Introduction

This document describes the technical aspects of the testing integrated into Dspace. In it we describe the tools used as well as how to use them and the solutions applied to some issues found during development. It's intended to serve as a reference for the community so more test cases can be created.

Note the document is a work in progress and will change as the implementation evolves.

Issues Found

During implementation we found several issues, which are described in this section, along the solution implemented to work around them. 

...

The functional tests implementation is in processbeing done.

Structural Issues

During the development the following issues have been detected in the code, which make Unit Testing harder and impact the maintainability of the code:

...

• A mock of BrowseCreateDAOOracle has been done due to an incompatibility between H2 and the "upper" function call. This will affect tests related to case sensitivity in indexes.
• Many objects (like SupervisedItem) lack a proper definition of the "equals" method, which makes comparison between objects and unit testing harder
• Update method of many objects doesn't provide any feedback, we can only test if it raises an exception or not, but we can't be 100% sure if it worked 
• Many objects have methods to change the policies related to the object or children objects (like Item), it would be good to have some methods to retrieve these policies also in the same object (code coherence)
• There are some inconsistencies in the calls to certain methods. For example getName returns an empty String in a Collection where the name is not set, but a null in an Item without name
• DCDate: the tests raise many errors. I can't be sure if it's due to misunderstanding of the purpose of the methods or due to faulty implementation (probably the previous). In some cases extra encapsulation of the internals of the class would be advisable, to hide the complexities of the Calendars (months starting by 0, etc)
• The Authorization system gets a bit confusing. We have AuthorizationManager, AuthorizationUtils, methods that raise exceptions and methods that return booleans. Given the number of checks we have to do for permissions, and that some classes call methods that require extra permissions not declared or visible at first, this makes creation of tests (and usage of the API) a bit complex. I know we can ignore all permissions via context (turning on and off authorizations) but usually we don't want that
• Community: set logo checks for authorization, but set metadata doesn't. It's in purpose?
• Collection: methods create and delete don't check for authorization
• Item: there is no authorization check for changing policies, no need to be an administrator
• ItemIterator: it uses ArrayLists in the methods instead of List
• ItemIterator: we can't verify if the Iterator has been closed
• Metadata classes: usually most classes have a static method to create a new instance of the class. Instead, for the metadata classes (Schema, Field and Value) the method create is part of the object, thus requiring you to first create an instance via new and then calling create. This should be changed to follow the convention established in other objects (or the other objects should be amended to behave like the Metadata classes)
• Site: this class extends DSpaceObject. With it being a Singleton, it creates potential problems, for example when we use DSpaceObject methods to store details in the object. It's this Site: this class extends DSpaceObject. With it being a Singleton, it creates potential problems, for example when we use DSpaceObject methods to store details in the object. It's this relation necessary?

Fixes done:

- Bitstream: added an "isDeleted" method to verify if a bitstream has been deleted
- Bundle: added methods to check the policies of bundle and its bitstreams
- Collection: just a comment: delete requires authorization to remove the template Item and write, not to remove. Is that correct?
- Community: when a community is created with a parent, it's added as a child community immediately
- DCLanguage: added checks for null name in languages
- FormatIdentifier: fixed the check for filename == null in guessFormat 
- SiteTest: the test is in the abstract DSpaceObjectTest so I've make it inherit AbstractUnitTest.  I see the class has almost no usage so we could remove the inheritance from DSpaceObject, but I'm not sure if to do this. It's something that we should ask the developers?
- Several equals and hashCode methods added for other issues in tests
Pending of a review by a DSpace developer:
- DCDate: Here many tests fail because I'm not sure of the purpose of the class. I would expect it to hide the implementation of Calendar (with all those things like months starting by 0 and other odd stuff) so it's easier to use, but it seems that's not the case...

