View source for SMILA/Documentation/HowTo/How to integrate the HelloWorld webservice as a Pipelet

This page illustrates all steps that need to be performed in order to integrate the HelloWorld web service as a [[SMILA/Glossary#P|pipelet]] in SMILA. For general information on how to integrate components and add functionality to SMILA refer to [[SMILA/Howto_integrate_a_component_in_SMILA| How to integrate a component in SMILA]].

== Preparations ==

It may be helpful to first take a look at the SMILA [[SMILA/Development_Guidelines|Development guidelines]] as many topics that are beyond the scope of this tutorial are illustrated there.

=== Create new bundle ===

* Create a new bundle that should contain your pipelet. Follow the instructions on [[SMILA/Development_Guidelines/Create_a_bundle_(plug-in)|How to create a bundle]] and use the following settings:
<pre>
Project name: org.eclipse.smila.sample.pipelet
Plug-in ID: org.eclipse.smila.sample.pipelet
Plug-in Version: 1.0.0
Plug-in Name: Sample Pipelet Bundle
Plug-in Provider: your name or company
</pre>
* Edit the file <tt>META-INF/MANIFEST.MF</tt> and add the following import-package dependencies as those are required to implement the basic functionalities of your pipelet:
<pre>
Import-Package: org.apache.commons.logging;version="1.1.1",
 org.eclipse.smila.blackboard;version="0.8.0",
 org.eclipse.smila.datamodel;version="0.8.0",
 org.eclipse.smila.processing;version="0.8.0"
</pre>
* To make sure that the <tt>PipeletTrackerService</tt> detects your new pipelet, create a folder <tt>SMILA-INF</tt> in the bundle and add a file <tt>HelloWorldPipelet.json</tt> to this folder:
<pre>
{
  "class": "org.eclipse.smila.sample.pipelet.HelloWorldPipelet",
  "parameters": [        
    {
      "name": "IN_ATT_NAME",
      "type": "string"            
    },
    {
      "name": "OUT_ATT_NAME",
      "type": "string"            
    }
  ],
  "description": "Hello World pipelet. Modifies the content of the attribut denoted by the parameter IN_ATT_NAME to the attribute denoted by the parameter OUT_ATT_NAME."
}
</pre>
*Now add the folder <tt>SMILA-INF</tt> to the build.properties (or just check it in the <tt>Build</tt> view of the <tt>MANIFEST.MF</tt> file in your IDE.

=== Create Java classes from WSDL using Axis2 ===

* Install Axis2 1.4.1: Download from http://ws.apache.org/axis2/download/1_4_1/download.cgi and unpack into any directory.
* Open a shell in the Axis2 directory and execute <tt>wsdl2java</tt> similar to this example - replace the WSDL-URL with that of the Webservice you want to use after <tt>-uri</tt>, change the package name after <tt>-p</tt> and the output directory after <tt>-o</tt>:
<source lang="text">
bin\wsdl2java -uri http://localhost:8081/axis2/services/HelloWorld.HelloWorldImplPort?wsdl 
  -d xmlbeans -p org.eclipse.smila.sample.helloworld -s -o helloworld-ws
</source>
: This creates two folders inside <tt>helloworld-ws</tt>: <tt>src</tt> and <tt>resources</tt>. 
: If you do not want to run the generator inside the Axis2 installation you must set an environment variable <tt>AXIS2_HOME</tt> to the Axis2 installation directory.
* Add Import-Package declarations with minimum versions as available in your target platform (they will be set automatically if you use the Manifest editor's Dependencies tab to add them). To run this example at least these are needed (with valid versions at the time of writing):
<source lang="text">
javax.xml.stream;version="1.0.1",
org.apache.axiom.om;version="1.2.7",
org.apache.axiom.om.impl;version="1.2.7",
org.apache.axiom.om.impl.llom;version="1.2.7",
org.apache.axiom.soap;version="1.2.7",
org.apache.axis2;version="1.4.1",
org.apache.axis2.addressing;version="1.4.1",
org.apache.axis2.client;version="1.4.1",
org.apache.axis2.context;version="1.4.1",
org.apache.axis2.description;version="1.4.1",
org.apache.axis2.transport;version="1.4.1",
org.apache.axis2.transport.http;version="1.4.1",
org.apache.axis2.wsdl;version="1.4.1",
org.apache.xmlbeans;version="2.3.0",
org.apache.xmlbeans.impl.schema;version="2.3.0",
org.apache.xmlbeans.impl.values;version="2.3.0",
org.apache.xmlbeans.xml.stream;version="2.3.0"
</source>
: You will not get compile errors if the import for <tt>org.apache.xmlbeans.impl.schema</tt> is missing, but it is needed during runtime. 
: For more complex webservices, additional imports may be required. Check the imported generated client code for compile errors.
* Create a source folder <tt>code/gen</tt> in your bundle and move the '''content''' of the generated <tt>src</tt> folder into it.
* Create a folder <tt>lib</tt> in your bundle, create a zip file from the '''content''' of  the generated <tt>resources</tt> folder, change the suffix to <tt>jar</tt> and move it to <tt>lib</tt>. Refresh the bundle in your Eclipse workspace, and add this jar to the Bundle-Classpath of your bundle (Manifest editor, tab Runtime, Classpath setting).


