我的理解
- 在不发生错误时尝试 Catch 不重
- 出错的惩罚很大
- 最好在设计时考虑输入的不可靠程度和执行频率。
- 我惊讶地发现仅仅创建一个 try catch 块会花费很多。
⇒ 毕竟，“不要猜测，测量”是最好的
想确认
- System.Text 中的 Encoding.GetEncoding(string) 没有 TryGet
- 我想从用户输入字符串中获取编码，所以输入被认为是不安全的。
- 尤其是对于日语，类似“ISO-2022-JP”之类的东西可能会或可能不会在执行环境中工作
- 很容易将“shift_jis”拼写为“shift-jis”。
- Try Catch 印象很深，于是想到了可以用GetEncodings()获取的LINQing EncodingInfo[]
⇒ 我想知道到底发生了什么
基准
测量环境
- BenchmarkDotNet
- M2 MacBook Air
具体如下环境
BenchmarkDotNet=v0.13.2, OS=macOS Monterey 12.6 (21G115) [Darwin 21.6.0]
Apple M2, 1 CPU, 8 logical and 8 physical cores
.NET SDK=6.0.402
[Host] : .NET 6.0.7 (6.0.722.32202), Arm64 RyuJIT AdvSIMD
DefaultJob : .NET 6.0.10 (6.0.1022.47605), Arm64 RyuJIT AdvSIMD
-
请小心，因为除非您这样做，否则无法使用 BenchmarkDotNet
brew install --cask dotnet-sdk
结果
-
如果
GetEncoding(string)成功，即使使用 try catch 块也非常快- 另一方面，它在失败和捕获的情况下慢了大约 700 倍。
-
使用
GetEncodings().FirstOrDefault()，无论您在寻找什么，它几乎都是一样的（EncodingInfo[] 是一个数组并且大小足够小，因此找到 First 并中断扫描似乎不太可能是一个优势）- 但比
GetEncoding(string)成功时慢约 45 倍
- 但比
-
第一次
| Method | Mean | Error | StdDev |
|-------------------------------- |-------------:|----------:|----------:|
| GetValidEncodingWithTryCatch | 22.26 ns | 0.035 ns | 0.032 ns |
| GetInvalidEncodingWithTryCatch | 15,183.12 ns | 37.232 ns | 34.827 ns |
| GetValidEncodingWithIteration | 899.17 ns | 1.814 ns | 1.697 ns |
| GetInvalidEncodingWithIteration | 894.75 ns | 6.936 ns | 5.792 ns |
- 第二次
| Method | Mean | Error | StdDev |
|-------------------------------- |-------------:|-----------:|-----------:|
| GetValidEncodingWithTryCatch | 23.11 ns | 0.445 ns | 0.416 ns |
| GetInvalidEncodingWithTryCatch | 16,103.76 ns | 316.660 ns | 311.003 ns |
| GetValidEncodingWithIteration | 949.95 ns | 18.979 ns | 31.183 ns |
| GetInvalidEncodingWithIteration | 942.63 ns | 15.243 ns | 13.512 ns |
代码
- 比较代码
using System.Text;
using BenchmarkDotNet.Attributes;
namespace Benchy;
public class GetEncodingTest
{
private string validEncode { get; } = "UTF-8";
private string invalidEncode { get; } = "utf8";
public GetEncodingTest()
{
}
[Benchmark]
public Encoding GetValidEncodingWithTryCatch()
{
Encoding enc;
try
{
enc = Encoding.GetEncoding(validEncode);
}
catch (ArgumentException e)
{
enc = Encoding.UTF8;
}
return enc;
}
[Benchmark]
public Encoding GetInvalidEncodingWithTryCatch()
{
Encoding enc;
try
{
enc = Encoding.GetEncoding(invalidEncode);
}
catch (ArgumentException e)
{
enc = Encoding.UTF8;
}
return enc;
}
[Benchmark]
public Encoding GetValidEncodingWithIteration()
{
return Encoding.GetEncodings().FirstOrDefault(o => o.Name == validEncode)?.GetEncoding() ?? Encoding.UTF8;
}
[Benchmark]
public Encoding GetInvalidEncodingWithIteration()
{
return Encoding.GetEncodings().FirstOrDefault(o => o.Name == invalidEncode)?.GetEncoding() ?? Encoding.UTF8;
}
}
- 入口点
using BenchmarkDotNet.Running;
using Benchy;
BenchmarkRunner.Run<GetEncodingTest>();
多余
我觉得太快了，还以为是在IL级别优化的，所以从文件中读取，但是没有用。
| Method | Mean | Error | StdDev |
|-------------------------------- |-------------:|----------:|----------:|
| GetValidEncodingWithTryCatch | 22.88 ns | 0.069 ns | 0.061 ns |
| GetInvalidEncodingWithTryCatch | 15,525.24 ns | 20.935 ns | 18.558 ns |
| GetValidEncodingWithIteration | 918.30 ns | 1.785 ns | 1.670 ns |
| GetInvalidEncodingWithIteration | 916.27 ns | 1.619 ns | 1.435 ns |
using System.Text;
using BenchmarkDotNet.Attributes;
namespace Benchy;
public class GetEncodingTest
{
private string validEncode { get; }
private string invalidEncode { get; }
public GetEncodingTest()
{
validEncode = System.IO.File.ReadAllText("valid.txt").Trim();
invalidEncode = System.IO.File.ReadAllText("invalid.txt").Trim();
}
[Benchmark]
public Encoding GetValidEncodingWithTryCatch()
{
Encoding enc;
try
{
enc = Encoding.GetEncoding(validEncode);
}
catch (ArgumentException e)
{
enc = Encoding.UTF8;
}
return enc;
}
[Benchmark]
public Encoding GetInvalidEncodingWithTryCatch()
{
Encoding enc;
try
{
enc = Encoding.GetEncoding(invalidEncode);
}
catch (ArgumentException e)
{
enc = Encoding.UTF8;
}
return enc;
}
[Benchmark]
public Encoding GetValidEncodingWithIteration()
