目录
背景
思路
实现
背景
前段时间的任务中,遇到了需要识别两个对象不同属性的场景,如果使用传统的一个个属性比对equals方法,会存在大量的重复工作,而且为对象新增了属性后,比对方法也需要同步修改,不方便维护,于是便打算构造一个通用的比对方法,需要支持对象嵌套,并支持过滤比对逻辑的策略,适用大部分比对场景
思路
一个java对象,他的属性是多样的,会有集合类型(如Collection,Map,Array),基本类型(如String,int,double,long等认为是最基本对象的比较),或者对象类型,本质上是一个树状的结构
怎么去遍历对象树
常用的方法就是递归,那怎么定义递归的终点?基于上面三种类型,定为基本类型的比对为递归的终点,对于集合类型,那就再对每一个元素进行比较,对于对象类型,遍历属性进行比较
为了达成通用目的,这里使用java反射实现,基于约定和性能考虑,只通过无参的getter方法获取属性进行比较,另外如何记录当前属性所在的路径?这里参考spel表达式,相对轻量直观
实现
首先定义递归过程记录的上下文信息
这里记录了根对象,当前处理的对象,以及反射获取的对象信息
protected static class CompareContext {//对比结果private List<CompareResult.CompareInfo> resCollect;//当前遍历的对象信息,用于定位private ArrayDeque<String> pathMessageStack;protected Object rootSourceObj;protected Object rootOtherObj;protected Object sourceObj;protected Object otherObj;protected Object sourceInvokeVal;protected Object otherInvokeVal;//过滤耗费时间private long filterCost;private String checkCycle() {Set<String> set = new LinkedHashSet<>();//出现重复退出while (set.add(pathMessageStack.removeLast())) {}String[] elements = new String[set.size()];Iterator<String> iterator = set.iterator();int index = 0;while (iterator.hasNext()) {elements[set.size() - 1 - index++] = iterator.next();}return getPath(elements);}protected String getPath(String[] elements) {Object obj = this.rootSourceObj == null ? this.rootOtherObj : this.rootSourceObj;String simpleName = obj.getClass().getSimpleName();StringBuilder builder = new StringBuilder(simpleName);if (elements == null) {elements = this.pathMessageStack.toArray(new String[0]);}for (int i = elements.length - 1; i >= 0; i--) {String cur = elements[i];String value = cur.substring(2);if (cur.startsWith(FIELD_SIGN)) {builder.append(".").append(value);} else {builder.append("[").append(value).append("]");}}return builder.toString();}}
基于这个上下文,定义过滤接口,因为有些属性可能不用比对,如果命中过滤接口,则跳过
@FunctionalInterface
public interface MethodFilter {boolean isSkipCompare(Method method, String methodKey, CompareUtil.CompareContext compareContext);
}
比对结果类,主要存放每个比对不通过属性的值及其路径信息
@Getter
@ToString
public class CompareResult {protected boolean same;protected Object source;protected Object other;protected List<CompareInfo> diffList;//=====额外信息,可以不关注====//整体对比耗时private long totalCostMill;//方法过滤耗时protected long methodFilterCostMill;protected void genCost(long start) {this.totalCostMill = System.currentTimeMillis() - start;}@Getterpublic static class CompareInfo {protected Object sourceVal;protected Object otherVal;protected String path;protected boolean isShow() {if (sourceVal == null || otherVal == null) {return true;}if (CompareUtil.getObjType(sourceVal) != Object.class) {return true;}if (sourceVal instanceof Collection) {return ((Collection<?>) sourceVal).size() != ((Collection<?>) otherVal).size();}if (sourceVal instanceof Map) {return ((Map<?, ?>) sourceVal).size() != ((Map<?, ?>) otherVal).size();}if (sourceVal.getClass().isArray()) {return ((Object[]) sourceVal).length != ((Object[]) otherVal).length;}return false;}@Overridepublic String toString() {return String.format("path:%s,source:[%s],other:[%s]", path, sourceVal, otherVal);}}public String getBaseObjDiffInfo() {return getBaseObjDiffInfo("\n");}/*** 过滤出为空的父对象或基本对象,原生的diffList会包含父对象,不方便查看** @param seperator* @return*/public String getBaseObjDiffInfo(String seperator) {if (!same) {StringBuilder builder = new StringBuilder();diffList.stream().filter(CompareInfo::isShow).forEach(v -> builder.append(v).append(seperator));return builder.toString();}return "";}}
随后定义入口方法
