gpt4 book ai didi

c# - 构建线程安全的 GUID 增量器

转载 作者:太空狗 更新时间:2023-10-29 23:05:42 25 4
gpt4 key购买 nike

在我下面的代码中,我锁定了 guid,以尝试使其线程安全。在我的示例应用程序中,大约每运行 10 次我就会得到一个“重复 key ”。也就是说,我得到了一个副本,这不是我需要的。

有什么方法可以使“.NextGuid”线程安全吗?

using System;    
namespace MyConsoleOne.BAL
{
public class GuidStore
{
private static object objectlock = new object();
private Guid StartingGuid { get; set; }
private Guid? LastGuidHolder { get; set; }
public GuidStore(Guid startingGuid)
{
this.StartingGuid = startingGuid;
}

public Guid? GetNextGuid()
{
lock (objectlock)
{
if (this.LastGuidHolder.HasValue)
{
this.LastGuidHolder = Increment(this.LastGuidHolder.Value);
}
else
{
this.LastGuidHolder = Increment(this.StartingGuid);
}
}
return this.LastGuidHolder;
}

private Guid Increment(Guid guid)
{
byte[] bytes = guid.ToByteArray();
byte[] order = { 15, 14, 13, 12, 11, 10, 9, 8, 6, 7, 4, 5, 0, 1, 2, 3 };
for (int i = 0; i < 16; i++)
{
if (bytes[order[i]] == byte.MaxValue)
{
bytes[order[i]] = 0;
}
else
{
bytes[order[i]]++;
return new Guid(bytes);
}
}
throw new OverflowException("Guid.Increment failed.");
}
}
}

using MyConsoleOne.BAL;
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;

namespace MyConsoleOne
{
class Program
{
static void Main(string[] args)
{
GuidStore gs = new GuidStore(Guid.NewGuid());

for (int i = 0; i < 1000; i++)
{
Console.WriteLine(i);
Dictionary<Guid, int> guids = new Dictionary<Guid, int>();
Parallel.For(0, 1000, j =>
{
Guid? currentGuid = gs.GetNextGuid();
guids.Add(currentGuid.Value, j);
Console.WriteLine(currentGuid);
}); // Parallel.For
}
Console.WriteLine("Press ENTER to Exit");
Console.ReadLine();
}
}
}

我的代码是以下内容的组合:

由于我收到“为什么不使用 Guid.NewGuid”的问题,我将在此处提供原因:

我有一个父进程,它有一个由 Guid.NewGuid() 创建的唯一标识符。我将把它称为“父 guid”。该父进程将创建 N 个文件。如果我从头开始编写,我会在文件名末尾附加“N”。因此,如果父 Guid 是“11111111-1111-1111-1111-111111111111”,例如,我会写文件

"11111111-1111-1111-1111-111111111111_1.txt"
"11111111-1111-1111-1111-111111111111_2.txt"
"11111111-1111-1111-1111-111111111111_3.txt"

等。但是,通过与客户端的现有“契约(Contract)”:::文件名中必须有一个(唯一的)Guid,并且没有那个“N”(1、2 等)值在文件名中(这个“契约(Contract)”已经存在多年,所以它几乎是一成不变的)。有了这里列出的功能,我将能够保留“契约(Contract)”,但文件名松散地绑定(bind)到“父”Guid(父级再次由 Guid.NewGuid() 生成)。冲突不是文件名的问题(它们被放在一个不同的文件夹下用于“进程”执行)。碰撞是“父”Guid 的一个问题。但同样,这已经用 Guid.NewGuid 处理了。

因此,使用“11111111-1111-1111-1111-111111111111”的起始 Guid,我将能够编写如下文件名:

OTHERSTUFF_111111111-1111-1111-1111-111111111112_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-111111111113_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-111111111114_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-111111111115_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-111111111116_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-111111111117_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-111111111118_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-111111111119_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-11111111111a_MORESTUFF.txt
OTHERSTUFF_111111111-1111-1111-1111-11111111111b_MORESTUFF.txt

