scala and spring
DESCRIPTION
This presentation shows how Spring can be used with Scala.TRANSCRIPT
Eberhard Wolff Architecture and Technology Manager adesso AG, Germany
Spring und Scala
About me • Eberhard Wolff • Architecture & Technology Manager at adesso • adesso is a leading IT consultancy in Germany • Speaker • Author (e.g. first German Spring book) • Blog: http://ewolff.com • Twitter: @ewolff • http://slideshare.com/ewolff • [email protected]
Why Scala and Spring?
• Scala – Strongly typed
language – Elegant – Functional
programming – Focus on
Concurrency – Lack of enterprise
frameworks
• Spring – The tools for
enterprise apps – Well established – Lots of know how – Very flexible
Spring‘s Core Elements
• Dependency Injection – Organize the collaboration of objects
• Aspect Oriented Programming – Handle cross cutting concerns like security or
transactions • Portable Service Abstraction
– Easy, unified APIs for JMS, JDBC, tx … • Testing • How can they be used with Scala?
DEPENDENCY INJECTION
Dependency Injection
• Depended objects are injected
• Advantages: – Better handling of dependencies – Easier testability
– Easier configuration
Dependency Injection
• Dependency Injection is a Pattern • i.e. you can implement it in code • …and therefore in plain Scala • Configuration in a file: more flexibility
– No compile / redeploy – Configure values, not just references
• Spring offers a lot of approaches to DI • XML, JavaConfig and Annotations
Example • DAO depends on a DataSource• Injected in the constructor • Matches Scala’s immutability approach
class CustomerDAO(dataSource: DataSource) { val jdbcTemplate = new JdbcTemplate(dataSource)...}
On Singletons
• Scala introduces objects as Singletons • Example uses Scala classes • Spring needs to do the creation so
Dependency Injection can be done • Might consider @Configurable but that
adds AspectJ Load Time Weaving… • More flexibility concerning scopes
Spring XML Configuration <beans ...> <jdbc:embedded-database type="HSQL" id="dataSource" /> <bean id="customerDAO" class="de.adesso.scalaspring.dao.CustomerDAO"> <constructor-arg ref="dataSource" /> </bean></beans>
Spring XML Configuration
• Very easy and little difference to Java • Scala won’t create getter and setter for
<property > • So: Use Scala’s @BeanProperty to
generate getters and setters • Marks property as configurable by Spring • Might want to create your own Conversions
to configure Scala types
Spring XML & Scala Collections • Scala has its own collection classes • Cannot be configured with Spring XML
out of the box • Need Conversions • Or create custom namespace
<bean class="de.adesso....ScalaBean"> <property name="list" > <scala:list > <value type="java.lang.Integer">42</value> </scala:list> </property></bean>
Spring JavaConfig
• Allows the definition of Spring Beans using Java classes
• Classes contain code to create Spring Beans • Still conforms to Spring Bean rules
– Singleton, AOP, autowiring etc • Can be used with Scala
Spring JavaConfig with Scala @Configurationclass ScalaConfig { @Autowired var dataSource: DataSource = _ @Bean def transactionManager() = new DataSourceTransactionManager(dataSource) @Bean def customerDAO() = new CustomerDAO(dataSource)}
Defined in XML
Not really elegant..
