

and ind.O_TMST = ( select max(inind.O_TMST) from dummy inind

where inind.O_TMST < outd.O_TMST ) ), outd.From_Dept ),

outd.To_Dept

from

dummy outd

By the way, I think that isnull(b.to_dept, '') should be isnull(b.to_dept, a.from_dept) in Peter's left join approach..

/rocmoose

You must unlearn what You have learnt

ghughes SSCrazy Points: 2201 More actions · Answer 5

Rockmoose, The subquery returns more than one record which causes it to fail.

I have spent many hours on this already and the the self join is the only way to go so far.

Peter your solution is working about 90%, I am still trying to figure out under what conditions it fails. I will post again when I do.

Thanks everyone for there help! Keep suggestions coming all have been helpful and got me closer to the solution. Thanks again.

rockmoose SSCarpal Tunnel Points: 4601 More actions · Answer 6

Yeah, see this.

Try:

select

outd.O_ID,

outd.O_TMST,

isnull(

( select ind.To_Dept from dummy ind where ind.O_ID = outd.O_ID

and ind.O_TMST = ( select max(inind.O_TMST) from dummy inind

where inind.O_TMST < outd.O_TMST and inind.O_ID = outd.O_ID&nbsp ), outd.From_Dept ),

outd.To_Dept

from

dummy outd

You must unlearn what You have learnt

Peter Tillotson Hall of Fame Points: 3325 More actions · Answer 7

I know my query will fail of there are any records with duplicate o_id and o_tmst, as per your sample data, but the primary key should stop the duplicates being added. If you want to have records with duplicate o_id and o_tmst values you will have to add an identity field to separately identify the records. In which case use :

on a.o_id = b.o_id and b.ID = ( select max(ID)

if ID is the new field and the records are added to the table in datetime order. Instead of

on a.o_id = b.o_id and b.o_tmst = ( select max( o_tmst)

regards

Peter

rockmoose SSCarpal Tunnel Points: 4601 More actions · Answer 8

Should be no duplicate data. The sample data was corrupt.

( 31175, '2003/11/25 13:08' ) is duplicate in sample.

CONSTRAINT [PK_FSA_ORDER_CHNG_HIST] PRIMARY KEY CLUSTERED

( [O_ID], [O_TMST] )

You do have this on your table, ghughes ?

/rockmoose

You must unlearn what You have learnt

ghughes SSCrazy Points: 2201 More actions · Answer 9

There are no dups in the data, I just cut and pasted one record too many. I have 300,000 records in this table. Sorry about that.

The table is indexed in O_id and O_TMST.

Rockmoose, your subquery still returns more than one value and causes it to fail.

I appreciate this help alot.

David Waxberg SSC Journeyman Points: 78 More actions · Answer 10

If you could add an identity column, this query would be simple.

select *

from table as t1 join table as t2 on t1.id = t2.id + 1

dw

David Waxberg SSC Journeyman Points: 78 More actions · Answer 11

Or if you can not add a identity column to the base table, insert the data into a temp table with an identity column and then use the query from above.

dw

rockmoose SSCarpal Tunnel Points: 4601 More actions · Answer 12

Ok, Identity columns are only necessary if you don't have a primary key or unique index.

If you have that there is no point in adding an identity column.

This got me a bit perplexed , so try this sql:

-- yet another sql...

select

a.O_ID,

a.O_TMST,

isnull( b.To_Dept, a.From_Dept ),

a.To_Dept

from

dummy a

join

( select a.O_ID, a.O_TMST, max(b.O_TMST) as my_prev_O_TMST

from dummy a left outer join dummy b

on a.O_ID = b.O_ID

and a.O_TMST > b.O_TMST

group by a.O_ID, a.O_TMST ) deedum

on a.O_ID = deedum.O_ID and a.O_TMST = deedum.O_TMST

left join dummy b

on b.O_ID = deedum.O_ID and b.O_TMST = deedum.my_prev_O_TMST

And also, please double check that O_ID and O_TMST is Unique!,

Provided are 2 methods.

-- double check dups...

select O_ID, O_TMST from dummy group by O_ID, O_TMST having count(*) > 1

-- assert duplicates

if( select count(*) from dummy ) <>

( select count(*) from dummy a join dummy b on a.O_ID = b.O_ID and a.O_TMST = b.O_TMST )

print 'duplicates'

else

print 'no duplicates'

--/rockmoose

You must unlearn what You have learnt

rockmoose SSCarpal Tunnel Points: 4601 More actions · Answer 13

Assuming correct ordering of the rows !

And also not taking into account that the data is grouped on O_ID, which must be included in the join criteria.

/rockmoose

You must unlearn what You have learnt

Kathi Kellenberger SSChampion Points: 11811 More actions · Answer 14

Try this:

update a

set a.from_dept = b.to_dept

from dummy a join dummy b on

a.o_id = b.o_id and b.o_tmst = (select max(o_tmst) from dummy c where c.o_id = a.o_id and c.o_tmst < a.o_tmst)

Kathi

Aunt Kathi Data Platform MVP
Author of Expert T-SQL Window Functions
Simple-Talk Editor