Creating a Custom Reader Class for Ogg Vorbis/FLAC Comments

27 Jul 2004 15:49

In Dino Esposito’s “Cutting Edge” column in MSDN Magazine, he talks about the various readers (BinaryReader, XmlReader, etc.) available in .NET.

At the end of the column, he says a couple of things about how to design and implement a custom reader.

Here’s how I went about implementing a custom reader for reading Vorbis comments from FLAC files.

The first thing that we need to do is decide on what methods and properties the custom reader should support. We’re not tied to a specific interface, because Reader is a .NET pattern, rather than a specified type.

As it’s a reader, we’ll probably want to construct it with a stream. Most of the other readers defined in .NET allow this, so it seems like a good idea.

public class VorbisCommentReader
    public VorbisCommentReader(Stream stream)

What about the rest of the methods and properties? Vorbis comments begin with a vendor ID, which is followed by a sequence of tag=value comments.

It seems sensible, then, to add a VendorString property to our class:

    public string VendorString
        get { return this.vendorString; }

Interface Options

As for the tag=value comments, we could just suck them up into a collection, and iterate over them with foreach, like this:

VorbisCommentReader r = new VorbisCommentReader(stream);
foreach (UserComment uc in r.UserComments)
    string fieldName = uc.FieldName;
    string fieldValue = uc.FieldValue;

That’s not bad. It requires that we load all the comments at once, unless we want to do something a little more complicated in our IEnumerator implementation.

We could do this:

VorbisCommentReader r = new VorbisCommentReader(stream);
for (int i = 0; i < r.ReadCommentCount(); ++i)
    string fieldName = r.ReadFieldName();
    string fieldValue = r.ReadFieldValue();

That’s not bad, but the user might accidentally reverse the two calls, which makes it more fragile than it could be. For something like this, I’ve started opting for a DataReader-style interface:

VorbisCommentReader r = new VorbisCommentReader(stream);
while (r.Read())
    string fieldName = r.GetFieldName();
    string fieldValue = r.GetFieldValue();

Implementing the constructor

The constructor will assume that the stream is correctly positioned at the start of the block. Because Stream only defines some very basic reading methods (Read, which reads an array of bytes; ReadByte which reads a single byte), we’ll need to use a BinaryReader internally:

public VorbisCommentReader(Stream stream)
    this.binaryReader = new BinaryReader(stream);
    int vendorLength = binaryReader.ReadInt32();

    // The next thing is a UTF8 string.
    byte[] vendorBytes =
    this.vendorString =

    this.userCommentListLength = binaryReader.ReadInt32();
    this.userCommentListIndex = 0;

We read the various fixed values from the comment block, and reset our list index.

Implementing Read

We can then implement the Read method like this:

    public bool Read()
        if (userCommentListIndex < userCommentListLength)
            int userCommentLength = binaryReader.ReadInt32();

            byte[] userCommentBytes =
            string userCommentString =

            int pos = userCommentString.IndexOf('=');
            this.fieldName =
                userCommentString.Substring(0, pos);
            this.fieldValue =
                userCommentString.Substring(pos + 1);


            return true;

        return false;

If we’re being nitpicky, this isn’t completely correct. Only the field value (after the =) is allowed to be UTF8, but since 7-bit field names are a subset of UTF8, we can probably get away with it.

More implementation

With this in place, the other methods are a doddle:

    public string GetFieldName()
        return this.fieldName;

    public string GetFieldValue()
        return this.fieldValue;


Writing custom reader objects is easy, and it makes your code feel a bit more idiomatic; it feels like the other .NET readers.

I’ve also implemented a custom reader to go with this one that allows reading of the metadata blocks in a FLAC file. Source code for both will appear at some point.