Script takes long time to run

Question

Hi Friends,
We have a leave record in the following manner:
name date leaves
ABC 01-oct-08 1
ABC 02-oct-08 1
ABC 03-oct-08 1
ABC 04-oct-08 1
ABC 05-oct-08 1
ABC 10-oct-08 1
ABC 18-oct-08 1
ABC 19-oct-08 1
ABC 20-oct-08 1
ABC 25-oct-08 1
ABC 26-oct-08 1
ABC 27-oct-08 1
ABC 28-oct-08 1
and we need an output in the following manner:
output
Name FromDate ToDate Leaves
ABC 01-oct-08 05-oct-08 5
ABC 10-oct-08 10-oct-08 1
ABC 18-oct-08 20-oct-08 3
ABC 25-oct-08 28-oct-08 4
The code for the above mentioned logic is as follows:
=======================================================
=======================================================
procedure KPM_AB_LWP_PROC(p_err_buf OUT VARCHAR2, p_ret_code OUT NUMBER) is
diff number;
prev_date date;
p_prev_date date;
first_date date;
last_date date;
dt date;
total_leaves number := 0;
ecode varchar2(20);
ename varchar2(240);
cursor ab_lwp_cur(process_from_date date, process_to_date date, e_code varchar2) is
select
ab.EMPCODE,
ab.RECORD_DATE,
ab.FIRST_HALF,
ab.SECOND_HALF,
(em.last_name || ', ' || em.TITLE || ' '|| em.FIRST_NAME || em.MIDDLE_NAME) EMP_NAME
from
kpm_hr_absent_record ab, kpm_hr_emp_mst em
where
ab.EMPCODE = em.EMPCODE
and (ab.FIRST_HALF in ('AB','LWP') or ab.SECOND_HALF in ('AB','LWP'))
and ab.EMPCODE like e_code
and ab.record_date between process_from_date and process_to_date;
cursor active_emp is
select empcode
from kpm_hr_emp_mst
where status='Active'
order by empcode;
begin
for emp in active_emp loop
begin
select min(ab.RECORD_DATE), max(ab.RECORD_DATE)
into prev_date, last_date
from kpm_hr_absent_record ab
where (ab.FIRST_HALF in ('AB','LWP') or ab.SECOND_HALF in ('AB','LWP'))
and ab.EMPCODE like emp.empcode;
exception
when others then
prev_date := null;
last_date := null;
end;
dt := prev_date;
FND_FILE.PUT_LINE(FND_FILE.OUTPUT,'Employee' || chr(9) || chr(9) || 'Name' || chr(9) || chr(9) || 'From Date' || chr(9) || chr(9) || 'To Date' || chr(9) || chr(9) || 'Total Leaves');
FND_FILE.PUT_LINE(FND_FILE.OUTPUT,'---------' || chr(9) || chr(9) || '-----' || chr(9) || chr(9) || '----------' || chr(9) || chr(9) || '---------' || chr(9) || chr(9) || '-------------');
while dt <= last_date loop
first_date := dt;
total_leaves := 0;
for m in ab_lwp_cur(prev_date, last_date, emp.empcode) loop
ecode := m.empcode;
ename := m.emp_name;
diff := m.record_date - prev_date;
if diff = 0 then
if m.first_half in ('AB','LWP') then
total_leaves := total_leaves + 0.5;
end if;
if m.second_half in ('AB','LWP') then
total_leaves := total_leaves + 0.5;
end if;
prev_date := prev_date + 1;
else
prev_date := m.record_date;
goto print_leave;
end if;
p_prev_date := prev_date -1;
end loop;
<<print_leave>>
if total_leaves > 0 then
FND_FILE.PUT_LINE(FND_FILE.OUTPUT,ecode || chr(9) || chr(9) || ename || chr(9) || chr(9) || first_date || chr(9) || chr(9) || p_prev_date || chr(9) || chr(9) || total_leaves);
end if;
dt := prev_date;
end loop;
end loop;
exception
when others then
FND_FILE.PUT_LINE(FND_FILE.LOG,'Error: ' || sqlerrm);
end KPM_AB_LWP_PROC;
=======================================================
=======================================================
The problem is that this code takes about 24hrs to run which is not accepted.
Kindly suggest us some other technique to implement the same logic.
For your reference, the KPM_HR_ABSENT_RECORD table has about 3,75,000 records in it.
Also we have created some indexes on certain column as recommended by explain plans.
Thanks in advance
Ankur

Answer 1

with t as (select 'ABC' as nm, to_date('01-oct-08','dd-mon-yy') as dt, 1 as leaves from dual union all
               select 'ABC',to_date('02-oct-08','dd-mon-yy'), 1 from dual union all
               select 'ABC',to_date('03-oct-08','dd-mon-yy'), 1 from dual union all
               select 'ABC',to_date('04-oct-08','dd-mon-yy'), 1 from dual union all
               select 'ABC',to_date('05-oct-08','dd-mon-yy'), 1 from dual union all
               select 'ABC',to_date('10-oct-08','dd-mon-yy'), 1 from dual union all
              select 'ABC',to_date('18-oct-08','dd-mon-yy'), 1 from dual union all
               select 'ABC',to_date('19-oct-08','dd-mon-yy'), 1 from dual union all
               select 'ABC',to_date('20-oct-08','dd-mon-yy'), 1 from dual union all
              select 'ABC',to_date('25-oct-08','dd-mon-yy'), 1 from dual union all
              select 'ABC',to_date('26-oct-08','dd-mon-yy'), 1 from dual union all
              select 'ABC',to_date('27-oct-08','dd-mon-yy'), 1 from dual union all
              select 'ABC',to_date('28-oct-08','dd-mon-yy'), 1 from dual)
   -- end of sample data
select nm, min(dt), max(dt), sum(leaves)
from
   select nm, dt, max(group_id) over(partition by nm order by dt) group_id
            , leaves
   from
      select nm, dt, case when trunc(dt) - 1 != lag(trunc(dt),1,trunc(dt))
                                                                 over(partition by nm order by dt)
                           then rownum
                           end group_id
             , leaves
      from t
group by nm, group_id
order by 1,2;Hope this helps.
@Blushadow, thanks very much for the test data.
Regards
Raj
P.S : For more information check this link
http://www.oracle.com/technology/oramag/oracle/04-mar/o24asktom.html
and search for "Analytics to the Rescue (Again)" in that page.
Edited by: R.Subramanian on Oct 22, 2008 4:52 AM
Changed count function to Sum function