如何创建命令行预测器

项目
02/05/2025

PSReadLine 2.1.0 通过实现预测 IntelliSense 功能引入了智能命令行预测器的概念。 PSReadLine 2.2.2 通过添加插件模型来扩展该功能，以便创建自己的命令行预测器。

预测 IntelliSense 会在你键入时在命令行上提供建议，以此增强 Tab 自动补全功能。预测建议在光标后显示为彩色文本。这样，便可以根据命令历史记录或其他特定于域的插件的匹配预测发现、编辑和执行完整命令。

系统要求

若要创建和使用插件预测器，必须使用以下版本的软件：

PowerShell 7.2 （或更高版本） - 提供创建命令预测器所需的 API
PSReadLine 2.2.2 （或更高版本） - 允许配置 PSReadLine 以使用插件

预测器概述

预测器是 PowerShell 二进制模块。模块必须实现 System.Management.Automation.Subsystem.Prediction.ICommandPredictor 接口。此接口声明用于查询预测结果和提供反馈的方法。

预测器模块必须在加载时使用 PowerShell 的 CommandPredictor 注册 SubsystemManager 子系统，并在卸载时取消注册。

下图显示了 PowerShell 中预测器的体系结构。

体系结构

创建代码

若要创建预测器，必须为平台安装 .NET 6 SDK。有关 SDK 的详细信息，请参阅下载 .NET 6.0。

按照以下步骤创建新的 PowerShell 模块项目：

使用 dotnet 命令行工具创建初学者 classlib 项目。
```
dotnet new classlib --name SamplePredictor
```

编辑 SamplePredictor.csproj 以包含以下信息：

<Project Sdk="Microsoft.NET.Sdk">

  <PropertyGroup>
    <TargetFramework>net6.0</TargetFramework>
  </PropertyGroup>

  <ItemGroup>
    <PackageReference Include="Microsoft.PowerShell.SDK" Version="7.2.0" />
  </ItemGroup>

</Project>

删除 dotnet 创建的默认 Class1.cs 文件，并将以下代码复制到项目文件夹中的 SamplePredictorClass.cs 文件。

using System;
using System.Collections.Generic;
using System.Threading;
using System.Management.Automation;
using System.Management.Automation.Subsystem;
using System.Management.Automation.Subsystem.Prediction;

namespace PowerShell.Sample
{
    public class SamplePredictor : ICommandPredictor
    {
        private readonly Guid _guid;

        internal SamplePredictor(string guid)
        {
            _guid = new Guid(guid);
        }

        /// <summary>
        /// Gets the unique identifier for a subsystem implementation.
        /// </summary>
        public Guid Id => _guid;

        /// <summary>
        /// Gets the name of a subsystem implementation.
        /// </summary>
        public string Name => "SamplePredictor";

        /// <summary>
        /// Gets the description of a subsystem implementation.
        /// </summary>
        public string Description => "A sample predictor";

        /// <summary>
        /// Get the predictive suggestions. It indicates the start of a suggestion rendering session.
        /// </summary>
        /// <param name="client">Represents the client that initiates the call.</param>
        /// <param name="context">The <see cref="PredictionContext"/> object to be used for prediction.</param>
        /// <param name="cancellationToken">The cancellation token to cancel the prediction.</param>
        /// <returns>An instance of <see cref="SuggestionPackage"/>.</returns>
        public SuggestionPackage GetSuggestion(PredictionClient client, PredictionContext context, CancellationToken cancellationToken)
        {
            string input = context.InputAst.Extent.Text;
            if (string.IsNullOrWhiteSpace(input))
            {
                return default;
            }

            return new SuggestionPackage(new List<PredictiveSuggestion>{
                new PredictiveSuggestion(string.Concat(input, " HELLO WORLD"))
            });
        }

        #region "interface methods for processing feedback"

        /// <summary>
        /// Gets a value indicating whether the predictor accepts a specific kind of feedback.
        /// </summary>
        /// <param name="client">Represents the client that initiates the call.</param>
        /// <param name="feedback">A specific type of feedback.</param>
        /// <returns>True or false, to indicate whether the specific feedback is accepted.</returns>
        public bool CanAcceptFeedback(PredictionClient client, PredictorFeedbackKind feedback) => false;

