JDBCTemplate set nested POJO with BeanPropertyRowMapper

Question

Given the following example POJO's: (Assume Getters and Setters for all properties)

class User {
    String user_name;
    String display_name;
}

class Message {
    String title;
    String question;
    User user;
}

One can easily query a database (postgres in my case) and populate a list of Message classes using a BeanPropertyRowMapper where the db field matched the property in the POJO: (Assume the DB tables have corresponding fields to the POJO properties).

NamedParameterDatbase.query("SELECT * FROM message", new BeanPropertyRowMapper(Message.class));

I'm wondering - is there a convenient way to construct a single query and / or create a row mapper in such a way to also populate the properties of the inner 'user' POJO within the message.

That is, Some syntatical magic where each result row in the query:

SELECT * FROM message, user WHERE user_id = message_id

Produce a list of Message with the associated User populated

---

Use Case:

Ultimately, the classes are passed back as a serialised object from a Spring Controller, the classes are nested so that the resulting JSON / XML has a decent structure.

At the moment, this situation is resolved by executing two queries and manually setting the user property of each message in a loop. Useable, but I imagine a more elegant way should be possible.

---

Update : Solution Used -

Kudos to @Will Keeling for inspiration for the answer with use of the custom row mapper - My solution adds the addition of bean property maps in order to automate the field assignments.

The caveat is structuring the query so that the relevant table names are prefixed (however there is no standard convention to do this so the query is built programatically):

SELECT title AS "message.title", question AS "message.question", user_name AS "user.user_name", display_name AS "user.display_name" FROM message, user WHERE user_id = message_id

The custom row mapper then creates several bean maps and sets their properties based on the prefix of the column: (using meta data to get the column name).

public Object mapRow(ResultSet rs, int i) throws SQLException {

    HashMap<String, BeanMap> beans_by_name = new HashMap();

    beans_by_name.put("message", BeanMap.create(new Message()));
    beans_by_name.put("user", BeanMap.create(new User()));

    ResultSetMetaData resultSetMetaData = rs.getMetaData();

    for (int colnum = 1; colnum <= resultSetMetaData.getColumnCount(); colnum++) {

        String table = resultSetMetaData.getColumnName(colnum).split("\\.")[0];
        String field = resultSetMetaData.getColumnName(colnum).split("\\.")[1];

        BeanMap beanMap = beans_by_name.get(table);

        if (rs.getObject(colnum) != null) {
            beanMap.put(field, rs.getObject(colnum));
        }
    }

    Message m = (Task)beans_by_name.get("message").getBean();
    m.setUser((User)beans_by_name.get("user").getBean());

    return m;
}

Again, this might seem like overkill for a two class join but the IRL use case involves multiple tables with tens of fields.

Answer 1

Perhaps you could pass in a custom RowMapper that could map each row of an aggregate join query (between message and user) to a Message and nested User. Something like this:

List<Message> messages = jdbcTemplate.query("SELECT * FROM message m, user u WHERE u.message_id = m.message_id", new RowMapper<Message>() {
    @Override
    public Message mapRow(ResultSet rs, int rowNum) throws SQLException {
        Message message = new Message();
        message.setTitle(rs.getString(1));
        message.setQuestion(rs.getString(2));

        User user = new User();
        user.setUserName(rs.getString(3));
        user.setDisplayName(rs.getString(4));

        message.setUser(user);

        return message;
    }
});

Answer 2

A bit late to the party however I found this when I was googling the same question and I found a different solution that may be favorable for others in the future.

Unfortunately there is not a native way to achieve the nested scenario without making a customer RowMapper. However I will share an easier way to make said custom RowMapper than some of the other solutions here.

Given your scenario you can do the following:

class User {
    String user_name;
    String display_name;
}

class Message {
    String title;
    String question;
    User user;
}

public class MessageRowMapper implements RowMapper<Message> {

    @Override
    public Message mapRow(ResultSet rs, int rowNum) throws SQLException {
        User user = (new BeanPropertyRowMapper<>(User.class)).mapRow(rs,rowNum);
        Message message = (new BeanPropertyRowMapper<>(Message.class)).mapRow(rs,rowNum);
        message.setUser(user);
        return message;
     }
}

The key thing to remember with BeanPropertyRowMapper is that you have to follow the naming of your columns and the properties of your class members to the letter with the following exceptions (see Spring Documentation):

• column names are aliased exactly
• column names with underscores will be converted into "camel" case (ie. MY_COLUMN_WITH_UNDERSCORES == myColumnWithUnderscores)

Answer 3

Spring introduced a new AutoGrowNestedPaths property into the BeanMapper interface.

