Spring CDI Mocks: Separating Unit Tests and Integration Tests. The right things in the right place

INTRO

In the last post I clearly became more independently of Spring, not because I hate him, in fact I like Spring a lot. But to postpone the choice of the wiring/infrastructure technology until the production stage, and to encourage standardization and simplicity.

Just for that? No!

In many domains, all bleeding-edge Java framework/toolkit seems to honor the JSR specifications, at the same time JSR specifications very often receive continuous feedback from concrete implementations, some examples are: JPA (JSR-220) with Hibernate, Bean Validation (JSR-303) with Hibernate Validator and Apache Beans Validator, and our main focus CDI (JSR-299) and Spring. Spring actually honors CDI annotations, and if the leading framework on DI (Spring) honors the standard, then there is a big chance for both to be aligned with the future.

As you may noticed I used Spring [before] on the testing phase to wire up the SUT and dependencies with a relative small complexity. Nevertheless simple and small is better, that's why I'll show you how achieve a very similar wiring scenario w/o Spring (in testing phase) with a more simple approach IMO.

There's an open question: Do we still need Spring? Where does it fit?

The trivial answer is: Yes there's no a DI framework as mature as Spring. It fits in the Integration-Test (IT) phase very well.

NOTE: Take these things into account when you read this post: 1) it's a simplified scenario, 2) there MANY other phases besides of compilation, test, integration tests, etc. where Spring also shines, 3) Spring is much more than an DI framework.

HANDS ON LAB

So we need to replace the wiring benefits of Spring in the early testing stage (don't confuse with IT) showed on the [last post] with a worthy substitute. What we mean with worthy? Two things: 1) it solves the wiring at this stage, and 2) it's easy to use and reproduce.

Testing Phase

In the Java world many mocking frameworks exists, I performed a research on this topic and found up a mighty combination in the mocking area: PowerMock w/ Mockito.

PowerMock is the nearest similar thing to a silver bullet I have ever seen in the mocking area, you can combine it with Mockito. By the other hand Mockito is very user friendly, simple and easy to learn. I recommend both in combination for more elaborated tests, and only Mockito for simple scenarios like this.

pom.xml

      org.mockito
      mockito-all
      1.9.5
      test
    

To achieve the same goal simplifying things we need to inject the SUT's dependencies:

EditorTest.java

package demos.sf.editor.test;

import static org.junit.Assert.assertEquals;
import static org.mockito.Mockito.only;
import static org.mockito.Mockito.verify;

import org.junit.Test;
import org.junit.runner.RunWith;
import org.mockito.InjectMocks;
import org.mockito.Mock;
import org.mockito.runners.MockitoJUnitRunner;

import demos.sf.editor.SpellChecker;
import demos.sf.editor.impl.Editor;

@RunWith(MockitoJUnitRunner.class)
public class EditorTest {

 @InjectMocks
 private Editor editor;

 @Mock
 private SpellChecker spellChecker;

 @Test
 public void testPasteSuccess() {
  editor.paste("Hello everybody!");

  String expected = "Hello everybody!";
  String actual = editor.getText();
  assertEquals(expected, actual);
 }

 @Test
 public void testAddParagraphSpellIsChecked() {
  editor.paste("Hello everybody!");
  verify(spellChecker, only()).check("Hello everybody!");
 }
}

Observe the introduction of:

@Mock: annotates a dependency to be mocked.

@InjectMocks: annotates the SUT to be instantiated and its @Mocks dependencies to get injected.

@RunWith(MockitoJUnitRunner.class): uses the Mockito runner for tests.

 

Therefore, there's no need for DI framework (neither application context configuration file) in test phase at least for scenarios like this. Of course the injection and mocking capacities of Mockito are limited compared to Spring.

IT Phase

In the IT phase you replace mock objects with real ones, followed by a verification of the real objects' combination performing as you expect. Let's show one of such possible object graph where I use an English implementation for spell checking, dictionary and tokenizer:

By replacing every mock for a real object at time, you gradually introduce a full IT. You can achieve a complete mock-replacement using a top-down approach, there may be other ways of course but I took this one:

  (1) test Editor using SpellChecker's mock

  (2) replace SpellChecker's mock for EnglishSpellChecker and use Dictionary and Tokenizer mocks, then test Editor

  (3) replace Dictionary's mock for OxfordDictionary, then test Editor

  (4) replace Tokenizer's mock for EnglishTokenizer, then test Editor

Two new interfaces appear in the demos.sf.editor package, they represent the abstractions mentioned above:

public interface Dictionary {
 String lookup(String word);
}

public interface Tokenizer {
 Collection tokenize(String text);
}

Three new classes appear in the demos.sf.editor.english package, they represent and English wiring combination:

