Mysql JOIN with extra priority column - mysql

Question

I have two days trying to do this query with no luck.
I have two tables 'DEMAND' and 'DEMAND_STATE' (one to many relation). The table DEMAND_STATE have millions entries.
CREATE TABLE DEMAND
(
ID INT NOT NULL,
DESTINY_ID INT NOT NULL
)
CREATE TABLE DEMAND_STATE
(
ID INT NOT NULL,
PRIORITY INT NOT NULL,
QUANTITY DOUBLE NOT NULL,
CASE_ID INT NOT NULL,
DEMAND_ID INT NOT NULL,
PHASE_ID INT NOT NULL
)
The QUANTITY of the DEMAND_STATE is given according to a CASE_ID and PHASE_ID. We have 'N' PHASES in 'M' CASES. Always the same number of Phases in all Cases. We always have a initial Base Quantity called 'BASE CASE' in the Case with CASE_ID = 1.
For example to obtain quantity for Case (id=2) and Case Base (id=1)
select D.*, S.PRIORITY, S.QUANTITY, S.CASE_ID, S.DEMAND_ID, S.PHASE_ID
FROM DEMAND D
join DEMAND_STATE S on (D.ID = S.DEMAND_ID)
WHERE (S.CASE_ID = 2 OR S.CASE_ID = 1)
(paste only for id=8)
ID PRIORITY QUANTITY CASE_ID DEMAND_ID PHASE_ID
8 0 85 1 8 1
8 0 83 1 8 2
8 0 88 1 8 3
8 0 89 1 8 4
8 10 85 2 8 1
8 10 84 2 8 2
8 10 86 2 8 3
8 10 89 2 8 4
We need to obtain for all Demand in 'DEMAND' only the Quantity for Each Phase with MAX priority. The idea is no duplicate DEMAND_STATE data for each new Case creation. Only create new state rows when Demand-Case-Phase is different to Case Base. This is a new project and we accept changes in model for better performance.
I also tried with the MAX calculation. This query over DEMAND_STATE works fine but only obtain data for a concrete DEMAND_ID. Further i think this solution can be so expensive.
SELECT P.ID, P.QUANTITY, P.CASE_ID, P.DEMAND_ID, P.PHASE_ID
FROM DEMAND_STATE P
JOIN (
SELECT PHASE_ID, MAX(PRIORITY) max_priority, S.DEMAND_ID
from DEMAND_STATE S
WHERE S.DEMAND_ID = 1
AND (S.CASE_ID=1 OR S.CASE_ID=2)
GROUP BY S.PHASE_ID
) SUB
ON (SUB.PHASE_ID = P.PHASE_ID AND SUB.max_priority = P.PRIORITY)
WHERE P.DEMAND_ID = 1
GROUP BY P.PHASE_ID
The result:
ID QUANTITY CASE_ID DEMAND_ID PHASE_ID
1 86 1 1 1
2 85 1 1 2
3 81 1 1 3
8 500 2 1 4
This is the result expected:
ID ID PRIORITY QUANTITY CASE_ID PHASE_ID
8 1 0 86 1 1 (data from Case Base id=1 priority 0)
8 2 10 85 1 2 (data from Case Baseid=1 priority 0)
8 3 10 81 1 3 (data from Case Base id=1 priority 0)
8 64 10 500 2 4 (data from Case id=2 priority 10)
thank for help :)
Edit:
Result of Simon proposal:
ID QUANTITY CASE_ID DEMAND_ID PHASE_ID
1 86 1 1 1
2 85 1 1 2
3 81 1 1 3
4 84 1 1 4 (this row shouldnt exist)
8 500 2 1 4 (this is the correct row)
Also would have to join it with DEMAND
#didierc response:
ID ID MAX(S.PRIORITY) QUANTITY CASE_ID PHASE_ID
1 8 10 500 2 4
2 13 10 81 2 1
2 14 10 83 2 2
2 15 10 84 2 3
3 21 10 81 2 1
4 31 10 86 2 3
4 32 10 80 2 4
4 29 10 85 2 1
4 30 10 81 2 2
we need for each DEMAND four rows with the quantity Value. In Case Base we have four quantity and in Case 2 we only change the quantity for phase 4. We need always four rows for each demand.
Database DEMAND_STATE data:
ID PRIORITY QUANTITY CASE_ID DEMAND_ID PHASE_ID
1 0 86 1 1 1
2 0 85 1 1 2
3 0 81 1 1 3
4 0 84 1 1 4
8 10 500 2 1 4