{
return Encoding.GetEncodings().FirstOrDefault(o => o.Name == validEncode)?.GetEncoding() ?? Encoding.UTF8;
}
[Benchmark]
public Encoding GetInvalidEncodingWithIteration()
{
return Encoding.GetEncodings().FirstOrDefault(o => o.Name == invalidEncode)?.GetEncoding() ?? Encoding.UTF8;
}
}
后记
我从@albireo 的评论中确认了，但是在这个过程中缓存肯定是有利的。
即使您每次都生成一个 Dictionary，它也只会比每次执行 FirstOrDefault 快 1.5 倍左右，因此即使迭代次数很少，缓存也可能是有利的。
| Method | Mean | Error | StdDev |
|-------------------------------- |--------------:|-----------:|-----------:|
| GetValidEncodingWithTryCatch | 22.295 ns | 0.0540 ns | 0.0479 ns |
| GetInvalidEncodingWithTryCatch | 15,081.121 ns | 20.7711 ns | 19.4293 ns |
| GetValidEncodingWithIteration | 894.545 ns | 2.0607 ns | 1.9276 ns |
| GetInvalidEncodingWithIteration | 905.887 ns | 16.0912 ns | 15.8037 ns |
| GetValidEncodingWithCache | 7.990 ns | 0.0057 ns | 0.0045 ns |
| Method | Mean | Error | StdDev |
|-------------------------------- |--------------:|-----------:|-----------:|
| GetValidEncodingWithTryCatch | 22.312 ns | 0.0384 ns | 0.0341 ns |
| GetInvalidEncodingWithTryCatch | 15,092.002 ns | 33.9453 ns | 30.0916 ns |
| GetValidEncodingWithIteration | 893.170 ns | 1.2647 ns | 1.1830 ns |
| GetInvalidEncodingWithIteration | 848.864 ns | 1.5862 ns | 1.4837 ns |
| GetInvalidEncodingWithCache | 7.993 ns | 0.0045 ns | 0.0042 ns |
| Method | Mean | Error | StdDev |
|-------------------------------- |-------------:|----------:|----------:|
| GetValidEncodingWithTryCatch | 22.30 ns | 0.031 ns | 0.028 ns |
| GetInvalidEncodingWithTryCatch | 15,119.33 ns | 15.193 ns | 13.468 ns |
| GetValidEncodingWithIteration | 896.30 ns | 1.740 ns | 1.627 ns |
| GetInvalidEncodingWithIteration | 897.54 ns | 2.107 ns | 1.971 ns |
| GetValidEncodingWithDic | 1,124.98 ns | 1.886 ns | 1.764 ns |
- 验证码
using System.Text;
using BenchmarkDotNet.Attributes;
namespace Benchy;
public class GetEncodingTest
{
private string validEncode { get; }
private string invalidEncode { get; }
private static Dictionary<string, Encoding> encodingCache = null;
public GetEncodingTest()
{
validEncode = System.IO.File.ReadAllText("valid.txt").Trim();
invalidEncode = System.IO.File.ReadAllText("invalid.txt").Trim();
}
[Benchmark]
public Encoding GetValidEncodingWithTryCatch()
{
Encoding enc;
try
{
enc = Encoding.GetEncoding(validEncode);
}
catch (ArgumentException e)
{
enc = Encoding.UTF8;
}
return enc;
}
[Benchmark]
public Encoding GetInvalidEncodingWithTryCatch()
{
Encoding enc;
try
{
enc = Encoding.GetEncoding(invalidEncode);
}
catch (ArgumentException e)
{
enc = Encoding.UTF8;
}
return enc;
}
[Benchmark]
public Encoding GetValidEncodingWithIteration()
{
return Encoding.GetEncodings().FirstOrDefault(o => o.Name == validEncode)?.GetEncoding() ?? Encoding.UTF8;
}
[Benchmark]
public Encoding GetInvalidEncodingWithIteration()
{
return Encoding.GetEncodings().FirstOrDefault(o => o.Name == invalidEncode)?.GetEncoding() ?? Encoding.UTF8;
}
public Encoding GetValidEncodingWithCache()
{
if (encodingCache is null)
{
encodingCache = new Dictionary<string, Encoding>(
Encoding.GetEncodings().Select(
encInf => new KeyValuePair<string, Encoding>(encInf.Name, encInf.GetEncoding())
)
);
}
return encodingCache.TryGetValue(validEncode, out Encoding enc) ? enc : Encoding.UTF8;
}
public Encoding GetInvalidEncodingWithCache()
{
if (encodingCache is null)
{
encodingCache = new Dictionary<string, Encoding>(
Encoding.GetEncodings().Select(
encInf => new KeyValuePair<string, Encoding>(encInf.Name, encInf.GetEncoding())
)
);
}
return encodingCache.TryGetValue(invalidEncode, out Encoding enc) ? enc : Encoding.UTF8;
}
[Benchmark]
public Encoding GetValidEncodingWithDic()
{
encodingCache = new Dictionary<string, Encoding>(
Encoding.GetEncodings().Select(
encInf => new KeyValuePair<string, Encoding>(encInf.Name, encInf.GetEncoding())
)
);
return encodingCache.TryGetValue(validEncode, out Encoding enc) ? enc : Encoding.UTF8;
}
}
原创声明：本文系作者授权爱码网发表，未经许可，不得转载;
原文地址：https://www.likecs.com/show-308631685.html