Arithmetic encoding implementation outputting chars

Question

I am looking at the arithmetic coding implementation on https://www.codeabbey.com/index/task_view/adaptive-arithmetic-coding

My understanding is that when you're using floating point precision, the compressed result will be some floating point value. However, when using infinite precision with integers and shifting, I thought the output are binary values. In the above link, there's a function

def remove_first_digit(v):     # returns truncated value and digit which was truncated
    d = v // tail
    v %= tail
    return v, d

and d is appended to output which is the compressed result. However d here is a char. I don't quite understanding how this implementation works. Why is there a conversion to char going on here?

Answer 1

Here d is not a "char". It is a base 27 digit, i.e. an integer in the range 0..26. code_to_char() then converts the digit into an upper-case letter or a period to write out.

This is not an arithmetic encoder that produces binary output. It is an arithmetic encoder that produces output consisting of those 27 characters.

What is going on there is that the high digit of the low and high values are equal, so that digit is written out, and then removed from both the low and the high values. Then new digits are shifted in to low and high -- a 0 digit into low and a 26 digit into high.