Hi all,

We need to track work items across several existing applications. Items can
start as a document on disk, can be renamed or moved, converted to a task
in an application, until they reach a certain state. The applications know
nothing about preceding or following applications and there is no common
identifier for the items, which has always been a good thing.

My idea is to create a table for all events. Each application will record
their
own events. A minimum is one event per application with a starting and
finishing situation for that application. Example: "file \\folder1\x.doc was
moved to \\folder2\y.doc". Another application may follow up with
"\\folder2\y.doc was archived with id 12345". Each event will have a date
and time.

My table will probably have the following columns:

event_id number(10)
start_item varchar2(255)
end_item varchar2(255)
event_date date

Sample data:

1 A B 2009-10-11
2 G H 2009-10-12
3 B C 2009-10-13
4 H I 2009-10-17

Records can be linked from end_item to start_item (the next application in
line will continue where the previous application stops). So in this sample
data I have two sets of records (1 with 3, 2 with 4) and recognize two
starting points A and G with two end points C and I. From A to C takes
two days, from G to I takes 5 days.

Data entry is no problem. Reporting is. My resultset (a view, another table)
should be this:

A C 2
G I 5

How can I achieve this? I have looked at the connect by, but I get to many
records. So basically I have no idea where to start.

Thanks in advance,
PJ

On Sep 23, 8:22*am, "PJ" <pjv_dev.nos... (AT) geenspam (DOT) hotmail.com> wrote:
You're making this too difficult:

SQL> create table event_track(
2 event_id number(10),
3 start_item varchar2(255),
4 end_item varchar2(255),
5 event_date date
6 );

Table created.

SQL>
SQL>
SQL> insert all
2 into event_track
3 values(1,'A','B',to_date('2009-10-11', 'RRRR-MM-DD'))
4 into event_track
5 values(2,'G','H',to_date('2009-10-12', 'RRRR-MM-DD'))
6 into event_track
7 values(3,'B','C',to_date('2009-10-13', 'RRRR-MM-DD'))
8 into event_track
9 values(4,'H','I',to_date('2009-10-17', 'RRRR-MM-DD'))
10 select * from dual;

4 rows created.

SQL>
SQL> commit;

Commit complete.

SQL>
SQL> column start_item format a10
SQL> column end_item format a10
SQL>
SQL> select e1.event_id, e1.start_item, e2.end_item, round
(e2.event_date - e1.event_date, 0) event_days
2 from event_track e1, event_track e2
3 where e2.start_item = e1.end_item;

EVENT_ID START_ITEM END_ITEM EVENT_DAYS
---------- ---------- ---------- ----------
1 A C 2
2 G I 5

SQL>


David Fitzjarrell

I should have known, the sample data is too easy. It won't work with more
than two records per item. Real world will have a varying number of records
per item, probably with an average of 6.

PJ

"ddf" <oratune (AT) msn (DOT) com> wrote

On Sep 23, 8:22 am, "PJ" <pjv_dev.nos... (AT) geenspam (DOT) hotmail.com> wrote:
You're making this too difficult:

SQL> create table event_track(
2 event_id number(10),
3 start_item varchar2(255),
4 end_item varchar2(255),
5 event_date date
6 );

Table created.

SQL>
SQL>
SQL> insert all
2 into event_track
3 values(1,'A','B',to_date('2009-10-11', 'RRRR-MM-DD'))
4 into event_track
5 values(2,'G','H',to_date('2009-10-12', 'RRRR-MM-DD'))
6 into event_track
7 values(3,'B','C',to_date('2009-10-13', 'RRRR-MM-DD'))
8 into event_track
9 values(4,'H','I',to_date('2009-10-17', 'RRRR-MM-DD'))
10 select * from dual;

4 rows created.

SQL>
SQL> commit;

Commit complete.

SQL>
SQL> column start_item format a10
SQL> column end_item format a10
SQL>
SQL> select e1.event_id, e1.start_item, e2.end_item, round
(e2.event_date - e1.event_date, 0) event_days
2 from event_track e1, event_track e2
3 where e2.start_item = e1.end_item;

EVENT_ID START_ITEM END_ITEM EVENT_DAYS
---------- ---------- ---------- ----------
1 A C 2
2 G I 5

SQL>


David Fitzjarrell

On Sep 23, 12:36*pm, "PJ" <pjv_dev.nos... (AT) geenspam (DOT) hotmail.com> wrote:
Provide more accurate sample data, then. We work with what we're
given; if that sample is not representative of the actual data how can
anyone expect an accurate solution? Remember, please, "Garbage In,
Garbage Out".


David Fitzjarrell

On Sep 23, 12:36*pm, "PJ" <pjv_dev.nos... (AT) geenspam (DOT) hotmail.com> wrote:
I gave this a try and it appears to be working:

SQL> --
SQL> -- Create table
SQL> --
SQL> create table event_track(
2 event_id number(10),
3 start_item varchar2(255),
4 end_item varchar2(255),
5 event_date date
6 );

Table created.

SQL>
SQL> --
SQL> -- Load data
SQL> --
SQL> insert all
2 into event_track
3 values(1,'A','B',to_date('2009-10-11', 'RRRR-MM-DD'))
4 into event_track
5 values(2,'G','H',to_date('2009-10-12', 'RRRR-MM-DD'))
6 into event_track
7 values(3,'B','C',to_date('2009-10-13', 'RRRR-MM-DD'))
8 into event_track
9 values(4,'H','I',to_date('2009-10-17', 'RRRR-MM-DD'))
10 into event_track
11 values(5,'C','D',to_date('2009-10-19', 'RRRR-MM-DD'))
12 into event_track
13 values(6,'I','J',to_date('2009-10-22', 'RRRR-MM-DD'))
14 into event_track
15 values(7,'J','K',to_date('2009-10-23', 'RRRR-MM-DD'))
16 into event_track
17 values(8,'D','E',to_date('2009-10-27', 'RRRR-MM-DD'))
18 select * from dual;

8 rows created.