Answer 1

We need to obtain for all Demand in 'DEMAND' only the Quantity for Each Phase with MAX priority
I translate the above, according to your sample result set, as:
SELECT
D.ID, S.ID, MAX(S.PRIORITY), S.QUANTITY, S.CASE_ID, S.PHASE_ID
FROM DEMAND D
LEFT JOIN DEMAND_STATE S
ON D.ID = S.DEMAND_ID
GROUP BY S.PHASE_ID, S.DEMAND_ID
Update:
To get the maximum priority for each pair(demand_id,phase_id)n we use the following query:
SELECT
DEMAND_ID, PHASE_ID, MAX(PRIORITY) AS PRIORITY
FROM DEMAND_STATE
GROUP BY DEMAND_ID, PHASE_ID
Next, to retrieve the set of phases for a given demand, just make an inner join on demand state:
SELECT S.* FROM DEMAND_STATE S
INNER JOIN (
SELECT
DEMAND_ID, PHASE_ID, MAX(PRIORITY) AS PRIORITY
FROM DEMAND_STATE
GROUP BY DEMAND_ID, PHASE_ID
) S2
USING (DEMAND_ID,PHASE_ID, PRIORITY)
WHERE DEMAND_ID = 1
If you want to limit the possible cases, include a where clause in the query S2:
SELECT S.* FROM DEMAND_STATE S
INNER JOIN (
SELECT
DEMAND_ID, PHASE_ID, MAX(PRIORITY) AS PRIORITY
FROM DEMAND_STATE
WHERE CASE_ID IN (1,2)
GROUP BY DEMAND_ID, PHASE_ID
) S2
USING (DEMAND_ID,PHASE_ID, PRIORITY)
WHERE DEMAND_ID = 1
However, your comments and update indicates that MAX(PRIORITY) does not seem very relevant after all. My understanding is that you have a base case, which may be overriden by another case in a given scenario (that scenario is the pair base case + some other case). Clarify that point in your question body if this is incorrect. If that is the case, you may change the above query by replacing PRIORITY by CASE_ID:
SELECT S.* FROM DEMAND_STATE S
INNER JOIN (
SELECT
DEMAND_ID, PHASE_ID, MAX(CASE_ID) AS CASE_ID
FROM DEMAND_STATE
WHERE CASE_ID IN (1,2)
GROUP BY DEMAND_ID, PHASE_ID
) S2
USING (DEMAND_ID,PHASE_ID, CASE_ID)
WHERE DEMAND_ID = 1
The only reason I see from having a priority is if you wish to combine more than 2 cases, and use priority to select which case will prevail depending on the phase.
You may of course prepend an inner join on DEMAND to include the related demand data.

Answer 2

Use of subqueries should be able to do as you wish, if I understand your question correctly. Something along the lines of the following:
SELECT
P.ID,
P.QUANTITY,
P.CASE_ID,
P.DEMAND_ID,
P.PHASE_ID
FROM DEMAND_STATE P
INNER JOIN (
-- Next level up groups it down and so gets the rows first returned for each PHASE_ID, which is the highest priority due to the subquery
SELECT
D.PHASE_ID,
D.PRIORITY,
D.DEMAND_ID
FROM (
-- Top level query to get all rows and order them in desc priority order
SELECT
S.PHASE_ID,
S.PRIORITY,
S.DEMAND_ID
FROM DEMAND_STATE S
WHERE S.DEMAND_ID IN (1) -- Update this to be whichever DEMAND_IDs you are interested in
AND S.CASE_ID IN (1,2)
ORDER BY
S.PHASE_ID ASC,
S.DEMAND_ID ASC,
S.PRIORITY DESC
) D
GROUP BY
D.PHASE_ID,
S.DEMAND_ID
) SUB
ON SUB.PHASE_ID = P.PHASE_ID
AND SUB.DEMAND_ID = P.DEMAND_ID
The top level subquery exists to get the rows you are interested in and order them in an order which allows predictable results when they are then grouped down by PHASE_ID and DEMAND_ID. This in turn allows a simple INNER JOIN to DEMAND_STATE hopefully (unless I have misunderstood your query)
This may still be expensive though depending on how much data is within that top level query.