QueryRunner run = new QueryRunner(dataSource);// Use the BeanHandler implementation to convert the first// ResultSet row into a Person JavaBean.ResultSetHandler<Person> h = new BeanHandler<Person>(Person.class);// Execute the SQL statement with one replacement parameter and// return the results in a new Person object generated by the BeanHandler.Person p = run.query( "SELECT * FROM Person WHERE name=?", h, "John Doe");q里有个地方有约束，是要求CZ中的JavaBeancPerson中的字段定义要和数据库的字段定义一致。Java的命名习惯一般是骆峰写法Q例如userIdQ那么数据库中就必须定义为userIdQ而问题在于：有时候我们需要数据库中字D늚定义格式与JavaBean的命名不一P比如数据库定义ؓQuser_idQ而JavaBean则定义ؓuserId
看源代码可能有点Ҏ_在官方的example面的最下面果然有一D关于自定义BeanProcessor的指引。摘录出来：
BasicRowProcessor uses a BeanProcessor to convert ResultSet columns into JavaBean properties. You can subclass and override processing steps to handle datatype mapping specific to your application. The provided implementation delegates datatype conversion to the JDBC driver.
BeanProcessor maps columns to bean properties as documented in the BeanProcessor.toBean() javadoc. Column names must match the bean's property names case insensitively. For example, the firstname column would be stored in the bean by calling its setFirstName() method. However, many database column names include characters that either can't be used or are not typically used in Java method names. You can do one of the following to map these columns to bean properties:
1. Alias the column names in the SQL so they match the Java names: select social_sec# as socialSecurityNumber from person
2. Subclass BeanProcessor and override the mapColumnsToProperties() method to strip out the offending characters.
大概意思就是提供二U方式：一U就是最直接的，用as关键字把colName重命名，另一U方式就是承BeanProcessorc，重写mapColumnsToProperties()Ҏ?br /> 那当然是W二U方式更加具有代表性。尝试了一下。代码如下：
1
public class CustomBeanProcessor extends BeanProcessor {
2
3 @Override
4 protected int[] mapColumnsToProperties(ResultSetMetaData rsmd,
5
PropertyDescriptor[] props) throws SQLException {
6 int cols = rsmd.getColumnCount();
7 int columnToProperty[] = new int[cols + 1];
8 Arrays.fill(columnToProperty, PROPERTY_NOT_FOUND);
9
10 for (int col = 1; col <= cols; col++) {
11 String columnName = rsmd.getColumnLabel(col);
12 if (null == columnName || 0 == columnName.length()) {
13 columnName = rsmd.getColumnName(col);
14
}
15 columnName = colNameConvent(columnName); // 在这里进行数据库表columnName的特D处?/span>
16 for (int i = 0; i < props.length; i++) {
17
18 if (columnName.equalsIgnoreCase(props[i].getName())) {
19 columnToProperty[col] = i;
20 break;
21 }
22 }
23 }
24 return columnToProperty;
25 }
26
27 /**
28 * 数据库列名重新约?br />
29 * @param columnName
30 * @return
31 */
32 private String colNameConvent(String columnName) {
33 String[] strs = columnName.split("_");
34 String conventName = "";
35 for (int i = 0; i < strs.length; i++) {
36 conventName += StringUtils.capitalize(strs[i]);
37 }
38 StringUtils.uncapitalize(conventName);
39 return conventName;
40 }
41
}
public class CustomBeanProcessor extends BeanProcessor {
2
3
@Override4
protected int[] mapColumnsToProperties(ResultSetMetaData rsmd,5
PropertyDescriptor[] props) throws SQLException {
PropertyDescriptor[] props) 6
int cols = rsmd.getColumnCount();7
int columnToProperty[] = new int[cols + 1];8
Arrays.fill(columnToProperty, PROPERTY_NOT_FOUND);9
10
for (int col = 1; col <= cols; col++) {11
String columnName = rsmd.getColumnLabel(col); 12
if (null == columnName || 0 == columnName.length()) {13
columnName = rsmd.getColumnName(col);14
}15
columnName = colNameConvent(columnName); // 在这里进行数据库表columnName的特D处?/span>16
for (int i = 0; i < props.length; i++) {17
18