As long as the SQL query formats the column names with a . separator (as before) then the Row mapper will automatically target inner objects.

With this, I created a new generic row mapper as follows:

QUERY:

SELECT title AS "message.title", question AS "message.question", user_name AS "user.user_name", display_name AS "user.display_name" FROM message, user WHERE user_id = message_id

---

ROW MAPPER:

package nested_row_mapper;

import org.springframework.beans.*;
import org.springframework.jdbc.core.RowMapper;
import org.springframework.jdbc.support.JdbcUtils;

import java.sql.ResultSet;
import java.sql.ResultSetMetaData;
import java.sql.SQLException;

public class NestedRowMapper<T> implements RowMapper<T> {

  private Class<T> mappedClass;

  public NestedRowMapper(Class<T> mappedClass) {
    this.mappedClass = mappedClass;
  }

  @Override
  public T mapRow(ResultSet rs, int rowNum) throws SQLException {

    T mappedObject = BeanUtils.instantiate(this.mappedClass);
    BeanWrapper bw = PropertyAccessorFactory.forBeanPropertyAccess(mappedObject);

    bw.setAutoGrowNestedPaths(true);

    ResultSetMetaData meta_data = rs.getMetaData();
    int columnCount = meta_data.getColumnCount();

    for (int index = 1; index <= columnCount; index++) {

      try {

        String column = JdbcUtils.lookupColumnName(meta_data, index);
        Object value = JdbcUtils.getResultSetValue(rs, index, Class.forName(meta_data.getColumnClassName(index)));

        bw.setPropertyValue(column, value);

      } catch (TypeMismatchException | NotWritablePropertyException | ClassNotFoundException e) {
         // Ignore
      }
    }

    return mappedObject;
  }
}

Answer 4

Update: 10/4/2015. I typically don't do any of this rowmapping anymore. You can accomplish selective JSON representation much more elegantly via annotations. See this gist.

---

I spent the better part of a full day trying to figure this out for my case of 3-layer nested objects and just finally nailed it. Here's my situation:

Accounts (i.e. users) --1tomany--> Roles --1tomany--> views (user is allowed to see)

(These POJO classes are pasted at the very bottom.)

And I wanted the controller to return an object like this:

[ {
  "id" : 3,
  "email" : "[email protected]",
  "password" : "sdclpass",
  "org" : "Super-duper Candy Lab",
  "role" : {
    "id" : 2,
    "name" : "ADMIN",
    "views" : [ "viewPublicReports", "viewAllOrders", "viewProducts", "orderProducts", "viewOfferings", "viewMyData", "viewAllData", "home", "viewMyOrders", "manageUsers" ]
  }
}, {
  "id" : 5,
  "email" : "[email protected]",
  "password" : "stereopass",
  "org" : "Stereolab",
  "role" : {
    "id" : 1,
    "name" : "USER",
    "views" : [ "viewPublicReports", "viewProducts", "orderProducts", "viewOfferings", "viewMyData", "home", "viewMyOrders" ]
  }
}, {
  "id" : 6,
  "email" : "[email protected]",
  "password" : "ukmedpass",
  "org" : "University of Kentucky College of Medicine",
  "role" : {
    "id" : 2,
    "name" : "ADMIN",
    "views" : [ "viewPublicReports", "viewAllOrders", "viewProducts", "orderProducts", "viewOfferings", "viewMyData", "viewAllData", "home", "viewMyOrders", "manageUsers" ]
  }
} ]

A key point is to realize that Spring doesn't just do all this automatically for you. If you just ask it to return an Account item without doing the work of nested objects, you'll merely get:

{
  "id" : 6,
  "email" : "[email protected]",
  "password" : "ukmedpass",
  "org" : "University of Kentucky College of Medicine",
  "role" : null
}

So, first, create your 3-table SQL JOIN query and make sure you're getting all the data you need. Here's mine, as it appears in my Controller:

@PreAuthorize("hasAuthority('ROLE_ADMIN')")
@RequestMapping("/accounts")
public List<Account> getAllAccounts3() 
{
    List<Account> accounts = jdbcTemplate.query("SELECT Account.id, Account.password, Account.org, Account.email, Account.role_for_this_account, Role.id AS roleid, Role.name AS rolename, role_views.role_id, role_views.views FROM Account JOIN Role on Account.role_for_this_account=Role.id JOIN role_views on Role.id=role_views.role_id", new AccountExtractor() {});
    return accounts;
}

Note that I'm JOINing 3 tables. Now create a RowSetExtractor class to put the nested objects together. The above examples show 2-layer nesting... this one goes a step further and does 3 levels. Note that I'm having to maintain the second-layer object in a map as well.

public class AccountExtractor implements ResultSetExtractor<List<Account>>{

    @Override
    public List<Account> extractData(ResultSet rs) throws SQLException, DataAccessException {

        Map<Long, Account> accountmap = new HashMap<Long, Account>();
        Map<Long, Role> rolemap = new HashMap<Long, Role>();

        // loop through the JOINed resultset. If the account ID hasn't been seen before, create a new Account object. 
        // In either case, add the role to the account. Also maintain a map of Roles and add view (strings) to them when encountered.

