Library to make PostgreSQL stored procedures(SProcs) available through simple Java "SProcService" interfaces including automatic object serialization and deserialization (using typemapper and convention-over-configuration).
Supports horizontal database sharding (partition/access logic lies within application), easy use of pg_advisory_lock
through annotations to ensure single Java node execution, configurable statement timeouts per stored procedure, and
PostgreSQL types including enums and hstore.
To use SProcWrapper, add the following lines to your pom.xml:
<dependency>
<groupId>org.zalando</groupId>
<artifactId>zalando-sprocwrapper</artifactId>
<version>${zalando-sprocwrapper.version}</version>
</dependency>
SProcWrapper provides an efficient and easy-to-use mechanism for translating values from database to Java objects and vice-versa. It allows us to map not only primitive types, but also complex types (Java domain objects).
Here are some examples!
Using basic types:
@SProcService
public interface CustomerSProcService {
@SProcCall
int registerCustomer(@SProcParam String name, @SProcParam String email);
}
CREATE FUNCTION register_customer(p_name TEXT, p_email TEXT)
RETURNS INT AS
$$
INSERT INTO z_data.customer(c_name, c_email)
VALUES (p_name, p_email)
RETURNING c_id
$$
LANGUAGE 'sql'
SECURITY DEFINER;
And a complex type:
@SProcService
public interface OrderSProcService {
@SProcCall
List<Order> findOrders(@SProcParam String email);
}
CREATE FUNCTION find_orders(p_email TEXT,
OUT order_id INT,
OUT order_date TIMESTAMP,
OUT shipping_address ORDER_ADDRESS)
RETURNS SETOF RECORD AS
$$
SELECT o_id,
o_date,
ROW (oa_street, oa_city, oa_country)::ORDER_ADDRESS
FROM z_data.order
JOIN z_data.order_address
ON oa_order_id = o_id
JOIN z_data.customer
ON c_id = o_customer_id
WHERE c_email = p_email
$$
LANGUAGE 'sql'
SECURITY DEFINER;
Please check unit/integration tests for more examples.
The following table shows the mapping between a database type and a Java type:
Database | Java Type |
---|---|
smallint | int |
integer | int |
bigint | long |
decimal | java.math.BigDecimal |
numeric | java.math.BigDecimal |
real | float |
double precision | double |
serial | int |
bigserial | long |
varchar | java.lang.String |
char | char |
text | java.lang.String |
timestamp | java.sql.Timestamp |
timestamptz | java.sql.Timestamp |
date | java.sql.Timestamp |
time | java.sql.Timestamp |
timetz | java.sql.Timestamp |
boolean | boolean |
enum | java.lang.Enum |
array | java.util.List / java.util.Set |
hstore | java.util.Map<java.lang.String, java.lang.String> |
Note: SProcwrapper doesn't support functions returning arrays as a single output. If one wants to return a collection, please return a SETOF instead.
Global value transformer loader loads global value transformers from the classpath. By default, it searches value
transformers in org.zalando
package. You can change this behaviour using
the global.value.transformer.search.namespace
environment variable:
EXPORT global.value.transformer.search.namespace="my.package"
or by calling a method:
GlobalValueTransformerLoader.changeNamespaceToScan("my.package");
It is also possible to provide multiple scan packages using ;
separator:
EXPORT global.value.transformer.search.namespace="my.package;another.package;third.package"
- Java 11
- To compile, one should use Maven 3.0.0 or above
- Spring Framework 5.x
- PostgreSQL JDBC driver ;)
- Google Guava
See pom.xml for the full list of dependencies.
The provided helper script will start a PostgreSQL instance with Docker on port 5432 and run integration tests:
./test.sh
- PostgreSQL JDBC driver does not honor identical type names in different schemas, this may lead to issues if typemapper is used where types are present with equal name in more than one schema (this problem is solved now with the commit 3ca94e64d6322fa91c477200bfb3719deaeac153 to pgjdbc driver);
- PostgreSQL domains are not supported as for now;
- PostgreSQL
hstore
type is mapped from and toMap<String,String>
, there is no way to useMap
of different types for now; - Two different datasources with the same JDBC URL and different search paths might not work properly when we have types with the identical name;
- SProcWrapper relies on the search path to resolve conflicting types with the same name (right now, we are not checking
the schema). If one specifies the schema of the stored procedure's return type, SProcWrapper might end up using the
wrong one, because it will use the
search_path
to resolve the conflict. For more info check test:SimpleIT.testTypeLookupBugWithSchema
; - For integration with Spring's transaction management use the
TransactionAwareDataSourceProxy
as the data source injected into the data source provider.
You can find some more information about the SProcWrapper in our various Zalando Technology blog posts:
- http://tech.zalando.com/blog/goto-2013-why-zalando-trusts-in-postgresql/
- http://tech.zalando.com/blog/zalando-stored-procedure-wrapper-part-i/
- http://tech.zalando.com/blog/files/2013/04/jug_dortmund_april_2013.pdf
🚨 Upgrading from 2.x to 3.x? Please be aware that default scanning package for custom objectTransformer has changed from de.zalando
to org.zalando
.
MIT license. See license file.