Archive

Archive for the ‘Spring’ Category

Working with Freebase – Part 2

February 14, 2013 Leave a comment

In part 1 I demonstrated a method to generate a Freebase MQL query from an annotated class that is intended to mimic JPA’s ORM behavior. Next I’ll show how to use the Spring RestTemplate to execute the query and create an array of objects from the results. The following code fragment illustrates the steps.

RestTemplate restTemplate = new RestTemplate();
restTemplate.getMessageConverters().add(new MappingJacksonHttpMessageConverter());

URI uri = new URI("https", "www.googleapis.com", "/freebase/v1/mqlread/", Util.createQuery(variables), null);
ResponseEntity entity = restTemplate.getForEntity(uri, PersonsResponse.class);
PersonsResponse response = entity.getBody();
Person[] persons = response.getResult();

The first step is the create a RestTemplate and then add a JSON message converter (MappingJacksonHttpMessageConverter) to it. Freebase returns the its results in JSON format. An array of results will look like this …

{
"code":          "/api/status/ok",
"result": [
{
"type": "/people/person",
"id": "/en/dan_milbrath",
"gender": {
"type": "/people/gender",
"id": "/en/male",
"name": "Male"
},
"name": "Dan Milbrath"
},
{
"type": "/people/person",
"id": "/en/david_safavian",
"gender": {
"type": "/people/gender",
"id": "/en/male",
"name": "Male"
},
"name": "David Safavian"
},
{
"type": "/people/person",
"id": "/en/robert_cook",
"gender": {
"type": "/people/gender",
"id": "/en/male",
"name": "Male"
},
"name": "Robert Cook"
}
],
"status":        "200 OK",
"transaction_id": "cache;cache02.p01.sjc1:8101;2013-02-14T07:31:07Z;0064"
}

ResponseEntity holds the code, status, and transaction_id properties. Typing it to an appropriate class allows it to hold the result returned in the JSON format. In this case it is PersonsResponse which is a subclass of MQLMultipleResultResponse typed to Person.

class PersonsResponse extends MQLMultipleResultResponse<Person>  {
}

public class MQLMultipleResultResponse<T> {
	private String cursor;
	private T[] result;

	public T[] getResult() {
		return result;
	}
	public void setResult(T[] result) {
		this.result = result;
	}
	public String getCursor() {
		return cursor;
	}
	public void setCursor(String cursor) {
		this.cursor = cursor;
	}
}

The code is available at GitHub.

Categories: Bruce's Posts, Java, Spring

REST Service Using Spring Data

October 24, 2012 1 comment

I was intrigued by Spring Data’s REST project which promises to, “… make it easy to expose JPA based repositories as RESTful endpoints.” To try it out I decided to create a service which provides famous quotes such as Albert Einstein’s.

Insanity: doing the same thing over and over again and expecting different results.

Consider a JPA entity that holds a quote.

@Entity
@Table(name = "quote")
public class Quote {
	@Id
	@GeneratedValue
	private Long id;
	private String quote;

	public Long getId() {
		return id;
	}
	public void setId(Long id) {
		this.id = id;
	}

	public String getQuote() {
		return quote;
	}
	public void setQuote(String quote) {
		this.quote = quote;
	}

}

This is a very simple example of an entity. Just a id and text string.

Now create a repository service to expose CRUD style access to the entity. For now I’m just going to read a quote so all that’s needed is the following interface.

@RestResource(path = "quote")
public interface QuoteRepository extends PagingAndSortingRepository<Quote, Long> {
}

Spring Data provide a paging repository (PagingAndSortingRepository) and the simpler CrudRepository.

All the magic is in the @RestResource annotation. When the Spring REST Exporter starts up it will expose the QuoteRespository as a REST endpoint.

The Spring REST Exporter is a web application build on top of Spring MVC. The documentation is sparse but examples are pretty straight forward.

Also the Spring REST Exporter example uses gradle. I want to use a combination of Maven and Eclipse. It uses Java classes to configure the application environment. I wanted to use XML application context files. I also wanted multiple projects to encapsulate various applications and libraries. The entire source code is available here at github. Getting all the pieces just right took some time. I ran the quote war using Tomcat 7.

