CREATE TABLE person (
id INT PRIMARY KEY,
name VARCHAR NOT NULL
);
CREATE TABLE pet (
id INT PRIMARY KEY,
name VARCHAR NOT NULL,
owner_id INT NOT NULL
);
It is common to want to join them, like so: SELECT owner.id, owner.name, COUNT(pet.id) AS numberOfPets
FROM person AS owner
LEFT OUTER JOIN pet ON owner.id = pet.owner_id
GROUP BY owner.id
This doesn't work in standard SQL, because all columns in the SELECT list have to either be aggregated or included in the GROUP BY. owner.name is neither. That is a bit silly though because we know each result row will only have one unambiguous value for the owner name, since the GROUP BY is on a unique column from the same table as the owner name. SELECT owner.id, ANY_VALUE(owner.name) AS name, COUNT(pet.id) AS numberOfPets
FROM person AS owner
LEFT OUTER JOIN pet ON owner.id = pet.owner_id
GROUP BY owner.id
It does seem a bit of a mixed message to enable this by default while still saying the functionality is experimental. I fear many people will take this as a signal to start using the feature in production, then find a later Node.js update breaks things if the behaviour changes.