Streaming a zip file over http in .net with SharpZ

问题:

I'm making a simple download service so a user can download all his images from out site. To do that i just zip everything to the http stream.

However it seems everything is stored in memory, and the data isn't sent til zip file is complete and the output closed. I want the service to start sending at once, and not use too much memory.

public void ProcessRequest(HttpContext context)
{
    List<string> fileNames = GetFileNames();
    context.Response.ContentType = "application/x-zip-compressed";
    context.Response.AppendHeader("content-disposition", "attachment; filename=files.zip");
    context.Response.ContentEncoding = Encoding.Default;
    context.Response.Charset = "";

    byte[] buffer = new byte[1024 * 8];

    using (ICSharpCode.SharpZipLib.Zip.ZipOutputStream zipOutput = new ICSharpCode.SharpZipLib.Zip.ZipOutputStream(context.Response.OutputStream))
    {
        foreach (string fileName in fileNames)
        {
            ICSharpCode.SharpZipLib.Zip.ZipEntry zipEntry = new ICSharpCode.SharpZipLib.Zip.ZipEntry(fileName);
            zipOutput.PutNextEntry(zipEntry);
            using (var fread = System.IO.File.OpenRead(fileName))
            {
                ICSharpCode.SharpZipLib.Core.StreamUtils.Copy(fread, zipOutput, buffer);
            }
        }
        zipOutput.Finish();
    }

    context.Response.Flush();
    context.Response.End();
}

I can see the the worker process memory growing while it makes the file, and then releases the memory when its done sending. How do i do this without using too much memory?

回答1:

Disable response buffering with context.Response.BufferOutput = false; and remove the Flush call from the end of your code.

回答2:

use Response.BufferOutput = false; at start of ProcessRequest and flush response after each file.

回答3:

Check this post: http://www.nablasoft.com/alkampfer/index.php/2008/08/26/use-icsharpcodesharpziplib-to-write-directly-to-responseoutputstream-in-a-aspnet-application/

回答4:

FYI. This is working code to recursively add an entire tree of files, with streaming to browser:

string path = @"c:\files";

Response.Clear();
Response.ContentType = "application/zip";
Response.AddHeader("Content-Disposition", string.Format("attachment; filename=\"{0}\"", "hive.zip"));
Response.BufferOutput = false;

byte[] buffer = new byte[1024 * 1024];
using (ZipOutputStream zo = new ZipOutputStream(Response.OutputStream, 1024 * 1024)) {
    zo.SetLevel(0);
    DirectoryInfo di = new DirectoryInfo(path);
    foreach (string file in Directory.GetFiles(di.FullName, "*.*", SearchOption.AllDirectories)) {
        string folder = Path.GetDirectoryName(file);
        if (folder.Length > di.FullName.Length) {
            folder = folder.Substring(di.FullName.Length).Trim('\\') + @"\";
        } else {
            folder = string.Empty;
        }
        zo.PutNextEntry(new ZipEntry(folder + Path.GetFileName(file)));
        using (FileStream fs = File.OpenRead(file)) {
            ICSharpCode.SharpZipLib.Core.StreamUtils.Copy(fs, zo, buffer);
        }
        zo.Flush();
        Response.Flush();
    }
    zo.Finish();
}

Response.Flush();