My quote endpoint will expose a list of quotes. In this case I’ve exposed paging and sorting possibilities.

http://localhost:8080/quote-json-server/quote produces a list of quotes stored in the repository.

{
  "links" : [ ],
  "content" : [ {
    "links" : [ {
      "rel" : "self",
      "href" : "http://localhost:8080/quote-json-server/quote/1"
    } ],
    "quote" : "Insanity: doing the same thing over and over again and expecting different results."
  }, {
    "links" : [ {
      "rel" : "self",
      "href" : "http://localhost:8080/quote-json-server/quote/2"
    } ],
    "quote" : "Better to remain silent and be thought a fool than to speak out and remove all doubt."
  } ],
  "page" : {
    "size" : 20,
    "totalElements" : 2,
    "totalPages" : 1,
    "number" : 1
  }
}

The output is a JSON serialization of the org.springframework.hateoas.PagedResources class. The serialization uses a HATEOAS (Hypermedia as the Engine of Application State) style format to provide a state representation of the resource.

In this case it is the state of the first page of the quotes list.

http://localhost:8080/quote-json-server/quote/1 provides the state of the first quote.

{
  "links" : [ {
    "rel" : "self",
    "href" : "http://localhost:8080/quote-json-server/quote/1"
  } ],
  "quote" : "Insanity: doing the same thing over and over again and expecting different results."
}

Having come from RPC as the method of choice REST takes a bit of getting used to. Roy T. Fielding’s REST APIs must be hypertext-driven discusses this. The biggest thing being individual resources are identified in requests. In this case a href (URL).

The next step is to create a client application to test the service. I created two clients. One a simple Java client and the other a simple Android application. REST clients are free to deserialize the response in any way they wish. I chose to use Spring’s RestTemplate API. Here’s the simple Java Client.

public class QuoteRestTemplateClient {
  private static final String URL = "http://localhost:8080/quote-json-server/quote/1";

  public static void main(String[] args) {
   RestTemplate restTemplate = new RestTemplate();

  HttpHeaders requestHeaders = new HttpHeaders();
  List<MediaType> acceptableMediaTypes = new ArrayList<MediaType>();
  acceptableMediaTypes.add(MediaType.APPLICATION_JSON);
  requestHeaders.setAccept(acceptableMediaTypes);
  HttpEntity<?> requestEntity = new HttpEntity<Object>(requestHeaders);
  restTemplate.getMessageConverters().add(new MappingJacksonHttpMessageConverter());

  System.out.printf("url: %s\n", URL);
  try {
            ResponseEntity<Quote> entity = restTemplate.exchange(URL, HttpMethod.GET,              requestEntity, Quote.class);
            System.out.printf("json: %s\n", entity.getBody().toString());

            Quote quote = restTemplate.getForObject(URL, Quote.class);
            System.out.printf("quote: %s\n", quote.getQuote());
      } catch (RestClientException e) {
            e.printStackTrace();
    }
  }
}

The biggest challenge was creating the client side Quote class. There were actually a couple of challenges.

The Spring REST Explorer uses Spring’s HATEOAS API to wrap the JPA entity (entities in the case of a list). The response is either an instance of Resource or Resources. While I could have included the Spring’s HATEOAS jar with my application that would pulled in baggage that I wanted to avoid and as I found out later would also be problematic with the Android client. So I punted and just copied the six source files I needed into the project and edited them accordingly.

It would have been nice to have just used the service side Quote class with the client. This too was problematic because it’s annotated with JPA. This again would pull in jars I don’t want or need on the client side. Also, its getId() clashes with one found in the HATEOAS Identifier’s getId(). Here’s the client side Quote class.

public class Quote extends ResourceSupport {
	private String quote;

	public String getQuote() {
		return quote;
	}
	public void setQuote(String quote) {
		this.quote = quote;
	}
}

Pretty simple but then the example is simple. I’d truly want a way to easily generate the server and client side versions of Quote in a repeatable manner. Something to investigate on another day.

