Hibernate generated query is slow in application but fast when executed manually - spring

Question

I'm using Spring DATA JPA with Hibernate + PostgreSQL, and I generate q query like this:
SELECT A.field1 AS field1_10_,
A.field2 AS field2_10_,
A.field3 AS field3_10_,
...
A.field10 as field10_10_
FROM mytable A
WHERE (A.field1 BETWEEN 1 AND 2)
AND ((cast(A.field1 AS varchar(255))||cast(A.field2 AS varchar(255))) IN
(SELECT (cast(B.field1 AS varchar(255))||cast(max(B.field2) AS varchar(255)))
FROM mytable B
WHERE B.field1 BETWEEN 1 AND 2
GROUP BY B.field2))
LIMIT 20
All A and B fields are numeric.
If I copy that query and execute it directly in Postgres (using Pgadmin), the result comes in less than a second. But in my application, it takes about a minute for Hibernate to retrieve the result. I have Hibernate statistics activated and the gross of the time is running JDBC statements.
I'm using Spring pagination, what generates a count query using the same specification. This query runs fast (less that a second):
SELECT count(*) AS col_0_0_
FROM mytable A
WHERE (A.field1 BETWEEN 1 AND 2)
AND ((cast(A.field1 AS varchar(255))||cast(A.field2 AS varchar(255))) IN
(SELECT (cast(B.field1 AS varchar(255))||cast(max(B.field2) AS varchar(255)))
FROM mytable B
WHERE B.field1 BETWEEN 1 AND 2
GROUP BY B.field2))
Any idea?
Update: I've just found out that the generatedstatement executed by PostgreSQL is the following:
SELECT A.field1 AS field1_10_,
A.field2 AS field2_10_,
A.field3 AS field3_10_,
...
A.field10 as field10_10_
FROM mytable A
WHERE (A.field1 BETWEEN ? AND ?)
AND ((cast(A.field1 AS varchar(255))||cast(A.field2 AS varchar(255))) IN
(SELECT (cast(B.field1 AS varchar(255))||cast(max(B.field2) AS varchar(255)))
FROM mytable B
WHERE B.field1 BETWEEN ? AND ?
GROUP BY B.field2))
LIMIT $1
Please note that the replacement character in the BETWEEN conditions is the question mark ? but in the limit is the dollar along with the parameter order $1. The limit is automatically added by Spring pagination, if I remove it the query is fast.
Update 2: I've just found out that the problem happens only when specifying a limit of 1, with bigger values all works as expected. Weird.

Answer 1

There are multiple possible causes.
JPA will fetch all rows before returning you the result. Make sure PGAdmin does the same by scrolling to the very end of the result before stopping your timer.
JPA (and in some cases Spring) will convert the ResultSet you get from the database into entities or some other Java objects. If you do non-trivial stuff in constructors or getters, this might cause a performance penalty. Also if your session is really big this might as well cause performance problems. These should become obvious, once you attach a profiler.
The query has to get created from a Specification (in your case), this might take some time as well. Again: attach a profiler to identify this.