ANCESTOR Query Parse Error: Using DISTINCT

Not this question (where the problem is using ANCESTOR = instead of ANCESTOR IS)

I used ANCESTOR IS properly:

SELECT DISTINCT batch_no 
FROM Entry 
WHERE ANCESTOR IS KEY('ag1kZXZ-dHRiLXRhbWVychkLEgdCcmV3ZXJ5IgxCUi1USEVSRURQRUEM')

But I still got an error:

Identifier is a reserved keyword as symbol ANCESTOR

Why is that?

EDIT

I'm getting the same error later years later as I try to update from db to ndb, but I think it's a different reason...

Identifier is a reserved keyword at symbol ANCESTOR

But in this case my query is not using DISTINCT

SELECT *
FROM Balance
WHERE ANCESTOR IS KEY('...')
AND first_balance_ever = TRUE
ORDER BY created_date

So I'm not sure the problem here, maybe it's ANCESTOR related, maybe it's KEY(...) related...

In this case, DISTINCT was causing the problem. Remove it and the query works as expected.

I think the query is implicitly projecting only that property you DISTINCT on, so you cannot access the key/ancestor for ANCESTOR IS. (But is this true even if you use SELECT *?)

But the error message is not very clear...

Google describes that DISTINCT is experimental. Mentions this in db.Query documentation:

Grouping^{(experimental)}

Projection queries can use the distinct keyword to ensure that only completely unique results will be returned in a result set. This will only return the first result for entities which have the same values for the properties that are being projected.

And in the GQL documentation

The optional DISTINCT^{(experimental)} clause specifies that only completely unique results will be returned in a result set. This will only return the first result for entities which have the same values for the properties that are being projected.