IBatis 的缓存机制
缓存机制，也是基于 Key-Value 的方式，确定了 Key 的来龙去脉能很好的认识缓存的生存周期。
从配置文件解析说起：
0 1 parser.addNodelet( "/sqlMap/cacheModel", new Nodelet() {
0 2 public void process(Node node) throws Exception ... {
03 Properties attributes = NodeletUtils.parseAttributes(node, state.getGlobalProps());
04 String id = state.applyNamespace(attributes.getProperty("id"));
05 String type = attributes.getProperty("type");
06 String readOnlyAttr = attributes.getProperty("readOnly");
07 Boolean readOnly = readOnlyAttr == null || readOnlyAttr.length() <= 0 ? null : new Boolean("true".equals(readOnlyAttr));
08 String serializeAttr = attributes.getProperty("serialize");
09 Boolean serialize = serializeAttr == null || serializeAttr.length() <= 0 ? null : new Boolean("true".equals(serializeAttr));
10 type = state.getConfig().getTypeHandlerFactory().resolveAlias(type);
11 Class clazz = Resources.classForName(type);
12 if (readOnly == null) ...{
13 readOnly = Boolean.TRUE;
14 }
15 if (serialize == null) ...{
16 serialize = Boolean.FALSE;
17 }
18 CacheModelConfig cacheConfig = state.getConfig().newCacheModelConfig(id, (CacheController) Resources.instantiate(clazz), readOnly.booleanValue(), serialize.booleanValue());
19 state.setCacheConfig(cacheConfig);
20 }
21 });
0 2 public void process(Node node) throws Exception ... {
03 Properties attributes = NodeletUtils.parseAttributes(node, state.getGlobalProps());
04 String id = state.applyNamespace(attributes.getProperty("id"));
05 String type = attributes.getProperty("type");
06 String readOnlyAttr = attributes.getProperty("readOnly");
07 Boolean readOnly = readOnlyAttr == null || readOnlyAttr.length() <= 0 ? null : new Boolean("true".equals(readOnlyAttr));
08 String serializeAttr = attributes.getProperty("serialize");
09 Boolean serialize = serializeAttr == null || serializeAttr.length() <= 0 ? null : new Boolean("true".equals(serializeAttr));
10 type = state.getConfig().getTypeHandlerFactory().resolveAlias(type);
11 Class clazz = Resources.classForName(type);
12 if (readOnly == null) ...{
13 readOnly = Boolean.TRUE;
14 }
15 if (serialize == null) ...{
16 serialize = Boolean.FALSE;
17 }
18 CacheModelConfig cacheConfig = state.getConfig().newCacheModelConfig(id, (CacheController) Resources.instantiate(clazz), readOnly.booleanValue(), serialize.booleanValue());
19 state.setCacheConfig(cacheConfig);
20 }
21 });
上面红色的代码很关键，它是 Statement 中，对应的Cache 的Key 就是它。返回的对象：CacheModelConfig,对应的部分代码如下：
0 1 private ErrorContext errorContext;
0 2 private CacheModel cacheModel;
0 3
0 4 CacheModelConfig(SqlMapConfiguration config, String id, CacheController controller, boolean readOnly, boolean serialize) ... {
05 this.errorContext = config.getErrorContext();
06 this.cacheModel = new CacheModel();
07 SqlMapClientImpl client = config.getClient();
08 errorContext.setActivity("building a cache model");
09 cacheModel.setReadOnly(readOnly);
10 cacheModel.setSerialize(serialize);
11 errorContext.setObjectId(id + " cache model");
12 errorContext.setMoreInfo("Check the cache model type.");
13 cacheModel.setId(id);
14 cacheModel.setResource(errorContext.getResource());
15 try ...{
16 cacheModel.setCacheController(controller);
17 } catch (Exception e) ...{
18 throw new RuntimeException("Error setting Cache Controller Class. Cause: " + e, e);
19 }
20 errorContext.setMoreInfo("Check the cache model configuration.");
21 if (client.getDelegate().isCacheModelsEnabled()) ...{
22 client.getDelegate().addCacheModel(cacheModel);
23 }
24 errorContext.setMoreInfo(null);
25 errorContext.setObjectId(null);
26 }
0 2 private CacheModel cacheModel;
0 3
0 4 CacheModelConfig(SqlMapConfiguration config, String id, CacheController controller, boolean readOnly, boolean serialize) ... {
05 this.errorContext = config.getErrorContext();
06 this.cacheModel = new CacheModel();
07 SqlMapClientImpl client = config.getClient();
08 errorContext.setActivity("building a cache model");
