基于HTTP的多文件上传问题

企业应用中会经常用到基于Http的文件上传，其中Multipart是多文件上传所使用的，在用JAVA开发时有时会用第三方类库来拼HTTP报文，有时则手动组装，

然后手动组装前先要了解一下报文的格式，如下是HTTP的报文头：

POST /test/upload HTTP/1.1Content-Type: multipart/form-data;boundary=0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64charset: utf-8Content-Length: 19520216Host: 10.19.220.234:8080Connection: Keep-AliveUser-Agent: Apache-HttpClient/4.2.5 (java 1.5)

重点是：

Content-Type: multipart/form-data;boundary=0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64

其中Content-Type决定了报文的类型，boundary则用于分隔各条目项。

示例如下：

--0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64Content-Disposition: form-data; name="messageId"Content-Type: text/plain; charset=US-ASCIIContent-Transfer-Encoding: 8bit1395271935--0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64Content-Disposition: form-data; name="name"Content-Type: text/plain; charset=US-ASCIIContent-Transfer-Encoding: 8bitimage--0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64Content-Disposition: form-data; name="type"Content-Type: text/plain; charset=US-ASCIIContent-Transfer-Encoding: 8bit3--0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64Content-Disposition: form-data; name="content"; filename="11.mp4"Content-Type: application/octet-streamContent-Transfer-Encoding: binary....ftypisom....isomavc1..4.moov...lmvhd.....1...1.....X........(省略文件)--0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64Content-Disposition: form-data; name="vimg"; filename="test.jpg"Content-Type: application/octet-streamContent-Transfer-Encoding: binary......JFIF.............C................................. $.' ",#..(7),01444.'9=82<.342...C........(省略文件)--0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64--

对于multipart的解析我用的是apache的commons-fileupload-1.3.jar，demo如下：

boolean isHaveData = ServletFileUpload.isMultipartContent(request);if (isHaveData) {FileItemFactory factory = new DiskFileItemFactory();ServletFileUpload upload = new ServletFileUpload(factory);try {List<?> items = upload.parseRequest(request);Iterator<?> iter = items.iterator();while (iter.hasNext()) {FileItem item = (FileItem) iter.next();if (item.isFormField()) {// 普通文本信息处理String paramName = item.getFieldName();String paramValue = item.getString();map.put(paramName, paramValue);} else {// 上传文件信息处理if("vimage".equalsIgnoreCase(item.getName()) || "vimg".equalsIgnoreCase(item.getFieldName())){vimg = item.get();} else {data = item.get();}}}} catch (FileUploadException e) {logger.error("FileUploadException : ", e);}}

其中判断类型的代码如下：

    public static final String MULTIPART = "multipart/";    public static final boolean isMultipartContent(RequestContext ctx) {        String contentType = ctx.getContentType();        if (contentType == null) {            return false;        }        if (contentType.toLowerCase(Locale.ENGLISH).startsWith(MULTIPART)) {            return true;        }        return false;    }

有一点需要注意的是，upload.parseRequest(request)方法里的源码是这样的：

                boolean nextPart;                if (skipPreamble) {                    nextPart = multi.skipPreamble();                } else {                    nextPart = multi.readBoundary();                }                if (!nextPart) {                    if (currentFieldName == null) {                        // Outer multipart terminated -> No more data                        eof = true;                        return false;                    }                    // Inner multipart terminated -> Return to parsing the outer                    multi.setBoundary(boundary);                    currentFieldName = null;                    continue;                }

也就是说如果两个boundary之间是空的，就认为到eof了，也就是结束了往下的遍历，因为在实际的使用当中，我在拼报文时不小心在两个条项目中多加了一个boundary，这时后面的条项目就取不到了，这个问题让我纠结了半天，觉得很诡异，后来读源码才发现，原来如此。

如果使用第三方的类库就不会有此烦恼了，我用的是apache的httpmime-4.2.2.jar，测试的代码如下：

        MultipartEntity entity = new MultipartEntity(null, "0xKhTmLbOuNdArY-35123DE7-B577-4495-AA1E-092BB0CCFC64", null);        File file = new File("E:\\ceshi\\11.flv");        File file2 = new File("E:\\ceshi\\test.jpg");        entity.addPart("messageId", new StringBody("1395271935"));        entity.addPart("name", new StringBody("image"));        entity.addPart("type", new StringBody("3"));        entity.addPart("to", new StringBody("6000028403"));        entity.addPart("from", new StringBody("6000608066"));        entity.addPart("content", new FileBody(file));        entity.addPart("vimage", new FileBody(file2));        httpost.setEntity(entity);

测试的过程中我用的抓包工具是Wireshark，界面如下：

                                                                       图1--Wireshark界面

查看到的TCP流如下：

                                                                                   图2--TCP流

最后总结一下：multipart/form-data对于多文件上传还是很好用的