我经常遇到这样的情况:我想在声明查询的地方对查询进行求值。这通常是因为我需要对它进行多次迭代，计算成本很高。例如:
string raw = "...";
var lines = (from l in raw.Split('\n')
let ll = l.Trim()
where !string.IsNullOrEmpty(ll)
select ll).ToList();
这很好。但是如果我不打算修改结果，那么我也可以调用ToArray()而不是ToList()。
然而，我想知道ToArray()是否通过首先调用ToList()来实现，因此内存效率比只调用ToList()低。
我疯了吗?我是否应该调用ToArray() -在知道内存不会被分配两次的情况下安全可靠?
当前回答
我发现人们在这里做的其他基准测试都有不足，所以这里是我的尝试。如果你发现我的方法有问题，请告诉我。
/* This is a benchmarking template I use in LINQPad when I want to do a
* quick performance test. Just give it a couple of actions to test and
* it will give you a pretty good idea of how long they take compared
* to one another. It's not perfect: You can expect a 3% error margin
* under ideal circumstances. But if you're not going to improve
* performance by more than 3%, you probably don't care anyway.*/
void Main()
{
// Enter setup code here
var values = Enumerable.Range(1, 100000)
.Select(i => i.ToString())
.ToArray()
.Select(i => i);
values.GetType().Dump();
var actions = new[]
{
new TimedAction("ToList", () =>
{
values.ToList();
}),
new TimedAction("ToArray", () =>
{
values.ToArray();
}),
new TimedAction("Control", () =>
{
foreach (var element in values)
{
// do nothing
}
}),
// Add tests as desired
};
const int TimesToRun = 1000; // Tweak this as necessary
TimeActions(TimesToRun, actions);
}
#region timer helper methods
// Define other methods and classes here
public void TimeActions(int iterations, params TimedAction[] actions)
{
Stopwatch s = new Stopwatch();
int length = actions.Length;
var results = new ActionResult[actions.Length];
// Perform the actions in their initial order.
for (int i = 0; i < length; i++)
{
var action = actions[i];
var result = results[i] = new ActionResult { Message = action.Message };
// Do a dry run to get things ramped up/cached
result.DryRun1 = s.Time(action.Action, 10);
result.FullRun1 = s.Time(action.Action, iterations);
}
// Perform the actions in reverse order.
for (int i = length - 1; i >= 0; i--)
{
var action = actions[i];
var result = results[i];
// Do a dry run to get things ramped up/cached
result.DryRun2 = s.Time(action.Action, 10);
result.FullRun2 = s.Time(action.Action, iterations);
}
results.Dump();
}
public class ActionResult
{
public string Message { get; set; }
public double DryRun1 { get; set; }
public double DryRun2 { get; set; }
public double FullRun1 { get; set; }
public double FullRun2 { get; set; }
}
public class TimedAction
{
public TimedAction(string message, Action action)
{
Message = message;
Action = action;
}
public string Message { get; private set; }
public Action Action { get; private set; }
}
public static class StopwatchExtensions
{
public static double Time(this Stopwatch sw, Action action, int iterations)
{
sw.Restart();
for (int i = 0; i < iterations; i++)
{
action();
}
sw.Stop();
return sw.Elapsed.TotalMilliseconds;