public static <T> CompareResult compareObjByGetter(T source, T other, MethodFilter... methodFilters) {List<CompareResult.CompareInfo> diffList = new ArrayList<>();CompareContext compareContext = new CompareContext();compareContext.resCollect = diffList;compareContext.pathMessageStack = new ArrayDeque<>(MAX_DEEP_SIZE);compareContext.rootSourceObj = source;compareContext.rootOtherObj = other;long start = System.currentTimeMillis();boolean res = compareObjByGetter(source, other, compareContext, methodFilters);CompareResult compareResult = new CompareResult();compareResult.genCost(start);compareResult.diffList = diffList;compareResult.same = res;compareResult.other = other;compareResult.source = source;compareResult.methodFilterCostMill = compareContext.filterCost;return compareResult;}
主流程方法,流程如下
- 校验
- 路径长度校验
- 类校验
- 递归终点,对于基本类型,直接走比对方法,并生成路径
- 对于复合类型,递归进行比对逻辑
- 对于对象类型,采用反射获取具体值,并通过方法过滤进行比较
- 首先判断是否有public的无参数getter方法
- 有的话采用调用该方法获取两边的值,设置上下文信息
- 过滤器过滤无需比对的字段
- 如果任一方为空,构造结果
- 对于集合类型的属性,需要进一步获取里面的元素进行递归处理
- 对于对象类型或基本类型,重复上述步骤
主流程代码如下
private static <T> boolean compareObjByGetter(T source, T other, CompareContext compareContext, MethodFilter... methodFilters) {//对比路径太深或出现循环引用不支持if (compareContext.pathMessageStack.size() > MAX_DEEP_SIZE) {String path = compareContext.checkCycle();if (!StringUtils.isEmpty(path)) {//路径仅供参考,不一定准确throw new IllegalStateException(String.format("reference cycle happen,please check your object,path:%s", path));}throw new IllegalStateException(String.format("compare path over max size:%s,please check your object", MAX_DEEP_SIZE));}if(source==other){return true;}if (source == null || other == null) {generateResult(source, other, compareContext);return false;}if (!source.getClass().equals(other.getClass())) {throw new IllegalArgumentException(String.format("not the same object,class source:%s,class other:%s,path:%s", source.getClass(), other.getClass(),compareContext.getPath(null)));}//基本类型不再对比getter方法if (getObjType(source) != Object.class) {boolean isSame = compareBaseObj(source, other);if (!isSame) {generateResult(source, other, compareContext);}return isSame;}//复合类型if (isCollectionOrMapOrArray(source)) {return dealWithCMA(source, other, compareContext, methodFilters);}//对象类型,遍历对应的方法final boolean[] val = new boolean[]{true};ReflectionUtils.doWithMethods(source.getClass(), new ReflectionUtils.MethodCallback() {@Overridepublic void doWith(Method method) throws IllegalArgumentException, IllegalAccessException {String name = method.getName();if (method.getModifiers() == Modifier.PUBLIC && (name.startsWith("get") || name.startsWith("is"))) {//有入参的getter不处理if (method.getParameterTypes().length != 0) {return;}String methodKey = method.getDeclaringClass().getName() + "#" + method.getName();Object sourceInvokeVal = null;Object otherInvokeVal = null;try {sourceInvokeVal = method.invoke(source);otherInvokeVal = method.invoke(other);} catch (Exception e) {throw new RuntimeException(e);}compareContext.otherObj = other;compareContext.otherInvokeVal = otherInvokeVal;compareContext.sourceObj = source;compareContext.sourceInvokeVal = sourceInvokeVal;//过滤,methodFilter是||关系long start = System.currentTimeMillis();try {for (MethodFilter methodFilter : methodFilters) {if (methodFilter.isSkipCompare(method, methodKey, compareContext)) {return;}}} finally {compareContext.filterCost += System.currentTimeMillis() - start;}if (sourceInvokeVal == null && otherInvokeVal == null) {return;}compareContext.pathMessageStack.push(String.format("%s%s", FIELD_SIGN, ColumnSelector.getFieldName(method.getName(), method.getDeclaringClass().getName())));if (sourceInvokeVal == null || otherInvokeVal == null) {generateResult(sourceInvokeVal, otherInvokeVal, compareContext);val[0] = false;compareContext.pathMessageStack.pop();return;}if (isCollectionOrMapOrArray(sourceInvokeVal)) {val[0] &= dealWithCMA(sourceInvokeVal, otherInvokeVal, compareContext, methodFilters);} else {//对象类型 or 基本类型val[0] &= compareObjByGetter(sourceInvokeVal, otherInvokeVal, compareContext, methodFilters);}compareContext.pathMessageStack.pop();}}});return val[0];}