所以在我上面的示例中,“父 guid”由“this.StartingGuid”表示......然后我从中得到“递增”的 guid。

还有。我可以编写更好的单元测试,因为现在我可以提前知道文件名。

追加:

最终代码版本:

public class GuidStore
{
private static object objectlock = new object();

private static int[] byteOrder = { 15, 14, 13, 12, 11, 10, 9, 8, 6, 7, 4, 5, 0, 1, 2, 3 };

private Guid StartingGuid { get; set; }

private Guid? LastGuidHolder { get; set; }

public GuidStore(Guid startingGuid)
{
this.StartingGuid = startingGuid;
}

public Guid GetNextGuid()
{
return this.GetNextGuid(0);
}

public Guid GetNextGuid(int firstGuidOffSet)
{
lock (objectlock)
{
if (this.LastGuidHolder.HasValue)
{
this.LastGuidHolder = Increment(this.LastGuidHolder.Value);
}
else
{
this.LastGuidHolder = Increment(this.StartingGuid);
for (int i = 0; i < firstGuidOffSet; i++)
{
this.LastGuidHolder = Increment(this.LastGuidHolder.Value);
}
}

return this.LastGuidHolder.Value;
}
}

private static Guid Increment(Guid guid)
{
var bytes = guid.ToByteArray();
var canIncrement = byteOrder.Any(i => ++bytes[i] != 0);
return new Guid(canIncrement ? bytes : new byte[16]);
}
}

和单元测试:

public class GuidStoreUnitTests
{
[TestMethod]
public void GetNextGuidSimpleTest()
{
Guid startingGuid = new Guid("11111111-1111-1111-1111-111111111111");
GuidStore gs = new GuidStore(startingGuid);


List<Guid> guids = new List<Guid>();

const int GuidCount = 10;

for (int i = 0; i < GuidCount; i++)
{
guids.Add(gs.GetNextGuid());
}

Assert.IsNotNull(guids);
Assert.AreEqual(GuidCount, guids.Count);
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-111111111112")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-111111111113")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-111111111114")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-111111111115")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-111111111116")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-111111111117")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-111111111118")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-111111111119")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-11111111111a")));
Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-11111111111b")));
}

[TestMethod]
public void GetNextGuidWithOffsetSimpleTest()
{
Guid startingGuid = new Guid("11111111-1111-1111-1111-111111111111");
GuidStore gs = new GuidStore(startingGuid);

List<Guid> guids = new List<Guid>();

const int OffSet = 10;

guids.Add(gs.GetNextGuid(OffSet));

Assert.IsNotNull(guids);
Assert.AreEqual(1, guids.Count);

Assert.IsNotNull(guids.FirstOrDefault(g => g == new Guid("11111111-1111-1111-1111-11111111111c")));
}

[TestMethod]
public void GetNextGuidMaxRolloverTest()
{
Guid startingGuid = new Guid("ffffffff-ffff-ffff-ffff-ffffffffffff");
GuidStore gs = new GuidStore(startingGuid);

List<Guid> guids = new List<Guid>();

const int OffSet = 10;

guids.Add(gs.GetNextGuid(OffSet));

Assert.IsNotNull(guids);
Assert.AreEqual(1, guids.Count);

Assert.IsNotNull(guids.FirstOrDefault(g => g == Guid.Empty));
}

[TestMethod]
public void GetNextGuidThreadSafeTest()
{
Guid startingGuid = Guid.NewGuid();
GuidStore gs = new GuidStore(startingGuid);

/* The "key" of the ConcurrentDictionary must be unique, so this will catch any duplicates */
ConcurrentDictionary<Guid, int> guids = new ConcurrentDictionary<Guid, int>();
Parallel.For(
0,
1000,
j =>
{
Guid currentGuid = gs.GetNextGuid();
if (!guids.TryAdd(currentGuid, j))
{
throw new ArgumentOutOfRangeException("GuidStore.GetNextGuid ThreadSafe Test Failed");
}
}); // Parallel.For
}

