Left outer join for the same table works wrong on redshift

Question

I need to perform left outer join on the same table. As a result I expect all requested columns from table 1 and only those from table 2 which joined.

Here is how to reproduce:

drop table nz_mri_survey_agg;

create table mri_survey_agg (
HOUSEHOLD_PERSON_HK integer,
MRI_DICTIONARY_ID   INTEGER,
rank integer);

insert into mri_survey_agg values (651694412, 2127115057, 36903);
insert into mri_survey_agg values (647638311, 1293574238, 35413);
insert into mri_survey_agg values (647638311, -2076426274, 35413);
insert into mri_survey_agg values (651694412, -2076426274, 35413);
insert into mri_survey_agg values (651694412, -2051582071, 35411);
insert into mri_survey_agg values (647638311, -1747375415, 35613);
insert into mri_survey_agg values (647638311, 1234567, 35610);

Here is the query:

select distinct t1.household_person_hk, t2.mri_dictionary_id 
from mri_survey_agg t1 
left outer join (
    select household_person_hk, mri_dictionary_id from mri_survey_agg 
    where mri_dictionary_id in 
    (-2076426274, -2051582071, -1747375415)) t2 
on t1.household_person_hk = t2.household_person_hk;

I expect the next output:

household_person_hk mri_dictionary_id
651694412           -2051582071
647638311           -2076426274
651694412           -2076426274
647638311           -1747375415
647638311            <NaN>

The output is:

household_person_hk mri_dictionary_id
651694412           -2051582071
647638311           -2076426274
651694412           -2076426274
647638311           -1747375415

It works perfect on Postgres, but doesn't give me expected results on Redshift.

Appreciate for any hints.

UPD : Actually, the actual output is correct!

Answer 1

I just ran your code on Postgres and it returns four rows, which is correct. Here is a SQL Fiddle, which is now working.

Note that your query can more easily be written as:

select distinct t1.household_person_hk, t2.mri_dictionary_id 
from mri_survey_agg t1 left outer join
     mri_survey_agg t2
     on t1.household_person_hk = t2.household_person_hk and
        t2.mri_dictionary_id in (-2076426274, -2051582071, -1747375415);