问题
Is there any way to get the functionality of the Sql Server 2005+ Sequential Guid generator without inserting records to read it back on round trip or invoking a native win dll call? I saw someone answer with a way of using rpcrt4.dll but I'm not sure if that would be able to work from my hosted environment for production.
Edit: Working with @John Boker's answer I attempted to turn it into more of a GuidComb generator instead of being dependent on the last generated Guid other than starting over. That for the seed instead of starting with Guid.Empty that I use
public SequentialGuid()
{
var tempGuid = Guid.NewGuid();
var bytes = tempGuid.ToByteArray();
var time = DateTime.Now;
bytes[3] = (byte) time.Year;
bytes[2] = (byte) time.Month;
bytes[1] = (byte) time.Day;
bytes[0] = (byte) time.Hour;
bytes[5] = (byte) time.Minute;
bytes[4] = (byte) time.Second;
CurrentGuid = new Guid(bytes);
}
I based that off the comments on
// 3 - the least significant byte in Guid ByteArray
[for SQL Server ORDER BY clause]
// 10 - the most significant byte in Guid ByteArray
[for SQL Server ORDERY BY clause]
SqlOrderMap = new[] {3, 2, 1, 0, 5, 4, 7, 6, 9, 8, 15, 14, 13, 12, 11, 10};
Does this look like the way I'd want to seed a guid with the DateTime or does it look like I should do it in reverse and work backwards from the end of the SqlOrderMap indexes? I'm not too concerned about their being a paging break anytime an initial guid would be created since it would only occur during application recycles.
回答1:
this person came up with something to make sequential guids, here's a link
http://developmenttips.blogspot.com/2008/03/generate-sequential-guids-for-sql.html
relevant code:
public class SequentialGuid {
Guid _CurrentGuid;
public Guid CurrentGuid {
get {
return _CurrentGuid;
}
}
public SequentialGuid() {
_CurrentGuid = Guid.NewGuid();
}
public SequentialGuid(Guid previousGuid) {
_CurrentGuid = previousGuid;
}
public static SequentialGuid operator++(SequentialGuid sequentialGuid) {
byte[] bytes = sequentialGuid._CurrentGuid.ToByteArray();
for (int mapIndex = 0; mapIndex < 16; mapIndex++) {
int bytesIndex = SqlOrderMap[mapIndex];
bytes[bytesIndex]++;
if (bytes[bytesIndex] != 0) {
break; // No need to increment more significant bytes
}
}
sequentialGuid._CurrentGuid = new Guid(bytes);
return sequentialGuid;
}
private static int[] _SqlOrderMap = null;
private static int[] SqlOrderMap {
get {
if (_SqlOrderMap == null) {
_SqlOrderMap = new int[16] {
3, 2, 1, 0, 5, 4, 7, 6, 9, 8, 15, 14, 13, 12, 11, 10
};
// 3 - the least significant byte in Guid ByteArray [for SQL Server ORDER BY clause]
// 10 - the most significant byte in Guid ByteArray [for SQL Server ORDERY BY clause]
}
return _SqlOrderMap;
}
}
}
回答2:
You could just use the same Win32 API function that SQL Server uses:
UuidCreateSequential
and apply some bit-shifting to put the values into big-endian order.
And since you want it in C#:
private class NativeMethods
{
[DllImport("rpcrt4.dll", SetLastError=true)]
public static extern int UuidCreateSequential(out Guid guid);
}
public static Guid NewSequentialID()
{
//Code is released into the public domain; no attribution required
const int RPC_S_OK = 0;
Guid guid;
int result = NativeMethods.UuidCreateSequential(out guid);
if (result != RPC_S_OK)
return Guid.NewGuid();
//Endian swap the UInt32, UInt16, and UInt16 into the big-endian order (RFC specified order) that SQL Server expects
//See https://stackoverflow.com/a/47682820/12597
//Short version: UuidCreateSequential writes out three numbers in litte, rather than big, endian order
var s = guid.ToByteArray();
var t = new byte[16];
//Endian swap UInt32
t[3] = s[0];
t[2] = s[1];
t[1] = s[2];
t[0] = s[3];
//Endian swap UInt16
t[5] = s[4];
t[4] = s[5];
//Endian swap UInt16
t[7] = s[6];
t[6] = s[7];
//The rest are already in the proper order
t[8] = s[8];
t[9] = s[9];
t[10] = s[10];
t[11] = s[11];
t[12] = s[12];
t[13] = s[13];
t[14] = s[14];
t[15] = s[15];
return new Guid(t);
}
See also
- Is there a .NET equalent to SQL Servers newsequentialid()
Microsoft's UuidCreateSequential
is just an implementation of a type 1 uuid from RFC 4122.
