Does this C# routine to delete all files that are NOT in an array look alright?

Question

I have a DB that contains a list of paths to files. I want to build a routine to cleanup the folders, removing files in the directories if there is not a db record for it (for temp ajax file uploads, in cases where the user doesn't complete the form, etc...).

I'm thinking something like this:

var dbFiles = db.allPaths();
var allFiles = Directory.EnumerateFiles(path);

foreach (var f in allFiles) {
  if (!dbFiles.Contains(f) {
    File.Delete(f);
  }
}

Any "Gotchas" waiting for me? The routine will be set to run once a week at first, more often if temp files become a problem. It will be run during a time when there are nearly no users on, so performance - while important - is not paramount.

Thanks in advance!

UPDATE

Wow, lots of great answers. This bit of code is turning into something "share" worthy. ;D My code above was just a simple, quick placeholder bit... but it's transformed into solid code. Thank you!

Answer 1

Looks good to me; however I've never deleted files within C#, just VB. However, you might want to throw that into a Try/Catch loop, as if the file isn't able to be deleted (read-only, currently in use, no longer exists, etc.), it will throw an exception.

EDIT: How are the paths stored? Remember, in C# you need to escape out paths "//" instead of using "\" IIRC.

EDIT 2: Scratch that last edit out lol.

Answer 2

Looks okay, but you can make it simpler:

foreach (var file in allFiles.Except(dbFiles))
{
    File.Delete(file);
}

You've got to make sure that the paths are in exactly the same format though. If one list has relative files and the other has absolute files, or if one uses "/" and the other uses "\" you'll end up deleting things you don't expect to.

Ideally you'd canonicalise the files explicitly first, but I can't see a nice way of getting a canonical file name in .NET...

EDIT: Note that Path.GetFullPath does not canonicalize. It fixes slashes and makes it absolute, but it doesn't address case: "c:/users" becomes "c:\users", but "c:/Users" becomes "c:\Users".

This could be fixed by using a string comparer in the call to Except:

var dbFiles = db.AllPaths().Select(Path.GetFullPath));
var allFiles = Directory.EnumerateFiles(path).Select(Path.GetFullPath));

foreach (var file in allFiles.Except(dbFiles, StringComparer.OrdinalIgnoreCase))
{
    File.Delete(file);
}

Now that's ignoring case - but in an "ordinal" manner. I don't know what the Windows file system really does in terms of its case sensitivity.

Answer 3

I think it's alright in spirit, though it would be closer to:

List<string> dbFiles = db.allPaths();
string[] allFiles = Directory.GetFiles(path);

foreach (string f in allFiles)
    if (!dbFiles.Contains(f))
        File.Delete(f);

Answer 4

To combine all the suggestions into one:

// canonicalize paths
var dbFiles = db.allPaths().Select(Path.GetFullPath);
var allFiles = Directory.EnumerateFiles(Path.GetFullPath(path))

foreach (var file in allFiles.Except(dbFiles, StringComparer.OrdinalIgnoreCase))
{
    try {
        File.Delete(file);
    } catch (IOException) {
        // handle exception here
    }
}