}
}
#endregion
你可以在这里下载LINQPad脚本。
结果:
调整上面的代码，你会发现:
当处理较小的数组时，差异就不那么显著了。 在处理整型而不是字符串时，这种差异不太显著。 使用大型结构体而不是字符串通常会花费更多的时间，但并不会真正改变比例。
这与投票最多的答案的结论一致:
除非您的代码经常生成许多大型数据列表，否则不太可能注意到性能上的差异。(当创建1000个包含100K字符串的列表时，只有200ms的差异。) ToList()始终运行得更快，如果不打算长时间保留结果，那么它是一个更好的选择。
更新
@JonHanna指出，根据Select的实现，ToList()或ToArray()实现可以提前预测结果集合的大小。将上面代码中的. select (i => i)替换为Where(i => true)会产生非常相似的结果，并且更有可能这样做，而不管. net实现如何。
其他回答
编辑2:(这是对原始答案的更正)
使用基准。NET，我们可以通过性能测量来确认，公认的答案实际上是正确的:ToList在一般情况下更快，因为它不需要从已分配的缓冲区中修剪空空间。ToArray可能会执行额外的分配和复制操作，以使缓冲区的大小精确到元素的数量。
为了确认这一点，使用下面的基准测试。
[MemoryDiagnoser]
[ShortRunJob]
public class Benchmarks
{
[Params(0, 1, 6, 10, 42, 100, 1337, 10000)]
public int Count { get; set; }
public IEnumerable<int> Items => Enumerable.Range(0, Count).Where(i => i > 0);
[Benchmark(Baseline = true)]
public int[] ToArray() => Items.ToArray();
[Benchmark]
public List<int> ToList() => Items.ToList();
}
结果证实，在大多数情况下，ToList要快10% - 15%。
BenchmarkDotNet=v0.13.1, OS=Windows 10.0.22000
Intel Core i9-10885H CPU 2.40GHz, 1 CPU, 16 logical and 8 physical cores
.NET SDK=6.0.302
[Host] : .NET 6.0.7 (6.0.722.32202), X64 RyuJIT
DefaultJob : .NET 6.0.7 (6.0.722.32202), X64 RyuJIT
| Method | Count | Mean | Error | StdDev | Ratio | RatioSD | Gen 0 | Gen 1 | Allocated |
|-------- |------ |-------------:|-----------:|-----------:|------:|--------:|--------:|-------:|----------:|
| ToArray | 0 | 29.73 ns | 0.546 ns | 0.536 ns | 1.00 | 0.00 | 0.0067 | - | 56 B |
| ToList | 0 | 31.51 ns | 0.485 ns | 0.405 ns | 1.06 | 0.02 | 0.0105 | - | 88 B |
| | | | | | | | | | |
| ToArray | 1 | 37.36 ns | 0.314 ns | 0.294 ns | 1.00 | 0.00 | 0.0114 | - | 96 B |
| ToList | 1 | 36.75 ns | 0.605 ns | 0.537 ns | 0.98 | 0.01 | 0.0153 | - | 128 B |
| | | | | | | | | | |
| ToArray | 6 | 100.05 ns | 1.522 ns | 1.349 ns | 1.00 | 0.00 | 0.0286 | - | 240 B |
| ToList | 6 | 85.16 ns | 0.808 ns | 0.756 ns | 0.85 | 0.01 | 0.0267 | - | 224 B |
| | | | | | | | | | |
| ToArray | 10 | 137.20 ns | 2.766 ns | 2.452 ns | 1.00 | 0.00 | 0.0372 | - | 312 B |
| ToList | 10 | 123.05 ns | 2.198 ns | 1.949 ns | 0.90 | 0.01 | 0.0372 | - | 312 B |
| | | | | | | | | | |
| ToArray | 42 | 398.25 ns | 6.583 ns | 5.836 ns | 1.00 | 0.00 | 0.0877 | - | 736 B |
| ToList | 42 | 352.04 ns | 4.976 ns | 4.411 ns | 0.88 | 0.02 | 0.0887 | - | 744 B |
| | | | | | | | | | |
| ToArray | 100 | 730.80 ns | 6.501 ns | 6.081 ns | 1.00 | 0.00 | 0.1488 | - | 1,248 B |
| ToList | 100 | 705.49 ns | 9.947 ns | 9.305 ns | 0.97 | 0.01 | 0.1526 | - | 1,280 B |
| | | | | | | | | | |
| ToArray | 1337 | 8,023.57 ns | 147.388 ns | 137.867 ns | 1.00 | 0.00 | 1.6785 | 0.0458 | 14,056 B |
| ToList | 1337 | 7,980.27 ns | 138.469 ns | 122.749 ns | 1.00 | 0.02 | 1.9989 | 0.1221 | 16,736 B |
| | | | | | | | | | |
| ToArray | 10000 | 57,037.19 ns | 510.492 ns | 452.538 ns | 1.00 | 0.00 | 12.6343 | 1.7700 | 106,280 B |
