gitea/vendor/github.com/rs/xid/id.go

366 lines
11 KiB
Go

// Package xid is a globally unique id generator suited for web scale
//
// Xid is using Mongo Object ID algorithm to generate globally unique ids:
// https://docs.mongodb.org/manual/reference/object-id/
//
// - 4-byte value representing the seconds since the Unix epoch,
// - 3-byte machine identifier,
// - 2-byte process id, and
// - 3-byte counter, starting with a random value.
//
// The binary representation of the id is compatible with Mongo 12 bytes Object IDs.
// The string representation is using base32 hex (w/o padding) for better space efficiency
// when stored in that form (20 bytes). The hex variant of base32 is used to retain the
// sortable property of the id.
//
// Xid doesn't use base64 because case sensitivity and the 2 non alphanum chars may be an
// issue when transported as a string between various systems. Base36 wasn't retained either
// because 1/ it's not standard 2/ the resulting size is not predictable (not bit aligned)
// and 3/ it would not remain sortable. To validate a base32 `xid`, expect a 20 chars long,
// all lowercase sequence of `a` to `v` letters and `0` to `9` numbers (`[0-9a-v]{20}`).
//
// UUID is 16 bytes (128 bits), snowflake is 8 bytes (64 bits), xid stands in between
// with 12 bytes with a more compact string representation ready for the web and no
// required configuration or central generation server.
//
// Features:
//
// - Size: 12 bytes (96 bits), smaller than UUID, larger than snowflake
// - Base32 hex encoded by default (16 bytes storage when transported as printable string)
// - Non configured, you don't need set a unique machine and/or data center id
// - K-ordered
// - Embedded time with 1 second precision
// - Unicity guaranteed for 16,777,216 (24 bits) unique ids per second and per host/process
//
// Best used with xlog's RequestIDHandler (https://godoc.org/github.com/rs/xlog#RequestIDHandler).
//
// References:
//
// - http://www.slideshare.net/davegardnerisme/unique-id-generation-in-distributed-systems
// - https://en.wikipedia.org/wiki/Universally_unique_identifier
// - https://blog.twitter.com/2010/announcing-snowflake
package xid
import (
"bytes"
"crypto/md5"
"crypto/rand"
"database/sql/driver"
"encoding/binary"
"errors"
"fmt"
"hash/crc32"
"io/ioutil"
"os"
"sort"
"sync/atomic"
"time"
)
// Code inspired from mgo/bson ObjectId
// ID represents a unique request id
type ID [rawLen]byte
const (
encodedLen = 20 // string encoded len
rawLen = 12 // binary raw len
// encoding stores a custom version of the base32 encoding with lower case
// letters.
encoding = "0123456789abcdefghijklmnopqrstuv"
)
var (
// ErrInvalidID is returned when trying to unmarshal an invalid ID
ErrInvalidID = errors.New("xid: invalid ID")
// objectIDCounter is atomically incremented when generating a new ObjectId
// using NewObjectId() function. It's used as a counter part of an id.
// This id is initialized with a random value.
objectIDCounter = randInt()
// machineId stores machine id generated once and used in subsequent calls
// to NewObjectId function.
machineID = readMachineID()
// pid stores the current process id
pid = os.Getpid()
nilID ID
// dec is the decoding map for base32 encoding
dec [256]byte
)
func init() {
for i := 0; i < len(dec); i++ {
dec[i] = 0xFF
}
for i := 0; i < len(encoding); i++ {
dec[encoding[i]] = byte(i)
}
// If /proc/self/cpuset exists and is not /, we can assume that we are in a
// form of container and use the content of cpuset xor-ed with the PID in
// order get a reasonable machine global unique PID.
b, err := ioutil.ReadFile("/proc/self/cpuset")
if err == nil && len(b) > 1 {
pid ^= int(crc32.ChecksumIEEE(b))
}
}
// readMachineId generates machine id and puts it into the machineId global
// variable. If this function fails to get the hostname, it will cause
// a runtime error.
