Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 1 | // Copyright 2011 The Go Authors. All rights reserved. |
| 2 | // Use of this source code is governed by a BSD-style |
| 3 | // license that can be found in the LICENSE file. |
| 4 | |
| 5 | package binary |
| 6 | |
| 7 | // This file implements "varint" encoding of 64-bit integers. |
| 8 | // The encoding is: |
| 9 | // - unsigned integers are serialized 7 bits at a time, starting with the |
| 10 | // least significant bits |
| 11 | // - the most significant bit (msb) in each output byte indicates if there |
| 12 | // is a continuation byte (msb = 1) |
| 13 | // - signed integers are mapped to unsigned integers using "zig-zag" |
| 14 | // encoding: Positive values x are written as 2*x + 0, negative values |
| 15 | // are written as 2*(^x) + 1; that is, negative numbers are complemented |
| 16 | // and whether to complement is encoded in bit 0. |
| 17 | // |
| 18 | // Design note: |
| 19 | // At most 10 bytes are needed for 64-bit values. The encoding could |
| 20 | // be more dense: a full 64-bit value needs an extra byte just to hold bit 63. |
| 21 | // Instead, the msb of the previous byte could be used to hold bit 63 since we |
| 22 | // know there can't be more than 64 bits. This is a trivial improvement and |
| 23 | // would reduce the maximum encoding length to 9 bytes. However, it breaks the |
| 24 | // invariant that the msb is always the "continuation bit" and thus makes the |
| 25 | // format incompatible with a varint encoding for larger numbers (say 128-bit). |
| 26 | |
| 27 | import ( |
Russ Cox | c2049d2 | 2011-11-01 22:04:37 -0400 | [diff] [blame] | 28 | "errors" |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 29 | "io" |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 30 | ) |
| 31 | |
| 32 | // MaxVarintLenN is the maximum length of a varint-encoded N-bit integer. |
| 33 | const ( |
| 34 | MaxVarintLen16 = 3 |
| 35 | MaxVarintLen32 = 5 |
| 36 | MaxVarintLen64 = 10 |
| 37 | ) |
| 38 | |
| 39 | // PutUvarint encodes a uint64 into buf and returns the number of bytes written. |
Brad Fitzpatrick | 2c2c202 | 2012-01-24 14:19:59 -0800 | [diff] [blame] | 40 | // If the buffer is too small, PutUvarint will panic. |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 41 | func PutUvarint(buf []byte, x uint64) int { |
Robert Griesemer | 61cc872 | 2011-09-29 09:40:59 -0700 | [diff] [blame] | 42 | i := 0 |
| 43 | for x >= 0x80 { |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 44 | buf[i] = byte(x) | 0x80 |
| 45 | x >>= 7 |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 46 | i++ |
| 47 | } |
Robert Griesemer | 61cc872 | 2011-09-29 09:40:59 -0700 | [diff] [blame] | 48 | buf[i] = byte(x) |
| 49 | return i + 1 |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 50 | } |
| 51 | |
| 52 | // Uvarint decodes a uint64 from buf and returns that value and the |
| 53 | // number of bytes read (> 0). If an error occurred, the value is 0 |
| 54 | // and the number of bytes n is <= 0 meaning: |
| 55 | // |
| 56 | // n == 0: buf too small |
| 57 | // n < 0: value larger than 64 bits (overflow) |
| 58 | // and -n is the number of bytes read |
| 59 | // |
| 60 | func Uvarint(buf []byte) (uint64, int) { |
| 61 | var x uint64 |
| 62 | var s uint |
| 63 | for i, b := range buf { |
| 64 | if b < 0x80 { |
| 65 | if i > 9 || i == 9 && b > 1 { |
| 66 | return 0, -(i + 1) // overflow |
| 67 | } |
| 68 | return x | uint64(b)<<s, i + 1 |
| 69 | } |
| 70 | x |= uint64(b&0x7f) << s |
| 71 | s += 7 |
| 72 | } |
| 73 | return 0, 0 |
| 74 | } |
| 75 | |
| 76 | // PutVarint encodes an int64 into buf and returns the number of bytes written. |
Brad Fitzpatrick | 2c2c202 | 2012-01-24 14:19:59 -0800 | [diff] [blame] | 77 | // If the buffer is too small, PutVarint will panic. |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 78 | func PutVarint(buf []byte, x int64) int { |
| 79 | ux := uint64(x) << 1 |
| 80 | if x < 0 { |
| 81 | ux = ^ux |
| 82 | } |
| 83 | return PutUvarint(buf, ux) |
| 84 | } |
| 85 | |
| 86 | // Varint decodes an int64 from buf and returns that value and the |
| 87 | // number of bytes read (> 0). If an error occurred, the value is 0 |
| 88 | // and the number of bytes n is <= 0 with the following meaning: |
| 89 | // |
| 90 | // n == 0: buf too small |
| 91 | // n < 0: value larger than 64 bits (overflow) |
| 92 | // and -n is the number of bytes read |
| 93 | // |
| 94 | func Varint(buf []byte) (int64, int) { |
| 95 | ux, n := Uvarint(buf) // ok to continue in presence of error |
| 96 | x := int64(ux >> 1) |
| 97 | if ux&1 != 0 { |
| 98 | x = ^x |
| 99 | } |
| 100 | return x, n |
| 101 | } |
| 102 | |
Russ Cox | c2049d2 | 2011-11-01 22:04:37 -0400 | [diff] [blame] | 103 | var overflow = errors.New("binary: varint overflows a 64-bit integer") |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 104 | |
| 105 | // ReadUvarint reads an encoded unsigned integer from r and returns it as a uint64. |
Russ Cox | c2049d2 | 2011-11-01 22:04:37 -0400 | [diff] [blame] | 106 | func ReadUvarint(r io.ByteReader) (uint64, error) { |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 107 | var x uint64 |
| 108 | var s uint |
| 109 | for i := 0; ; i++ { |
| 110 | b, err := r.ReadByte() |
| 111 | if err != nil { |
| 112 | return x, err |
| 113 | } |
| 114 | if b < 0x80 { |
| 115 | if i > 9 || i == 9 && b > 1 { |
| 116 | return x, overflow |
| 117 | } |
| 118 | return x | uint64(b)<<s, nil |
| 119 | } |
| 120 | x |= uint64(b&0x7f) << s |
| 121 | s += 7 |
| 122 | } |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 123 | } |
| 124 | |
Shenghou Ma | f9902c7 | 2012-10-26 21:14:34 +0800 | [diff] [blame] | 125 | // ReadVarint reads an encoded signed integer from r and returns it as an int64. |
Russ Cox | c2049d2 | 2011-11-01 22:04:37 -0400 | [diff] [blame] | 126 | func ReadVarint(r io.ByteReader) (int64, error) { |
Robert Griesemer | f30719d | 2011-09-28 22:36:52 -0700 | [diff] [blame] | 127 | ux, err := ReadUvarint(r) // ok to continue in presence of error |
| 128 | x := int64(ux >> 1) |
| 129 | if ux&1 != 0 { |
| 130 | x = ^x |
| 131 | } |
| 132 | return x, err |
| 133 | } |