== Implementation ==

* Create the package <tt>org.eclipse.smila.sample.pipelet</tt> and the Java class <tt>HelloWorldPipelet</tt>.
* Use the following code as a template for your new class. It contains empty method bodies and a reference to the logger. In the following we are going to gradually replace the comments in this file by the corresponding code snippets. For your convenience you may also download the complete zipped source file from [[Media:HelloWorldPipelet_0.9.zip|HelloWorldPipelet.zip]].

<source lang="Java">
package org.eclipse.smila.sample.pipelet

import org.apache.axis2.transport.http.HTTPConstants;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.eclipse.smila.blackboard.Blackboard;
import org.eclipse.smila.datamodel.AnyMap;
import org.eclipse.smila.datamodel.Value;
import org.eclipse.smila.processing.Pipelet;
import org.eclipse.smila.processing.ProcessingException;
import org.eclipse.smila.processing.parameters.ParameterAccessor;
import org.eclipse.smila.processing.util.ProcessingConstants;
import org.eclipse.smila.processing.util.ResultCollector;

import com.empolis.smila.sample.helloworld.HelloWorldStub;

import demo.hw.server.SayHi;
import demo.hw.server.SayHiDocument;
import demo.hw.server.SayHiResponse;
import demo.hw.server.SayHiResponseDocument;
 
public class HelloWorldPipelet implements Pipelet {
 
  // additional member variables or constants
  private final Log _log = LogFactory.getLog(getClass());

  public HelloWorldPipelet(){
  }
 
  public void configure(final AnyMap configuration) throws ProcessingException {
    // read the configuration properties
  }
 
  public String[] process(final Blackboard blackboard, final String[] recordIds) throws ProcessingException {
    // process the recordIds and create a result
    return null;
  }
}
</source>

=== Read PipeletConfiguration ===
* First let's create two constants for the property names used in the configuration (or the parameters section of the records to be processed) to retrieve the names of the source and target attribute. Replace the comment "<tt>// additional member variables or constants</tt>" with the following code snippet.
<source lang="Java">
  private final String PROP_IN_ATT_NAME = "IN_ATT_NAME";

  private final String PROP_OUT_ATT_NAME = "OUT_ATT_NAME";

  private AnyMap _config;
</source>

* Then we are going to store the the <tt>PipeletConfiguration</tt> in method <tt>configure(final AnyMap configuration)</tt> for later evalutaion in <tt>process(final Blackboard blackboard, final String[] recordIds)</tt>. So we will allow the user of this pipelet to either use the pipelet configuration to configure the attributes as well as the records themselves (e.g. the administrator could define the attributes in a job, these job properties can override default pipelet configuration properties when using the <tt>ParameterAccessor</tt> in the process method).

<source lang="Java">
@Override
public void configure(final AnyMap configuration) throws ProcessingException {
  _config = configuration;
}
</source>

'''Note''': It would also be possible to use and configure member variables directly in the configure method and not use the ParameterAccessor to retrieve configuration parameters in the process method. You can do so for properties that won't change during operation or will always stay the same for each record, no matter what the parameters of the record contain. Or for lengthy initialization like reading and parsing configuration from files and such. In these cases you should use member variables that are initialized in the configuration method using only the information from the PipeletConfiguration. but you should clearly document which parameters can only be defined with the PipeletConfiguration and which can be overridden in the records.

=== Process IDs and implement exception handling ===
The method <tt>process(Blackboard blackboard, String[] recordIds)</tt> has two parameters:
* a reference to the [[SMILA/Glossary#B|blackboard service]] that allows access on [[SMILA/Glossary#R|records]] and
* a list of record IDs to process.
<The HelloWorld pipelet should therefore iterate over the IDs in the parameter <tt>recordIds</tt>, get the required data from the record identified by the ID, process this data, and store the result in the record.