09 cacheModel.setReadOnly(readOnly);
10 cacheModel.setSerialize(serialize);
11 errorContext.setObjectId(id + " cache model");
12 errorContext.setMoreInfo("Check the cache model type.");
13 cacheModel.setId(id);
14 cacheModel.setResource(errorContext.getResource());
15 try ...{
16 cacheModel.setCacheController(controller);
17 } catch (Exception e) ...{
18 throw new RuntimeException("Error setting Cache Controller Class. Cause: " + e, e);
19 }
20 errorContext.setMoreInfo("Check the cache model configuration.");
21 if (client.getDelegate().isCacheModelsEnabled()) ...{
22 client.getDelegate().addCacheModel(cacheModel);
23 }
24 errorContext.setMoreInfo(null);
25 errorContext.setObjectId(null);
26 }
注意红色的代码，再来看看 CacheModel,如下：
1 
public class CacheModel implements ExecuteListener
public class CacheModel implements ExecuteListener
ExecuteListener 是一个监听器，它在 Update、insert,delete 操作执行后，触发些事件：源头如下：
0 1 com.ibatis.sqlmap.engine.mapping.statement.MappedStatement
0 2 public int executeUpdate(StatementScope statementScope, Transaction trans, Object parameterObject)
0 3 throws SQLException ... {
04 ErrorContext errorContext = statementScope.getErrorContext();
05 errorContext.setActivity("preparing the mapped statement for execution");
06 errorContext.setObjectId(this.getId());
07 errorContext.setResource(this.getResource());
08
09 statementScope.getSession().setCommitRequired(true);
10
11try ...{
12 parameterObject = validateParameter(parameterObject);
13
14 Sql sql = getSql();
15
16 errorContext.setMoreInfo("Check the parameter map.");
17 ParameterMap parameterMap = sql.getParameterMap(statementScope, parameterObject);
18
19 errorContext.setMoreInfo("Check the result map.");
20 ResultMap resultMap = sql.getResultMap(statementScope, parameterObject);
21
22 statementScope.setResultMap(resultMap);
23 statementScope.setParameterMap(parameterMap);
24
25int rows = 0;
26
27 errorContext.setMoreInfo("Check the parameter map.");
28 Object[] parameters = parameterMap.getParameterObjectValues(statementScope, parameterObject);
29
30 errorContext.setMoreInfo("Check the SQL statement.");
31 String sqlString = sql.getSql(statementScope, parameterObject);
32
33 errorContext.setActivity("executing mapped statement");
34 errorContext.setMoreInfo("Check the statement or the result map.");
35 rows = sqlExecuteUpdate(statementScope, trans.getConnection(), sqlString, parameters);
36
37 errorContext.setMoreInfo("Check the output parameters.");
38if (parameterObject != null) ...{
39 postProcessParameterObject(statementScope, parameterObject, parameters);
40 }
41
42 errorContext.reset();
43 sql.cleanup(statementScope);
44notifyListeners();
45return rows;
46 } catch (SQLException e) ...{
47 errorContext.setCause(e);
48throw new NestedSQLException(errorContext.toString(), e.getSQLState(), e.getErrorCode(), e);
49 } catch (Exception e) ...{
50 errorContext.setCause(e);
51throw new NestedSQLException(errorContext.toString(), e);
52 }
53 }
54
55 public void notifyListeners() ... {
56for (int i = 0, n = executeListeners.size(); i < n; i++) ...{
57 ((ExecuteListener) executeListeners.get(i)).onExecuteStatement(this);
58 }
59 }
com.ibatis.sqlmap.engine.mapping.statement.MappedStatement 0 2
public int executeUpdate(StatementScope statementScope, Transaction trans, Object parameterObject) 0 3
throws SQLException ... { 04
ErrorContext errorContext = statementScope.getErrorContext(); 05
errorContext.setActivity("preparing the mapped statement for execution"); 06
errorContext.setObjectId(this.getId()); 07
errorContext.setResource(this.getResource()); 08
09
statementScope.getSession().setCommitRequired(true); 10
11
try ...{ 12
parameterObject = validateParameter(parameterObject); 13
14
Sql sql = getSql(); 15
16
errorContext.setMoreInfo("Check the parameter map."); 17
ParameterMap parameterMap = sql.getParameterMap(statementScope, parameterObject); 18
19