| ToList | 10000 | 57,728.15 ns | 583.353 ns | 517.127 ns | 1.01 | 0.01 | 15.5640 | 5.1270 | 131,496 B |
作为参考，下面是原始答案，不幸的是，它只在一个非常特殊的情况下执行基准测试，避免了中间的调整大小和复制操作。
最初的回答:
现在已经是2020年了，每个人都在使用。net Core 3.1，所以我决定用Benchmark.NET运行一些基准测试。
TL;DR: ToArray()在性能方面更好，如果不打算改变集合，则可以更好地传达意图。
编辑:从注释中可以看出，这些基准测试可能不是指示性的，因为Enumerable.Range(…)返回一个IEnumerable<T>，其中包含关于序列大小的信息，随后在ToArray()的优化中使用它来预分配正确大小的数组。考虑为您的具体场景手动测试性能。
[MemoryDiagnoser]
public class Benchmarks
{
[Params(0, 1, 6, 10, 39, 100, 666, 1000, 1337, 10000)]
public int Count { get; set; }
public IEnumerable<int> Items => Enumerable.Range(0, Count);
[Benchmark(Description = "ToArray()", Baseline = true)]
public int[] ToArray() => Items.ToArray();
[Benchmark(Description = "ToList()")]
public List<int> ToList() => Items.ToList();
public static void Main() => BenchmarkRunner.Run<Benchmarks>();
}
结果如下:
BenchmarkDotNet=v0.12.0, OS=Windows 10.0.14393.3443 (1607/AnniversaryUpdate/Redstone1)
Intel Core i5-4460 CPU 3.20GHz (Haswell), 1 CPU, 4 logical and 4 physical cores
Frequency=3124994 Hz, Resolution=320.0006 ns, Timer=TSC
.NET Core SDK=3.1.100
[Host] : .NET Core 3.1.0 (CoreCLR 4.700.19.56402, CoreFX 4.700.19.56404), X64 RyuJIT
DefaultJob : .NET Core 3.1.0 (CoreCLR 4.700.19.56402, CoreFX 4.700.19.56404), X64 RyuJIT
| Method | Count | Mean | Error | StdDev | Median | Ratio | RatioSD | Gen 0 | Gen 1 | Gen 2 | Allocated |
|---------- |------ |--------------:|------------:|------------:|--------------:|------:|--------:|--------:|------:|------:|----------:|
| ToArray() | 0 | 7.357 ns | 0.2096 ns | 0.1960 ns | 7.323 ns | 1.00 | 0.00 | - | - | - | - |
| ToList() | 0 | 13.174 ns | 0.2094 ns | 0.1958 ns | 13.084 ns | 1.79 | 0.05 | 0.0102 | - | - | 32 B |
| | | | | | | | | | | | |
| ToArray() | 1 | 23.917 ns | 0.4999 ns | 0.4676 ns | 23.954 ns | 1.00 | 0.00 | 0.0229 | - | - | 72 B |
| ToList() | 1 | 33.867 ns | 0.7350 ns | 0.6876 ns | 34.013 ns | 1.42 | 0.04 | 0.0331 | - | - | 104 B |
| | | | | | | | | | | | |
| ToArray() | 6 | 28.242 ns | 0.5071 ns | 0.4234 ns | 28.196 ns | 1.00 | 0.00 | 0.0280 | - | - | 88 B |
| ToList() | 6 | 43.516 ns | 0.9448 ns | 1.1949 ns | 42.896 ns | 1.56 | 0.06 | 0.0382 | - | - | 120 B |
| | | | | | | | | | | | |
| ToArray() | 10 | 31.636 ns | 0.5408 ns | 0.4516 ns | 31.657 ns | 1.00 | 0.00 | 0.0331 | - | - | 104 B |
| ToList() | 10 | 53.870 ns | 1.2988 ns | 2.2403 ns | 53.415 ns | 1.77 | 0.07 | 0.0433 | - | - | 136 B |
| | | | | | | | | | | | |
| ToArray() | 39 | 58.896 ns | 0.9441 ns | 0.8369 ns | 58.548 ns | 1.00 | 0.00 | 0.0713 | - | - | 224 B |
| ToList() | 39 | 138.054 ns | 2.8185 ns | 3.2458 ns | 138.937 ns | 2.35 | 0.08 | 0.0815 | - | - | 256 B |
| | | | | | | | | | | | |
| ToArray() | 100 | 119.167 ns | 1.6195 ns | 1.4357 ns | 119.120 ns | 1.00 | 0.00 | 0.1478 | - | - | 464 B |
