下面的类是一个老系统的代码,如今放到sonar上面进行扫描,扫出来的结果发现复杂度超过了30。java
代码复杂度是指代码中的分支数量,好比有一个if分支,代码复杂度就加1,若是if中有“||”或者“&&”那么代码复杂度就加2,for和while同理。通常复杂度超过10的类就算是比较复杂的了,而这个类的复杂度居然达到了30,代码的糟糕程度可见一斑,如今咱们就来重构一下这个类的代码。git
原始文件在这里。
重构开始吧!github
多处String类型非空判断
1 2 3 4 5 6 |
if (StringUtil.isEmpty(username)) throw new ICRClientException("username can not be null"); if (StringUtil.isEmpty(password)) throw new ICRClientException("password can not be null"); if (udto == null) throw new ICRClientException("ICRUploadDTO can not be null"); |
重构以后:apache
1 2 3 4 5 6 7 8 9 10 11 |
//将原来的地方替换为 checkStringParamEmpty(username, "username"); checkStringParamEmpty(password, "password"); checkStringParamEmpty(udto.getUrlPath(), "urlPath"); ... //新增一个方法 private void checkStringParamEmpty(String value, String name) throws ICRClientException { if (StringUtil.isEmpty(value)) { throw new ICRClientException(name + " can not be null"); } } |
原代码中不止这3个参数的校验,还有不少,越多参数的校验,咱们重构后的复杂度就会越低。post
代码复杂度变化:原来是3,修改后为1。单元测试
多String值判断
1 2 3 |
if (!udto.getPriority().equals("0") && !udto.getPriority().equals("1") && !udto.getPriority().equals("2") && !udto.getPriority().equals("3")) throw new ICRClientException("priority must be 0/1/2/3"); |
重构以后:测试
1 2 3 4 5 6 7 8 9 |
//将原来代码替换为 checkValueWithinList(udto.getPriority()); ... //新增一个方法: private void checkValueWithinList(String priority) throws ICRClientException { if (!Arrays.asList("0", "1", "2", "3").contains(priority)) { throw new ICRClientException("priority must be 0/1/2/3"); } } |
代码复杂度变化:原来是4,修改后为1。url
对list的非空判断
1 2 |
if (list == null || list.size() == 0) throw new ICRClientException("list can not be null"); |
重构以后:spa
1 2 3 4 5 6 7 |
//将原来的代码替换为 checkValueWithinList(udto.getPriority()); ... //新增一个方法 private void checkListNoNull(List list) throws ICRClientException { if (list.isEmpty()) throw new ICRClientException("list can not be null"); } |
代码复杂度变化:原来是2,修改后为1。code
多个catch的内容相同
1 2 3 4 5 6 7 8 |
int code = 0; try { code = httpClient.executeMethod(post); } catch (HttpException e) { throw new ICRClientException(e.getMessage(), e); } catch (IOException e) { throw new ICRClientException(e.getMessage(), e); } |
重构以后:
1 2 3 4 5 6 7 8 9 10 11 12 13 |
//将原来的地方替换为 int code = executeHttpClient(httpClient, post); ... //新增一个方法 private int executeHttpClient(HttpClient httpClient, PostMethod post) throws ICRClientException { int code; try { code = httpClient.executeMethod(post); } catch (Exception e) { throw new ICRClientException(e.getMessage(), e); } return code; } |
代码复杂度变化:原来是2,修改后为1。
if判断结果复杂化
1 2 3 4 5 6 7 8 9 10 11 12 13 14 |
if (code == 200) { try { if (post.getResponseBodyAsString().equals("ok")) { return true; } } catch (IOException e) { throw new ICRClientException(e.getMessage(), e); } return false; } else if (code == 500) { throw new ICRClientException(post.getResponseBodyAsString()); } else { throw new ICRClientException(code + ":" + post.getStatusText()); } |
重构以后:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 |
//将原来代码替换为 return returnFinialResult(post, code); ... //新增一个方法 private boolean returnFinialResult(PostMethod post, int code) throws ICRClientException, IOException { if (code == 500) throw new ICRClientException(post.getResponseBodyAsString()); if (code != 200) throw new ICRClientException(code + ":" + post.getStatusText()); try { return post.getResponseBodyAsString().equals("ok"); } catch (IOException e) { throw new ICRClientException(e.getMessage(), e); } } |
代码复杂度变化:原来是4,修改后为3。
本地变量始终不为null
1 2 3 4 5 6 7 8 9 10 11 12 13 14 |
public boolean uploadToICR(String username, String password, ICRUploadDTO udto) throws ICRClientException { HttpClient httpClient = null; PostMethod post = null; httpClient = new HttpClient(); //some code here … } finally { if (post != null) { post.releaseConnection(); } if (httpClient != null) { httpClient.getHttpConnectionManager().closeIdleConnections(0); } } |
重构以后:
1 2 3 4 5 6 7 8 9 10 11 |
public boolean uploadToICR(String username, String password, ICRUploadDTO udto) throws ICRClientException { HttpClient httpClient = new HttpClient(); PostMethod post = null; //some code here … } finally { if (post != null) { post.releaseConnection(); } } } |
代码复杂度变化:原来是1,修改后为0。
读取IO流的方法,为何要本身实现?
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 |
private byte[] readData(InputStream ins) throws IOException { byte[] buf = new byte[2048]; int count = 0; int len = 0; byte data[] = new byte[2048]; byte[] result = null; try { while ((len = ins.read(data, 0, 2048)) != -1) { int newcount = count + len; if (newcount > buf.length) { byte newbuf[] = new byte[Math .max(buf.length << 1, newcount)]; System.arraycopy(buf, 0, newbuf, 0, count); buf = newbuf; } System.arraycopy(data, 0, buf, count, len); count = newcount; } result = new byte[count]; System.arraycopy(buf, 0, result, 0, count); } finally { ins.close(); } return result; } |
在原代码里面本身实现了一个对读取IO流字节的方法,这个能够使用apache-io或者guava的API代替:
1 2 3 4 |
//使用apache io API的实现: byte[] bytes = IOUtils.toByteArray(inputStream); //使用guava API的实现: byte[] bytes1 = ByteStreams.toByteArray(inputStream); |
代码复杂度变化:原来是不少,修改后为0。
最终重构后的版本见这里,最后的代码复杂度从原来的30降到了3。 代码写的比较仓促,没有写单元测试,其实最好的作法是在重构以前先写好单元测试,而后再慢慢修改原来的代码,每修改一处地方跑一遍单元测试,这样能够保证你的重构没有破坏原来的代码逻辑。