Join two tables in Maria DB generating new columns

Question

I need to join two tables in MariaDB in a specific way. For each project in table A I need to add one column for objective 1 and one column for objective 2. The value of objective 1 and objective 2 is the value of the field is_core associated with that objective in the table B in the highest phase as in the example.

In my real problem there will be more objectives but for simplicity I have used only 2 here.

Can anyone help with this? I am not very experienced with DB and my head is burning.

Table A

id	name
1	project A
2	project B
3	project C

Table B

id	project_id	objective	is_core	phase
1	1	objective1	0	0
2	1	objective1	0	1
3	1	objective1	1	2
4	1	objective2	0	0
5	1	objective2	0	1
6	1	objective2	0	2
7	2	objective1	1	0
8	2	objective2	0	0
9	3	objective1	0	0
10	3	objective2	0	0

TableA.id links to TableB.projectId

Desired result:

project_id	name	objective1	objective2
1	project A	1	0
2	project B	1	0
3	project B	0	0

Many thanks!

Answer 1

SELECT *,
       EXISTS ( SELECT NULL
                FROM tableB 
                WHERE tableB.project_id = tableA.project_id
                   AND tableB.objective = 'objective1' ) objective1,
       EXISTS ( SELECT NULL
                FROM tableB 
                WHERE tableB.project_id = tableA.project_id
                   AND tableB.objective = 'objective2' ) objective2
FROM tableA;

If the amount of objectives (and hence the amount of final columns) is indefinite (i.e. you need in PIVOT) then use dynamic SQL.

---

our MariaDB version is 10.6.11-MariaDB-log – Jesus Paradinas

WITH cte AS (
  SELECT *, 
         ROW_NUMBER() OVER (PARTITION BY project_id, objective ORDER BY phase DESC) rn
  FROM B
)
SELECT A.id, A.name,
       MAX((objective = 'objective1') * (is_core)) objective1,
       MAX((objective = 'objective2') * (is_core)) objective2
FROM A
JOIN cte ON A.id = cte.project_id
WHERE rn = 1
GROUP BY cte.project_id

https://dbfiddle.uk/M6GMRIMH

PS. The same solution (which uses not CTE but subquery) was already provided by @nnichols..

Answer 2

My apologies! I have just reread your question and I obviously completely missed the explanation of the values for the objective columns. The highest phase requirement turns this into a fairly typical groupwise-maximum problem. Here are three of the many possible solutions.

This first query is just using two correlated subqueries in the SELECT list:

SELECT a.id AS project_id, a.name,
    (
        SELECT is_core
        FROM tableB
        WHERE project_id = a.id
        AND objective = 'objective1'
        ORDER BY phase DESC
        LIMIT 1
    ) AS objective1,
    (
        SELECT is_core
        FROM tableB
        WHERE project_id = a.id
        AND objective = 'objective2'
        ORDER BY phase DESC
        LIMIT 1
    ) AS objective2
FROM tableA a;

This next query uses NOT EXISTS to make sure we have the latest phase and conditional aggregation:

SELECT b.project_id, a.name, 
    MAX(objective = 'objective1' AND is_core) AS objective1,
    MAX(objective = 'objective2' AND is_core) AS objective2
FROM tableA a
JOIN tableB b ON a.id = b.project_id
              AND NOT EXISTS (
                  SELECT 1 FROM tableB
                  WHERE project_id = b.project_id
                  AND objective = b.objective
                  AND phase > b.phase
              )
GROUP BY b.project_id;

This query checks the value is_core for the latest phase by using ROW_NUMBER() and conditional aggregation:

SELECT project_id, name,
    MAX(objective = 'objective1' AND is_core) AS objective1,
    MAX(objective = 'objective2' AND is_core) AS objective2
FROM (
    SELECT
        a.name, b.project_id, b.objective, b.is_core,
        ROW_NUMBER() OVER (PARTITION BY project_id, objective ORDER BY phase DESC) rn
    FROM tableA a
    JOIN tableB b ON a.id = b.project_id
) t
WHERE rn = 1
GROUP BY project_id;

All three of these queries will probably perform best with a full covering index on (project_id, objective, phase, is_core) but should at least have an index on (project_id, objective). YMMV!

Here's a fiddle

Answer 3

You can make two left join, one for objective1 and the other for objective2.

SELECT tableA.id AS project_id,
       tableA.name,
       MAX(tableB.is_core) AS "objective1",
       MAX(tableB2.is_core) AS "objective2"
FROM tableA
LEFT JOIN tableB ON tableB.project_id = tableA.id AND tableB.objective = "objective1"
LEFT JOIN tableB AS tableB2 ON tableB2.project_id = tableA.id AND tableB2.objective = "objective2"
GROUP BY tableA.id