Simply, we take the bases or digits in the number and sort them in buckets, the memory might be high, so that’s why its done per digit to avoid billions of buckets being created for each number.
I like the approach of floating point numbers, the guy really thought it through, was really clever of using a format that would sort the number as a digit. Absolutely fantastic mathematics in this.
Will try to implement in golang to get a feel for it.
