IBM i5OS SQL: Forecasting periodic events - sql

Question

I have a situation where I have about 4000 tasks that all have different periodic rules for occurences.
They are preventive maintenance tasks. The table I get them from only provides me the start date and frequency of occurence in units of weeks.
Example:
Task (A) is scheduled to occur every two weeks, starting on week 1 of 2015.
Task (B) is scheduled to occur every 6 weeks, starting on week 2 of 2011.
And so on...
What I need to do is produce a resultset that contains a record for each occurence since the start point, for each task.
It's like generating a sequence.
Example:
Task | Year | Week
------|-------|-------
A | 2015 | 1
A | 2015 | 3
A | 2015 | 5
A | 2015 | 7
[...]
B | 2011 | 2
B | 2011 | 8
And so on...
You probably think "hey, that is simple, just put it in a loop then your good."
Not so fast!
The trick is that I need this to be within one SQL query.
I know I probably should be doing it in a stored procedure or a function. But I can't, for now. I could also do it in some VbA code since it will go in an Excel spreadsheet. But Excel has become an unstable product lately and I do not want to risk my code to fail after an update from Microsoft. So I try as much as possible to stay within the limits of IBM i5OS SQL queries.
I know the answer could be that it is impossible. But I believe in this community.
Thanks in advance,
EDIT :
I have found this post where it shows how to list dates within a range.
IBM DB2: Generate list of dates between two dates
I tried to generate a list of dates based on periodicity and it worked.
I am still struggling on the generation of multiple sequences based on multiple periodicity.
Here's the code I have so far:
SELECT d.min + num.n DAYS AS DATES
FROM
(VALUES(DATE('2017-01-01'), DATE('2017-03-01'))) AS d(min, max)
JOIN
(
-- Creates a table of numbers based on periodicity
SELECT
n1.n + n10.n + n100.n AS n
FROM
(VALUES(0),(1),(2),(3),(4),(5),(6),(7),(8),(9)) AS n1(n)
CROSS JOIN
(VALUES(0),(10),(20),(30),(40),(50),(60),(70),(80),(90)) AS n10(n)
CROSS JOIN
(VALUES(0),(100),(200),(300),(400),(500),(600),(700),(800),(900)) AS n100(n)
-- I just need to replace the 2nd argument by the desired frequency */
WHERE MOD(n1.n+n10.n+n100.n, 6 )=0
ORDER BY n1.n + n10.n + n100.n
) num
ON
d.min + num.n DAYS<= d.max
ORDER BY num.n
In other words, I need the dates in table d to be dynamic as well as the periodicity (6) in num's table WHERE clause.
Should I be using a WITH statement? If so, can someone please guide me because I am not very used to this kind of statement.
EDIT#2:
Here is the table structure I'm working with:
TABLE NAME: SGTRCDP (Programmed Tasks):
| | Start | Start | Freq.
Asset | Task | Year | Week | (week)
--------------|------------|----------|----------|----------
TMPC531 | VER0560 | 2011 | 10 | 26
BAT0404 | IPNET030 | 2011 | 2 | 4
B-EXTINCT-151 | 001H-0011 | 2014 | 15 | 17
[...] | [...] | [...] | [...] | [...]
4000 more like these, the unique key being combination of `Asset` and `Task` fields.
What I would like to have is this:
Asset | Task | Year | Week
--------------|------------|----------|----------
TMPC531 | VER0560 | 2011 | 10
TMPC531 | VER0560 | 2011 | 36
TMPC531 | VER0560 | 2012 | 10
TMPC531 | VER0560 | 2012 | 36
TMPC531 | VER0560 | 2013 | 10
TMPC531 | VER0560 | 2013 | 36
TMPC531 | VER0560 | 2014 | 10
TMPC531 | VER0560 | 2014 | 36
TMPC531 | VER0560 | 2015 | 10
TMPC531 | VER0560 | 2015 | 36
TMPC531 | VER0560 | 2016 | 10
TMPC531 | VER0560 | 2016 | 36
TMPC531 | VER0560 | 2017 | 10
TMPC531 | VER0560 | 2017 | 36
BAT0404 | IPNET030 | 2011 | 2
BAT0404 | IPNET030 | 2011 | 6
BAT0404 | IPNET030 | 2011 | 10
BAT0404 | IPNET030 | 2011 | 14
BAT0404 | IPNET030 | 2011 | 18
BAT0404 | IPNET030 | 2011 | 22
BAT0404 | IPNET030 | 2011 | 26
BAT0404 | IPNET030 | 2011 | 30
BAT0404 | IPNET030 | 2011 | 34
BAT0404 | IPNET030 | 2011 | 38
[...] | [...] | [...] | [...]
BAT0404 | IPNET030 | 2017 | 34
BAT0404 | IPNET030 | 2017 | 38
B-EXTINCT-151 | 001H-0011 | 2014 | 15
B-EXTINCT-151 | 001H-0011 | 2014 | 32
B-EXTINCT-151 | 001H-0011 | 2014 | 49
B-EXTINCT-151 | 001H-0011 | 2015 | 14
B-EXTINCT-151 | 001H-0011 | 2015 | 31
[...] | [...] | [...] | [...]