| ToList() | 100 | 274.053 ns | 5.1073 ns | 4.7774 ns | 272.242 ns | 2.30 | 0.06 | 0.1578 | - | - | 496 B |
| | | | | | | | | | | | |
| ToArray() | 666 | 569.920 ns | 11.4496 ns | 11.2450 ns | 571.647 ns | 1.00 | 0.00 | 0.8688 | - | - | 2728 B |
| ToList() | 666 | 1,621.752 ns | 17.1176 ns | 16.0118 ns | 1,623.566 ns | 2.85 | 0.05 | 0.8793 | - | - | 2760 B |
| | | | | | | | | | | | |
| ToArray() | 1000 | 796.705 ns | 16.7091 ns | 19.8910 ns | 796.610 ns | 1.00 | 0.00 | 1.2951 | - | - | 4064 B |
| ToList() | 1000 | 2,453.110 ns | 48.1121 ns | 65.8563 ns | 2,460.190 ns | 3.09 | 0.10 | 1.3046 | - | - | 4096 B |
| | | | | | | | | | | | |
| ToArray() | 1337 | 1,057.983 ns | 20.9810 ns | 41.4145 ns | 1,041.028 ns | 1.00 | 0.00 | 1.7223 | - | - | 5416 B |
| ToList() | 1337 | 3,217.550 ns | 62.3777 ns | 61.2633 ns | 3,203.928 ns | 2.98 | 0.13 | 1.7357 | - | - | 5448 B |
| | | | | | | | | | | | |
| ToArray() | 10000 | 7,309.844 ns | 160.0343 ns | 141.8662 ns | 7,279.387 ns | 1.00 | 0.00 | 12.6572 | - | - | 40064 B |
| ToList() | 10000 | 23,858.032 ns | 389.6592 ns | 364.4874 ns | 23,759.001 ns | 3.26 | 0.08 | 12.6343 | - | - | 40096 B |
// * Hints *
Outliers
Benchmarks.ToArray(): Default -> 2 outliers were removed (35.20 ns, 35.29 ns)
Benchmarks.ToArray(): Default -> 2 outliers were removed (38.51 ns, 38.88 ns)
Benchmarks.ToList(): Default -> 1 outlier was removed (64.69 ns)
Benchmarks.ToArray(): Default -> 1 outlier was removed (67.02 ns)
Benchmarks.ToArray(): Default -> 1 outlier was removed (130.08 ns)
Benchmarks.ToArray(): Default -> 1 outlier was detected (541.82 ns)
Benchmarks.ToArray(): Default -> 1 outlier was removed (7.82 us)
// * Legends *
Count : Value of the 'Count' parameter
Mean : Arithmetic mean of all measurements
Error : Half of 99.9% confidence interval
StdDev : Standard deviation of all measurements
Median : Value separating the higher half of all measurements (50th percentile)
Ratio : Mean of the ratio distribution ([Current]/[Baseline])
RatioSD : Standard deviation of the ratio distribution ([Current]/[Baseline])
Gen 0 : GC Generation 0 collects per 1000 operations
Gen 1 : GC Generation 1 collects per 1000 operations
Gen 2 : GC Generation 2 collects per 1000 operations
Allocated : Allocated memory per single operation (managed only, inclusive, 1KB = 1024B)
1 ns : 1 Nanosecond (0.000000001 sec)
内存总是会被分配两次——或者类似的情况。由于不能调整数组的大小，这两种方法都将使用某种机制在不断增长的集合中收集数据。(好吧，这个名单本身就是一个不断增长的集合。)
List使用数组作为内部存储，并在需要时将容量增加一倍。这意味着平均2/3的项目至少被重新分配过一次，其中一半至少被重新分配过两次，一半至少被重新分配过三次，以此类推。这意味着每个项目平均被重新分配了1.3次，这并不是很大的开销。
还要记住，如果你在收集字符串，集合本身只包含对字符串的引用，字符串本身不会被重新分配。
一种选择是添加自己的扩展方法，该方法返回一个只读的ICollection<T>。当您既不想使用数组/列表的索引属性，也不想从列表中添加/删除时，这可能比使用ToList或ToArray更好。
public static class EnumerableExtension
{
/// <summary>
/// Causes immediate evaluation of the linq but only if required.