It is suggested that you use the <tt>org.eclipse.smila.processing.util.ResultCollector</tt> utility class to cope with result id collection that also provides a configurable exception handling approach. When creating the ResultCollector, you have to decide whether records that cause an exception will be excepted from the result set or if they will stay in the result set. We will use the system wide default <tt>ProcessingConstants.DROP_ON_ERROR_DEFAULT</tt> which is set to <tt>false</tt>. The ResultCollector will also check the ParameterAccessor for the parameter <tt>_failOnError</tt> (default: <tt>false</tt>).

Let's place a <tt>try ... catch()</tt> block in the <tt>for</tt> loop to ensure that errors do only interrupt the processing of the current ID. The comments in the code serve as placeholders for the functionality described in the following sections. At the end we ask the ResulotCollector for the set of <tt>recordIds</tt> as the result of the pipelet. Replace the comment "<tt>// process the recordIds and create a result</tt>" with the following code snippet.
<source lang="Java">
final ParameterAccessor paramAccessor = new ParameterAccessor(blackboard, _config);
final ResultCollector resultCollector =
      new ResultCollector(paramAccessor, _log, ProcessingConstants.DROP_ON_ERROR_DEFAULT);
for (String id : recordIds) {
    try {
        // read your configuration using the parameteraccessor
        paramAccessor.setCurrentRecord(id);
        // read configuration from the accessor

        // Read Input Data

        // Process Input Data

        // Write Output Data

        // add the id for a successful operation
        resultCollector.addResult(id);
    } catch (final Exception ex) {
        // mark the id for a failed record and let the result collector handle the exception as configured
        resultCollector.addFailedResult(id, e);
    }
} // for

// let the ResultColletor decide which ids to return:
return resultCollector.getResultIds();
</source>
'''Note''': Most of the time the return value of a pipelet is the same set of record ids as was processed (<tt>recordIds</tt>). However, in some cases a pipelet may filter record IDs or even create new records. Then the record IDs of the records to be filtered out should not be added to the ResultCollector and new record IDs have to be added to the ResultCollector in order to get the correct set of IDs as the result of the process method.

=== evaluate configuration parameters ===
Now we have to determine the source an target attribute names that have to be provided with the configuration parameters <tt>PROP_IN_ATT_NAME</tt> and <tt>PROP_OUT_ATT_NAME</tt>. Therefore we first have to determine the attribute names using the parameter accessor (Note: if we didn't want to let job parameters change these attributes, we could have evaluated the piplet configuration in the configure method and stored the result in member variables, but we want to be flexible here in this example).
Replace the comment <tt>// read configuration from the accessor</tt> with the following snippet:
<source lang="Java">
final String inAttName = paramAccessor.getRequiredParameter(PROP_IN_ATT_NAME);
if (inAttName.trim().length() == 0) {
  throw new ProcessingException("Property " + PROP_IN_ATT_NAME + " must not be an empty String");
}
final String outAttName = paramAccessor.getRequiredParameter(PROP_OUT_ATT_NAME);
if (outAttName.trim().length() == 0) {
  throw new ProcessingException("Property " + PROP_OUT_ATT_NAME + " must not be an empty String");
}
</source>

=== Read input data ===
Now we want to read the data of the attribute we stored in <tt>inAttName</tt>.
Replace the comment "<tt>// Read Input Data</tt>" with the following code snippet.
<source lang="Java">
String inputValue = "";
if (blackboard.getMetadata(id).containsKey(inAttName)) {
  inputValue = blackboard.getMetadata(id).getStringValue(inAttName);
}
</source>
'''Note''': Accessing attribute values can be achieved more generically. Therefore you have to check what data type a certain attribute contains using the method <tt>getValueType()</tt> (or the checking methods <tt>isBoolean()</tt>... etc.). Then you can use the appropriate getter method to access the raw data.

=== Process input data ===
Now we will call the HelloWorld web service with the parameter <tt>inputValue</tt> and store the result in variable <tt>outputValue</tt>. Therefore we use the classes generated from WSDL by Axis2. The HelloWorld web service will return a String message in the format <tt>"Hello "</tt> + the content of variable <tt>inputValue</tt>. Replace the comment "<tt>// Process Input Data</tt>" with the following code snippet.
<source lang="Java">
HelloWorldStub ws = new HelloWorldStub("http://localhost:8081/axis2/services/HelloWorld.HelloWorldImplPort");
ws._getServiceClient().getOptions().setProperty(HTTPConstants.CHUNKED, Boolean.FALSE);
SayHiDocument sayHiDoc = SayHiDocument.Factory.newInstance();
SayHi sayHi = sayHiDoc.addNewSayHi();
sayHi.setArg0(inputValue);
SayHiResponseDocument respDoc = ws.sayHi(sayHiDoc);
SayHiResponse response = respDoc.getSayHiResponse();        
String outputValue = response.getReturn();
</source>