• Bitstream: added an "isDeleted" method to verify if a bitstream has been deleted
• Bundle: added methods to check the policies of bundle and its bitstreams
• Collection: just a comment: delete requires authorization to remove the template Item and write, not to remove. Is that correct?
• Community: when a community is created with a parent, it's added as a child community immediately
• DCLanguage: added checks for null name in languages
• FormatIdentifier: fixed the check for filename == null in guessFormat 
• SiteTest: the test is in the abstract DSpaceObjectTest so I've make it inherit AbstractUnitTest.  I see the class has almost no usage so we could remove the inheritance from DSpaceObject, but I'm not sure if to do this. It's something that we should ask the developers?
• Several equals and hashCode methods added for other issues in tests

Proposals:

To solve the previous issues, some proposals are done:

• Database dependency causes too many issues, making unit testing much harder and increasing the complexity of the code. Refactoring to a database-neutral system should be a priority
• A release could be done (1.8?) centered on cleaning code, improving stability and coherency and refactoring unit tests, as well as replacing the database system. No new functionalities. This would make future work much easier.

...

Dependencies

There is a set of tools used by all the tests. These tools will be described in this section.

...

JMockit 0.998 has been added to the project to provide a mocking framework to the tests.

ContiPerf

ContiPerf is a lightweight testing utility that enables the user to easily leverage JUnit 4 test cases as performance tests e.g. for continuous performance testing.

The project makes use of ContiPerf 1.06.

Unit Tests Implementation

H2

H2 is an in-memory database that has been used 

The project makes use of H2 version 1.2.137

Unit Tests Implementation

These are tests which These are tests which test just how one object works. Typically test each method on an object for expected output in several situations. They are executed exclusively at the API level.

...

To achieve this we mock DatabaseManager and we replace the connector to point to our in-memory database. In this class we also initialise the replica with the proper data.

Structure

There is a base class called "AbstractUnitTest". This class contains a series of mocks and references which are necessary to run the tests in DSpace, like mocks of the DatabaseManager object. All Unit Tests should inherit this class, located under the package "org.dspace" in the test folder of DSpace-test.

About the implementation, several objects only offer a hidden constructor and a factory method to create an instance of the object. This means we effectively have to create them using available factory methods. Other specifics have been commented above, like:

• Usage of a temporal file system
• Usage of an in-memory database (h2)
• Mocking the DatabaseManager class

Integration Tests

These tests work at the API level and test the interaction of components within the system. Some examples, placing an item into a collection, or creating a new metadata schema and adding some fields. Primarily these tests operate at the API level ignoring the interface components above it.

The main difference between these and the unit tests is in the test implemented, not in the infrastructure required, as these tests will use several classes at once to emulate a user action.

The integration tests also make use of ContiPerf to evaluate the performance of the system. We believe it doesn't make sense to add this layer to the unit tests, as they are tested in isolation and we care about performance not on individual calls but on certain tasks that can only be emulated by integration testing.

Structure

Integration tests use the same structure as Unit tests. A class has been created, called AbstractIntegrationTest, that inherits from AbstractUnitTest. This provides the integration tests with the same temporal file system and in-memory database as the unit tests. The class AbstractIntegrationTest is created just in case we may need some extra scaffolding for these tests. All integration tests should inherit from it to both distinguish themselves from unit tests and in case we require specific changes for them.

Classes that contain teh code for Integration Tests are named <class>IntegrationTest.java.

Events Concurrency Issues

We have detected an issue with the integration tests, related to the Context class. In this class, the List of events was implemented as an ArrayList<Event>. The issue here is that ArrayList is not a safe class for concurrency. Although this would not be a problem while running the application in a JEE container, as there will be a unique thread per request (at least in normal conditions), we can't be sure of the kind of calls users may do to the API while extending DSpace.

To avoid the issue we have to wrap the List into a synchronized stated via Collections.synchronizedList . This, along a synchronized block, will ensure the code behaves as expected.

We have detected the following classes affected by this behavior:

• BasicDispatcher.java

In fact any class that calls Context.getEvents() may be affected by this. A comment has been added in the javadoc of this class (alongside a TODO tag) to warn about the issue.

Context Concurrency Issues

There is another related issue in the Context class. Context establishes locks in the tables when doing some modifications, locks that are not lifted until the context is committed or completed. The consequence is that some methods can't be run in parallel or some executions will fail due to table locks. This can be solved, in some cases, by running context.commit() after a method that modifies the database, but this doesn't work in all cases. For example, in the CommunityCollection Integration Test, the creation of a community can mean the modification of 2 rows (parent and new community). This causes this kind of locks, but as it occurs during the execution of the method create() it can't be solved by context.commit().

