將 Playwright 網頁自動操作程式部署到客戶端執行

2023-02-18 10:44 PM

上回提到我想用 Playwright for .NET 也可以用來開發網頁操作自動化機器人，但部署到客戶端可能會是問題。Playwright 原本被設計用於 End-To-End 測試，多在開發測試環境或 Build 主機上執行，程式庫下載安裝不用太講究 User Friendly。我想把它放到客戶端應用，得研究所需檔案數量、大小、目錄位置要求，評估可行性。

首先，我想借用客戶端上現有的 Chrome 或 Edge，不要再另外下載安裝瀏覽器軟體，這不是難事。以 Chrome 為例，Chromium.LaunchAsync() 時可查詢 Registry 找到 Chrome 安裝路徑，傳入 ExecutablePath。

using System.Text.RegularExpressions;
using Microsoft.Playwright;

using var playwright = await Playwright.CreateAsync();
await using var browser = await playwright.Chromium.LaunchAsync(new BrowserTypeLaunchOptions {
    Headless = false,
    ExecutablePath = GetChromePath()
});
var page = await browser.NewPageAsync();
await page.GotoAsync("https://playwright.dev/dotnet");
await page.ScreenshotAsync(new()
{
    Path = "screenshot.png"
});

// get Chrome installed path from registry (Windows only)
string GetChromePath()
{
    var path = Microsoft.Win32.Registry.GetValue(
        @"HKEY_CLASSES_ROOT\ChromeHTML\shell\open\command", null, null) as string;
    if (string.IsNullOrEmpty(path))
        throw new ApplicationException("Chrome not installed");
    var m = Regex.Match(path, "\"(?<p>.+?)\"");
    if (!m.Success)
        throw new ApplicationException($"Invalid Chrome path - {path}");
    return m.Groups["p"].Value;
}

實測 dotnet publish -r Release -r win-x64 --no-self-contained，除了 dll 外，還有一個 .playwright 目錄，其下包含 node.exe 及 TypeScript、JavaScript 程式及設定檔。由此可知 Playwright 本體仍是跑在 Node.js 平台，Playwright for .NET 只是溝通整合的中介層。

但有點要注意，Microsoft.Playwright.dll 必須獨立存在，若使用 -p:PublishSignleFile=true 整併進 .exe，執行時會噴出錯誤：

Unhandled exception. System.NotSupportedException: CodeBase is not supported on assemblies loaded from a single-file bundle.
   at System.Reflection.RuntimeAssembly.get_CodeBase()
   at Microsoft.Playwright.Helpers.Driver.GetExecutablePath() in /_/src/Playwright/Helpers/Driver.cs:line 43
   at Microsoft.Playwright.Transport.StdIOTransport.GetProcess() in /_/src/Playwright/Transport/StdIOTransport.cs:line 114
   at Microsoft.Playwright.Transport.StdIOTransport..ctor() in /_/src/Playwright/Transport/StdIOTransport.cs:line 44
   at Microsoft.Playwright.Playwright.CreateAsync() in /_/src/Playwright/Playwright.cs:line 44
   at Program.<Main>$(String[] args) in X:\Github\MiscLabs\playwright-dep-test\Program.cs:line 4
   at Program.<Main>(String[] args)

全部程式共 310 個檔案，近 70 MB，大小還可被接受：

將整包程式丟到客戶端，實測能成功執行!

如此，至少驗證未來將發佈將整包程式複製過去並裝好 .NET SDK，要在客戶端跑 Playwright for .NET 是可行的。

還有一個待優化的地方 - 程式附帶的 68MB .playwright 資料夾內容都是相同的，每支程式都複製一份是種浪費，有沒有可能多支程式共用一份呢？追進原始碼，GetExecutablePath() 假設 .playwright 資料夾放在執行程式所在目錄或是 Microsoft.Playwright.dll 的同目錄下，目前並未提供自訂路徑的選項。有人提了 PR 想比照 PuppeteerSharp 為 Playwright.CreateAsync(driversPath: "path/.playwright") 加個自訂路徑參數但沒被接受。現階段要解決，可考慮 Fork 一個魔改版本或用 Hacking 技巧搞定，有空再來想想。

將 Playwright 網頁自動操作程式部署到客戶端執行

Comments

Post a comment