The Android application, if one discounts the necessary Android code, is identical to the Java client. It also needs the HATEOAS source files and the client side Quote class. Here’s the main Activity class.

public class MainActivity extends Activity {
    private static final String URL = "http://192.168.2.201:8080/quote-json-server/quote/1";

    @Override
    public void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        setContentView(R.layout.main);
        TextView t1 =(TextView)findViewById(R.id.textView1);
        t1.setText(URL);

        RestTemplate restTemplate = new RestTemplate();

        HttpHeaders requestHeaders = new HttpHeaders();
        List<MediaType> acceptableMediaTypes = new ArrayList<MediaType>();
        acceptableMediaTypes.add(MediaType.APPLICATION_JSON);
        requestHeaders.setAccept(acceptableMediaTypes);
        HttpEntity<?> requestEntity = new HttpEntity<Object>(requestHeaders);
        restTemplate.getMessageConverters().add(new MappingJacksonHttpMessageConverter());

        String string;

        try {
        	ResponseEntity<Quote> entity = restTemplate.exchange(URL, HttpMethod.GET, requestEntity, Quote.class);
            string = entity.getBody().toString();
        } catch (RestClientException e) {
        	string = e.getMessage();
        }

        TextView t2 = (TextView)findViewById(R.id.textView2);
        t2.setText(string);

        Quote quote = restTemplate.getForObject(URL, Quote.class);
        TextView t3 =(TextView)findViewById(R.id.textView3);
        t3.setText(quote.getQuote());
    }
}

Here’s what it looks like.

 

Categories: Bruce's Posts, JPA, Spring Data

Open Lane Changes

April 21, 2012 3 comments

My Open Lane application has undergone some significant changes. I’ve added:

  1. Entity classes for a swim meet and events within the meet.
  2. Enumerators for Gender and Stroke.
  3. An AgeGroup class to encapsulate an event’s age category.
  4. JSF DataModel subclasses for Meet and Event along with a refactored SignUp backing bean.
  5. JPA services for all data input and output.
  6. JSF Converter classes for Gender, Stroke, and AgeGroup.
  7. CSV file initialization functions to populate the Swimmer, Meet, Event, and User tables.

Gender and stroke each represent a fixed set of constants which can be objectified in Java as an enumerator. An enumerator provides a clean, self-documenting, and an efficient way of coding an object model. Java enumerators can also be used to encapsulated conversion functionality.

In the case of this application the meet, event, swimmer, and entry records are imported from a CSV file. Within the event record is a stroke (i.e. Freestyle, Backstroke) filed that is coded using a number.  A “1” represents Freestyle, a “2” represents Backstroke, and so on. Stroke has been coded so that each enumerator has its import field number code associated with it. During the import process a simple call to Stroke.parse(string) generates the correct enumerator.

Instead of putting parse into Stroke I could have created a distinct converter. That would have been a cleaner separation of the code. But I don’t anticipate the conversion changing or needing to be adaptable. This way the code is simpler and straight forward.

Stroke.java

package org.bwgz.swim.openlane.data.model;

public enum Stroke {
    FREE("1"),
    BACK("2"),
    BREAST("3"),
    FLY("4"),
    IM("5");

    private String code;

    Stroke(String code) {
    	this.code = code;
    }

     public String getCode() {
          return code;
     }

     public void setCode(String code) {
          this.code = code;
     }

    public static Stroke parse(String string) {
    	Stroke result = null;

    	for (Stroke stroke : Stroke.values() ) {
    		if (stroke.getCode().equalsIgnoreCase(string)) {
				result = stroke;
				break;
			}
		}

		return result;
    }
}

This is an example of using the Gender enumerator within a JPA query to abstract away the codes used in the database. I don’t have to concern myself with how the data is coded once it is imported into the database.

em.createQuery("select e from Event e where e.meet = :meet and (e.gender = :mixed or e.gender = :gender")
.setParameter("meet", meet)
.setParameter("mixed", Gender.MIXED)
.setParameter("gender", gender)
.getResultList();
 

Later in this article is an example of how the enumerator’s parse method is used.

Eventually the JSF tables will support sorting and paging. In preparation for that I’ve moved from List and Collection to JSF’s DataModel. DataModel is a wrapper that abstracts the underlying data. I’ve also used it to store which record (object) may have been selected by the user. Previously there was an independent bean to handle that. Putting it here keeps thing a bit tidier. I created DataModel’s for Event and Meet. Later I’ll reuse the pattern for a user’s open lane applications.