Static Analysis

Due to comments in the GSoC meetings, some static analysis tools have been added to the project. The tools are just a complement, a platform like Sonar should be used as it integrates better with the structure of DSpace and we could have the reports linked to Jira.

We have added the following reports:

• FindBugs : static code bug analyser
• PMD and CPD: static analyser and "copy-and-paste" detector
• TagList: finds comments with a certain annotation (like XXX or TODO)
• Testability Explorer: detects issues in classes that difficult the creation of unit tests

They can be generated by launching:

mvn site

from the main folder. Be aware this will take a long time!

Functional Tests

...

Due to the Dspace Maven structure discussed in previous sections, all the tests belonging to any module (dspace-api, dspace-xmlui-api, etc) must be stored in the module dspace-test. This module enables us to apply common configuration, required by all tests, in a single area thus avoiding duplication of code. Related to this point is the requirement for Dspace to run using a database and a certain file system structure. We have created a base class that initializes this structure via a in-memory database (using H2) and a temporary copy of the required file system.

The described base class is called "AbstractUnitTest". This class contains a series of mocks and references which are necessary to run the tests in DSpace, like mocks of the DatabaseManager object. All Unit Tests should inherit this class, located under the package "org.dspace" in the test folder of dspace-test. There is an exception with classes that originally inherit DSpaceObject, its tests should inherit AbstractDSpaceObjectTest class.

Several mocks are used in the tests. The more relevant ones are:

• MockDatabaseManager: a mock of the database manager that launches H2 instead of PostgreSQL/Oracle and creates the basic structure of tables for DSpace in memory
• MockBrowseCreateDAOOracle: due to the strong link between DSpace and the databases, there are some classes that have specific implementations if we are using Oracle or PostgreSQL, like this one. In this case we've had to create a mock class that overrides the functionality of MockBrowseCreateDAOOracle so we are able to run the Browse related tests.

You may need to create new mocks to be able to test certain areas of code. Creation of the Mock goes beyond the scope of this document, but you can see the mentioned classes as an example. BAsically it consists on adding annotations to a copy of the existing class to indicate a method is a mock of the original implementation and modifying the code as required for our tests.

Limitations

The solution to copy the file system is not a very elegant one, so we appreciate any insight that can help us to replicate the required files appropriately.

The fact that we load the tests configuration from a dspace-test.cfg file means we are only testing the classes against a specific set of configurations. We probably would like to have tests that runs with multiple settings for the specific piece of code we are testing. This will require some extra classes to modify the configuration system and the way this is accessed by DSpace.

How to build new tests

To build a new Unit Test, create the corresponding class in the project dspace-test, under the test folder, in the package where the original class belongs. Tests for all the projects (dspace-api, dspace-jspui-api, etc) are stored in this project, to avoid duplication of code. Name the class following the format <OriginalClass>Test.java.

There are some common imports and structure, you can use the following code as a template:

//Add DSpace licensing here at the top!
package org.dspace.content;

import java.sql.SQLException;
import org.dspace.core.Context;
import org.junit.*;
import static org.junit.Assert.* ;
import static org.hamcrest.CoreMatchers.*;
import mockit.*;
import org.apache.log4j.Logger;
import org.dspace.core.Constants;

/**
 * Unit Tests for class <OriginalClass>Test
 * @author you name
 */
public class <OriginalClass>Test extends AbstractUnitTest
{

    /** log4j category */
    private static final Logger log = Logger.getLogger(<OriginalClass>Test.class);

    /**
     * <OriginalClass> instance for the tests
     */
    private <OriginalClass> c;

    /**
     * This method will be run before every test as per @Before. It will
     * initialize resources required for the tests.
     *
     * Other methods can be annotated with @Before here or in subclasses
     * but no execution order is guaranteed
     */
    @Before
    @Override
    public void init()
    {
        super.init();
        try
        {
            //we have to create a new community in the database
            context.turnOffAuthorisationSystem();
            this.c = <OriginalClass>.create(null, context);


