重构之美之一改造面向过程式设计
使用面向对象语言进行过程式设计的例子，可谓俯拾皆是。看这段代码：
public class SyncExecutor {
public void executeSync() {
syncSchools();
syncGrades();
syncFaculties();
}
}
这段代码很清晰，分别执行了对学校、年级与教师信息的同步。一目了然，似乎没有什么问题。然而，如果深入阅读各个同步子方法，就会发现某种坏味道，那就是重复代码。
private void syncSchools() {
List<School> sourceSchools = getSourceSchools();
List<School> targetSchools = new ArrayList<School>();
List<String> sourceSchoolCodes = getSchoolCodes(sourceSchools);
Map<String,School> targetSchoolWithCodeMapping = schoolService.getSchoolWithCodeMapping(sourceSchoolCodes);
for (School sourceSchool:sourceSchools) {
String schoolCode = sourceSchool.getSchoolCode();
School targetSchool = targetSchoolWithCodeMapping.get(schoolcode);
if (targetSchool == null) {
targetSchool = new School(
sourceSchool.getSchoolCode(),
sourceSchool.getSchoolName(),
sourceSchool.getProvinceCode(),
sourceSchool.getSchoolAddress(),
sourceSchool.getSchoolZip(),
sourceSchool.getSchoolTel());
} else if (isCover) {
targetSchool.setSchoolCode(sourceSchool.getSchoolCode());
targetSchool.setSchoolName(sourceSchool.getSchoolName());
targetSchool.setProvinceCode(sourceSchool.getProvinceCode());
targetSchool.setSchoolAddress(sourceSchool.getSchoolAddress());
targetSchool.setSchoolZip(sourceSchool.getSchoolZip());
targetSchool.setSchoolTel(sourceSchool.getSchoolTel());
}
targetSchools.add(targetSchool);
}
syncService.saveOrUpdate(targetSchools);
}
private void syncGrades() {
List<Grade> sourceGrades = getSourceGrades();
List<Grade> targetGrades = new ArrayList<Grade>();
List<String> sourceGradeCodes = getGradeCodes(sourceGrades);
Map<String,Grade> targetGradeWithCodeMapping = gradeService.getGradeWithCodeMapping(sourceGradeCodes);
for (Grade sourceGrade:sourceGrades) {
String gradeCode = sourceGrade.getGradeCode();
Grade targetGrade = targetGradeWithCodeMapping.get(gradeCode);
if (targetGrade == null) {
targetGrade = new Grade(
sourceGrade.getGradeCode(),
sourceGrade.getName(),
sourceGrade.getEntranceDay(),
sourceGrade.getGraduateDay(),
sourceGrade.getSchoolCode(),
sourceGrade.getSchoolProperty());
} else if (isCover) {
targetGrade.setGradeCode(sourceGrade.getGradeCode());
targetGrade.setName(sourceGrade.getName());
targetGrade.setEntranceDay(sourceGrade.getEntranceDay());
targetGrade.setGraduateDay(sourceGrade.getGraduateDay());
targetGrade.setSchoolCode(sourceGrade.getSchoolCode());
targetGrade.setSchoolProperty(sourceGrade.getSchoolProperty());
}
targetGrades.add(targetGrade);
}
syncService.saveOrUpdate(targetGrades);
}
当然，真实的代码更加复杂与混乱，但如果经过一系列重构，例如Rename Method，Extract Method之后，就会变得逐渐清晰，大体结构如上述展示的代码。阅读这样的代码，是否发现各个同步子方法均有似曾相识的感觉呢？究其原因，在于同步的执行逻辑大体相似，换言之，它们具有相似的模板。我们需要改善其结构，实现代码的重用。然而，在方法层面上，我们已很难实现这一点。事实上，当我们在编写同步方法时，已经落入了过程式设计的窠臼。我们首先想到的是执行的过程，而非对象。现在，我们需要将这些执行过程封装为对象，充分地利用继承等机制实现类级别的重用。显然，这里可以运用Form Template Method重构。当然，在此之前，我们还需要运用Extract Superclass，对School、Grade等类进行一系列重构，例如为它们建立共同的父类Entity，提供getCode()方法。并运用Rename Method，将原来各自实体类的相关方法，例如getSchoolCode()、getGradeCode()等，更名为getCode()。
现在，我们需要为同步操作定义一个共同的抽象类DataSynchronizer，然后利用Move Method重构，将原有SyncExecutor的相关代码搬移到DataSynchronizer中：
public abstract class DataSynchronizer {
public void execute() {
List<Entity> sourceEntities = getSourceEntities();
List<Entity> targetEntities = new ArrayList<Entity>();
List<String> sourceEntityCodes = getEntityCodes(sourceEntities);
Map<String,Entity> targetEntityWithCodeMapping = getEntityWithCodeMapping(sourceEntityCodes);
for (Entity sourceEntity:sourceEntities) {
String entityCode = sourceEntity.getCode();
Entity targetEntity = targetEntityWithCodeMapping.get(entityCode);
if (targetEntity == null) {
targetGrade = createEntity(sourceEntity);
} else if (isCover) {
updateEntity(targetEntity,sourceEntity);
}
targetEntities.add(targetEntity);
}
syncService.saveOrUpdate(targetEntities);
}
protected abstract List<Entity> getSourceEntities();
protected abstract List<String> getEntityCodes(List<Entity> entities);
protected abstract Map<String,Entity> getEntityWithCodeMapping(List<String> entityCodes);
protected abstract Entity createEntity(Entity sourceEntity);
protected abstract void updateEntity(Entity target, Entity source);
}
注意，在获得Entity与Code的Map对象时，我对原有的代码实现进行了封装，因为不同的实体同步类，所要调用的Service对象是不一样的。因此，需要将调用Service相关方法的实现留给子类。现在，只需要定义各个同步类继承DataSynchronizer，重写相关的受保护抽象方法即可：
public class SchoolSynchronizer extends DataSynchronizer{}
public class GradeSynchronizer extends DataSynchronizer{}
public class FacultySynchronizer extends DataSynchronizer{}
接着，修改SyncExecutor类的实现。为方便调用同步子类的相关方法，我定义了一个Factory Method：
public class SyncExecutor {
public void executeSync() {
for (DataSynchronizer dataSync:createSynchronizers()) {
dataSync.execute();
}
}
protected List<DataSynchronizer> createSynchronizers() {
List< DataSynchronizer> synchronizers =
new ArrayList< DataSynchronizer();
synchronizers.add(new SchoolSynchronizer());
synchronizers.add(new GradeSynchronizer());
synchronizers.add(new FacultySynchronizer());
return synchronizers;
}
}
以真正面向对象的方式来完成上述功能，无论在代码结构、重用性还是扩展性方面，比诸之前的实现，都有了长足的改善。这就是面向对象设计的优雅之处。
纵观整个重构过程，实际上，我在运用Convert Procedural Design to Objects重构时，大量运用了Rename Method、Extract Method、Move Method、Extract Superclass、Form Template Method等重构手法。这是合乎常情的。当我们在对程序进行重构时，往往需要运用各种重构手法，才能达到最终的重构目的。对于大型重构而言，这种特征尤其明显。
