Can't delete temporary files after returning FileStream

Question

I have a function in a C# MVC application that creates a temporary directory and a temporary file, then opens the file using FileStream, returns the FileStream to a calling function, and then needs to delete the temporary files. However, I do not know how to delete the temp directory and file because it always errors out saying "the process cannot access the file because it is being used by another process." This is what I tried, but the FileStream is still using the temp file in the finally block. How can I return the FileStream and delete the temporary files?

public FileStream DownloadProjectsZipFileStream()
{
    Directory.CreateDirectory(_tempDirectory);
    // temporary file is created here
    _zipFile.Save(_tempDirectory + _tempFileName);

    try
    {
        FileStream stream = new FileStream(_tempDirectory + _tempFileName, FileMode.Open);
        return stream;
    }
    finally
    {
        File.Delete(_tempDirectory + _tempFileName);
        Directory.Delete(_tempDirectory);
    }
}

The function the FileStream is returned to looks like this:

public ActionResult DownloadProjects ()
{
    ProjectDownloader projectDownloader = new ProjectDownloader();

    FileStream stream = projectDownloader.DownloadProjectsZipFileStream();
    return File(stream, "application/zip", "Projects.zip");
}

Update: I forgot to mention the zip file is 380 MB. I get a system out of memory exception when using a MemoryStream.

Answer 1

Here's a cutdown version of the above that I use:

public class DeleteAfterReadingStream : FileStream
{
    public DeleteAfterReadingStream(string path)
        : base(path, FileMode.Open)
    {
    }

    protected override void Dispose(bool disposing)
    {
        base.Dispose(disposing);
        if (File.Exists(Name))
            File.Delete(Name);
    }
}

Answer 2

You could create a wrapper class that implements the Stream contract and that contains the FileStream internally, as well as maintaining the path to the file.

All of the standard Stream methods and properties would just be passed to the FileStream instance.

When this wrapper class is Disposed, you would (after Disposeing the wrapped FileStream) then delete the file.

Answer 3

I used Damien_The_Unbeliever's (accepted answer)'s advice, wrote it up, and it worked beautifully. Just thought I'd share the class:

public class BurnAfterReadingFileStream : Stream
{
    private FileStream fs;

    public BurnAfterReadingFileStream(string path) { fs = System.IO.File.OpenRead(path); }

    public override bool CanRead { get { return fs.CanRead; } }

    public override bool CanSeek { get { return fs.CanRead; } }

    public override bool CanWrite { get { return fs.CanRead; } }

    public override void Flush() { fs.Flush(); }

    public override long Length { get { return fs.Length; } }

    public override long Position { get { return fs.Position; } set { fs.Position = value; } }

    public override int Read(byte[] buffer, int offset, int count) { return fs.Read(buffer, offset, count); }

    public override long Seek(long offset, SeekOrigin origin) { return fs.Seek(offset, origin); }

    public override void SetLength(long value) { fs.SetLength(value); }

    public override void Write(byte[] buffer, int offset, int count) { fs.Write(buffer, offset, count); }

    protected override void Dispose(bool disposing)
    {
        base.Dispose(disposing);
        if (Position > 0) //web service quickly disposes the object (with the position at 0), but it must get rebuilt and re-disposed when the client reads it (when the position is not zero)
        {
            fs.Close();
            if (System.IO.File.Exists(fs.Name))
                try { System.IO.File.Delete(fs.Name); }
                finally { }
        }
    }
}

Answer 4

The problem is that you can only delete the file after it has been written to the response and the file is written by the FileStreamResult only after it is returned from the action.

One way to handle is to create a subclass of FileResult that will delete the file.

It's easier to subclass FilePathResult so that the class has access to the filename.

public class FilePathWithDeleteResult : FilePathResult
{
    public FilePathResult(string fileName, string contentType)
        : base(string fileName, string contentType)
    {
    }

    protected override void WriteFile(HttpResponseBase response)
    {
        base.WriteFile(response);
        File.Delete(FileName);
        Directory.Delete(FileName);
    }
}

Note: I haven't tested the above. Remove all it's bugs before using it.

Now change the controller code to something like:

public ActionResult DownloadProjects ()
{
    Directory.CreateDirectory(_tempDirectory);
    // temporary file is created here
    _zipFile.Save(_tempDirectory + _tempFileName);

    return new FilePathWithDeleteResult(_tempDirectory + _tempFileName, "application/zip") { FileDownloadName = "Projects.zip" };
}

Answer 5

I use the method suggested by @hwiechers, but the only way to make it work is to close the response stream before deleting the file.

Here is the source code, note that I flush the stream before deleting it.

public class FilePathAutoDeleteResult : FilePathResult
{
    public FilePathAutoDeleteResult(string fileName, string contentType) : base(fileName, contentType)
    {
    }

    protected override void WriteFile(HttpResponseBase response)
    {
        base.WriteFile(response);
        response.Flush();
        File.Delete(FileName);
    }

}

And here is how the Controller supposed to call it:

public ActionResult DownloadFile() {

    var tempFile = Path.GetTempFileName();

    //do your file processing here...
    //For example: generate a pdf file

    return new FilePathAutoDeleteResult(tempFile, "application/pdf")
    {
        FileDownloadName = "Awesome pdf file.pdf"
    };
}