Postgres Partition Performance Tuning - postgresql

Question

Select query scans all the child tables. can we make the query-optimizer to scan the right child table?
Example:
Created a parent table and two child table using the inheritance concept in Postgres 9.6, ignored constraints to make it simple
create table student(id INTEGER, name varchar(10), result varchar(1) );
create table student_pass() inherits (student);
create table student_fail() inherits (student);
Index
create index student_result_idx on student (result);
create index student_result_idx2 on student_pass (result) where result='P';
create index student_result_idx3 on student_fail (result) where result='F';
Procedure
CREATE OR REPLACE FUNCTION student_partition()
RETURNS TRIGGER AS $$
BEGIN
IF (new.result = 'P')THEN
INSERT INTO student_pass VALUES (NEW.*);
ELSE
INSERT INTO student_fail VALUES (NEW.*);
END IF;
RETURN NULL;
END;
$$
LANGUAGE plpgsql;
Trigger
CREATE TRIGGER insert_trigger BEFORE INSERT ON student
FOR EACH ROW EXECUTE procedure student_partition();
Insert
insert into student values
(1,'aaa','P'),
(2,'bbb','F');
insert happens in their respective tables as expected
Select
select * from student where result='P';
The problem here is when I do select it scans all the tables. How to make the query-optimizer smart enough to pick the right child table?
Do we need the where condition in the index as the entire table is going to be either 'P' or 'F'?
Output of EXPLAIN(analyze, buffers) select * from student where result='P'
Append (cost=0.00..37.94 rows=11 width=50) (actual time=0.016..0.042 rows=2 loops=1)
Buffers: shared hit=4
-> Seq Scan on student (cost=0.00..2.30 rows=1 width=50) (actual time=0.015..0.017 rows=1 loops=1)
Filter: ((result)::text = 'P'::text)
Rows Removed by Filter: 1
Buffers: shared hit=1
-> Bitmap Heap Scan on student_pass (cost=4.17..12.64 rows=5 width=50) (actual time=0.013..0.014 rows=1 loops=1)
Recheck Cond: ((result)::text = 'P'::text)
Heap Blocks: exact=1
Buffers: shared hit=2
-> Bitmap Index Scan on student_result_idx2 (cost=0.00..4.17 rows=5 width=0) (actual time=0.007..0.007 rows=1 loops=1)
Buffers: shared hit=1
-> Seq Scan on student_fail (cost=0.00..23.00 rows=5 width=50) (actual time=0.007..0.007 rows=0 loops=1)
Filter: ((result)::text = 'P'::text)
Rows Removed by Filter: 1
Buffers: shared hit=1
Planning time: 0.447 ms
Execution time: 0.120 ms

Answer 1

Adding constraints helps
alter table student_pass add constraint pass_cst check (result ='P');
alter table student_fail add constraint fail_cst check (result not in ('P'));
Output of EXPLAIN(analyze, buffers) select * from student where result='P'
Append (cost=0.00..23.00 rows=6 width=50) (actual time=0.299..0.303 rows=2 loops=1)
Buffers: shared read=1
-> Seq Scan on student (cost=0.00..0.00 rows=1 width=50) (actual time=0.004..0.004 rows=0 loops=1)
Filter: ((result)::text = 'P'::text)
-> Seq Scan on student_pass (cost=0.00..23.00 rows=5 width=50) (actual time=0.294..0.296 rows=2 loops=1)
Filter: ((result)::text = 'P'::text)
Buffers: shared read=1
Planning time: 10.488 ms
Execution time: 0.361 ms
Query-optimiser skipped student_fail table