[TestMethod]
public void GetNextGuidTwoRunsProduceSameResultsTest()
{
Guid startingGuid = Guid.NewGuid();

GuidStore gsOne = new GuidStore(startingGuid);

/* The "key" of the ConcurrentDictionary must be unique, so this will catch any duplicates */
ConcurrentDictionary<Guid, int> setOneGuids = new ConcurrentDictionary<Guid, int>();
Parallel.For(
0,
1000,
j =>
{
Guid currentGuid = gsOne.GetNextGuid();
if (!setOneGuids.TryAdd(currentGuid, j))
{
throw new ArgumentOutOfRangeException("GuidStore.GetNextGuid ThreadSafe Test Failed");
}
}); // Parallel.For

gsOne = null;

GuidStore gsTwo = new GuidStore(startingGuid);

/* The "key" of the ConcurrentDictionary must be unique, so this will catch any duplicates */
ConcurrentDictionary<Guid, int> setTwoGuids = new ConcurrentDictionary<Guid, int>();
Parallel.For(
0,
1000,
j =>
{
Guid currentGuid = gsTwo.GetNextGuid();
if (!setTwoGuids.TryAdd(currentGuid, j))
{
throw new ArgumentOutOfRangeException("GuidStore.GetNextGuid ThreadSafe Test Failed");
}
}); // Parallel.For

bool equal = setOneGuids.Select(g => g.Key).OrderBy(i => i).SequenceEqual(
setTwoGuids.Select(g => g.Key).OrderBy(i => i), new GuidComparer<Guid>());

Assert.IsTrue(equal);
}
}

internal class GuidComparer<Guid> : IEqualityComparer<Guid>
{
public bool Equals(Guid x, Guid y)
{
return x.Equals(y);
}

public int GetHashCode(Guid obj)
{
return 0;
}
}

最佳答案

这里有两个问题:

  1. Dictionary.Add() 不是线程安全的。请改用 ConcurrentDictionary.TryAdd()
  2. 您的 GetNextGuid() 实现有一个竞争条件,因为您在锁外返回 this.LastGuidHolder,所以它可能在它返回之前被另一个线程修改返回。

一个明显的解决方案是将返回移动到锁内:

public Guid? GetNextGuid()
{
lock (objectlock)
{
if (this.LastGuidHolder.HasValue)
{
this.LastGuidHolder = Increment(this.LastGuidHolder.Value);
}
else
{
this.LastGuidHolder = Increment(this.StartingGuid);
}

return this.LastGuidHolder;
}
}

但是,我会将返回类型更改为 Guid - 返回 Guid? 似乎没有任何作用 - 这是应该隐藏的东西类内:

public Guid GetNextGuid()
{
lock (objectlock)
{
if (this.LastGuidHolder.HasValue)
{
this.LastGuidHolder = Increment(this.LastGuidHolder.Value);
}
else
{
this.LastGuidHolder = Increment(this.StartingGuid);
}

return this.LastGuidHolder.Value;
}
}

这是使用 ConcurrentDictionary 的测试方法的一个版本:

static void Main(string[] args)
{
GuidStore gs = new GuidStore(Guid.NewGuid());

for (int i = 0; i < 1000; i++)
{
Console.WriteLine(i);
ConcurrentDictionary<Guid, int> guids = new ConcurrentDictionary<Guid, int>();
Parallel.For(0, 1000, j =>
{
Guid currentGuid = gs.GetNextGuid();
if (!guids.TryAdd(currentGuid, j))
{
Console.WriteLine("Duplicate found!");
}
}); // Parallel.For
}

Console.WriteLine("Press ENTER to Exit");
Console.ReadLine();
}

说了这么多,我不明白为什么你不只是使用 Guid.NewGuid()...

关于c# - 构建线程安全的 GUID 增量器,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/40885081/

25 4 0
Copyright 2021 - 2024 cfsdn All Rights Reserved 蜀ICP备2022000587号
广告合作:1813099741@qq.com 6ren.com