            //we need to commit the changes so we don't block the table for testing
            context.restoreAuthSystemState();
            context.commit();
        }
        catch (AuthorizeException ex)
        {
            log.error("Authorization Error in init", ex);
            fail("Authorization Error in init");
        }
        catch (SQLException ex)
        {
            log.error("SQL Error in init", ex);
            fail("SQL Error in init");
        }
    }

    /**
     * This method will be run after every test as per @After. It will
     * clean resources initialized by the @Before methods.
     *
     * Other methods can be annotated with @After here or in subclasses
     * but no execution order is guaranteed
     */
    @After
    @Override
    public void destroy()
    {
        c = null;
        super.destroy();
    }

    /**
     * Test of XXXX method, of class <OriginalClass>
     */
    @Test
    public void testXXXX() throws Exception
    {
        int id = c.getID();
        <OriginalClass> found =  <OriginalClass>.find(context, id);
        assertThat("testXXXX 0", found, notNullValue());
        assertThat("testXXXX 1", found.getID(), equalTo(id));
        assertThat("testXXXX 2", found.getName(), equalTo(""));
    }

   [... more tests ...]
}

The sample code contains common imports for the tests and common structure (init and destroy methods as well as the log). You should add any initialization required for the test in the init method, and free the resources in the destroy method. 

The sample test shows the usage of the assertThat clause. This clause (more information in JUnit help) allows you to check for condition that, if not true, will cause the test to fail. We name every condition via a simple schema (method name plus an integer indicating order) as the first parameter. This allows you to identify which specific assert if failing whenever a test returns an error.

Please be aware methods init and destroy will run once per test, which means that if you create a new instance every time you run init, you may end up with several instances in the database. This can be confusing when implementing tests, specially when using methods like findAll.

If you want to add code that it's executed once per test class, edit the parent AbstractUnitTest and its methods initOnce and destroyOnce. Be aware these methods contain code used to recreate the structure needed to run DSpace tests, so be careful when adding or removing code there. Our suggestion is to add code at the end of initOnce and at the beginning of destroyOnce, to minimize the risk of interferences between components.

Be aware that tests of classes that extend DSpaceObject should extend AbstractDSpaceObjectTest instead due to some extra methods and requirements implemented in there.

How to run the tests

The tests can be activated using the commands

mvn package -Dmaven.test.skip=false  //builds DSpace and runs tests

  or
mvn test -Dmaven.test.skip=false     //just runs the tests

or by changing the property "activeByDefault" at the profile (skiptests) in the main pom.xml file, at the root of the project and then running

mvn package  //builds DSpace and runs tests
  or
mvn test     //just runs the tests

Be aware that this command will launch both unit and integration tests.

Integration Tests

These tests work at the API level and test the interaction of components within the system. Some examples are placing an item into a collection or creating a new metadata schema and adding some fields. Primarily these tests operate at the API level ignoring the interface components above it.

The main difference between these and the unit tests is in the test implemented, not in the infrastructure required, as these tests will use several classes at once to emulate a user action.

The integration tests also make use of ContiPerf to evaluate the performance of the system. We believe it doesn't make sense to add this layer to the unit tests, as they are tested in isolation and we care about performance not on individual calls but on certain tasks that can only be emulated by integration testing.

Structure

Integration tests use the same structure as Unit tests. A class has been created, called AbstractIntegrationTest, that inherits from AbstractUnitTest. This provides the integration tests with the same temporal file system and in-memory database as the unit tests. The class AbstractIntegrationTest is created just in case we may need some extra scaffolding for these tests. All integration tests should inherit from it to both distinguish themselves from unit tests and in case we require specific changes for them.

Classes that contain the code for Integration Tests are named <class>IntegrationTest.java.

The only difference right now between Unit Tests and Integration Tests is that the later include configuration settings for ContiPerf. These is a performance testing suite that allows us to reuse the same methods we use for integration testing as performance checks. Due to limitations mentioned in the following section we can't make use of all the capabilities of ContiPerf (namely, multiple threads to run the tests) but they can be still be useful.