A uuid has three important parts:
node
: (6 bytes) - the computer's MAC addresstimestamp
: (7 bytes) - number of 100 ns intervals since 00:00:00.00, 15 October 1582 (the date of Gregorian reform to the Christian calendar)clockSequenceNumber
(2 bytes) - counter in case you generate a guid faster than 100ns, or you change your mac address
The basic algorithm is:
- obtain a system-wide lock
- read the last
node
,timestamp
andclockSequenceNumber
from persistent storage (registry/file) - get the current
node
(i.e. MAC address) - get the current
timestamp
- a) if the saved state was not available or corrupted, or the mac address has changed, generate a random
clockSequenceNumber
- b) if the state was available, but the current
timestamp
is the same or older than the saved timestamp, increment theclockSequenceNumber
- a) if the saved state was not available or corrupted, or the mac address has changed, generate a random
- save
node
,timestamp
andclockSequenceNumber
back to persistent storage - release the global lock
- format the guid structure according to the rfc
There is a 4-bit version number, and 2 bit variant that also need to be ANDed into the data:
guid = new Guid(
timestamp & 0xFFFFFFFF, //timestamp low
(timestamp >> 32) & 0xFFFF, //timestamp mid
((timestamp >> 40) & 0x0FFF), | (1 << 12) //timestamp high and version (version 1)
(clockSequenceNumber & 0x3F) | (0x80), //clock sequence number and reserved
node[0], node[1], node[2], node[3], node[4], node[5], node[6]);
Note: Completely untested; i just eyeballed it from the RFC.
- the byte order might have to be changed (Here is byte order for sql server)
- you might want to create your own version, e.g. Version 6 (version 1-5 are defined). That way you're guaranteed to be universally unique
回答3:
Here is how NHibernate implements the Guid.Comb algorithm:
private Guid GenerateComb()
{
byte[] guidArray = Guid.NewGuid().ToByteArray();
DateTime baseDate = new DateTime(1900, 1, 1);
DateTime now = DateTime.Now;
// Get the days and milliseconds which will be used to build the byte string
TimeSpan days = new TimeSpan(now.Ticks - baseDate.Ticks);
TimeSpan msecs = now.TimeOfDay;
// Convert to a byte array
// Note that SQL Server is accurate to 1/300th of a millisecond so we divide by 3.333333
byte[] daysArray = BitConverter.GetBytes(days.Days);
byte[] msecsArray = BitConverter.GetBytes((long) (msecs.TotalMilliseconds / 3.333333));
// Reverse the bytes to match SQL Servers ordering
Array.Reverse(daysArray);
Array.Reverse(msecsArray);
// Copy the bytes into the guid
Array.Copy(daysArray, daysArray.Length - 2, guidArray, guidArray.Length - 6, 2);
Array.Copy(msecsArray, msecsArray.Length - 4, guidArray, guidArray.Length - 4, 4);
return new Guid(guidArray);
}
回答4:
A Sequential guid that updates often (at least 3 times per milisecond), can be found here. It is create with regular C# code (no native code call).
