releases/src/Paca.Browser/Content/MhtmlWebArchivePackager.cs

using System;
using System.Collections.Generic;
using System.Text;
using System.Text.RegularExpressions;

namespace Paca.Browser.Content;

public sealed class MhtmlWebArchivePackager
{
    public string Package(WebArchivePackageRequest request)
    {
        if (request == null)
            throw new ArgumentNullException(nameof(request));

        var boundary = "----=_NextPart_" + Guid.NewGuid().ToString("N");
        var sb = new StringBuilder();

        sb.AppendLine("From: <Saved by PACA Browser>");
        sb.AppendLine($"Snapshot-Content-Location: {request.Url ?? string.Empty}");
        sb.AppendLine($"Subject: {request.Title ?? "Untitled"}");
        sb.AppendLine($"Date: {DateTime.UtcNow:R}");
        sb.AppendLine("MIME-Version: 1.0");
        sb.AppendLine($"Content-Type: multipart/related; boundary=\"{boundary}\"; type=\"text/html\"");
        sb.AppendLine();

        // HTML Part
        sb.AppendLine($"--{boundary}");
        sb.AppendLine("Content-Type: text/html; charset=\"utf-8\"");
        sb.AppendLine("Content-Transfer-Encoding: 8bit");
        sb.AppendLine($"Content-Location: {request.Url ?? string.Empty}");
        sb.AppendLine();
        sb.AppendLine(request.HtmlContent ?? string.Empty);
        sb.AppendLine();

        // Resources
        if (request.Resources != null)
        {
            foreach (var res in request.Resources)
            {
                if (res == null) continue;
                var data = res.Data ?? Array.Empty<byte>();
                sb.AppendLine($"--{boundary}");
                sb.AppendLine($"Content-Type: {res.MimeType ?? "application/octet-stream"}");
                sb.AppendLine("Content-Transfer-Encoding: base64");
                sb.AppendLine($"Content-Location: {res.Url ?? string.Empty}");
                sb.AppendLine();
                sb.AppendLine(Convert.ToBase64String(data));
                sb.AppendLine();
            }
        }

        sb.AppendLine($"--{boundary}--");
        return sb.ToString();
    }

    public WebArchiveUnpackResult Unpack(string mhtmlContent)
    {
        if (string.IsNullOrWhiteSpace(mhtmlContent))
        {
            return new WebArchiveUnpackResult("Untitled", string.Empty, string.Empty, Array.Empty<WebArchiveResource>());
        }

        var locMatch = Regex.Match(mhtmlContent, @"Snapshot-Content-Location:\s*([^\r\n]+)", RegexOptions.IgnoreCase);
        var locationUrl = locMatch.Success ? locMatch.Groups[1].Value.Trim() : string.Empty;

        var subjMatch = Regex.Match(mhtmlContent, @"Subject:\s*([^\r\n]+)", RegexOptions.IgnoreCase);
        var title = subjMatch.Success ? subjMatch.Groups[1].Value.Trim() : "Untitled";

        var boundMatch = Regex.Match(mhtmlContent, @"boundary=""?([^"";\r\n]+)""?", RegexOptions.IgnoreCase);
        if (!boundMatch.Success)
        {
            return new WebArchiveUnpackResult(title, locationUrl, mhtmlContent, Array.Empty<WebArchiveResource>());
        }

        var boundary = boundMatch.Groups[1].Value;
        var parts = mhtmlContent.Split(new[] { "--" + boundary }, StringSplitOptions.RemoveEmptyEntries);

        string htmlContent = string.Empty;
        var resources = new List<WebArchiveResource>();

        foreach (var part in parts)
        {
            if (part.StartsWith("--") || string.IsNullOrWhiteSpace(part)) continue;

            var doubleLine = part.IndexOf("\r\n\r\n", StringComparison.Ordinal);
            if (doubleLine < 0) doubleLine = part.IndexOf("\n\n", StringComparison.Ordinal);
            if (doubleLine < 0) continue;

            var headers = part.Substring(0, doubleLine);
            var body = part.Substring(doubleLine).Trim();

            var typeMatch = Regex.Match(headers, @"Content-Type:\s*([^;\r\n]+)", RegexOptions.IgnoreCase);
            var mimeType = typeMatch.Success ? typeMatch.Groups[1].Value.Trim() : "text/html";

            if (mimeType.Contains("multipart/related", StringComparison.OrdinalIgnoreCase) ||
                headers.Contains("boundary=", StringComparison.OrdinalIgnoreCase))
            {
                continue;
            }

            var urlMatch = Regex.Match(headers, @"Content-Location:\s*([^\r\n]+)", RegexOptions.IgnoreCase);
            var resUrl = urlMatch.Success ? urlMatch.Groups[1].Value.Trim() : string.Empty;

            if (mimeType.Contains("text/html", StringComparison.OrdinalIgnoreCase))
            {
                htmlContent = body;
            }
            else
            {
                byte[] data;
                if (headers.Contains("base64", StringComparison.OrdinalIgnoreCase))
                {
                    try
                    {
                        data = Convert.FromBase64String(body);
                    }
                    catch
                    {
                        data = Encoding.UTF8.GetBytes(body);
                    }
                }
                else
                {
                    data = Encoding.UTF8.GetBytes(body);
                }

                resources.Add(new WebArchiveResource(resUrl, mimeType, data));
            }
        }

        return new WebArchiveUnpackResult(title, locationUrl, htmlContent, resources.AsReadOnly());
    }
}