func readMachineID() []byte {
id := make([]byte, 3)
hid, err := readPlatformMachineID()
if err != nil || len(hid) == 0 {
hid, err = os.Hostname()
}
if err == nil && len(hid) != 0 {
hw := md5.New()
hw.Write([]byte(hid))
copy(id, hw.Sum(nil))
} else {
// Fallback to rand number if machine id can't be gathered
if _, randErr := rand.Reader.Read(id); randErr != nil {
panic(fmt.Errorf("xid: cannot get hostname nor generate a random number: %v; %v", err, randErr))
}
}
return id
}
// randInt generates a random uint32
func randInt() uint32 {
b := make([]byte, 3)
if _, err := rand.Reader.Read(b); err != nil {
panic(fmt.Errorf("xid: cannot generate random number: %v;", err))
}
return uint32(b[0])<<16 | uint32(b[1])<<8 | uint32(b[2])
}
// New generates a globally unique ID
func New() ID {
return NewWithTime(time.Now())
}
// NewWithTime generates a globally unique ID with the passed in time
func NewWithTime(t time.Time) ID {
var id ID
// Timestamp, 4 bytes, big endian
binary.BigEndian.PutUint32(id[:], uint32(t.Unix()))
// Machine, first 3 bytes of md5(hostname)
id[4] = machineID[0]
id[5] = machineID[1]
id[6] = machineID[2]
// Pid, 2 bytes, specs don't specify endianness, but we use big endian.
id[7] = byte(pid >> 8)
id[8] = byte(pid)
// Increment, 3 bytes, big endian
i := atomic.AddUint32(&objectIDCounter, 1)
id[9] = byte(i >> 16)
id[10] = byte(i >> 8)
id[11] = byte(i)
return id
}
// FromString reads an ID from its string representation
func FromString(id string) (ID, error) {
i := &ID{}
err := i.UnmarshalText([]byte(id))
return *i, err
}
// String returns a base32 hex lowercased with no padding representation of the id (char set is 0-9, a-v).
func (id ID) String() string {
text := make([]byte, encodedLen)
encode(text, id[:])
return string(text)
}
// MarshalText implements encoding/text TextMarshaler interface
func (id ID) MarshalText() ([]byte, error) {
text := make([]byte, encodedLen)
encode(text, id[:])
return text, nil
}
// MarshalJSON implements encoding/json Marshaler interface
func (id ID) MarshalJSON() ([]byte, error) {
if id.IsNil() {
return []byte("null"), nil
}
text, err := id.MarshalText()
return []byte(`"` + string(text) + `"`), err
}
// encode by unrolling the stdlib base32 algorithm + removing all safe checks
func encode(dst, id []byte) {
dst[0] = encoding[id[0]>>3]
dst[1] = encoding[(id[1]>>6)&0x1F|(id[0]<<2)&0x1F]
dst[2] = encoding[(id[1]>>1)&0x1F]
dst[3] = encoding[(id[2]>>4)&0x1F|(id[1]<<4)&0x1F]
dst[4] = encoding[id[3]>>7|(id[2]<<1)&0x1F]
dst[5] = encoding[(id[3]>>2)&0x1F]
dst[6] = encoding[id[4]>>5|(id[3]<<3)&0x1F]
dst[7] = encoding[id[4]&0x1F]
dst[8] = encoding[id[5]>>3]
dst[9] = encoding[(id[6]>>6)&0x1F|(id[5]<<2)&0x1F]
dst[10] = encoding[(id[6]>>1)&0x1F]
dst[11] = encoding[(id[7]>>4)&0x1F|(id[6]<<4)&0x1F]
dst[12] = encoding[id[8]>>7|(id[7]<<1)&0x1F]
dst[13] = encoding[(id[8]>>2)&0x1F]
dst[14] = encoding[(id[9]>>5)|(id[8]<<3)&0x1F]
dst[15] = encoding[id[9]&0x1F]
dst[16] = encoding[id[10]>>3]
dst[17] = encoding[(id[11]>>6)&0x1F|(id[10]<<2)&0x1F]
dst[18] = encoding[(id[11]>>1)&0x1F]
dst[19] = encoding[(id[11]<<4)&0x1F]
}
// UnmarshalText implements encoding/text TextUnmarshaler interface
func (id *ID) UnmarshalText(text []byte) error {
if len(text) != encodedLen {
return ErrInvalidID
}
for _, c := range text {
if dec[c] == 0xFF {
return ErrInvalidID
}
}
decode(id, text)
return nil
}
// UnmarshalJSON implements encoding/json Unmarshaler interface