Limitations

Tests structure

These limitations are shared with the unit tests.

The solution to copy the file system is not a very elegant one, so we appreciate any insight that can help us to replicate the required files appropriately.

The fact that we load the tests configuration from a dspace-test.cfg file means we are only testing the classes against a specific set of configurations. We probably would like to have tests that runs with multiple settings for the specific piece of code we are testing. This will require some extra classes to modify the configuration system and the way this is accessed by DSpace.

Events Concurrency Issues

We have detected an issue with the integration tests, related to the Context class. In this class, the List of events was implemented as an ArrayList<Event>. The issue here is that ArrayList is not a safe class for concurrency. Although this would not be a problem while running the application in a JEE container, as there will be a unique thread per request (at least in normal conditions), we can't be sure of the kind of calls users may do to the API while extending DSpace.

To avoid the issue we have to wrap the List into a synchronized stated via Collections.synchronizedList . This, along a synchronized block, will ensure the code behaves as expected.

We have detected the following classes affected by this behavior:

• BasicDispatcher.java

In fact any class that calls Context.getEvents() may be affected by this. A comment has been added in the javadoc of this class (alongside a TODO tag) to warn about the issue.

Context Concurrency Issues

There is another related issue in the Context class. Context establishes locks in the tables when doing some modifications, locks that are not lifted until the context is committed or completed. The consequence is that some methods can't be run in parallel or some executions will fail due to table locks. This can be solved, in some cases, by running context.commit() after a method that modifies the database, but this doesn't work in all cases. For example, in the CommunityCollection Integration Test, the creation of a community can mean the modification of 2 rows (parent and new community). This causes this kind of locks, but as it occurs during the execution of the method create() it can't be solved by context.commit().

Due to these concurrency issues, ContiPerf can only be run with one thread. This slows the process considerably, but until the concurrency issue is solved this can't be avoided.

How to build new tests

To build a new Integration Test, create the corresponding class in the project dspace-test, under the test folder, in the package where the original class belongs. Tests for all the projects (dspace-api, dspace-jspui-api, etc) are stored in this project, to avoid duplication of code. Name the class following the format <RelatedClasses>IntegrationTest.java.

There are some common imports and structure, you can use the following code as a template:

//Add DSpace licensing here at the top!
package org.dspace.content;

import java.sql.SQLException;
import org.dspace.core.Context;
import org.junit.*;
import static org.junit.Assert.* ;
import static org.hamcrest.CoreMatchers.*;
import mockit.*;
import org.apache.log4j.Logger;
import org.dspace.core.Constants;
/**
 * This is an integration test to validate the metadata classes
 * @author pvillega
 */
public class MetadataIntegrationTest  extends AbstractIntegrationTest
{
    /** log4j category */
    private static final Logger log = Logger.getLogger(MetadataIntegrationTest.class);


    /**
     * This method will be run before every test as per @Before. It will
     * initialize resources required for the tests.
     *
     * Other methods can be annotated with @Before here or in subclasses
     * but no execution order is guaranteed
     */
    @Before
    @Override
    public void init()
    {
        super.init();
    }

    /**
     * This method will be run after every test as per @After. It will
     * clean resources initialized by the @Before methods.
     *
     * Other methods can be annotated with @After here or in subclasses
     * but no execution order is guaranteed
     */
    @After
    @Override
    public void destroy()
    {
        super.destroy();
    }

    /**
     * Tests the creation of a new metadata schema with some values
     */
    @Test
    @PerfTest(invocations = 50, threads = 1)
    @Required(percentile95 = 500, average= 200)
    public void testCreateSchema() throws SQLException, AuthorizeException, NonUniqueMetadataException, IOException
    {
        String schemaName = "integration";

        //we create the structure
        context.turnOffAuthorisationSystem();
        Item it = Item.create(context);

        MetadataSchema schema = new MetadataSchema("htpp://test/schema/", schemaName);
        schema.create(context);
        [...]
        