SQL>
SQL> commit;

Commit complete.

SQL>
SQL> column start_item format a10
SQL> column end_item format a10
SQL>
SQL> --
SQL> -- Display source data
SQL> --
SQL> select event_id, start_item, end_item ,event_date
2 from event_track;

EVENT_ID START_ITEM END_ITEM EVENT_DATE
---------- ---------- ---------- --------------------
1 A B 11-OCT-2009 00:00:00
2 G H 12-OCT-2009 00:00:00
3 B C 13-OCT-2009 00:00:00
4 H I 17-OCT-2009 00:00:00
5 C D 19-OCT-2009 00:00:00
6 I J 22-OCT-2009 00:00:00
7 J K 23-OCT-2009 00:00:00
8 D E 27-OCT-2009 00:00:00

8 rows selected.

SQL>
SQL> --
SQL> -- Generate desired report
SQL> --
SQL> with etrack as(
2 select e1.event_id, e1.start_item, e2.end_item,
e2.event_date edate2, e1.event_date edate1
3 from event_track e1, event_track e2
4 where e2.start_item = e1.end_item
5 )
6 select e11.event_id, e11.start_item, e21.end_item, round
(e21.edate2 - e11.edate1,0) evt_days
7 from etrack e11, etrack e21
8 where e21.start_item = e11.end_item
9 order by 1;

EVENT_ID START_ITEM END_ITEM EVT_DAYS
---------- ---------- ---------- ----------
1 A E 16
2 G K 11

SQL>

Test it on your data to see if the query returns correct results.


David Fitzjarrell

ddf schreef:
This does not work for a sequence of more than 4 events. All solutions
of this type can only work for a fixed max. number of events.
Proof:
insert into event_track values (8,'E','F',to_date('2009-11-27', 'RRRR-MM-DD
'));
commit;

and try again.

I think pl/sql is your friend here. Create a function that calculates
the end-point for a given start point (that one could use a connect by
or a cursor loop) and use it in your select statement:
select start_item, f_calc_end_item(startitem) end_item from event_track.
Only problem remaining is: what are the start items? (I leave that to
you..., hint: use a not exists clause.

I hope performance will not be an issue ;-)


Shakespeare

On Sep 23, 2:55*pm, Shakespeare <what... (AT) xs4all (DOT) nl> wrote:
I never said it was an absolute answer.


David Fitzjarrell

ddf schreef:
I just responded to your challenge:
"> Test it on your data to see if the query returns correct results."

;-)

Shakespeare

Shakespeare schreef:
I just could not leave it alone:

CREATE OR REPLACE FUNCTION f_endpoint(p_startpoint IN NUMBER) RETURN
NUMBER IS
CURSOR c_nextpoint(b_startpoint NUMBER) IS
SELECT t2.event_id
FROM event_track t1, event_track t2
WHERE t1.event_id = b_startpoint
AND t1.end_item = t2.start_item;
r_nextpoint c_nextpoint%ROWTYPE;
BEGIN
OPEN c_nextpoint(p_startpoint);
FETCH c_nextpoint
INTO r_nextpoint;
IF c_nextpoint%NOTFOUND THEN
CLOSE c_nextpoint;
RETURN p_startpoint;
ELSE
CLOSE c_nextpoint;
RETURN f_endpoint(r_nextpoint.event_id);
END IF;
-- this could be done by a connect by and taking the last record
-- as well, but I love recursion ....
END;


SELECT t3.event_id, t3.start_item, t4.end_item
FROM (SELECT t.event_id, t.start_item, f_endpoint(t.event_id) end_event
FROM event_track t
WHERE NOT EXISTS
(SELECT 1 FROM event_track t2 WHERE t2.end_item =
t.start_item)) t3,
event_track t4
WHERE t4.event_id = t3.end_event
ORDER BY 1;

EVENT_ID START_ITEM END_ITEM
1 A E
2 G K

Not the most optimal version I'm sure, but it works. Tested with more
than 4 steps as well.

Shakespeare

PJ (pjv_dev.nospam (AT) geenspam (DOT) hotmail.com) wrote:
: Hi all,

: We need to track work items across several existing applications. Items can
: start as a document on disk, can be renamed or moved, converted to a task
: in an application, until they reach a certain state. The applications know
: nothing about preceding or following applications and there is no common
: identifier for the items, which has always been a good thing.

: My idea is to create a table for all events. Each application will record
: their
: own events. A minimum is one event per application with a starting and
: finishing situation for that application. Example: "file \\folder1\x.doc was
: moved to \\folder2\y.doc". Another application may follow up with
: "\\folder2\y.doc was archived with id 12345". Each event will have a date
: and time.


Use connect by to join the entries and sys_connect_by_path to show each
complete path.

You will get extra entries such as
e.g.

A->B
A->B->D
A->B->D->X

Filter that list so you only see the results where the end (e.g. X) does
not have a next connection.

CONNECT_BY_ISLEAF might help do that.

CONNECT_BY_ROOT might be useful when displaying the end points.

sys_connect_by_path might not even be required here.