B-EXTINCT-151 | 001H-0011 | 2017 | 8
B-EXTINCT-151 | 001H-0011 | 2017 | 24
I was able to make it using CTE, but it generates so many records that whenever I want to filter or order data, it takes forever. Same goes for downloading the whole resultset.
And I wouldn't risk creating a temporary table and bust the disk space.
Another caveat of CTE, is that It cannot be referenced as a subquery.
And guess what, my plan was to use it as a subquery in FROM clause of a SELECT joining it with the actual work orders table and do Asset-Task-Year-Week matching to see if the programmed tasks were executed as planned or not.
Anyway, here is the CTE I used to get it:
WITH PPM (EQ, TASK, FREQ, OCCYR, OCCWK, OCCDAT, NXTDAT) AS
(
SELECT
TRCD.DLACCD EQ,
TRCD.DLJ1CD TASK,
INT(SUBSTR(TRCD.DLL1TX,9,3)) FREQ,
AOAGNB OCCYR,
AOAQNB OCCWK,
CASE
WHEN aoaddt/1000000 >= 1 THEN
DATE('20'||substr(aoaddt,2,2)||'-'||substr(aoaddt,4,2)||'-'||substr(aoaddt,6,2))
ELSE
DATE('19'||substr(aoaddt,1,2)||'-'||substr(aoaddt,3,2)||'-'||substr(aoaddt,5,2))
END OCCDAT,
(CASE
WHEN aoaddt/1000000 >= 1 THEN
DATE('20'||substr(aoaddt,2,2)||'-'||substr(aoaddt,4,2)||'-'||substr(aoaddt,6,2))
ELSE
DATE('19'||substr(aoaddt,1,2)||'-'||substr(aoaddt,3,2)||'-'||substr(aoaddt,5,2))
END + (INT(SUBSTR(TRCD.DLL1TX,9,3)) * 7) DAYS) NXTDAT
FROM
(SELECT * FROM SGTRCDP WHERE DLIMST<>'H' AND TRIM(DLK5Cd)='S') TRCD
JOIN
(
SELECT
AOAGNB,
AOAQNB,
min(AOADDT) aoaddt
FROM SGCALDP
GROUP BY AOAGNB, AOAQNB
) CLND
ON AOAGNB=SUBSTR(TRCD.DLL1TX,1,4) AND AOAQNB=INT(SUBSTR(TRCD.DLL1TX,12,2))
WHERE DLACCD='CON0539' AND DLJ1CD='CON0539-04'
UNION ALL
SELECT
PPMNXT.EQ,
PPMNXT.TASK,
PPMNXT.FREQ,
AOAGNB OCCYR,
AOAQNB OCCWK,
CASE
WHEN aoaddt/1000000 >= 1 THEN
DATE('20'||substr(aoaddt,2,2)||'-'||substr(aoaddt,4,2)||'-'||substr(aoaddt,6,2))
ELSE
DATE('19'||substr(aoaddt,1,2)||'-'||substr(aoaddt,3,2)||'-'||substr(aoaddt,5,2))
END OCCDAT,
(CASE
WHEN aoaddt/1000000 >= 1 THEN
DATE('20'||substr(aoaddt,2,2)||'-'||substr(aoaddt,4,2)||'-'||substr(aoaddt,6,2))
ELSE
DATE('19'||substr(aoaddt,1,2)||'-'||substr(aoaddt,3,2)||'-'||substr(aoaddt,5,2))
END + (PPMNXT.FREQ * 7) DAYS) NXTDAT
FROM
PPM
PPMNXT
JOIN
(
SELECT
AOAGNB,
AOAQNB,
min(AOADDT) aoaddt
FROM SGCALDP
GROUP BY AOAGNB, AOAQNB
) CLND
ON AOAGNB=YEAR(PPMNXT.NXTDAT) AND AOAQNB=WEEK_ISO(PPMNXT.NXTDAT)
WHERE
YEAR(CASE
WHEN aoaddt/1000000 >= 1 THEN
DATE('20'||substr(aoaddt,2,2)||'-'||substr(aoaddt,4,2)||'-'||substr(aoaddt,6,2))
ELSE
DATE('19'||substr(aoaddt,1,2)||'-'||substr(aoaddt,3,2)||'-'||substr(aoaddt,5,2))
END + (PPMNXT.FREQ * 7) DAYS) <= YEAR(CURRENT_DATE)
)
SELECT EQ, TASK, OCCYR, OCCWK, OCCDAT FROM PPM
That was the best I could do.
You will notice that I set a root to a specific Asset and Task:
WHERE DLACCD='CON0539' AND DLJ1CD='CON0539-04'
Normally I would not filter data in order to retrieve all the scheduled weeks for each tasks. I had to filter on one root key to avoid the query to eventually eat up resources make our AS/400 crash.
Again, I am not an expert in CTEs, there might be a better solution.
Thanks