Multiple Signatures and Offsets

[justin-dx]

I've been adding my own file extensions, mainly for raw camera images (cr2, nef, orf, dng etc). Some of these raw file types use magic numbers at multiple offsets e.g.: 08 00 00 00 offset 4 AND 2D 00 FE 00 offset 8. Both sets of these bytes must match at the specified offsets to be considered a match for the file type. Is it possible to handle multiple signatures for a single stream or is this handled by the package already?

[tiesont]

I handled this for the AVI video type by overriding the IsMatch(Stream) method - if you know where the "fixed" magic bytes are, you can just check those byte ranges. I don't know that what I did was the most efficient way (I assume it's not), but it seems to work. If it helps:

public abstract class Video : FileFormat
{
    protected Video(byte[] signature, string mediaType, string extension, int offset = 0) : base(signature, mediaType, extension, offset)
    {
    }
}

public class AviVideo : Video
{
    // bytes 4 to 7 are variable --> we're using a placeholder byte (00) for those values
    private static readonly byte[] expected = new byte[] { 0x52, 0x49, 0x46, 0x46, 00, 00, 00, 00, 0x41, 0x56, 0x49, 0x20 };
    private const string mediaType = "video/x-msvideo";
    private const string extension = ".avi";

    public AviVideo()
        : base(expected, mediaType, extension)
    {
    }

    public override bool IsMatch(Stream stream)
    {
        byte[] header;

        if (stream != null)
        {
            using (var ms = new MemoryStream())
            {
                stream.CopyTo(ms);
                header = ms.ToArray();

                // We need at least 12 bytes of data to check for an AVI header
                if (header.Length > 11)
                {
                    // bytes 0 - 3 must match expected values
                    for (int i = 0; i < 4; i++)
                    {
                        if (expected[i] != header[i])
                        {
                            return false;
                        }
                    }

                    // bytes 8 - 11 must match expected values
                    for (int i = 8; i < 12; i++)
                    {
                        if (expected[i] != header[i])
                        {
                            return false;
                        }
                    }

                    return true;
                }
            }
        }

        return false;
    }
}

You don't really need the base class - I created that because I needed to add multiple video formats and wanted to be able to something like

var isVideo = format is Video;

[neilharvey]

Overriding IsMatch like @tiesont described is the way to do it.

Multiple signatures within a file extension usually indicates that a format is a specialisation of something else - e.g. CR2 is apparently based on TIFF so if I were doing it in the library I would inherit from the base class and override IsMatch but if it's just for your own use just do whatever works :)

(While looking at this I realised that I'm including the TIFF byte order marks in the file signature so I should go and fix that)