回答5:
C# Version
public static Guid ToSeqGuid()
{
Int64 lastTicks = -1;
long ticks = System.DateTime.UtcNow.Ticks;
if (ticks <= lastTicks)
{
ticks = lastTicks + 1;
}
lastTicks = ticks;
byte[] ticksBytes = BitConverter.GetBytes(ticks);
Array.Reverse(ticksBytes);
Guid myGuid = new Guid();
byte[] guidBytes = myGuid.ToByteArray();
Array.Copy(ticksBytes, 0, guidBytes, 10, 6);
Array.Copy(ticksBytes, 6, guidBytes, 8, 2);
Guid newGuid = new Guid(guidBytes);
string filepath = @"C:\temp\TheNewGuids.txt";
using (StreamWriter writer = new StreamWriter(filepath, true))
{
writer.WriteLine("GUID Created = " + newGuid.ToString());
}
return newGuid;
}
}
}
回答6:
Maybe interesting to compare with the other suggestions:
EntityFramework Core also implements a sequentialGuidValueGenerator. They generate randoms guids for each value and only change the most significant bytes based on a timestamp and thread-safe increments for sorting in SQL Server.
source link
This leads to values that are all very different but with a timestamp sortable.
回答7:
My solution (in VB but easy to convert). It changes the most significant (for SQL Server sorting) first 8 bytes of the GUID to DateTime.UtcNow.Ticks and also has extra code to help the issue of getting the same Ticks multiple times if you call for a new GUID faster than the system clock updates.
Private ReadOnly _toSeqGuidLock As New Object()
''' <summary>
''' Replaces the most significant eight bytes of the GUID (according to SQL Server ordering) with the current UTC-timestamp.
''' </summary>
''' <remarks>Thread-Safe</remarks>
<System.Runtime.CompilerServices.Extension()> _
Public Function ToSeqGuid(ByVal guid As Guid) As Guid
Static lastTicks As Int64 = -1
Dim ticks = DateTime.UtcNow.Ticks
SyncLock _toSeqGuidLock
If ticks <= lastTicks Then
ticks = lastTicks + 1
End If
lastTicks = ticks
End SyncLock
Dim ticksBytes = BitConverter.GetBytes(ticks)
Array.Reverse(ticksBytes)
Dim guidBytes = guid.ToByteArray()
Array.Copy(ticksBytes, 0, guidBytes, 10, 6)
Array.Copy(ticksBytes, 6, guidBytes, 8, 2)
Return New Guid(guidBytes)
End Function
回答8:
I just took the NHibernate based answer by Moslem Ben Dhaou and made it an extension function:
using System;
namespace Atlas.Core.Kernel.Extensions
{
public static class Guids
{
public static Guid Comb(this Guid source)
{
byte[] guidArray = source.ToByteArray();
DateTime baseDate = new DateTime(1900, 1, 1);
DateTime now = DateTime.Now;
// Get the days and milliseconds which will be used to build the byte string
TimeSpan days = new TimeSpan(now.Ticks - baseDate.Ticks);
TimeSpan msecs = now.TimeOfDay;
// Convert to a byte array
// Note that SQL Server is accurate to 1/300th of a millisecond so we divide by 3.333333
byte[] daysArray = BitConverter.GetBytes(days.Days);
byte[] msecsArray = BitConverter.GetBytes((long)(msecs.TotalMilliseconds / 3.333333));
// Reverse the bytes to match SQL Servers ordering
Array.Reverse(daysArray);
Array.Reverse(msecsArray);
// Copy the bytes into the guid
Array.Copy(daysArray, daysArray.Length - 2, guidArray, guidArray.Length - 6, 2);
Array.Copy(msecsArray, msecsArray.Length - 4, guidArray, guidArray.Length - 4, 4);
return new Guid(guidArray);
}
}
}
回答9:
As far I know NHibernate have special generator, called GuidCombGenerator. You can look on it.
回答10:
Not specifically guid but I now normally use a Snowflake style sequential id generator. The same benefits of a guid while having even better clustered index compatibility than a sequential guid.
Flakey for .NET Core
IdGen for .NET Framework
回答11:
I just saw this question... I happen to be the author of a small open-source .NET library for generating COMB-style GUIDs.
The library supports both the original method (compatible with SQL Server's datetime
type) and one using Unix timestamps, which have more time precision. It also includes a variant that works better for PostgrSQL:
https://github.com/richardtallent/RT.Comb
来源:https://stackoverflow.com/questions/1752004/sequential-guid-generator