Spring JavaConfig • Almost like a Spring Configuration DSL • No need for Spring Scala DSL (?) • Full power of Scala for creating objects • Can also add configuration for value from
properties files etc • Also nice for infrastructure
• But reconfiguration = recompiling and redeployment
Annotations
• Annotate classes • Classpath scanned for annotated classes • These become Spring beans
Annotations Code
<beans ... ><context:component-scan base-package="de.adesso.scalaspring.dao" />
@Componentclass CustomerDAO { @Autowired var dataSource: DataSource = _ }
Annotations Code
<beans ... default-autowire="constructor"><context:component-scan base-package="de.adesso.scalaspring.dao" />
@Componentclass CustomerDAO(dataSource: DataSource) {}
Naming Convention
<context:component-scan base-package="de.adesso.scalaspring.dao" use-default-filters="false"> <context:include-filter type="regex" expression=".*DAO" /></context:component-scan>
class CustomerDAO(dataSource: DataSource) {}
No annotations – just a naming convention
SERVICE ABSTRACTION
Service Abstraction
• Example: JDBC • Common advantages:
– Runtime exceptions instead of checked exceptions
– Uniform API (e.g. transactions) – Resource handling solved
Service Abstraction: Code • Works out of the box • However, needs Java type issues (Integer)
class CustomerDAO(dataSource: DataSource) { val jdbcTemplate = new JdbcTemplate(dataSource) def deleteById(id: Int) = jdbcTemplate.update( "DELETE FROM CUSTOMER WHERE ID=?", id : java.lang.Integer)}
More Complex
• How can one access a ResultSet? • Resource handled by JDBC • Cannot return it – it has to be closed • Solution: callback • …and inner class
Callbacks in Java public class CustomerDAO extends SimpleJdbcDaoSupport { private static final class CustomerResultSetRowMapper implements ParameterizedRowMapper<Customer> { public Customer mapRow(ResultSet rs, int rowNum) { Customer customer = new Customer(rs.getString(1), rs.getString(2), rs.getDouble(4)); customer.setId(rs.getInt(3)); return customer;
} } public List<Customer> getByName(String name) { return getSimpleJdbcTemplate() .query( "SELECT * FROM T_CUSTOMER WHERE NAME=?", new CustomerResultSetRowMapper(), name); }}
Callbacks in Scala
• Callbacks are really functions • Called on each row
• Use template with Scala function?
Callback in Scala
• Why can the function be converted in a RowMapper?
def findById(id: Int): Option[Customer] = { val result: Buffer[Customer] = jdbcTemplate.query( "SELECT * FROM CUSTOMER C WHERE C.ID=?", (rs: ResultSet) => { Customer(rs.getInt(1), rs.getString(2), rs.getString(3), rs.getDouble(4)) } : RowMapper[Customer] ,id : java.lang.Integer) result.headOption}
Implicit Conversions in Scala
• Implicits allow conversion from one type to another
• Example: Double to Int • Used any time an Int is needed and a Double
is provided • Can we use implicits to convert a function to
a callback class?
implicit def double2int(d:Double): Int = d.toInt
Implicit for Function to Callback
• Converts a function into a callback object • Transparently behind the scenes
implicit def rowMapperImplicit[T]( func: (ResultSet) => T) = { new RowMapper[T] { def mapRow(rs: ResultSet, rowNum: Int) = func(rs) }}
Some Problems
• Scala value types and collections must be converted to Java objects (i.e. Int to Integer)
• null instead of Option[T] • classOf[T] instead of plain type
• Wrapper would be more natural but more effort
ASPECT ORIENTED PROGRAMMING
Why AOP?
• Centralized implementation of cross cutting concerns
• E.g. security, transactions, tracing.. • Aspect =
– Advice : executed code – Pointcut : where the code is executed
• Let’s see some Pointcut expressions…
execution(void hello())
Execution of method hello, no parameters, void return type
execution(int com.ewolff.Service.hello(int))
Execution of method hello in class Service in package com.ewolff one int as parameters, int return type
execution(* *..*Service.*(..))
Execution of any method in class in any package with suffix Service Any number of parameters, any return type
Any Service i.e. add behavior to every service
(security, transaction)
Defines what constitutes a service
Proper and orderly usage of AOP
AOP Example @Aspectpublic class TracingAspect {
@Before("execution(* com.ewolff.highscore..*.*(..))")public void traceEnter(JoinPoint joinPoint) { System.out.println("enter "+joinPoint);}
@After("execution(* com.ewolff.highscore..*.*(..))")public void traceExit(JoinPoint joinPoint) { System.out.println("exit "+joinPoint);}
}
Problems
• Must provide parameter less constructor • Pointcut depends on Java type system • Scala has a different type system • Can combine Scala + Spring AOP
– Use bean Pointcut: bean(aVerySpecificBean) bean(*DAO)
– Or Annotations: execution(@retry.Retry * *(..))