        //commit to free locks on tables
        context.commit();

        //verify it works as expected
        assertThat("testCreateSchema 0", schema.getName(), equalTo(schemaName));
        assertThat("testCreateSchema 1", field1.getSchemaID(), equalTo(schema.getSchemaID()));
        assertThat("testCreateSchema 2", field2.getSchemaID(), equalTo(schema.getSchemaID()));              [...]
        //clean database
        value1.delete(context);
        [...]

        context.restoreAuthSystemState();
        context.commit();
    }

}

The sample code contains common imports for the tests and common structure (init and destroy methods as well as the log). You should add any initialization required for the test in the init method, and free the resources in the destroy method. 

The sample test shows the usage of the assertThat clause. This clause (more information in JUnit help) allows you to check for condition that, if not true, will cause the test to fail. We name every condition via a simple schema (method name plus an integer indicating order) as the first parameter. This allows you to identify which specific assert if failing whenever a test returns an error.

Please be aware methods init and destroy will run once per test, which means that if you create a new instance every time you run init, you may end up with several instances in the database. This can be confusing when implementing tests, specially when using methods like findAll.

If you want to add code that it's executed once per test class, edit the parent AbstractUnitTest and its methods initOnce and destroyOnce. Be aware these methods contain code used to recreate the structure needed to run DSpace tests, so be careful when adding or removing code there. Our suggestion is to add code at the end of initOnce and at the beginning of destroyOnce, to minimize the risk of interferences between components.

How to run the tests

The tests can be activated using the commands

mvn package -Dmaven.test.skip=false  //builds DSpace and runs tests
  or
mvn test -Dmaven.test.skip=false     //just runs the tests

or by changing the property "activeByDefault" at the profile (skiptests) in the main pom.xml file, at the root of the project and then running

mvn package  //builds DSpace and runs tests  
 or
mvn test     //just runs the tests

Be aware that this command will launch both unit and integration tests.

Code Analysis Tools

Due to comments in the GSoC meetings, some static analysis tools have been added to the project. The tools are just a complement, a platform like Sonar should be used as it integrates better with the structure of DSpace and we could have the reports linked to Jira.

We have added the following reports:

• FindBugs : static code bug analyser
• PMD and CPD: static analyser and "copy-and-paste" detector
• TagList: finds comments with a certain annotation (like XXX or TODO)
• Testability Explorer: detects issues in classes that difficult the creation of unit tests

These reports can't replace a Quality Management tool but can give you an idea of the status of the project and of issues to be solved.

The reports can be generated by launching:

mvn site

from the main folder. Be aware this will take a long time, probably more than 20 minutes.

Functional Tests

These are tests which come from user-based use cases. Such as a user wants to search DSpace to find material and download a pdf. Or something more complex like a user wants to submit their thesis to DSpace and follow it through the approval process. These are stories about how users would preform tasks and cover a wide array of components within Dspace.

Be aware that in this section we don't focus on testing the layout of the UI, as this has to be done manually to ensure we see exactly the same in different browsers. In this section we only consider tests that replicate a process via the UI, to ensure it is not broken due to some links missing, unexpected errors or similar issues.

Choices taken

To decide on a specific implementation of these tests some choices have been taken. They may not be the best but at this time they seemed the appropriate ones. Contributions and criticism are welcomed.

On one hand, functional tests run against a live instance of the application. This means we need a full working environment with the database and file system. It also means we are running them against the modified UI of a specific installation. As a consequence, we want the tests to be very generic or easily customizable, but we have to be aware that due to way Maven (and particularly its packaging system) works it isn't possible to run the tests as a step of the unit and integration testing described in the above sections.

On the other hand, if we focus on the tools available, the best choice we have available is Selenium, a suite of tools to automate testing of web applications. Selenium can be run in two modes: as a Firefox plug in that will allow us to record test scripts and run them later, or as a distributed system that allows us to run the tests against several browsers in different platforms.

We have to choose a way to run the tests that is easy to set up and adapt to a particular project by the developers, while limited by the options that maven and selenium provide. The decision has been to use the Selenium IDE to run the tests. This means the tests can only be run in Firefox and they have to be launched manually, but on the other hand they are easily customizable and runnable.

