Wrong result for Agg.regrR2Double() about jool HOT 6 CLOSED

Thanks, can confirm. Will look into this soon

from jool.

The test is correct. The expected result is 1.0:

select regr_r2(x, y)
from (values (1, 1), (1, 2)) t (x, y)

Produces:

|regr_r2|
|-------|
|1      |

from jool.

jOOQ/jOOQ#11547 defines REGR_R2 as:

REGR_R2(x, y) = CASE 
  WHEN VAR_POP(NVL2(x, y, NULL)) = 0 THEN NULL 
  WHEN VAR_POP(NVL2(y, x, NULL)) = 0 THEN 1 
  ELSE CORR(x, y)^2 
END

from jool.

The formula is implemented correctly, but the internal mapAll(Tuple3<Optional<T1>, Optional<T2>, Optional<T3>>) isn't. It's requires all optionals to be non-empty in order to map anything, when in fact this isn't what the CASE expression suggests.

from jool.

Well, mapAll() is implemented correctly, but using it is wrong.

from jool.

The reason why this hasn't been detected in our CI is because we recently migrated CI and jOOλ wasn't part of the new installation...

from jool.