@Named
public class EnglishSpellChecker implements SpellChecker {

 @Resource
 private Tokenizer tokenizer;

 @Resource
 private Dictionary dictionary;

 public String check(String text) {
        ...
 }
}

@Named
public class EnglishTokenizer implements Tokenizer {

 public Collection tokenize(String text) {
  // TODO Auto-generated method stub
  return null;
 }
}

@Named
public class OxfordDictionary implements Dictionary {

 public String lookup(String word) {
  // TODO Auto-generated method stub
  return null;
 }
}

The beauty behind is that out Editor remains the same:

@Named
public class Editor {

 @Getter
 protected String text;

 @Resource
 private SpellChecker spellChecker;

 public void paste(String cut) {
  if (text == null) {
   text = "";
  }
  text += cut;
  spellChecker.check(cut);
 }
}

And there is a change for introducing and test for integration by only changing the IT application context, observe the component-scan with demos.sf.editor.english package which instantiates the desired combination of beans:

it-applicationContext.xml

There are additional aliasing entries due to the fact that the injection is performed on-resource-basis (by bean name), remember @Resource and @Named annotations:

@Named
public class Editor {

 // the following injects a bean named spellChecker
 @Resource
 private SpellChecker spellChecker;
 ...
}

// the following produces a bean named englishSpellChecker
@Named
public class EnglishSpellChecker implements SpellChecker {

 // the following injects a bean named tokenizer
 @Resource
 private Tokenizer tokenizer;

 // the following injects a bean named dictionary
 @Resource
 private Dictionary dictionary;
 ...
}

// the following produces a bean named englishTokenizer
@Named
public class EnglishTokenizer implements Tokenizer {
 ...
}

// the following produces a bean named oxfordDictionary
@Named
public class OxfordDictionary implements Dictionary {
 ...
}

You may want to inject based on type instead of name, it's easy to achieve by replacing the @Resource annotations with @Inject. Lets resume again the meaning of these annotations:

     * @Resource: 

• Matching: First matches by name, then matches by type and filters by qualifiers if not name-based match gets found.
• Applicable to: Fields, Classes and Methods.
• Package: javax.annotation  P { margin-bottom: 0.08in; }A:link { }

• @Inject: 
  • Matching:  First matches by type, filters by qualifiers, then matches be name.
  • Applicable to: Methods, Constructors and Fields.
  • Package: javax.inject

• @Named: 
  • Description:  Its a qualifier that denotes a named bean. Annotated classes are considered for auto-detection on classpath scanning. It's similar to Spring's @Component
  • Applicable to: Classes and Fields.
  • Package: javax.inject
    

 

How to change it to Spanish?

1- First implement the three corresponding classes in demos.sf.editor.spanish package, they will represent the desired wiring combination:

@Named
public class SpanishSpellChecker implements SpellChecker {

 @Resource
 private Tokenizer tokenizer;

 @Resource
 private Dictionary dictionary;

 public String check(String text) {
        ...
 }
}

@Named
public class SpanishTokenizer implements Tokenizer {

 public Collection tokenize(String text) {
  // TODO Auto-generated method stub
  return null;
 }
}

@Named
public class RAEDictionary implements Dictionary {

 public String lookup(String word) {
  // TODO Auto-generated method stub
  return null;
 }
}

Create a new IT application context, or change the existing one with a component-scan with demos.sf.editor.spanish package which instantiates the desired combination of Spanish beans :)

A final step (really the first one)

Create the integration test in the package demos.sf.editor.it, it's very similar to the unit test case in the [previous post], the main differences are the use of it-applicationContext.xml as context configuration, and the IT suffix on the test case name:

@RunWith(SpringJUnit4ClassRunner.class)
@ContextConfiguration(locations = { "/it-applicationContext.xml" })
@TestExecutionListeners({ DependencyInjectionTestExecutionListener.class,
  DirtiesContextTestExecutionListener.class })
@DirtiesContext(classMode = ClassMode.AFTER_EACH_TEST_METHOD)
public class EditorIT {

 @Resource
 private Editor editor;

 @Test
 public void testPasteWorks() {
  editor.paste("Hello everybody!");

  String expected = "Hello everybody!";
  String actual = editor.getText();
  assertEquals(expected, actual);
 }

}

To honor the IT test case only in the integration test phase the maven-failsafe-plugin may be used, hence the mvn test execution only runs the traditional JUnit test cases and ignores the IT test case classes suffixed with IT:
pom.xml

        maven-failsafe-plugin
        2.13
        
          
            failsafe-integration-tests
            integration-test
            
              integration-test
              verify
            
          
        
      

To run IT test then execute:

$ mvn verify

The source code can be found here [https://bitbucket.org/eduardo_lago_aguilar/spring-demos] under the directory: spring-hello-word-cdi-mocks.