The Selenium RC environment was discarded due to the complexity it would add to the testing process. If you are a DSpace developer and have the resources and expertise to set it up, we clearly recommend so. You can reuse the scripts generated by the Selenium IDE and add extra tests to be run with JUnit alongside the existing unit and integration tests, which allows you to build more detailed and accurate tests. Even if we recommend it and it is a more powerful (and desirable) option, we are aware of the extra complexity it would add to just run the functional tests, as not everybody has the resources to deploy the required applications. That's why we have decided to provide just the Selenium IDE scripts, as they require much less effort to be set up and will do the job.

Structure

Selenium tests consist of two components:

• The Selenium IDE, downloadable here , which is a Firefox addon that can record our actions and save them as a test
• The tests, which are HTML files that store a list of actions to be "replayed" by the browser. If any of the actions can't be executed (due to a field or link missing or some other reason) the test will fail.
  • The recommendation is to create one test per user-action (like creating an item). Several tests can be loaded at once in the IDE and run sequentially.

To install the Selenium IDE, first install Firefox and then try to download it from Firefox itself. The browser will recognize the file as an addon and it will install it.

Limitations

The resulting tests have several limitations:

• Tests are recorded against a specific run on a particular machine. This means some steps may include specific values for some variables (id numbers, paths to files, etc) that link the test to a particular installation and state in the system. As a consequence we have to ensure we have the same state in the system every time we run the tests. That may mean run the tests in a certain order and probably starting from a wiped Dspace installation.
• For the same reason as above, some scripts may require manual changes before being able to be run (to ensure the values expected by Selenium exist). This is specially important in tests which include a reference to a file (like when we create an item) as the path to the file is hard coded in the script.
• Tests have to be launched manually and can only be run in Firefox using the Eclipse IDE. We can launch all our tests at once, as a test suite, but we have to do so manually.
• Due to they way Selenium works (checking HTML code received and acting upon it) high network latency or slowness in the server may cause a test to fail when it shouldn't. To avoid this, it is recommended to run the tests at minimum speed and (if required) to increase the time Selenium waits for an answer (it can be set in the Options panel).
• Tests are linked to a language version of the system. As Selenium might reference, in some situations, a control by its text, a test will fail if we change the language of the application. Usually this is not a big problem as the functionality of the application will be the same independently of the language the user has selected, but if we want to test the application including its I18N components, we will need to record the actions once per language enabled in the system.

How to build new tests

Building a test in Selenium IDE is easy. Open the IDE (in Firefox, Tools > Selenium IDE) and navigate to your DSpace instance in a tab of Firefox (i.e.: http://localhost:8080/jspui/). Press the record button (red circle at top-right corner) in the Selenium IDE and navigate through your application. Selenium will record every click you do and text you write in the screen. If you do a mistake, you can right-click over an entry in the list and remove it.

Actions are stored by default as a reference to the control activated. This reference is generic, meaning that Selenium might look for an anchor link (<a>) that points to a certain url (like '/jspui/handle/1234/5678'), independently of its id, name or position in the application. This means that the test will not be affected (usually) by changes in the layout or some refactoring. That said, in some specific cases you may need to edit the test cases to change some values.

Once you are finished, press again the record button. Then, in the Selenium IDE, go to File > Save Test Case and save your test case.

In Selenium the tests cases are a HTML file that stores data in a table. The table contains a row per action, and each row has 3 columns :

1. An action to be run (mandatory). This can be an action like open, click, etc.
2. A path or control id against which the action is executed (mandatory). This can point to an URL, a control (input, anchor, etc) or similar.
3. A text to be added or selected in an input control (optional).