对于集合类型的处理如下
主要根据各自的特性取出元素,再给到上面的主流程去执行
private static boolean dealWithCMA(Object sourceObj, Object otherObj, CompareContext compareContext, MethodFilter... methodFilters) {if(sourceObj==otherObj){return true;}boolean isDiff = true;if (sourceObj instanceof Collection) {Collection<?> sourceCollection = ((Collection<?>) sourceObj);Collection<?> otherCollection = ((Collection<?>) otherObj);//要求顺序,这里不做排序if (sourceCollection.size() != otherCollection.size()) {isDiff = false;} else {Iterator<?> sourceI = sourceCollection.iterator();Iterator<?> otherI = otherCollection.iterator();int index = 0;while (sourceI.hasNext()) {Object sourceElement = sourceI.next();Object otherElement = otherI.next();//下一层不匹配的值compareContext.pathMessageStack.push(String.format("%s%s", COLLECTION_SIGN, index++));isDiff &= compareObjByGetter(sourceElement, otherElement, compareContext, methodFilters);compareContext.pathMessageStack.pop();}}}if (sourceObj.getClass().isArray()) {Object[] sourceArray = (Object[]) sourceObj;Object[] otherArray = (Object[]) otherObj;if (sourceArray.length != otherArray.length) {isDiff = false;} else {for (int i = 0; i < sourceArray.length; i++) {Object sourceElement = sourceArray[i];Object otherElement = otherArray[i];compareContext.pathMessageStack.push(String.format("%s%s", ARRAY_SIGN, i));isDiff &= compareObjByGetter(sourceElement, otherElement, compareContext, methodFilters);compareContext.pathMessageStack.pop();}}}if (sourceObj instanceof Map) {Map<?, ?> sourceMap = (Map) sourceObj;Map<?, ?> otherMap = (Map) otherObj;if (sourceMap.size() != otherMap.size()) {isDiff = false;} else {HashSet<?> otherKeySet = new HashSet<>(otherMap.keySet());for (Map.Entry<?, ?> entry : sourceMap.entrySet()) {Object sourceKey = entry.getKey();Object otherVal = otherMap.get(sourceKey);otherKeySet.remove(sourceKey);compareContext.pathMessageStack.push(String.format("%s%s", KEY_SIGN, sourceKey));isDiff &= compareObjByGetter(entry.getValue(), otherVal, compareContext, methodFilters);compareContext.pathMessageStack.pop();}if (!otherKeySet.isEmpty()) {for (Object otherKey : otherKeySet) {compareContext.pathMessageStack.push(String.format("%s%s", KEY_SIGN, otherKey));isDiff &= compareObjByGetter(null, otherMap.get(otherKey), compareContext, methodFilters);compareContext.pathMessageStack.pop();}}}}if (!isDiff) {generateResult(sourceObj, otherObj, compareContext);}return isDiff;}
对于基本对象判断,如果返回object类型需要进一步解析判断,后续如果有其他基本对象,比如业务自定义的基本对象也可以往里面加,或者再进行优化,通过本地线程的方式动态指定基本对象类型
/*** 后续比对有其他基本类型可以往里加* @param obj* @return*/protected static Class<?> getObjType(Object obj) {if (obj instanceof Integer) {return Integer.class;} else if (obj instanceof String) {return String.class;} else if (obj instanceof BigDecimal) {return BigDecimal.class;} else if (obj instanceof Long) {return Long.class;} else if (obj instanceof Enum) {return Enum.class;} else if (obj instanceof Double) {return Double.class;} else if (obj instanceof Boolean) {return Boolean.class;}return Object.class;}
对于基本类型的比较方法
private static <T> boolean compareBaseObj(T obj, T other) {//同一个对象直接通过if(obj==other){return true;}//数字类型if (obj instanceof Number && obj instanceof Comparable) {return ((Comparable<T>) obj).compareTo(other) == 0;}//其他类型return Objects.equals(obj, other);}
效果
Java对象比对功能实现大概如此,来看看效果
定义了比对对象
@Getterpublic static class ComposeObj {private String id;private Object[] arrays;private List<Object> list;private Map<String, Object> map;private ComposeObj son;private int idLength;public ComposeObj(String id) {this.id = id;this.idLength = id.length();}}
测试代码
