Getting "The process cannot access the file because it is being used by another process" when saving multiple files from the request stream

Question

I am using the HTML5 canvas element and the new HTML5 file i\o function to drop multiple files on it and have them upload. It works fine, but now I need to generate a new filename if no files are in the destination directory (It's a 7 digit integer) or get the name of the last uploaded file, convert it to int32 and increment that by one for every new file being uploaded to the same directory. This is where the GetFileName(dir); comes in. The first image always uploads fine but the problem begins once the second file is saved and the process hits ImageJob.Build(), I presume this is because once the new file is starting to write, the GetFile() method runs for second file in line simultaneously and is checking for last written file, which is still being written and this creates the conflict. How can I fix this, maybe I can somehow itterate with a foreach over the Request.InputStream data or implement some kind process watch that waits for the process to finish?

Update: I tried using TempData to store the generated filename, and just increment on the int value in TempData for all the next file names and it appears to do better, gets more images in but still errors at some point. But TempData is not for that as it gets erased after each read, reassigning to it again does not help. Maybe I'll try storing it in session.

The process cannot access the file 'C:\Users\Admin\Documents\Visual Studio 2010\Projects\myproj\myproj\Content\photoAlbums\59\31\9337822.jpg' because it is being used by another process.

public PartialViewResult Upload()
{          
    string fileName = Request.Headers["filename"];
    string catid = Request.Headers["catid"];
    string pageid = Request.Headers["pageid"];
    string albumname = Request.Headers["albumname"];


    var dir = "~/Content/photoAlbums/" + catid + "/" + pageid + "/" + (albumname ?? null);

    var noex = GetFileName(dir);

    var extension = ".jpg";

    string thumbFile = noex + "_t" + extension;
    fileName = noex + extension;

    byte[] file = new byte[Request.ContentLength];
    Request.InputStream.Read(file, 0, Request.ContentLength);

    string imgdir;
    string thumbimgdir;
    string imageurl;

    if (albumname != null) 
    { 
        imgdir = Server.MapPath("~/Content/photoAlbums/" + catid + "/" + pageid + "/" + albumname + "/" + fileName);
        thumbimgdir = Server.MapPath("~/Content/photoAlbums/" + catid + "/" + pageid + "/" + albumname + "/" + thumbFile);
        imageurl = "/Content/photoAlbums/" + catid + "/" + pageid + "/" + albumname + "/" + thumbFile;
    }
    else 
    { 
        imgdir = Server.MapPath("~/Content/photoAlbums/" + catid + "/" + pageid + "/" + fileName);
        thumbimgdir = Server.MapPath("~/Content/photoAlbums/" + catid + "/" + pageid + "/" + thumbFile);
        imageurl = "/Content/photoAlbums/" + catid + "/" + pageid + "/" + thumbFile;
    }

    ImageJob b = new ImageJob(file, imgdir, new ResizeSettings("maxwidth=1024&maxheight=768&format=jpg")); b.CreateParentDirectory = true; b.Build();
    ImageJob a = new ImageJob(file, thumbimgdir, new ResizeSettings("w=100&h=100&mode=crop&format=jpg")); a.CreateParentDirectory = true; a.Build();

    ViewBag.CatID = catid;
    ViewBag.PageID = pageid;
    ViewBag.FileName = fileName;

    return PartialView("AlbumImage", imageurl);
}

public string GetFileName(string dir)
{
    var FullPath = Server.MapPath(dir);
    var dinfo = new DirectoryInfo(FullPath);

    string FileName;
    if (dinfo.Exists)
    {           
       var Filex = dinfo.EnumerateFiles().OrderBy(x => x.Name).LastOrDefault();
       FileName = Filex != null ? Path.GetFileNameWithoutExtension(Filex.Name) : null;

       if (FileName != null)
       {
           FileName = FileName.Contains("_t") ? FileName.Substring(0, FileName.Length - 2) : FileName;
           int fnum;
           Int32.TryParse(FileName, out fnum);
           FileName = (fnum + 1).ToString();

           if (fnum > 999999) { return FileName; } //Check that TryParse produced valid int
           else 
           {
              var random = new Random();
              FileName = random.Next(1000000, 9999000).ToString();
           }
       }
       else
       {
           var random = new Random();
           FileName = random.Next(1000000, 9999000).ToString();
       }
    }
    else
    {        
           var random = new Random();
           FileName = random.Next(1000000, 9999000).ToString();
    }
    return FileName;
}

Answer 1

You simply cannot use the Random class if you want to generate unique filenames. It uses the current time as the seed, so two exactly concurrent requests will always produce the same 'random' number.

You could use a cryptographic random number generator, but you would still have to ensure that (a) only one thread would generate it at a time, and (b) you used a sufficiently long identifier to prevent the Birthday paradox.

Thus, I suggest that everyone use GUID identifiers for their uploads, as they solve all of the above issues inherently (I believe an OS-level lock is used to prevent duplicates).

Your method also doesn't handle multiple file uploads per-request, although that may be intentional. You can support those by looping through Request.Files and passing each HttpPostedFile instance directly into the ImageJob.

Here's a simplified version of your code that uses GUIDs and won't encounter concurrency issues.

public PartialViewResult Upload()
{          

    string albumname = Request.Headers["albumname"];
    string baseDir = "~/Content/photoAlbums/" + Request.Headers["catid"] + "/" + Request.Headers["pageid"] + "/" (albumname != null ?  albumname + "/" : "");

    byte[] file = new byte[Request.ContentLength];
    Request.InputStream.Read(file, 0, Request.ContentLength);

    ImageJob b = new ImageJob(file, baseDir + "<guid>.<ext>", new ResizeSettings("maxwidth=1024&maxheight=768&format=jpg")); b.CreateParentDirectory = true; b.Build();
    ImageJob a = new ImageJob(file, baseDir + "<guid>_t.<ext>", new ResizeSettings("w=100&h=100&mode=crop&format=jpg")); a.CreateParentDirectory = true; a.Build();

    //Want both the have the same GUID? Pull it from the previous job.
    //string ext = PathUtils.GetExtension(b.FinalPath);
    //ImageJob a = new ImageJob(file, PathUtils.RemoveExtension(a.FinalPath) + "_t." + ext, new ResizeSettings("w=100&h=100&mode=crop&format=jpg")); a.CreateParentDirectory = true; a.Build();


    ViewBag.CatID = Request.Headers["catid"];
    ViewBag.PageID = Request.Headers["pageid"];
    ViewBag.FileName = Request.Headers["filename"];

    return PartialView("AlbumImage", PathUtils.GuessVirtualPath(a.FinalPath));
}

Answer 2

If the process is relatively quick (small files) you could go in a loop, check for that exception, sleep the thread for a couple of seconds, and try again (up to a maximum number of iterations). One caveat is that if the upload is asynchronous you might miss a file.

A couple of other suggestions:

• Make the GetFileName to be a private method so that it doesn't get triggered from the web.
• The OrderBy in the Filex query might not do what you expect once the it goes to 8 digits (possible if the first Random() is a very high number).
• The Random() should probably be seeded to produce better randomness.