errorContext.setMoreInfo("Check the result map."); 20
ResultMap resultMap = sql.getResultMap(statementScope, parameterObject); 21
22
statementScope.setResultMap(resultMap); 23
statementScope.setParameterMap(parameterMap); 24
25
int rows = 0; 26
27
errorContext.setMoreInfo("Check the parameter map."); 28
Object[] parameters = parameterMap.getParameterObjectValues(statementScope, parameterObject); 29
30
errorContext.setMoreInfo("Check the SQL statement."); 31
String sqlString = sql.getSql(statementScope, parameterObject); 32
33
errorContext.setActivity("executing mapped statement"); 34
errorContext.setMoreInfo("Check the statement or the result map."); 35
rows = sqlExecuteUpdate(statementScope, trans.getConnection(), sqlString, parameters); 36
37
errorContext.setMoreInfo("Check the output parameters."); 38
if (parameterObject != null) ...{ 39
postProcessParameterObject(statementScope, parameterObject, parameters); 40
} 41
42
errorContext.reset(); 43
sql.cleanup(statementScope); 44
notifyListeners(); 45
return rows; 46
} catch (SQLException e) ...{ 47
errorContext.setCause(e); 48
throw new NestedSQLException(errorContext.toString(), e.getSQLState(), e.getErrorCode(), e); 49
} catch (Exception e) ...{ 50
errorContext.setCause(e); 51
throw new NestedSQLException(errorContext.toString(), e); 52
} 53
} 54
55
public void notifyListeners() ... { 56
for (int i = 0, n = executeListeners.size(); i < n; i++) ...{ 57
((ExecuteListener) executeListeners.get(i)).onExecuteStatement(this); 58
} 59
}
再看 CacheModel 是如何处理的：
0 1 public void onExecuteStatement(MappedStatement statement) {
0 2 flush();
0 3 }
0 4
0 5 public void flush() ... {
06 synchronized (this) ...{
07controller.flush(this);
08 lastFlush = System.currentTimeMillis();
09if ( log.isDebugEnabled() ) ...{
10 log("flushed", false, null);
11 }
12 }
13 }
public void onExecuteStatement(MappedStatement statement) { 0 2
flush(); 0 3
} 0 4
0 5
public void flush() ... { 06
synchronized (this) ...{ 07
controller.flush(this); 08
lastFlush = System.currentTimeMillis(); 09
if ( log.isDebugEnabled() ) ...{ 10
log("flushed", false, null); 11
} 12
} 13
}
controller 是实现以下接口的类
public interface CacheController
实现以下几种：
FifoCacheController
LruCacheController
MemoryCacheController
OSCacheController
再看看具体的实现，如MemoryCacheController的：
private Map cache = Collections.synchronizedMap(new HashMap());
1 public void flush(CacheModel cacheModel) {
2 cache.clear();
3 }
其它的几种实现，也基本大致如此，不在具体贴代码了。
从上面的代码分析可以看出 IBatis 的缓存的更新机制大概如下：
执行 update,insert,delete 等操作时，触发对应的事件，注册的事件响应此操作，根据XML配置的缓存，对其数据做清空操作，也就是这个缓存中的数据全部清空，而不是
清空某一个 Key对应的值，即如我更新了一个 id= 5 的数据，则这个缓存中的数据全部将清空。这样倒很简单也很彻底，由此带来的问题就很多了，如果你的 update 等相关的操作太频繁了，这里的缓存则失去意义，且加大了系统本身的开销。因此，我认为对于更新不是很频繁的数据可以用 IBatis 自身的缓存机制，如果是很频繁的数据就不要使用 IBatis 自身的缓存。举例说明：
0 1 < cacheModel id ="product-cache" imlementation ="LRU" >
0 2 < flushInterval hours ="24" />
0 3 < flushOnExecute statement ="insertProduct" />
0 4 < flushOnExecute statement ="updateProduct" />
0 5 < flushOnExecute statement ="deleteProduct" />
0 6 < property name ="size" value ="1000" />
0 7 </ cacheModel >
0 8 < statement id ="getProductList" parameterClass ="int" cacheModel ="product-cache" >
0 9 select * from PRODUCT where PRD_CAT_ID = #value#
10 </ statement >
0 2 < flushInterval hours ="24" />
0 3 < flushOnExecute statement ="insertProduct" />
0 4 < flushOnExecute statement ="updateProduct" />
0 5 < flushOnExecute statement ="deleteProduct" />
0 6 < property name ="size" value ="1000" />
0 7 </ cacheModel >
0 8 < statement id ="getProductList" parameterClass ="int" cacheModel ="product-cache" >
0 9 select * from PRODUCT where PRD_CAT_ID = #value#
10 </ statement >
如果你配置了以上操作，如果做了 insert_product,updateProduct,deleteProduct 中的任何一个操作，都将清空缓存 product-cache 中的所有数据，也就是getProductList 方法将执行数据库操作，不能从缓存中获取数据。
最后一句，大家不要对 IBatis 的缓存抱太大的希望，虽然我们的系统中使用的是 IBatis。
结束，下次再说IBatis 的 SQL 的缓存机制。http://lifei114.iteye.com/blog/583178