REFERENCES

CDI / JSR-299:

• http://jcp.org/en/jsr/detail?id=299
• http://docs.jboss.org/cdi/spec/1.0/html/

Creating a REST web service using webtoolkit (aka witty)

I previously introduced witty as a powerful web framework for C++ developers, mostly those one that don't want to re-invent the wheel and become productive.  Witty uses a resource-based approach to register services that listen on specific paths (resource paths), these resources are registered on the witty server first and the server is started later.

So the steps of an standalone witty w/ custom resource server:

1. create a server
2. register a custom resource on a desired path
3. start the server, hence its resources
4. wait for server shutdown signal
5. stop the server

Here is the snippet of the main.cpp,  ignore all log-related plumbing:

#include <Wt/WServer>
#include <iostream>
#include "MyResource.h"
#include "log.h"

using namespace std;
using namespace Wt;

int main(int argc, char **argv) {
 WLogger logger;
 configLogger(logger);
 try {
  WServer server(argv[0], "");
  try {
   server.setServerConfiguration(argc, argv);
   MyResource dr;
   server.addResource(&dr, "/resource");
   info(logger, "Starting resource server.");
   if (server.start()) {
    WServer::waitForShutdown();
    server.stop();
   } else {
    fatal(logger, "Fatal error starting resource server.");
    return 1;
   }
   return 0;
  } catch (std::exception& e) {
   fatal(logger, "Fatal error starting resource server.", e.what());
   return 1;
  }
 } catch (WServer::Exception& e) {
  fatal(logger, "Fatal error creating WServer.", e.what());
  return 1;
 } catch (exception& e) {
  fatal(logger, "Fatal error occurred.", e.what());
  return 1;
 }
}

You may notice the creation and registration of MyResource, this is place where you provide the service implementation, specifically on the handleRequest() method, take a look at MyResource.h:

#ifndef MYRESOURCE_H_
#define MYRESOURCE_H_

#include <Wt/WResource>

using namespace Wt;
using namespace Wt::Http;

class MyResource: public WResource {
 public:
  MyResource();
  virtual ~MyResource();
 protected:
  virtual void handleRequest(const Request &request, Response &response);
};

#endif /* MYPRESOURCE_H_ */

The implemantation is very trivial, it's just an echo of the request content also including some extra data like content type, content length and request method: GET, POST, etc. Enjoy MyResource.cpp:

#include <iostream>
#include <Wt/Http/Response>
#include "MyResource.h"

using namespace std;
using namespace Wt::Http;

MyResource::MyResource() {
}

MyResource::~MyResource() {
}

void MyResource::handleRequest(const Request& request, Response& response) {
 string method = request.method();
 string contentType = request.contentType();
 int contentLength = request.contentLength();
 char* buffer = new char[contentLength + 1];
 request.in().read(buffer, contentLength);
 buffer[contentLength] = 0;
 response.setMimeType("application/xml");
 ostream& out = response.out();
 out << "<?xml version='1.0' encoding='utf-8' ?>" << endl;
 out << "<reply>" << endl;
 out << "<method>" << method << "</method>" << endl; 
 out << "<contentType>" << contentType << "</contentType>" << endl;
 out << "<contentLenght>" << contentLength << "</contentLenght>" << endl;
 out << "<body>" << buffer << "</body>" << endl;
 out << "</reply>";
 delete[] buffer;
}

The reamining stuff is just plumbing as I said before, here you got log.h:

#ifndef LOG_H_
#define LOG_H_

#include <Wt/WLogger>

using namespace std;
using namespace Wt;

void info(WLogger& logger, const string& message);
void fatal(WLogger& logger, const string& message, const char* what);
void fatal(WLogger& logger, const string& message);
void configLogger(WLogger& logger);

#endif /* LOG_H_ */

And log.cpp:

#include "log.h"
#include <iostream>

void info(WLogger& logger, const string& message) {
 WLogEntry entry = logger.entry("info");
 entry << WLogger::timestamp << WLogger::sep << WLogger::sep << '[' << "info"
   << ']' << WLogger::sep << message;
}
void fatal(WLogger& logger, const string& message, const char* what) {
 WLogEntry entry = logger.entry("fatal");
 entry << WLogger::timestamp << WLogger::sep << WLogger::sep << '['
  << "fatal" << ']' << WLogger::sep << message << what;
}

void fatal(WLogger& logger, const string& message) {
 fatal(logger, message, "");
}

void configLogger(WLogger& logger) {
 logger.addField("datetime", false);
 logger.addField("type", false);
 logger.addField("message", true);
 logger.setFile("/var/log/resource.log");
}

Now compile an run the server, here in this case using standalone server mode instead of fast cgi mode, but also works w/ fast cgi variant:

$ g++ log.cpp MyResource.cpp main.cpp -lwthttp -oresource

Run the server:

$ ./resource --http-address 0.0.0.0 --http-port 80 --docroot=.
INFO: Opened log file (/var/log/resource.log).
[2013-May-07 19:52:11.046985] 9658 - [info] "config: reading Wt config file: /etc/wt/wt_config.xml (location = './resource')"
[2013-May-07 19:52:11.047757] 9658 - [info] "WServer/wthttp: initializing built-in wthttpd"
[2013-May-07 19:52:11.048027] 9658 - [info] "wthttp: started server: http://0.0.0.0:80"

Finally test the service using a simple call:

$ curl -X POST -H "Content-Type: application/xml" -d"<payload>PAYLOAD GOES HERE!</payload>" http://localhost/resource

<?xml version='1.0' encoding='utf-8' ?>
<reply>
<method>POST</method>
<contentType>application/xml</contentType>
<contentLenght>37</contentLenght>
<body><payload>PAYLOAD GOES HERE!</payload></body>
</reply>

Very simple as you can see!

ImageMagick convert to raw RGB format in C++

This is a code snippet to convert from any supported ImageMagick format to raw RGB format, it's just an example on how to use the conversion features of ImageMagick libraries. The example was tested on both: Windows and Linux.

#include <Magick++.h>
#include <iostream>
#include <boost/filesystem.hpp>

using namespace std;
using namespace Magick;
using namespace boost::filesystem;

int main(int argc, char **argv) {
 Image img;
 string src_path;
 string tgt_path;
 Blob blob;
 FILE* tgt_file;

 InitializeMagick(*argv);

 // read image
 src_path = argv[1];
 if(exists(src_path)) {
  img.read(src_path);
  cout << "Detected format: " << img.format() << endl;

  // set raw RGBS output format & convert it into a Blob
  img.magick("RGB");
  img.write(&blob);

  // dump blob to disk
  tgt_path = src_path + ".RGB.raw";
  tgt_file = fopen(tgt_path.c_str(), "wb");
  fwrite(blob.data(), blob.length(), 1, tgt_file);
  fclose(tgt_file);
  cout << "Converted to raw RGB: " << tgt_path << endl;
  exit(0);
 } else {
  cout << "Could not load image, file not found " << src_path << endl;
  exit(1);
 }
}

Recording dialog-based interactive user input using expect on Linux

This is a how to successfully combine the shell script dialog tool with expect in order to record all user interactions and repeat them later. You can apply expect to record almost anything on a terminal, I applied it on a very complex dialog-based installation wizard with 100% success. There's nothing special here, nor a magic spell. The solution is pretty straightforward: start recording, launch dialogs + interact, stop record and finally replay it on a hit.

Show the example dialogs below:

#!/bin/bash
#
# dlgs.sh : all dialogs in one script, 1- a radio list, 2- a Yes/No dialog, 3-An input box


  
radiolist_result=$(dialog --stdout --clear --backtitle "Combining dialog w/ expect" \
   --radiolist "Select distro:" 10 40 3 1 "CentOS" off 2 "Ubuntu" on 3 "Debian" off)

dialog --stdout --clear --title "Simple question" --backtitle "Combining dialog w/ expect" \
  --yesno "Are you having fun?" 6 25 && yesno_result=Yes || yesno_result=No

inputbox_result=$(dialog --stdout --clear  --backtitle "Combining dialog w/ expect" \
  --inputbox "Enter your name:" 8 40)

I use CentOS 6 for this demo, but it seems to work on other distros:

$ yum -y install expect

Assign execution permissions to dlgs.sh:

$ chmod +x dlgs.sh

Start recording using autoexpect:

$ autoexpect
autoexpect started, file is script.exp

Call dlgs.sh and interact:

$ ./dlgs.sh
...
3
No
Eduardo Lago Aguilar

By pressing Ctrl+D end the session watch. Then verify the recording by calling the generated script:

$ ./script.exp
...
3
No
Eduardo Lago Aguilar

Very simple isn't? See more.

