一般来说,通过c.Request.FormFile()获取文件的时候,所有内容都全部读到了内存。如果是个巨大的文件,则可能内存会爆掉;且,有的时候我们需要一边上传一边处理。
以下的代码实现了大文件流式上传。
还非常不完美,但是可以作为参考:
upload.html
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>upload file</title>
</head>
<body>
<form method="post" enctype="multipart/form-data" action="/gin_upload">
<input type="file" name="ff" multiple="multiple"/><br/>
<input type="submit" value="提交"/>
</form>
</body>
gin_stream_upload_file.go
/*
本例子实现了gin框架下的多个大文件流式上传,避免了文件内容存在内存而无法支持大文件的情况
*/
package main
import (
"fmt"
"github.com/gin-gonic/gin"
"os"
"bytes"
"io"
"log"
"strconv"
"strings"
)
/// 解析多个文件上传中,每个具体的文件的信息
type FileHeader struct{
ContentDisposition string
Name string
FileName string ///< 文件名
ContentType string
ContentLength int64
}
/// 解析描述文件信息的头部
/// @return FileHeader 文件名等信息的结构体
/// @return bool 解析成功还是失败
func ParseFileHeader(h []byte) (FileHeader, bool){
arr := bytes.Split(h, []byte("
"))
var out_header FileHeader
out_header.ContentLength = -1
const (
CONTENT_DISPOSITION = "Content-Disposition: "
NAME = "name=""
FILENAME = "filename=""
CONTENT_TYPE = "Content-Type: "
CONTENT_LENGTH = "Content-Length: "
)
for _,item := range arr{
if bytes.HasPrefix(item, []byte(CONTENT_DISPOSITION)){
l := len(CONTENT_DISPOSITION)
arr1 := bytes.Split(item[l:], []byte("; "))
out_header.ContentDisposition = string(arr1[0])
if bytes.HasPrefix(arr1[1], []byte(NAME)){
out_header.Name = string(arr1[1][len(NAME):len(arr1[1])-1])
}
l = len(arr1[2])
if bytes.HasPrefix(arr1[2], []byte(FILENAME)) && arr1[2][l-1]==0x22{
out_header.FileName = string(arr1[2][len(FILENAME):l-1])
}
} else if bytes.HasPrefix(item, []byte(CONTENT_TYPE)){
l := len(CONTENT_TYPE)
out_header.ContentType = string(item[l:])
} else if bytes.HasPrefix(item, []byte(CONTENT_LENGTH)){
l := len(CONTENT_LENGTH)
s := string(item[l:])
content_length,err := strconv.ParseInt(s, 10, 64)
if err!=nil{
log.Printf("content length error:%s", string(item))
return out_header, false
} else {
out_header.ContentLength = content_length
}
} else {
log.Printf("unknown:%s
", string(item))
}
}
if len(out_header.FileName)==0{
return out_header,false
}
return out_header,true
}
/// 从流中一直读到文件的末位
/// @return []byte 没有写到文件且又属于下一个文件的数据
/// @return bool 是否已经读到流的末位了
/// @return error 是否发生错误
func ReadToBoundary(boundary []byte, stream io.ReadCloser, target io.WriteCloser)([]byte, bool, error){
read_data := make([]byte, 1024*8)
read_data_len := 0
buf := make([]byte, 1024*4)
b_len := len(boundary)
reach_end := false
for ;!reach_end; {
read_len, err := stream.Read(buf)
if err != nil {
if err != io.EOF && read_len<=0 {
return nil, true, err
}
reach_end = true
}
//todo: 下面这一句很蠢,值得优化
copy(read_data[read_data_len:], buf[:read_len]) //追加到另一块buffer,仅仅只是为了搜索方便
read_data_len += read_len
if (read_data_len<b_len+4){
continue
}
loc := bytes.Index(read_data[:read_data_len], boundary)
if loc>=0{
//找到了结束位置
target.Write(read_data[:loc-4])
return read_data[loc:read_data_len],reach_end, nil
}
target.Write(read_data[:read_data_len-b_len-4])
copy(read_data[0:], read_data[read_data_len-b_len-4:])
read_data_len = b_len + 4
}
target.Write(read_data[:read_data_len])
return nil, reach_end, nil
}
/// 解析表单的头部
/// @param read_data 已经从流中读到的数据
/// @param read_total 已经从流中读到的数据长度
/// @param boundary 表单的分割字符串
/// @param stream 输入流
/// @return FileHeader 文件名等信息头
/// []byte 已经从流中读到的部分
/// error 是否发生错误
func ParseFromHead(read_data []byte, read_total int, boundary []byte, stream io.ReadCloser)(FileHeader, []byte, error){
buf := make([]byte, 1024*4)
found_boundary := false
boundary_loc := -1
var file_header FileHeader
for {
read_len, err := stream.Read(buf)
if err!=nil{
if err!=io.EOF{
return file_header, nil, err
}
break
}
if read_total+read_len>cap(read_data){
return file_header, nil, fmt.Errorf("not found boundary")
}
copy(read_data[read_total:], buf[:read_len])
read_total += read_len
if !found_boundary {
boundary_loc = bytes.Index(read_data[:read_total], boundary)
if -1 == boundary_loc {
continue
}
found_boundary = true
}
start_loc := boundary_loc+len(boundary)
file_head_loc := bytes.Index(read_data[start_loc:read_total], []byte("
"))
if -1==file_head_loc{
continue
}
file_head_loc += start_loc
ret := false
file_header,ret = ParseFileHeader(read_data[start_loc:file_head_loc])
if !ret{
return file_header,nil,fmt.Errorf("ParseFileHeader fail:%s", string(read_data[start_loc:file_head_loc]))
}
return file_header, read_data[file_head_loc+4:read_total], nil
}
return file_header,nil,fmt.Errorf("reach to sream EOF")
}
func main(){
log.SetFlags(log.LstdFlags | log.Lshortfile)
r := gin.Default()
r.StaticFile("/upload.html", "./upload.html")
r.POST("/gin_upload", func(c *gin.Context) {
var content_length int64
content_length = c.Request.ContentLength
if content_length<=0 || content_length>1024*1024*1024*2{
log.Printf("content_length error
")
return
}
content_type_,has_key := c.Request.Header["Content-Type"]
if !has_key{
log.Printf("Content-Type error
")
return
}
if len(content_type_)!=1{
log.Printf("Content-Type count error
")
return
}
content_type := content_type_[0]
const BOUNDARY string = "; boundary="
loc := strings.Index(content_type, BOUNDARY)
if -1==loc{
log.Printf("Content-Type error, no boundary
")
return
}
boundary := []byte(content_type[(loc+len(BOUNDARY)):])
log.Printf("[%s]
", boundary)
//
read_data := make([]byte, 1024*12)
var read_total int = 0
for {
file_header, file_data, err := ParseFromHead(read_data, read_total, append(boundary, []byte("
")...), c.Request.Body)
if err != nil {
log.Printf("%v", err)
return
}
log.Printf("file :%s
", file_header.FileName)
//
f, err := os.Create(file_header.FileName)
if err != nil {
log.Printf("create file fail:%v
", err)
return
}
f.Write(file_data)
file_data = nil
//需要反复搜索boundary
temp_data, reach_end, err := ReadToBoundary(boundary, c.Request.Body, f)
f.Close()
if err != nil {
log.Printf("%v
", err)
return
}
if reach_end{
break
} else {
copy(read_data[0:], temp_data)
read_total = len(temp_data)
continue
}
}
//
c.JSON(200, gin.H{
"message": fmt.Sprintf("%s", "ok"),
})
})
r.Run()
}