Transactions with AOP
• Spring’s @Transaction annotation • AOP is used for transactions and security • Mark a method or class as transactional • AOP behind the scenes
@Transactionalclass CustomerDAO(dataSource: DataSource) { @Transactional def save(customer: Customer): Customer = { }}
AOP and Scala: 2nd Thought
• Spring AOP is not efficient • Method calls are done dynamically • AspectJ will make project setup too complex • A modern programming language should
handle cross cutting concerns • E.g. meta programming in dynamic
languages • Can we do better?
Functions
• Can use functions to “wrap” methods, blocks and functions and do transactions
• Based on TransactionTemplate and callbacks
• TransactionTemplate executes callbacks in a transaction
Code
implicit def txCallbackImplicit[T](func: => T)…• Again implicit • Converts a function with return type T to a TransactionCallback
• Now we need to call the TransactionTemplate to provide the transaction
• A function will take the transaction configuration and call the passed in function
Code def transactional[T]( propagation: Propagation = Propagation.REQUIRED, …)(func: => T): T = { val txAttribute = new TransactionAttributeWithRollbackRules( propagation,…) val txTemplate = new TransactionTemplate(txManager,txAttribute) txTemplate.execute(func)}
1st parameter: transaction configuration
2nd parameter: function
Call TransactionTemplate with function (converted to TransactionCallback) and configuration
Usage
• Can wrap methods to make them transactional
@Componentclass TxCustomerDAO(dataSource: DataSource) { def save(customer: Customer): Customer = transactional() { jdbcTemplate.update(…); }}
Usage
• Can also be used to wrap any code block • But: No way to make a whole class /
system transactional
transactional(propagation=Propagation.REQUIRED){ customerDAO.save( Customer(0, "Wolff", "Eberhard", 42.0))}
TESTING
Testing in Spring
• Injection in Test classes
• Transaction handling – Start a transaction for each test method – At the end of the method: Rollback
• Benefit: No need to clean up the database • Good start: No production code in Scala
Testing with JUnit 4, Spring and Scala
@RunWith(classOf[SpringJUnit4ClassRunner])@Transactional@ContextConfiguration(Array("/spring/scalaSpringConfig.xml"))class CustomerDAOTest extends Config { @Autowired var customerDAO : CustomerDAO = null @Test def testSaveDelete() { val numberOfCustomersBefore = customerDAO.count() …}}
SUM UP
Sum Up • Scala and Spring are a good match • Spring is very adaptable • Dependency Injection
– XML works, some improvements possible – Annotations and JavaConfig: No problems
• Service Abstraction – Functions are a good fit
• AOP – Can work with Scala but not ideal – Scala can do similar things with functions
Potential Improvements
• Dependency Injection – Support for Scala collections mostly done – Support for Scala properties (no @BeanProperty) – Support for Scala Objects aka Singletons – Conversions for all basic Scala types – Spring configuration DSL (?)
• Service Abstraction – Provide implicits for non-JDBC callbacks
Potential Improvements
• AOP – Provide functions for all common aspects besides
transaction
• Testing – Support Scala test frameworks – http://www.cakesolutions.org/specs2-spring.html
Links • https://github.com/ewolff/scala-spring • Request for Scala version of Spring (only 16 votes)
https://jira.springsource.org/browse/SPR-7876 • Scala and AspectJ: Approaching modularization of crosscutting
functionalitieshttp://days2011.scala-lang.org/sites/days2011/files/52.%20AspectJ.pdf
• Sample for Spring Security and Scala https://github.com/tekul/scalasec
• Spring Integration Scala DSL https://github.com/SpringSource/spring-integration-scala
• (German) Thesis about Scala & Lift vs. Java EE: http://www.slideshare.net/adessoAG/vergleich-des-scala-webframeworks-lift-mit-dem-java-ee-programmiermodell
• (German) Thesis about Scala, JSF and Hibernate: http://www.slideshare.net/bvonkalm/thesis-5821628
www.AAAjobs.de
Wir suchen Sie als
Ø Software-Architekt (m/w) Ø Projektleiter (m/w) Ø Senior Software Engineer (m/w)
Ø Kommen Sie zum Stand und gewinnen Sie ein iPad 2!