UPDATE table JOIN table SET only updates unique id's - mysql

Question

I want to update Table B with sum(aa) from Table A where A.id=B.id and into the column set by A.d (E.g. A:223, d1, 5 updates B: id 223, column d1 = d1 + 5).
TABLE A TABLE B
id d aa id d1 d4 d7
--- +-------+---- ----+-------+-------+-----
223 | d1 | 5 221 | 0 | 5 | 0 <
224 | d1 | 5 222 | 0 | 5 | 0 <
225 | d1 | 5 223 | 5 | 0 < | 0 <
226 | d1 | 5 224 | 5 | 0 < | 0 <
| | 225 | 5 | 0 < | 0 <
221 | d4 | 5 226 | 5 | 0 < | 0 <
222 | d4 | 5 < = missing
223 | d4 | 5
224 | d4 | 5
225 | d4 | 5
226 | d4 | 5
| | EXPECTED:
221 | d7 | 5 id d1 d4 d7
222 | d7 | 4 ----+-------+-------+------
223 | d7 | 5 221 | 0 | 5 | 5
224 | d7 | 5 222 | 0 | 5 | 5
224 | d7 | 5 223 | 5 | 5 | 5
225 | d7 | 5 224 | 5 | 5 | 10
226 | d7 | 5 225 | 5 | 5 | 5
226 | d7 | 5 226 | 5 | 5 | 15
226 | d7 | 5
Query: (As I can't use dynamic column names I need a CASE WHEN THEN)
UPDATE `TABLE_B` `B`
JOIN(
SELECT `id`,`d`,SUM(`aa`)`aa`
FROM `TABLE_A` GROUP BY `id`,`d`
) `A`
ON `A`.`id`=`B`.`id`
SET
`d1`= CASE `B`.`d` WHEN "d1" THEN `A`.`d1`+`B`.`aa` ELSE `A`.`d1` END,
`d2`= CASE `B`.`d` WHEN "d2" THEN `A`.`d2`+`B`.`aa` ELSE `A`.`d2` END,
`d3`= CASE `B`.`d` WHEN "d3" THEN `A`.`d3`+`B`.`aa` ELSE `A`.`d3` END,
`d4`= CASE `B`.`d` WHEN "d4" THEN `A`.`d4`+`B`.`aa` ELSE `A`.`d4` END,
`d5`= CASE `B`.`d` WHEN "d5" THEN `A`.`d5`+`B`.`aa` ELSE `A`.`d5` END,
`d6`= CASE `B`.`d` WHEN "d6" THEN `A`.`d6`+`B`.`aa` ELSE `A`.`d6` END,
`d7`= CASE `B`.`d` WHEN "d7" THEN `A`.`d7`+`B`.`aa` ELSE `A`.`d7` END
The problem is this only updates unique id's. It takes 223, 224, 225, 226 from d1, then only 221, 222 from d4 and nothing from d7, instead of updating 16 values (4x d1, 6x d4, 6x d7).
So where in the JOIN am I doing wrong?
Fiddle

Answer 1

I got a solution by using two subqueries to first make a full table with all columns in it and only one row per id:
SELECT
`id`,
CASE `d` WHEN "d1" THEN SUM(`aa`) ELSE 0 END AS `a1`,
CASE `d` WHEN "d4" THEN SUM(`aa`) ELSE 0 END AS `a4`,
CASE `d` WHEN "d7" THEN SUM(`aa`) ELSE 0 END AS `a7`
FROM `TABLE_A`
GROUP BY `id`,`d`
Giving (example only for id's 221 and 226):
id d1 d4 d7
----+---+---+----
221 | 0 | 5 | 0
221 | 0 | 0 | 5
226 | 5 | 0 | 0
226 | 0 | 5 | 0
226 | 0 | 0 | 15
Then make sure there is only one row for each id:
SELECT
`id`,
SUM(`a1`) `a1`,
SUM(`a4`) `a4`,
SUM(`a7`) `a7`
FROM(
.... previous select ...
)
GROUP BY `id`
Giving (example only for id's 221 and 226):
id d1 d4 d7
----+---+---+----
221 | 0 | 5 | 5
226 | 5 | 5 | 15
Now I could update Table B:
UPDATE `TABLE_B` `B`
JOIN(
SELECT
`id`,
SUM(`a1`) `a1`,
SUM(`a4`) `a4`,
SUM(`a7`) `a7`
FROM(
SELECT
`id`,
CASE `d` WHEN "d1" THEN SUM(`aa`) ELSE 0 END AS `a1`,
CASE `d` WHEN "d4" THEN SUM(`aa`) ELSE 0 END AS `a4`,
CASE `d` WHEN "d7" THEN SUM(`aa`) ELSE 0 END AS `a7`
FROM `TABLE_A`
GROUP BY `id`,`d`
) `C`
GROUP BY `id`
)`A`
ON `A`.`id`=`B`.`id`
SET
`d1`=`d1`+`a1`,
`d4`=`d4`+`a4`,
`d7`=`d7`+`a7`
Maybe there are better solutions, but this works.