Lines Matching full:byte
38 * - byte stream (unsigned 8-bit bytes)
41 * - avoid byte values 0, 1, 2
51 * - Similar to a UTF, encoding the length of the byte sequence in the lead bytes.
53 * (trail byte values may overlap with lead/single byte values).
59 * Instead, we split the lead byte range in the middle and encode non-negative values
62 * For very small absolute values, the difference is added to a middle byte value
63 * for single-byte encoded differences.
65 * of byte values available, the modulo is used for one trail byte, and the remainder
66 * is added to a lead byte avoiding the single-byte range.
69 * This encoding does not use byte values 0, 1, 2, but uses all other byte values
72 * Note that the lead byte ranges overlap some, but that the sequences as a whole
73 * are well ordered. I.e., even if the lead byte is the same for sequences of different
84 /* Do not use byte values 0, 1, 2 because they are separators in sort keys. */
95 * 1 middle byte for 0
97 * 2*42=84 for double-byte values
98 * 2*3=6 for 3-byte values
99 * 2*1=2 for 4-byte values
104 * - There should be >=128 single-byte values to cover 128-blocks
106 * - There should be >=20902 single/double-byte values to cover Unihan.
107 * - It helps CJK Extension B some if there are 3-byte values that cover
110 * - Four-byte values are necessary to cover the rest of Unicode.
112 * Symmetrical lead byte counts are for convenience.
114 * no advantage to asymmetrical lead byte counts.
133 /* The lead byte start values. */
