An efficient LZW implementation

6. Making a gif codec

The LZW compression used in the GIF format is dynamic LZW as described here, with these details:

There's a field in the GIF format header called "code size" which tells the minimum bitsize for the dictionary. In other words, GIF always uses a dictionary sized as a multiple of 2, up to 256 elements.
The maximum bitsize is fixed to 12 (although in theory there shouldn't be any reason why higher bitsizes couldn't be used...)
The first index after the minimum dictionary size is a clear code. When encoding, whenever you reset the dictionary you should write this value to the output, and when decoding, you should reset the dictionary when this value is read. (Theoretically there shouldn't be any need for such code, but GIF uses it anyways.)
The second index after the minimum dictionary size is an end-of-input code.