postgres - single record with one to many join tables - postgresql

Question

I have two tables a and b. Where main table is 'a' where I want to select from and table 'b' is for filtering.
Below is sample table with some data.
create table a (
id varchar primary key,
name varchar,
p varchar[]
);
insert into a (id, name, p) values
('1', 'v1', array['p1']),
('2', 'v2', array['p1','p2']),
('3', 'v3', array['p2','p3']);
create table b (
p varchar,
x varchar
);
insert into b(p, x) values
('p1', 'x1'),
('p2', 'x2'),
('p3', 'x1'),
('p3', 'x3'),
('p1', 'x2');
I want only one row from table a based on join on column p and filter on x. I tried few options, it works when I have one to one record in a and b but when I have one to many then I get multiple records.
select a.* from a,b where b.p=any(a.p) and b.x='x2';
Output I get is:
id name p
-----------------
1 v1 p1
2 v2 p1,p2
2 v2 p1,p2
3 v3 p2,p3
What I want is
id name p
-----------------
1 v1 p1
2 v2 p1,p2
3 v3 p2,p3
Also I am expecting table 'a' to have millions of rows and 'b' will have only few, so query has to be perform effectively.

Answer 1

As you only want columns from table a use an exists condition rather than a join:
select a.*
from a
where exists (select *
from b
where b.p = any(a.p)
and b.x='x2');
That will however be quite hard to optimize.
Another option - if not too many rows match the join criteria - is to apply a distinct on the result of the join.
In that case a GIN index on the array column can be used:
create index on a using gin (p);
select distinct a.*
from a
join b on a.p #> array[b.p]
where b.x= 'x2';