Becoming unaware of the Dependency Injection Framework. Encouraging standarization

Intro

The previous post showed up how to combine Spring Framework, Mockito and Lombok in TDD fashion. But it's far from be the simplest and portable solution if such thing even exists. Let's try to make the whole example more unaware of the DI Framework, in this case Spring, but still use it for testing purposes.

Java 6 brought us several improvements, CDI among them. This refcardz is self-explanatory. IMO the main advantage of CDI is standardization, but CDI is pure specification, not an implementation, and its RI is Weld.

Results that Spring honors CDI since version 3.0, at least in Java Injection Standard (JSR-330). There is a bunch of posts and articles talking about the Spring support for CDI specifications, and which one is better to use. But this is not the goal of this post.

Goals

• Become unaware of the DI Framework at design time
• Support Spring at test time
• Support Spring at runtime
• Support others DI Framework at runtime

How to

• By replacing @Autowired and using injection by name with @Resource
• By naming beans using @Name and scanning packages for candidates instead of explicitly create the bean in application context configuration file.

Let's do it

As I said before, it's just an example. Real world scenarios are more complex, and you will end for sure combining many APIs from different sources.

1. Use the example from the previous post.
2. Include a couple of dependencies in your pom.xml.

  javax.annotation
  jsr250-api
  1.0


  javax.inject
  javax.inject
  1

JSR-250 provide us various standard annotations: @Resource, @PostConstruct, @PreDestroy , etc; these annotations are honored by Spring. In the other hand  with have JSR-333 (javax.inject) that give us @Inject, @Named, @Provider, @Qualifier, @Scope and @Singleton.
Lets explain some of these annotations:

• @Resource: 
• Matching: First matches by name, then matches by type and filters by qualifiers if not name-based match gets found.
• Applicable to: Fields, Classes and Methods.
• Package: javax.annotation
• @PostConstruct: 
• Runs: After the bean construction and wiring
• Applicable to: Methods.
• Package: javax.annotation
• @PreDestroy: 
• Runs: Before the release of the bean
• Applicable to: Methods.
• Package: javax.annotation
• @Inject: 
• Matching:  First matches by type, filters by qualifiers, then matches be name.
• Applicable to: Methods, Constructors and Fields.
• Package: javax.inject
• @Named: 
• Description:  Its a qualifier that denotes a named bean. Annotated classes are considered for auto-detection on classpath scanning. It's similar to Spring's @Component
• Applicable to: Classes and Fields.
• Package: javax.inject
3. Spring dependencies are now on test scope, since they are only required on test time. But in a real world scenario you will need a DI Framework, so don't forget to include Spring or Weld dependencies for production stage.

  ...
 
  org.springframework
  spring-test
  ${spring.version}
  test
 
 
  org.springframework
  spring-beans
  ${spring.version}
  test
 
 
  org.springframework
  spring-context
  ${spring.version}
  test
 
  ...

4. The test is an hybrid between Spring and more standard stuff, that is:

package demos.sf.editor.test;

import ... ;

@RunWith(SpringJUnit4ClassRunner.class)
@ContextConfiguration(locations = { "/test-applicationContext.xml" })
@TestExecutionListeners({ DependencyInjectionTestExecutionListener.class,
  DirtiesContextTestExecutionListener.class })
@DirtiesContext(classMode = ClassMode.AFTER_EACH_TEST_METHOD)
public class EditorTest {

 @Resource
 private Editor editor;

 @Resource
 private SpellChecker spellChecker;

 @Test
 public void testPasteSuccess() {
  editor.paste("Hello everybody!");

  String expected = "Hello everybody!";
  String actual = editor.getText();
  assertEquals(expected, actual);
 }

 @Test
 public void testAddParagraphSpellIsChecked() {
  editor.paste("Hello everybody!");
  verify(spellChecker, only()).check("Hello everybody!");
 }
}

5. The SUT (Editor) is now DI Framework unaware:

package demos.sf.editor.impl;

import ...;

@Named
public class Editor {

 @Getter
 protected String text;

 @Resource
 private SpellChecker spellChecker;

 public void paste(String cut) {
  if (text == null) {
   text = "";
  }
  text += cut;
  spellChecker.check(cut);
 }
}

6. The application context configuration file for testing stage file is now simplified. Only package scanning is needed to create the SUT (Editor):
NOTE:Observe that some minor package relocation were made to organize tests, interfaces and classes.

The final snapshot can be found here under the spring-hello-word-cdi directory. Enjoy it!