I’ve decided to build out services for each entity in the data model. This provides a level of encapsulation and organization that make the code more manageable. Using @Autowired I can also create services that are built upon other services. For example, the SignUpService uses the SwimmerService.

SignUpService.java snippet.

@Service("signupService")
@Repository
public class SignUpServiceImpl implements SignUpService, Serializable {

     private EntityManager em;
     @Autowired
     private SwimmerService swimmerService;

    @PersistenceContext
    public void setEntityManager(EntityManager em) {
    	this.em = em;
    }
….
    @Transactional
     public Boolean doSignUp(SignUp signUp) {
     Boolean result = Boolean.FALSE;

     Swimmer swimmer = swimmerService.findSwimmer(signUp.getUsasId());

As a general rule the code should have a clean separation between how data is represented internally and externally. JSF’s Converter class is one way to accomplish that. I needed a Converter for Gender, Stroke, and AgeGroup. Since Gender and Stroke were enumerators I create an abstract base class to simplify things. The Gender and Stroke converters need only supply the mappings between the internal representation (an enumerator) and string (display value).

AbstractEnumConverter.java

package org.bwgz.swim.openlane.faces.converter;

import java.util.Hashtable;
import java.util.Map;

import javax.faces.component.UIComponent;
import javax.faces.context.FacesContext;
import javax.faces.convert.Converter;

public abstract class AbstractEnumConverter<T> implements Converter {
	private final static int ASSOCIATION_ENUM	= 0;
	private final static int ASSOCIATION_STRING	= 1;

	private Class<T> clazz;
	private final Map<T, String> toStringMap = new Hashtable<T, String>();
	private final Map<String, T> toEnumMap = new Hashtable<String, T>();

	@SuppressWarnings("unchecked")
	public AbstractEnumConverter(Class<T> clazz, Object[][] associations) {
		this.clazz = clazz;

		for (Object[] association : associations) {
			toEnumMap.put((String) association[ASSOCIATION_STRING], (T) association[ASSOCIATION_ENUM]);
			toStringMap.put((T) association[ASSOCIATION_ENUM], (String) association[ASSOCIATION_STRING]);
		}
	}

	public Object getAsObject(FacesContext context, UIComponent component, String value) {
		return toEnumMap.get(value);
	}

	@SuppressWarnings("unchecked")
	public String getAsString(FacesContext context, UIComponent component, Object value) {
            if (value.getClass() == clazz) {
        	 return toStringMap.get((T) value);
            }
            else
            {
                throw new IllegalArgumentException(String.format("Cannot convert object - not of type %s", clazz.getSimpleName()));
            }
	}
}

StrokeConverter.java

package org.bwgz.swim.openlane.faces.converter;

import javax.faces.convert.FacesConverter;
import org.bwgz.swim.openlane.data.model.Stroke;

@FacesConverter(value="strokeConverter")
public class StrokeConverter extends AbstractEnumConverter<Stroke> {
	private final static Object associations[][] = {
		{ Stroke.FREE,		"Free" },
		{ Stroke.BACK,		"Back" },
		{ Stroke.BREAST,	"Breast" },
		{ Stroke.FLY,		"Fly" },
		{ Stroke.IM,		"IM" },
	};

	public StrokeConverter() {
		super(Stroke.class, associations);
	}
}

The AgeGroup converter encapsulates the four rules used to describe an age category.

AgeGroupConverter.java

package org.bwgz.swim.openlane.faces.converter;

import javax.faces.component.UIComponent;
import javax.faces.context.FacesContext;
import javax.faces.convert.Converter;
import javax.faces.convert.FacesConverter;

import org.bwgz.swim.openlane.data.model.AgeGroup;