A sample of a Selenium test is:

open          /xmlui 
clickAndWait  link=Subjects 
clickAndWait  //div[@id='aspect_artifactbrowser_Navigation_list_account']/ul/li[1]/a 
type          aspect_eperson_PasswordLogin_field_login_email                            admin@dspace.org 
type          aspect_eperson_PasswordLogin_field_login_password                         test 
clickAndWait  aspect_eperson_PasswordLogin_field_submit

The code generated by the Selenium IDE for this would be like:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
<head profile="http://selenium-ide.openqa.org/profiles/test-case">
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<link rel="selenium.base" href="" />
<title>Create_Comm_Coll_Item_XMLUI</title>
</head>
<body>
<table cellpadding="1" cellspacing="1" border="1">
<thead>
<tr><td rowspan="1" colspan="3">Create_Comm_Coll_Item_XMLUI</td></tr>
</thead><tbody>
<tr>
    <td>open</td>
    <td>/xmlui</td>
    <td></td>
</tr>
<tr>
    <td>clickAndWait</td>
    <td>link=Subjects</td>
    <td></td>
</tr>

[... more actions ...]

</tbody></table>
</body>
</html>Create_Comm_Coll_Item_XMLUI

You can use the Selenium IDE to generate the tests or write them manually by using the Selenese commands.

How to run the tests

To run the tests simply open the Selenium IDE (in Firefox, Tools > Selenium IDE) and navigate to your DSpace instance in a tab of Firefox (i.e.: http://localhost:8080/jspui/). Then, in the Selenium IDE, click on File > Open and select a test case. You can open as many files as you want, they will be run in the order you opened them.

Once you have selected the test cases to run, ensure the speed of Selenium is set to slow (use the slider) and press either the "Play entire test suite" or "Play current test case" button (the ones with a green arrow), according to your intentions. Selenium will run the actions one by one in the order recorded. If at some point it can't run an action, it will display an error and fail the test. You can see the reason of the error in log at the bottom of the Selenium IDE window.

A very common reason why a test fails is because the server returned the HTML slowly and Selenium was trying to locate and HTML element before having all the HTML. To avoid this make sure that Selenium speed is set to slow and increase the default timeout value in Options.

Provided tests

We have included some sample Selenium tests in Dspace so developers can experiment with them. The tests are located under "<dspace_root>/dspace-test/src/test/resources/Selenium scripts". They are HTML files following the format described above. They are doing some assumptions:

• Tests assume that you are running them against a vanilla environment with no previous data. They may work in an environment with data, but it's not assured
• Tests assume you are running the English UI, other languages may break some tests.
• Tests assume a user with user name admin@dspace.org and password test exists in the system and has administrator privileges
• Tests assume a file exists at /home/pvillega/Desktop/test.txt

You can edit the tests (see the format above) and change the required values (like user and path to a file) to values which are valid in your system.

Advanced Usage

If you set up Selenium RC, you can reuse the test scripts to be run as JUnit tests. Selenium can export them automatically to Java classes using JUnit. For this open the Selenium IDE (in Firefox, Tools > Selenium IDE), click on File > Open and select a test case. Once the test case is loaded, click on File > Export Test Case As > Java (JUnit) - Selenium RC. This will create a Java class that reproduces the test case, as the following:

package com.example.tests;

import com.thoughtworks.selenium.*;
import java.util.regex.Pattern;

public class CreateCommunity extends SeleneseTestCase {
    public void setUp() throws Exception {
        setUp("http://localhost:8080/", "*chrome");
    }
    public void testJunit() throws Exception {
        selenium.open("/jspui/");
        selenium.click("xpath=//a[contains(@href, '/jspui/community-list')]");
        selenium.waitForPageToLoad("30000");
        selenium.click("link=Issue Date");
        selenium.waitForPageToLoad("30000");
        selenium.click("link=Author");
        selenium.waitForPageToLoad("30000");
        
        [... ]
    }
}

As you can see in the code, this class suffers from the same problems as the Selenium IDE scripts (hardcoded values, etc) but can be run using Selenium RC in a distributed environment, alongside your JUnit tests.

Future Work

This project creates a structure for testing that can expanded. Future tasks would include:

• Integrating with a Quality Management tool like Sonar
• Integrating with a Continuous Integration tools 
• Adding Unit and Integration tests for the remaining classes
• Extending Functional tests
• Creating a "Code quality" release, where priority is not new functionalities but stability and quality of code

Thanks

This page has been created with help from Stuart Lewis, Scott Phillips and Gareth Waller. I want to thank them all for their comments. Some information has been taken from Wikipedia to make the text more complete. I'm to blame for errors in the text.

...