Retrieve n and n-1 rows on the same line

Question

In a database I am trying to get the latest employee information and the previous entry. The latest entry can be determined by the fact that DATEFIN is empty and the previous entry is returned by MAX(DATEFIN). All previous entries have a DATEFIN value
So I wrote a query which returned both the lines per employee. i.e. n and n-1
Actually what is needed is to compare certain fields of each line, so I need to show employee n and n-1 on the same line. This time I used LEAD so emp,date,field1,Lead(field1,1) etc etc.
As all the employee lines are returned, I used a minus function to remove all those where the DATEFIN IS NOT NULL, given that the LEAD line returned the information required BEFORE the minus removed it.
However, I am still getting n and n-1 rows, even though the minus part run by itself removes n-1,n-2... rows.
Question
How can I structure the minus or LEAD query just to get the n row and n-1 LEAD information
select trim(to_char(se.idexport,'0999'))
|| '|'
|| sa.referentiel_id
|| '|'
|| sa.idunique
|| '|'
|| sa.matricule
|| '|'
|| sr.niveau
|| '|'
|| LEAD(sr.niveau,1) OVER (ORDER BY sr.salarie_id,sr.datedebut desc )
|| '|'
|| sr.echelon
|| '|'
|| sr.coefficient
|| '|'
|| fo.code
|| '|'
|| sr.salairemensuel
|| '|'
|| sr.tauxhoraire
|| '|'
|| to_CHAR(ca.datevaleurrevisions,'DD/MM/YYYY')
|| '|'
|| em.code
|| '|'
|| fi.code
|| '|'
|| cc.code
|| '|'
|| sr.datedebut
|| '|'
|| LEAD(sr.datedebut,1) OVER (ORDER BY sr.salarie_id,sr.datedebut desc )
|| '|'
|| sr.datefin
|| '|'
|| LEAD(sr.datefin,1) OVER (ORDER BY sr.salarie_id,sr.datedebut desc )
AS Text
from salarie sa
inner join selection_export se on (se.IDUNIQUE = SA.IDUNIQUE AND se.idexport = 42 AND se.exporter = 1)
inner join salarierevision sr on (SR.SALARIE_ID = se.IDUNIQUE AND SR.ACCESRESTREINT = 0)
inner join campagne ca on CA.ISANNEEENCOURS = 1
inner join parambranche pb on (ca.organisationlds_id = pb.organisationlds_id)
inner join utilisateur u on (u.CODE = 'gagross' AND U.ORGANISATIONLDS_ID = CA.ORGANISATIONLDS_ID)
left outer join fonction fo on (fo.id = sr.fonction_id)
left outer join emploi em on (em.id = sr.emploi_id)
left outer join filiere fi on (fi.id = sr.filiere_id)
left outer join CONVENTIONCOLLECTIVE cc on (cc.id = sr.CONVENTIONCOLLECTIVE_ID)
minus
select trim(to_char(se.idexport,'0999'))
|| '|'
|| sa.referentiel_id
|| '|'
|| sa.idunique
|| '|'
|| sa.matricule
|| '|'
|| sr.niveau
|| '|'
|| LEAD(sr.niveau,1) OVER (ORDER BY sr.salarie_id,sr.datedebut desc )
|| '|'
|| sr.echelon
|| '|'
|| sr.coefficient
|| '|'
|| fo.code
|| '|'
|| sr.salairemensuel
|| '|'
|| sr.tauxhoraire
|| '|'
|| to_CHAR(ca.datevaleurrevisions,'DD/MM/YYYY')
|| '|'
|| em.code
|| '|'
|| fi.code
|| '|'
|| cc.code
|| '|'
|| sr.datedebut
|| '|'
|| LEAD(sr.datedebut,1) OVER (ORDER BY sr.salarie_id,sr.datedebut desc )
|| '|'
|| sr.datefin
|| '|'
|| LEAD(sr.datefin,1) OVER (ORDER BY sr.salarie_id,sr.datedebut desc )
AS Text
from salarie sa
inner join selection_export se on (se.IDUNIQUE = SA.IDUNIQUE AND se.idexport = 42 AND se.exporter = 1)
inner join salarierevision sr on (SR.SALARIE_ID = se.IDUNIQUE AND SR.ACCESRESTREINT = 0)
inner join campagne ca on CA.ISANNEEENCOURS = 1
inner join parambranche pb on (ca.organisationlds_id = pb.organisationlds_id)
inner join utilisateur u on (u.CODE = 'gagross' AND U.ORGANISATIONLDS_ID = CA.ORGANISATIONLDS_ID)
left outer join fonction fo on (fo.id = sr.fonction_id)
left outer join emploi em on (em.id = sr.emploi_id)
left outer join filiere fi on (fi.id = sr.filiere_id)
left outer join CONVENTIONCOLLECTIVE cc on (cc.id = sr.CONVENTIONCOLLECTIVE_ID)
where sr.datefin is not null
Thanks and I hope this is clear
Colin

Answer 1

Assuming that the sr.salarie_id column identifies the employee I think you need to partition by it, not order by it in the analytic functio. If it does not identify the employee, then you need to partition by whatever uniquely identifies an employee. I would likely structure it something like this:
SQL> with t as (
  2     SELECT 1 emp_id, TO_DATE('01-jul-2010', 'dd-mon-yyyy') dt, 'Jul' descr
  3     from dual union all
  4     SELECT 1, TO_DATE('01-aug-2010', 'dd-mon-yyyy'), 'Aug' from dual union all
  5     SELECT 1, TO_DATE('01-sep-2010', 'dd-mon-yyyy'), 'Sep' from dual union all
  6     SELECT 1, TO_DATE('01-oct-2010', 'dd-mon-yyyy'), 'Oct' from dual union all
  7     SELECT 1, TO_DATE(NULL), 'Current' from dual union all
  8     SELECT 2, TO_DATE('01-jan-2010', 'dd-mon-yyyy'), 'Jan' from dual union all
  9     SELECT 2, TO_DATE('01-mar-2010', 'dd-mon-yyyy'), 'Mar' from dual union all
10     SELECT 2, TO_DATE('01-may-2010', 'dd-mon-yyyy'), 'May' from dual union all
11     SELECT 2, TO_DATE('01-jul-2010', 'dd-mon-yyyy'), 'Jul' from dual union all
12     SELECT 2, TO_DATE(NULL), 'Current' from dual)
13  SELECT emp_id, curr_dt, last_dt, curr_descr, last_descr
14  FROM (SELECT emp_id, dt curr_dt,
15               LEAD(dt) OVER(PARTITION BY emp_id
16                             ORDER BY dt desc nulls first) last_dt,
17               descr curr_descr,
18               LEAD(descr) OVER(PARTITION BY emp_id
19                             ORDER BY dt desc nulls first) last_descr
20        FROM t)
21  WHERE curr_dt IS NULL
    EMP_ID CURR_DT     LAST_DT     CURR_DE LAST_DE
         1             01-OCT-2010 Current Oct
         2             01-JUL-2010 Current JulWhere my t is your set of joined tables without the where sr.datefin is not null predicate.
John