@FacesConverter(value="ageGroupConverter")
public class AgeGroupConverter implements Converter {
	private static final String SENIOR		= "Senior";
	private static final String UNDER		= "Under";
	private static final String OVER		= "Over";
	private static final String AMPERSAND	= "&";
	private static final String HYPHEN		= "-";

	private static final int LEFT	= 0;
	private static final int RIGHT	= 1;

	private Object getAsObject(String string) {
		long min = 0;
		long max = 0;

		if (string.equals(SENIOR)) {
			min = 0;
			max = 0;
		}
		else if (string.contains(AMPERSAND)) {
			String[] fields = string.split(AMPERSAND);

			if (fields[RIGHT].equals(UNDER)) {
				min = 0;
				max = Long.parseLong(fields[LEFT]);
			}
			else if (fields[RIGHT].equals(OVER)) {
				min = Long.parseLong(fields[LEFT]);
				max = 0;
			}
		}
		else if (string.contains(HYPHEN)) {
			String[] fields = string.split(HYPHEN);

			min = Long.parseLong(fields[LEFT]);
			max = Long.parseLong(fields[RIGHT]);
		}

		return new AgeGroup(min, max);
	}

	public Object getAsObject(FacesContext context, UIComponent component, String value) {
		return getAsObject(value);
	}

	private String getAsString(AgeGroup ageGroup) {
		String string;

		if (ageGroup.getMin() == 0 & ageGroup.getMax() == 0) {
			string = SENIOR;
		}
		else {
			String left;
			String seperator;
			String right;

			if (ageGroup.getMin() == 0) {
				left = String.valueOf(ageGroup.getMax());
				seperator = AMPERSAND;
				right = UNDER;
			}
			else if (ageGroup.getMax() == 0) {
				left = String.valueOf(ageGroup.getMin());
				seperator = AMPERSAND;
				right = OVER;
			}
			else {
				left = String.valueOf(ageGroup.getMin());
				seperator = HYPHEN;
				right = String.valueOf(ageGroup.getMax());
			}

			string = left + seperator + right;
		}

		return string;
	}

	public String getAsString(FacesContext context, UIComponent component, Object value) {
            if (value instanceof AgeGroup) {
                return getAsString((AgeGroup) value);
            }
            else
            {
                throw new IllegalArgumentException("Cannot convert object - not of type AgeGroup");
            }
	}
}

Meet, event, swimmer, and later entry records are imported from another system. Unfortunately the limitations of that system prevent the application from accessing them directly. Instead the data is exported to CSV files and then imported into the application.

Eventually I’ve incorporate a form of file upload within the application to provide live data import. For now some dummy (test) CSV files are included in the application and imported when the application is first accessed. Some services now have an initialize method. When called the method will read a CSV file and write the data to the database using JPA.

Using <on-start> within home-flow.xml I trigger the initializations. This is a hack for testing purposes only.

home-flow.xml snippet

<on-start>
        <evaluate expression="swimmerService.initialize()" />
        <evaluate expression="meetService.initialize()" />
        <evaluate expression="signupService.initialize()" />
</on-start>

I use FlatPack to process the CSV files. It’s a nice CSV library that I’ve used many times in the past. I particularly like that it allows you to create a map of the column names. On the other hand the map requires names for every column. This can be a bit tedious when you only need the first few columns. It can also choke when the file contains records with differing number of columns.

SwimmerServiceImpl.java snippet.

    @Transactional
    public void initialize() {
    if (!initialized) {
        DataSet dataSet;

        Parser parser;
        parser = DefaultParserFactory.getInstance().newDelimitedParser(
            new InputStreamReader(this.getClass().getClassLoader().getResourceAsStream("/test/data/athlete.pzmap.xml")), // xml column mapping
            new InputStreamReader(this.getClass().getClassLoader().getResourceAsStream("/test/data/athlete.csv")),  // csv file to parse
            ';', // delimiter
             '"', // text qualfier
             false); // ignore the first record (may need to be done if first record contain column names)

            dataSet = parser.parse();
            while (dataSet.next()) {
                Swimmer swimmer = new Swimmer();

                swimmer.setId(dataSet.getString("Reg_ID"));
                swimmer.setFirst(dataSet.getString("First_name"));
                swimmer.setLast(dataSet.getString("Last_name"));
                swimmer.setGender(Gender.parse(dataSet.getString("Ath_Sex")));
                swimmer.setId(dataSet.getString("Reg_ID"));
                swimmer.setBirthdate(stringToDate(dataSet.getString("Birth_date")));

                em.persist(swimmer);
                }

            initialized = true;
        }
     }
 

The code is available here on GitHub.