Building a recursive query using start with/connect by......

Question

Hi.
I have data in a table which looks like:
Top level     Step     3rd Level     4th Level Script Name
NORMAL_DAY     Step 1 of NORMAL_DAY     DATE_FEEDER     Step 1 of DATE_FEEDER      INT_EOD_DATE_FEEDER
NORMAL_DAY     Step 2 of NORMAL_DAY     EOD_FX_RATE_UPLOAD Step 3 of EOD_FX_RATE_UPLOAD SEND_MAIL_FXSPOTS
NORMAL_DAY     Step 2 of NORMAL_DAY     EOD_FX_RATE_UPLOAD Step 2 of EOD_FX_RATE_UPLOAD FXSPOTS
NORMAL_DAY     Step 2 of NORMAL_DAY     EOD_FX_RATE_UPLOAD Step 1 of EOD_FX_RATE_UPLOAD FX_FTPS_GET_EOD
NORMAL_DAY     Step 3 of NORMAL_DAY     CALENDAR_UPLOAD     Step 1 of CALENDAR_UPLOAD     CALENDAR
NORMAL_DAY     Step 3 of NORMAL_DAY     CALENDAR_UPLOAD      Step 2 of CALENDAR_UPLOAD     MDS_STOP
NORMAL_DAY     Step 3 of NORMAL_DAY     CALENDAR_UPLOAD     Step 3 of CALENDAR_UPLOAD     MDS_HOLIDAY
NORMAL_DAY     Step 3 of NORMAL_DAY     CALENDAR_UPLOAD     Step 4 of CALENDAR_UPLOAD     MDS_START
NORMAL_DAY     Step 4 of NORMAL_DAY     FA_VALUATIONS     Step 1 of FA_VALUATIONS     IMPORTMTM
NORMAL_DAY     Step 4 of NORMAL_DAY     FA_VALUATIONS     Step 2 of FA_VALUATIONS     INT_EOD_VALUATIONS
As you can see, it lends itself to a tree type structure. I am trying to display this information using the start with/connect by clauses.....
My initial query looks like:
select main.name "Top Level",
level2.name "Step",
level3.name "3rd Level",
level4.name "4th Level",
level5.name "Script Name",
level5.run_start,
level5.run_end,
(level5.run_end - level5.run_start) "Time difference",
to_char((level5.run_end - level5.run_start),'99999999999999999999999999990.0000000000')*1000 as Total_Run_Time
from jcs_jobs main,
jcs_jobs level2,
jcs_jobs level3,
jcs_jobs level4,
jcs_jobs level5
where main.name = 'NORMAL_DAY'
and main.PARENT_JOB_ID IS NULL
and main.job_id = 2253800
and main.job_id = level2.parent_job_id
and level2.job_id = level3.parent_job_id
and level3.job_id = level4.parent_job_id
and level4.job_id = level5.parent_job_id
order by level2.step_name;
This is a bit restrictive, so I need to make it more generic, because we can actually have more than 5 levels of depth.....
I haven't included the time element in the sample data. This also appears to be another conundrum.....how to display milliseconds in a reasonable date format?
Thank you very much in advance!
Dev

Answer 1

try this way:
SQL> select script_name, job,
  2         to_char(to_date(run_start,'Jhh24miss'),'dd-mon-yyyy hh24:mi:ss') run_start,
  3         to_char(to_date(run_end,'Jhh24miss'),'dd-mon-yyyy hh24:mi:ss') run_end,
  4         (to_date(run_end,'Jhh24miss')
  5                  - to_date(
  6                        substr(root_start,
  7                               2,
  8                               decode(instr(root_start,'/',1,2),
  9                                      0,
10                                      length(root_start)+1,
11                                      instr(root_start,'/',1,2)
12                                      )-2
13                               )
14                            ,'Jhh24miss')
15          ) "Time difference"
16  from (
17  select name script_name, lpad(' ',2*level,' ')||job_id job
18         ,to_char(run_start,'Jhh24miss') run_start, to_char(run_end,'Jhh24miss') run_end,
19         sys_connect_by_path(to_char(run_start,'Jhh24miss'),'/') root_start
20  from jcs_jobs
21  where name = 'NORMAL_DAY'
22  connect by prior job_id = parent_job_id
23  start with PARENT_JOB_ID IS NULL and job_id = 2253800
24  );
SCRIPT_NAM JOB                  RUN_START            RUN_END              Time difference
NORMAL_DAY   2253800            23-jan-2010 15:30:16 27-jan-2010 01:06:16             3,4
NORMAL_DAY     2253801          27-jan-2010 01:06:16 28-jan-2010 15:30:16               5Max
[My Italian Oracle blog|http://oracleitalia.wordpress.com/2010/01/23/la-forza-del-foglio-di-calcolo-in-una-query-la-clausola-model/]