=== Write output data ===
Finally, we want to store the content of the variable <tt>outputValue</tt> in the record attribute with the name contained in variable <tt>outAttName</tt>. Therefore we have to create a new <tt>Value</tt> object and set its value. Then we only need to set this <tt>Value</tt> for the current ID on the black board.
Replace the comment "<tt>// Write Output Data</tt>" with the following code snippet.
<source lang="Java">
final Value outLiteral = blackboard.getDataFactory().createStringValue(outputValue);
blackboard.getMetadata(id).put(outAttName, outLiteral);
</source>
'''Note''': The method <tt>commit(Id)</tt> of the blackboard service does not need to be called in each pipelet as it is automatically called at the end of the [[SMILA/Glossary/#p|pipeline]].

== Configuration and invocation in BPEL ==
In this tutorial we will integrate the HelloWorld pipelet in the SMILA indexing process just before the record is stored in the Solr core. With this configuration the input for the HelloWorld pipelet will be read from attribute ''Title'' and the modified output will be stored in the same attribute, overwriting the previous value.
* Edit the file <tt>configuration/org.eclipse.smila.processing.bpel/pipelines/addpipeline.bpel</tt> and add the following right between the <tt><extensionActivity name="convertDocument"></tt> and the <tt><extensionActivity name="SolrIndexPipelet"></tt> section.
<source lang="XML">
<extensionActivity>
    <proc:invokePipelet name="invokeHelloWorldPipelet">
        <proc:pipelet class="org.eclipse.smila.sample.pipelet.HelloWorldPipelet" />
        <proc:variables input="request" output="request" />
        <proc:configuration>
            <rec:Value name="IN_ATT_NAME">Title</rec:Value>
            <rec:Value name="OUT_ATT_NAME">Title</rec:Value>
            </proc:Property>
        </proc:configuration>       
    </proc:invokePipelet>
</extensionActivity>
</source>

== Build and Test your pipelet ==

Depending on the SMILA distribution you are using you have different options how to build and test your pipelet:

=== Source Code Distribution ===
You can simply integrate your pipelet in the SMILA build process. Refer to the instructions on [[SMILA/Development_Guidelines/How_to_integrate_new_bundle_into_build_process|How to integrate a new bundle into build process]] for details. You can build your SMILA application and run it as usual.

In addition to building the SMILA application you can also directly run SMILA within your eclipse IDE. To test your pipelet, you have to include the bundle in the OSGi launch configuration:
* Open ''Run > Open Run Dialog''. 
* In the left window select ''OSGi Framework > SMILA''. 
* In the right window expand ''Workspace'' and select ''org.eclipse.smila.sample.pipelet''.
* Set the ''Default Auto-Start'' option to ''true''.
* Click the ''Apply'' button.
* Launch SMILA by clicking the ''Run'' button.

=== Binary Distribution ===

To test your pipelet you have to add it as a plugin to your SMILA installation. 

* export your pipelet as a plugin using eclipse IDE wizards. Refer to the instructions on [[SMILA/Development_Guidelines/How_to_export_a_bundle|How to export a bundle]] for a step by step description.
* Copy your plugin to the directory <tt>%SMILA_HOME%/plugins</tt>.
* Add the following XML snippet to the file <tt>%SMILA_HOME%/features/org.eclipse.smila.feature_1.0.0/feature.xml</tt>:
<code lang="XML">
    <plugin
    id="org.eclipse.smila.sample.pipelet"
    download-size="0"
    install-size="0"
    version="1.0.0"
    unpack="false"/>
</code>
* Launch SMILA by starting <tt>SMILA.exe</tt>. 

If SMILA is running, you can start a crawling job as described in [[SMILA/Development_Guidelines#Run_and_manage_the_connectivity_framework|Run and manage the connectivity framework]] beginning at step 5.
While crawling your data source you can already search for indexed documents. Open your browser, navigate to [http://localhost:8080/SMILA/search http://localhost:8080/SMILA/search] and execute a query. In the result table take a look at the attribute '''Title'''. Every '''Title''' should now have the suffix <tt>"modified by HelloWorldPipelet"</tt>, as this was added by the pipelet.

=== Troubleshooting ===
If there are any problems please take a look at the log files <tt>SMILA.log</tt> and <tt>/workspace/.metadata/.log</tt> and feel free to ask for support at the [http://www.eclipse.org/smila/newsgroup.php SMILA Newsgroup].

[[Category:SMILA]]