        Set<String> views = null;
        while (rs.next()) 
        {
            Long id = rs.getLong("id");
            Account account = accountmap.get(id);
            if(account == null)
            {
                account = new Account();
                account.setId(id);
                account.setPassword(rs.getString("password"));
                account.setEmail(rs.getString("email"));
                account.setOrg(rs.getString("org"));
                accountmap.put(id, account);
            }

            Long roleid = rs.getLong("roleid");
            Role role = rolemap.get(roleid);
            if(role == null)
            {
                role = new Role();
                role.setId(rs.getLong("roleid"));
                role.setName(rs.getString("rolename"));
                views = new HashSet<String>();
                rolemap.put(roleid, role);
            }
            else
            {
                views = role.getViews();
                views.add(rs.getString("views"));
            }

            views.add(rs.getString("views"));
            role.setViews(views);
            account.setRole(role);
        }
        return new ArrayList<Account>(accountmap.values());
    }
}

And this gives the desired output. POJOs below for reference. Note the @ElementCollection Set views in the Role class. This is what automatically generates the role_views table as referenced in the SQL query. Knowing that table exists, its name and its field names is crucial to getting the SQL query right. It feels wrong to have to know that... it seems like this should be more automagic -- isn't that what Spring is for?... but I couldn't figure out a better way. You've got to do the work manually in this case, as far as I can tell.

@Entity
public class Account implements Serializable {
        private static final long serialVersionUID = 1L;

        @Id 
        @GeneratedValue(strategy=GenerationType.AUTO)
        private long id;
        @Column(unique=true, nullable=false)
        private String email;
        @Column(nullable = false) 
        private String password;
        @Column(nullable = false) 
        private String org;
        private String phone;

        @ManyToOne(fetch = FetchType.EAGER, optional = false)
        @JoinColumn(name = "roleForThisAccount") // @JoinColumn means this side is the *owner* of the relationship. In general, the "many" side should be the owner, or so I read.
        private Role role;

        public Account() {}

        public Account(String email, String password, Role role, String org) 
        {
            this.email = email;
            this.password = password;
            this.org = org;
            this.role = role;
        }
        // getters and setters omitted

    }

   @Entity
   public class Role implements Serializable {

        private static final long serialVersionUID = 1L;

        @Id 
        @GeneratedValue(strategy=GenerationType.AUTO)   
        private long id; // required

        @Column(nullable = false) 
        @Pattern(regexp="(ADMIN|USER)")
        private String name; // required

        @Column
        @ElementCollection(targetClass=String.class)
        private Set<String> views;

        @OneToMany(mappedBy="role")
        private List<Account> accountsWithThisRole;

        public Role() {}

        // constructor with required fields
        public Role(String name)
        {
            this.name = name;
            views = new HashSet<String>();
            // both USER and ADMIN
            views.add("home");
            views.add("viewOfferings");
            views.add("viewPublicReports");
            views.add("viewProducts");
            views.add("orderProducts");
            views.add("viewMyOrders");
            views.add("viewMyData");
            // ADMIN ONLY
            if(name.equals("ADMIN"))
            {
                views.add("viewAllOrders");
                views.add("viewAllData");
                views.add("manageUsers");
            }
        }

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

        public String getName() { return this.name; }
        public void setName(String name) { this.name = name; }

        public Set<String> getViews() { return this.views; }
        public void setViews(Set<String> views) { this.views = views; };
    }

Answer 5

I worked a lot on stuff like this and do not see an elegant way to achieve this without an OR mapper.

Any simple solution based on reflection would heavily rely on the 1:1 (or maybe N:1) relation. Further your columns returned are not qualified by their type, so you cannot say which columns matches which class.

You may get away with spring-data and QueryDSL. I did not dig into them, but I think you need some meta-data for the query that is later used to map back the columns from your database into a proper data structure.

You may also try the new PostgreSQL json support that looks promising.