if (columnName.equalsIgnoreCase(props[i].getName())) {19
columnToProperty[col] = i;20
break;21
}22
}23
}24
return columnToProperty;25
}26
27
/**28
* 数据库列名重新约?br />
29 * @param columnName30
* @return31
*/32
private String colNameConvent(String columnName) {33
String[] strs = columnName.split("_");34
String conventName = "";35
for (int i = 0; i < strs.length; i++) {36
conventName += StringUtils.capitalize(strs[i]);37
}38
StringUtils.uncapitalize(conventName);39
return conventName;40
}41
}注意mapColumnsToPropertiesҎ的逻辑是从父类的方法中直接复制出来的，然后在第15行那里变了个戏法Q这里的columnName是从数据库中读出来的，自定义一个privateҎ用于转换命名Q这里你可以添加自q命名U束。例如上面就是把 user_id 转化为Java的骆峰写法：userId
再深入一层思考，你可以在q里q行更多扩展Q以便让自己可以选择不同的命名{换方式。定义了q个Processor之后Q下面看看如何调用：
Connection conn = ConnectionManager.getInstance().getConnection();QueryRunner qr = new QueryRunner();CustomBeanProcessor convert = new CustomBeanProcessor();RowProcessor rp = new BasicRowProcessor(convert);BeanHandler<User> bh = new BeanHandler<User>(User.class, rp);User u = qr.query(conn, sql, bh, params);DbUtils.close(conn); 1protected <T> List<T> selectBeanList(Connection conn, String sql, Class<T> type,
2 Object[] params) throws Exception {
3 log.debug("select sql:[" + sql + "]");
4 QueryRunner qr = new QueryRunner();
5 CustomBeanProcessor convert = new CustomBeanProcessor();
6 RowProcessor rp = new BasicRowProcessor(convert);
7 ResultSetHandler<List<T>> bh = new BeanListHandler<T>(type, rp);
8 List<T> list = qr.query(conn, sql, bh, params);
9 return list;
10 }
protected <T> List<T> selectBeanList(Connection conn, String sql, Class<T> type,2
Object[] params) throws Exception {3
log.debug("select sql:[" + sql + "]");4
QueryRunner qr = new QueryRunner();5
CustomBeanProcessor convert = new CustomBeanProcessor();6
RowProcessor rp = new BasicRowProcessor(convert);7
ResultSetHandler<List<T>> bh = new BeanListHandler<T>(type, rp);8
List<T> list = qr.query(conn, sql, bh, params);9
return list;10
}至于Z么扩展这个方法就可以实现q个逻辑得去跟源代码看它的内部实现Q用了一些JavaBean的处理和反映的技巧来做的。具体就不说?br /> ȝQcommonslg都设计得非常好，可扩展性和实用性都非常高。虽然上面D例实C转换逻辑的替换，但是仍然需要开发h员在设计数据库的时候和写JavaBean旉要严格做好规范，避免产生不必要的问题。这斚wRuby On Railsq接内部实玎ͼ动态语a的优点特别能体现Q同时强制你在设计时必须用这U方式，典型的约定优于配|原则。当Ӟ在dbutils里你愿意二种字段名都一样也无可厚非?br /> ~点QBeanProcessor是不支持兌查询的，所以上面的方式也只能局限于单表的{换，q点׃如myBatis和HibernateQ当然用q二个就引入了一些复杂性，如何权衡需要自p量，哪个用得好都一栗本人就不喜ƢmyBatis那种把SQL写到XML中的方式Q见q太复杂的SQL最l在XML里面变得面目全非Q如果是接手别h的代码，是很痛苦的，而且你无法避免只修改XML而不改JavaQ既然二者都要改Q那直接写Java里又有什么区别？单就是美。格式和注释写好一点同样很Ҏ理解Q?
]]>
1public Map<String, Object> toMap(ResultSet rs) throws SQLException {
2 Map<String, Object> result = new CaseInsensitiveHashMap();
3 ResultSetMetaData rsmd = rs.getMetaData();
4 int cols = rsmd.getColumnCount();
5
6 for (int i = 1; i <= cols; i++) {
7 result.put(rsmd.getColumnName(i), rs.getObject(i));
8 }
9
10 return result;
11 }
CaseInsensitiveHashMap是dbutils自定义的一个MapQ忽略键大小写的K-V字典Q但是key使用的是ResultSetMetaData.getColumnName()Q我想问题大概出在这里，于是认真Mjava的api文档Q开发做久了Ҏ遗忘基础Q，果然Q原?strong>getColumnName()是：获取指定列的名称Q?/span>而as关键字之后，使列名称变成用于昄的意义，q个时候应该?strong>getColumnLabel()Q?span style="color: #339966">获取用于打印输出和显C的指定列的标题。徏议标题通常?SQL public Map<String, Object> toMap(ResultSet rs) throws SQLException {2
Map<String, Object> result = new CaseInsensitiveHashMap();3
ResultSetMetaData rsmd = rs.getMetaData();4
int cols = rsmd.getColumnCount();5
6
for (int i = 1; i <= cols; i++) {7
result.put(rsmd.getColumnName(i), rs.getObject(i));8
}9
10
return result;11
}AS 子句来指定。如果未指定 SQL ASQ则?getColumnLabel q回的值将?getColumnName Ҏq回的值相?/span>。自己手动试验了一下，果然如所料，问题出在这里?br />
所以呢Q如果想要dbutils在自动{换Map及MapList时能识别聚合函数的列名，那么最好的做法是重蝲q种方式Q懒一点的Q你干脆修改上面那D代码，让它判断是否使用了as关键字。个人暂时搞不清楚官方ؓ什么没有考虑q一步，有时间再思考一下！
]]>