compressing 2 streams into 1 BitOutputStream about reference-arithmetic-coding HOT 1 CLOSED

Adding any data after the end of an arithmetic-coded stream is tricky and easy to screw up. The problems are that the encoder needs to write a sufficient number of padding bits, and the decoder needs to read exactly the correct number of bits but no more and no less. Therefore my library does not attempt to implement or allow this functionality.

You cannot use my library to join two arithmetic-coded streams. You cannot put any bit data or byte data after an arithmetic-coded stream. You can work around this by methods such as:

• Adding an explicit framing around the two chunks of arithmetic-coded data. Make a header that describes the bit length. Make an input stream class that reads the bit length from the header, and gives EOF after all bits are read. In essence, create subblocks in the file to act like virtual files that each have an EOF.
• Make both vectors share the same arithmetic coder so that you only use one stream. There are many ways to do this - back-to-back, interleaved, etc.
• Modify the library, with deep low-level knowledge of arithmetic coding, to support stream concatenation without corrupting the data. This is theoretically feasible, but I have not put in the effort to explore this functionality. Hence I cannot tell you what a correct implementation looks like.

No further support will be provided from me.

from reference-arithmetic-coding.