/// As it returns a readonly ICollection, is better than using ToList or ToArray
/// when you do not want to use the indexing properties of an IList, or add to the collection.
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="enumerable"></param>
/// <returns>Readonly collection</returns>
public static ICollection<T> Evaluate<T>(this IEnumerable<T> enumerable)
{
//if it's already a readonly collection, use it
var collection = enumerable as ICollection<T>;
if ((collection != null) && collection.IsReadOnly)
{
return collection;
}
//or make a new collection
return enumerable.ToList().AsReadOnly();
}
}
单元测试:
[TestClass]
public sealed class EvaluateLinqTests
{
[TestMethod]
public void EvalTest()
{
var list = new List<int> {1, 2, 3};
var linqResult = list.Select(i => i);
var linqResultEvaluated = list.Select(i => i).Evaluate();
list.Clear();
Assert.AreEqual(0, linqResult.Count());
//even though we have cleared the underlying list, the evaluated list does not change
Assert.AreEqual(3, linqResultEvaluated.Count());
}
[TestMethod]
public void DoesNotSaveCreatingListWhenHasListTest()
{
var list = new List<int> {1, 2, 3};
var linqResultEvaluated = list.Evaluate();
//list is not readonly, so we expect a new list
Assert.AreNotSame(list, linqResultEvaluated);
}
[TestMethod]
public void SavesCreatingListWhenHasReadonlyListTest()
{
var list = new List<int> {1, 2, 3}.AsReadOnly();
var linqResultEvaluated = list.Evaluate();
//list is readonly, so we don't expect a new list
Assert.AreSame(list, linqResultEvaluated);
}
[TestMethod]
public void SavesCreatingListWhenHasArrayTest()
{
var list = new[] {1, 2, 3};
var linqResultEvaluated = list.Evaluate();
//arrays are readonly (wrt ICollection<T> interface), so we don't expect a new object
Assert.AreSame(list, linqResultEvaluated);
}
[TestMethod]
[ExpectedException(typeof (NotSupportedException))]
public void CantAddToResultTest()
{
var list = new List<int> {1, 2, 3};
var linqResultEvaluated = list.Evaluate();
Assert.AreNotSame(list, linqResultEvaluated);
linqResultEvaluated.Add(4);
}
[TestMethod]
[ExpectedException(typeof (NotSupportedException))]
public void CantRemoveFromResultTest()
{
var list = new List<int> {1, 2, 3};
var linqResultEvaluated = list.Evaluate();
Assert.AreNotSame(list, linqResultEvaluated);
linqResultEvaluated.Remove(1);
}
}
对于任何有兴趣在其他Linq-to-sql中使用此结果的人，例如
from q in context.MyTable
where myListOrArray.Contains(q.someID)
select q;
那么生成的SQL是相同的，无论你使用List或Array为myListOrArray。 现在我知道有些人可能会问为什么在这条语句之前枚举，但从IQueryable vs(列表或数组)生成的SQL之间是有区别的。
您应该根据理想的设计选择来决定使用ToList还是ToArray。如果您想要一个只能通过索引迭代和访问的集合，请选择ToArray。如果您希望以后能够轻松地从集合中添加和删除额外的功能，那么可以使用ToList(并不是说您不能添加到数组中，但这通常不是合适的工具)。
如果性能很重要，您还应该考虑哪些操作会更快。实际上，您不会调用ToList或ToArray一百万次，但可能会对获得的集合进行一百万次操作。在这方面[]更好，因为List<>是[]，有一些开销。查看这个线程的一些效率比较:List<int>或int[]
在我自己不久前的测试中，我发现ToArray更快。我不确定这些测试有多偏颇。然而，性能差异是如此微不足道，只有在循环运行这些查询数百万次时才能明显看出。
