Preamble: I have been sent here by @schauder in spring-data-relational in regards to https://github.com/spring-projects/spring-data-relational/issues/1300.
Currently JdbcTemplate
does not support SELECT
statements that use the VALUES
keyword. Here is an example from spring-data-jdbc that uses JdbcTemplate
under the hood:
@Repository
interface MyRepository {
@Query("SELECT * FROM my_entity WHERE id IN (VALUES (:ids))")
List<MyEntity> myQuery(List<String> ids)
}
:ids
is not properly expanded (it needs to put every list entry into ()
) and thus generates wrong SQL.
I know this would be complex to support for complex objects types as IN
-statements works with tuples in some databases but the simple, one-valued variant should be pretty straightforward.
At least PostgreSQL generates different plans for a simple IN
vs IN VALUES
clause, especially when the input list is big (>100) which perform quite differently (10-30% worse for us).
Using a VALUES
list is also interesting when using CTE (WITH
) to populate a temporary table with user provided input, this is not achievable with an IN
statement.
Disclaimer: I haven't checked if VALUES
expansion works in custom INSERTS
Comment From: jhoeller
I suppose this works fine with regular positional parameters (where we do not parse the provided SQL ourselves), and you are effectively asking for NamedParameterJdbcTemplate
support for VALUES
, expanding a nested named parameter accordingly?
Comment From: schauder
I suppose this works fine with regular positional parameters (where we do not parse the provided SQL ourselves), and you are effectively asking for
NamedParameterJdbcTemplate
support forVALUES
, expanding a nested named parameter accordingly?
Yes, this is correct. Spring Data JDBC uses NamedParameterJdbcTemplate
and it's named parameters exclusively.
Comment From: encircled
@jhoeller, I can prepare a PR for this, but I'm not sure if such a feature belongs in the framework, as it is not supported by jdbc statements.