func (id *ID) UnmarshalJSON(b []byte) error {
s := string(b)
if s == "null" {
*id = nilID
return nil
}
return id.UnmarshalText(b[1 : len(b)-1])
}
// decode by unrolling the stdlib base32 algorithm + removing all safe checks
func decode(id *ID, src []byte) {
id[0] = dec[src[0]]<<3 | dec[src[1]]>>2
id[1] = dec[src[1]]<<6 | dec[src[2]]<<1 | dec[src[3]]>>4
id[2] = dec[src[3]]<<4 | dec[src[4]]>>1
id[3] = dec[src[4]]<<7 | dec[src[5]]<<2 | dec[src[6]]>>3
id[4] = dec[src[6]]<<5 | dec[src[7]]
id[5] = dec[src[8]]<<3 | dec[src[9]]>>2
id[6] = dec[src[9]]<<6 | dec[src[10]]<<1 | dec[src[11]]>>4
id[7] = dec[src[11]]<<4 | dec[src[12]]>>1
id[8] = dec[src[12]]<<7 | dec[src[13]]<<2 | dec[src[14]]>>3
id[9] = dec[src[14]]<<5 | dec[src[15]]
id[10] = dec[src[16]]<<3 | dec[src[17]]>>2
id[11] = dec[src[17]]<<6 | dec[src[18]]<<1 | dec[src[19]]>>4
}
// Time returns the timestamp part of the id.
// It's a runtime error to call this method with an invalid id.
func (id ID) Time() time.Time {
// First 4 bytes of ObjectId is 32-bit big-endian seconds from epoch.
secs := int64(binary.BigEndian.Uint32(id[0:4]))
return time.Unix(secs, 0)
}
// Machine returns the 3-byte machine id part of the id.
// It's a runtime error to call this method with an invalid id.
func (id ID) Machine() []byte {
return id[4:7]
}
// Pid returns the process id part of the id.
// It's a runtime error to call this method with an invalid id.
func (id ID) Pid() uint16 {
return binary.BigEndian.Uint16(id[7:9])
}
// Counter returns the incrementing value part of the id.
// It's a runtime error to call this method with an invalid id.
func (id ID) Counter() int32 {
b := id[9:12]
// Counter is stored as big-endian 3-byte value
return int32(uint32(b[0])<<16 | uint32(b[1])<<8 | uint32(b[2]))
}
// Value implements the driver.Valuer interface.
func (id ID) Value() (driver.Value, error) {
if id.IsNil() {
return nil, nil
}
b, err := id.MarshalText()
return string(b), err
}
// Scan implements the sql.Scanner interface.
func (id *ID) Scan(value interface{}) (err error) {
switch val := value.(type) {
case string:
return id.UnmarshalText([]byte(val))
case []byte:
return id.UnmarshalText(val)
case nil:
*id = nilID
return nil
default:
return fmt.Errorf("xid: scanning unsupported type: %T", value)
}
}
// IsNil Returns true if this is a "nil" ID
func (id ID) IsNil() bool {
return id == nilID
}
// NilID returns a zero value for `xid.ID`.
func NilID() ID {
return nilID
}
// Bytes returns the byte array representation of `ID`
func (id ID) Bytes() []byte {
return id[:]
}
// FromBytes convert the byte array representation of `ID` back to `ID`
func FromBytes(b []byte) (ID, error) {
var id ID
if len(b) != rawLen {
return id, ErrInvalidID
}
copy(id[:], b)
return id, nil
}
// Compare returns an integer comparing two IDs. It behaves just like `bytes.Compare`.
// The result will be 0 if two IDs are identical, -1 if current id is less than the other one,
// and 1 if current id is greater than the other.
func (id ID) Compare(other ID) int {
return bytes.Compare(id[:], other[:])
}
type sorter []ID
func (s sorter) Len() int {
return len(s)
}
func (s sorter) Less(i, j int) bool {
return s[i].Compare(s[j]) < 0
}
func (s sorter) Swap(i, j int) {
s[i], s[j] = s[j], s[i]
}
// Sort sorts an array of IDs inplace.
// It works by wrapping `[]ID` and use `sort.Sort`.
func Sort(ids []ID) {
sort.Sort(sorter(ids))
}