When using the pull request I see a problem when there is more than attribute node. The same columns with the same aliases are repeated in the select clause.

I've added a FailureExpected test to master and 4.3 branches that reproduces the original bug: org.hibernate.jpa.test.graphs.queryhint.QueryHintEntityGraphTest.testLoadGraphWithRestriction().

After applying the pull request commit, the resulting SQL is:

select
company0_.id as id1_0_0_,
location1_.id as id1_7_1_,
company0_.location_id as location2_0_0_,
markets2_.Company_id as Company_1_0_0__,
markets2_.markets as markets2_2_0__,
location1_.address as address2_7_1_,
location1_.zip as zip3_7_1_,
markets2_.Company_id as Company_1_0_0__,
markets2_.markets as markets2_2_0__
from
Company company0_ cross
join
Location location1_
left outer join
Company_markets markets2_
on company0_.id=markets2_.Company_id
where
company0_.location_id=location1_.id
and location1_.zip=?
monospaced text

These are duplicated columns/aliases:

{{ markets2_.Company_id as Company_1_0_0__,
markets2_.markets as markets2_2_0__
}}

I don't think the pull request causes this duplication, but it seems to expose some other bug. I need to investigate further.