        /// <summary>
        /// One or more suggestions provided by the predictor were displayed to the user.
        /// </summary>
        /// <param name="client">Represents the client that initiates the call.</param>
        /// <param name="session">The mini-session where the displayed suggestions came from.</param>
        /// <param name="countOrIndex">
        /// When the value is greater than 0, it's the number of displayed suggestions from the list
        /// returned in <paramref name="session"/>, starting from the index 0. When the value is
        /// less than or equal to 0, it means a single suggestion from the list got displayed, and
        /// the index is the absolute value.
        /// </param>
        public void OnSuggestionDisplayed(PredictionClient client, uint session, int countOrIndex) { }

        /// <summary>
        /// The suggestion provided by the predictor was accepted.
        /// </summary>
        /// <param name="client">Represents the client that initiates the call.</param>
        /// <param name="session">Represents the mini-session where the accepted suggestion came from.</param>
        /// <param name="acceptedSuggestion">The accepted suggestion text.</param>
        public void OnSuggestionAccepted(PredictionClient client, uint session, string acceptedSuggestion) { }

        /// <summary>
        /// A command line was accepted to execute.
        /// The predictor can start processing early as needed with the latest history.
        /// </summary>
        /// <param name="client">Represents the client that initiates the call.</param>
        /// <param name="history">History command lines provided as references for prediction.</param>
        public void OnCommandLineAccepted(PredictionClient client, IReadOnlyList<string> history) { }

        /// <summary>
        /// A command line was done execution.
        /// </summary>
        /// <param name="client">Represents the client that initiates the call.</param>
        /// <param name="commandLine">The last accepted command line.</param>
        /// <param name="success">Shows whether the execution was successful.</param>
        public void OnCommandLineExecuted(PredictionClient client, string commandLine, bool success) { }

        #endregion;
    }

    /// <summary>
    /// Register the predictor on module loading and unregister it on module un-loading.
    /// </summary>
    public class Init : IModuleAssemblyInitializer, IModuleAssemblyCleanup
    {
        private const string Identifier = "843b51d0-55c8-4c1a-8116-f0728d419306";

        /// <summary>
        /// Gets called when assembly is loaded.
        /// </summary>
        public void OnImport()
        {
            var predictor = new SamplePredictor(Identifier);
            SubsystemManager.RegisterSubsystem(SubsystemKind.CommandPredictor, predictor);
        }

        /// <summary>
        /// Gets called when the binary module is unloaded.
        /// </summary>
        public void OnRemove(PSModuleInfo psModuleInfo)
        {
            SubsystemManager.UnregisterSubsystem(SubsystemKind.CommandPredictor, new Guid(Identifier));
        }
    }
}

以下示例代码返回所有用户输入的预测结果的字符串“HELLO WORLD”。由于示例预测器不处理任何反馈，代码不会从接口实现反馈方法。更改预测和反馈代码以满足预测器的需求。

注意

PSReadLine 的列表视图不支持多行建议。每个建议都应是一行。如果代码具有多行建议，则应将行拆分为单独的建议，或者将行与分号（;）联接。

运行 dotnet build 以生成程序集。可以在项目文件夹的 bin/Debug/net6.0 位置中找到已编译的程序集。

注意

为了确保响应式用户体验，ICommandPredictor 接口针对从预测器获取响应设定了 20 毫秒的超时时间。要显示您的预测器代码，必须在 20 毫秒以内返回结果。

使用预测器插件

若要试用新的预测器，请打开新的 PowerShell 7.2 会话并运行以下命令：

Set-PSReadLineOption -PredictionSource HistoryAndPlugin
Import-Module .\bin\Debug\net6.0\SamplePredictor.dll

在会话中加载程序集后，在终端中键入时，会看到文本“HELLO WORLD”。可以按 F2 在 Inline 视图和 List 视图之间切换。

有关 PSReadLine 选项的详细信息，请参阅 Set-PSReadLineOption。

可以使用以下命令获取已安装的预测器列表：

Get-PSSubsystem -Kind CommandPredictor

Kind              SubsystemType      IsRegistered Implementations
----              -------------      ------------ ---------------
CommandPredictor  ICommandPredictor          True {SamplePredictor}

注意

Get-PSSubsystem 是 PowerShell 7.1 中引入的实验性 cmdlet，必须启用 PSSubsystemPluginModel 实验性功能才能使用此 cmdlet。有关详细信息，请参阅使用实验性功能。

通过

如何创建命令行预测器

系统要求

预测器概述

创建代码

使用预测器插件

其他资源