ComposeObj sourceObj = new ComposeObj("source");ComposeObj arrayObj = new ComposeObj("A1");sourceObj.arrays = new Object[]{arrayObj, new ComposeObj("A22")};sourceObj.list = Arrays.asList(new ComposeObj("C1"), new ComposeObj("C11"));Map<String, Object> map = new HashMap<>();map.put("test", "test");map.put("test2", "test2");sourceObj.map = map;ComposeObj otherObj = new ComposeObj("other");ComposeObj arrayObj2 = new ComposeObj("A2");otherObj.arrays = new Object[]{arrayObj2, new ComposeObj("A22")};otherObj.list = Arrays.asList(new ComposeObj("C2"), new ComposeObj("C11"));Map<String, Object> map2 = new HashMap<>();map2.put("test", "test2");map2.put("test2", "test22");otherObj.map = map2;
测试方法
CompareResult compareResult = CompareUtil.compareObjByGetter(sourceObj, otherObj);assert checkPath(compareResult);assert compareResult.getDiffList().size() == 9;log.info(compareResult.getBaseObjDiffInfo());
结果
默认的支持的表达式过滤器如下
/*** 通过简单的表达式* "*abc|xxx" ->标识以abc结尾或xxx的属性忽略,不做比对,*代表通配符*/
public class RegexFieldFilter implements MethodFilter {private static final Cache<String, Pattern> REGEX_MAP = CacheBuilder.newBuilder().softValues().maximumSize(2048).build();private final List<String> rules;private static final String WILDCARD = "*";//是否缓存结果,如果调用次数较多(比如属性为list,且实际场景数量较多)可以用启用private boolean cacheResult = false;//方法keyprivate Cache<String, Boolean> resultMap;private final Set<String> equalsField = new HashSet<>();private final Map<String, String> prefixFieldMap = new HashMap<>();private final Map<String, String> suffixFieldMap = new HashMap<>();private RegexFieldFilter(String regex, boolean cacheResult) {this.cacheResult = cacheResult;rules = Splitter.on("|").splitToList(regex);for (String rule : rules) {//普通比对if (!rule.contains(WILDCARD)) {equalsField.add(rule);}//首尾通配符特殊逻辑if (onlyOneWildcard(rule)) {if (rule.startsWith(WILDCARD)) {suffixFieldMap.put(rule, rule.substring(1));}if (rule.endsWith(WILDCARD)) {prefixFieldMap.put(rule, rule.substring(0, rule.length() - 1));}}}if (cacheResult) {resultMap = CacheBuilder.newBuilder().softValues().maximumSize(1024).build();}}public static RegexFieldFilter of(String regex, boolean cacheResult) {return new RegexFieldFilter(regex, cacheResult);}public static RegexFieldFilter of(String regex) {return of(regex, false);}private boolean canSkip(String rule, String fieldName) {if (rule.contains(WILDCARD)) {if (suffixFieldMap.containsKey(rule)) {return fieldName.endsWith(suffixFieldMap.get(rule));}if (prefixFieldMap.containsKey(rule)) {return fieldName.startsWith(prefixFieldMap.get(rule));}//在中间或多个通配符String replace = StringUtils.replace(rule, WILDCARD, ".*");Pattern pattern = REGEX_MAP.asMap().computeIfAbsent(replace, Pattern::compile);return pattern.matcher(fieldName).matches();}return equalsField.contains(fieldName);}private boolean onlyOneWildcard(String rule) {if (!rule.contains(WILDCARD)) {return false;}return rule.indexOf(WILDCARD, rule.indexOf(WILDCARD) + 1) == -1;}@Overridepublic boolean isSkipCompare(Method method, String methodKey, CompareUtil.CompareContext context) {return cacheResult ? resultMap.asMap().computeIfAbsent(methodKey, s -> judgeSkip(method)) : judgeSkip(method);}private boolean judgeSkip(Method method) {if (!CollectionUtils.isEmpty(rules)) {String name = method.getName();String fieldName = CompareUtil.getFieldName(name, method.getDeclaringClass().getCanonicalName());for (String rule : rules) {if (canSkip(rule, fieldName)) {return true;}}}return false;}}
测试代码如下,定义这个过滤会把属性名为arrays,带有'l',或者ma开头的属性都过滤掉,不参与比对逻辑
boolean cacheResult = false;RegexFieldFilter of = RegexFieldFilter.of("arrays|*l*|ma*", cacheResult);CompareResult compareResult = CompareUtil.compareObjByGetter(source, other, of);assert checkPath(compareResult);assert compareResult.getDiffList().size() == 15;
ComopareUtil完整代码
https://github.com/97lele/redis-aux/tree/dev/common/src/main/java/com/xl/redisaux/common/utils/compare