| 摘要 |
An ascending numerical sequence, such as an index of word position values within a file, is variable length binary encoded, and subsequently decoded to derive the original sequence. Encoding involves obtaining certain properties of the sequence, including the first and last element values and the maximum and minimum gaps. Unless the maximum and minimum gaps are equal, each element value between the first and last element of the sequence is reduced and the reduced sequence is further encoded, all according to various rules. Reduction involves stripping the leading 0 bit's and the first 1 bit. Further encoding involves dropping repeating leading bits and storing the resulting binary values in a body having a bitstream format if an element value has the same number of bits as the number of bits in the preceding element value. If an element value has one more bit than the preceding element value and if the value of the first N bits are greater than the value of the previous N bits then a bump code is added before the resulting binary value is stored. If the value of the first N bits are less than the value of the previous N bits and the resulting binary value has N leading 0 bits then a bump code is added before the resulting binary value is stored. In most other circumstances, the resulting binary value is simply stored in the bitstream. If the maximum and minimum gaps are the same then the first and last values and the minimum and maximum gaps are simply stored in the header and no body is stored as the sequence is fully defined. When the length of the body is known then a header is stored that contains the properties of the sequence that permit the sequence to be decoded from the header and body. Corresponding rules permit the decoding of the sequence from the header and body.
|
