Wp2ghost/CSharpGhostImporter/HtmlToMarkdownConverter.cs

namespace wp2ghost
{
    // =============================================================================
    // HTML zu Markdown Konverter
    // =============================================================================
    public static class HtmlToMarkdownConverter
    {
        public static string Convert(string html)
        {
            if (string.IsNullOrEmpty(html))
                return "";

            string result = html;

            result = System.Text.RegularExpressions.Regex.Replace(result, "<!--.*?-->", "", System.Text.RegularExpressions.RegexOptions.Singleline);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<h([1-6])[^>]*>(.*?)</h\1>", m => new string('#', int.Parse(m.Groups[1].Value)) + " " + StripTags(m.Groups[2].Value) + Environment.NewLine + Environment.NewLine, System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<blockquote[^>]*>(.*?)</blockquote>", m => "> " + StripTags(m.Groups[1].Value).Replace(Environment.NewLine, Environment.NewLine + "> ") + Environment.NewLine, System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<pre[^>]*><code[^>]*>(.*?)</code></pre>", m => "```\n" + StripTags(m.Groups[1].Value) + "\n```\n\n", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<code[^>]*>(.*?)</code>", m => "`" + StripTags(m.Groups[1].Value) + "`", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<ul[^>]*>(.*?)</ul>", m => ConvertList(m.Groups[1].Value, "-"), System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);
            result = System.Text.RegularExpressions.Regex.Replace(result, @"<ol[^>]*>(.*?)</ol>", m => ConvertOrderedList(m.Groups[1].Value), System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = ConvertTables(result);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<p[^>]*>(.*?)</p>", m => CleanInnerHtml(m.Groups[1].Value) + Environment.NewLine + Environment.NewLine, System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<br\s*/?>", Environment.NewLine, System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<strong[^>]*>(.*?)</strong>", "**$1**", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);
            result = System.Text.RegularExpressions.Regex.Replace(result, @"<b[^>]*>(.*?)</b>", "**$1**", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<em[^>]*>(.*?)</em>", "*$1*", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);
            result = System.Text.RegularExpressions.Regex.Replace(result, @"<i[^>]*>(.*?)</i>", "*$1*", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<a\s+[^>]*href=[""']([^""']+)[""'][^>]*>(.*?)</a>", "[$2]($1)", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<img\s+[^>]*src=[""']([^""']+)[""'][^>]*alt=[""']([^""']*)[""'][^>]*>", m => "![" + m.Groups[2].Value + "](" + m.Groups[1].Value + ")", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);
            result = System.Text.RegularExpressions.Regex.Replace(result, @"<img\s+[^>]*src=[""']([^""']+)[""'][^>]*>", m => "![](" + m.Groups[1].Value + ")", System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"<div[^>]*class=[""'][^""']*wp-block-heading[^""']*[""'][^>]*>(.*?)</div>", m => "# " + StripTags(m.Groups[1].Value) + Environment.NewLine + Environment.NewLine, System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = System.Text.RegularExpressions.Regex.Replace(result,
                @"<figure[^>]*>(.*?)</figure>",
                m => ConvertFigure(m.Groups[1].Value),
                System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);

            result = StripTagsKeepFigure(result);

            result = System.Text.RegularExpressions.Regex.Replace(result, @"\n{3,}", "\n\n");

            result = result.Replace("&amp;", "&").Replace("&lt;", "<").Replace("&gt;", ">").Replace("&quot;", "\"").Replace("&#8217;", "'").Replace("&nbsp;", " ").Replace("&mdash;", "—").Replace("&ndash;", "–");

            return result.Trim();
        }

        private static string StripTags(string html) =>
            System.Text.RegularExpressions.Regex.Replace(html, "<[^>]*>", "");

        private static string CleanInnerHtml(string html)
        {
            string cleaned = StripTags(html);
            cleaned = System.Text.RegularExpressions.Regex.Replace(cleaned, @"\*\*(.+?)\*\*", "**$1**");
            return cleaned.Trim();
        }

        private static string ConvertFigure(string figureContent)
        {
            string imgSrc = "", altText = "", caption = "";

            var imgMatch = System.Text.RegularExpressions.Regex.Match(figureContent, @"<img\s+[^>]*src=[""']([^""']+)[""']");
            if (imgMatch.Success) imgSrc = imgMatch.Groups[1].Value;

            var altMatch = System.Text.RegularExpressions.Regex.Match(figureContent, @"alt=[""']([^""']*)[""']");
            if (altMatch.Success) altText = altMatch.Groups[1].Value;

            var figMatch = System.Text.RegularExpressions.Regex.Match(figureContent, @"<figcaption[^>]*>(.*?)</figcaption>", System.Text.RegularExpressions.RegexOptions.Singleline);
            if (figMatch.Success) caption = System.Text.RegularExpressions.Regex.Replace(figMatch.Groups[1].Value, "<[^>]*>", "").Trim();

            if (string.IsNullOrEmpty(imgSrc)) return "";

            string result = $"<figure class=\"kg-width-wide kg-card-width-wide\"><img src=\"{imgSrc}\" loading=\"lazy\"";
            if (!string.IsNullOrEmpty(altText)) result += $" alt=\"{altText}\"";
            result += ">";
            if (!string.IsNullOrEmpty(caption)) result += $"<figcaption>{caption}</figcaption>";
            result += "</figure>";
            return result + Environment.NewLine;
        }

        private static string StripTagsKeepFigure(string html)
        {
            var figureBlocks = new List<string>();
            html = System.Text.RegularExpressions.Regex.Replace(html, @"<figure[^>]*>.*?</figure>", m => { figureBlocks.Add(m.Value); return $"__FIGURE_{figureBlocks.Count - 1}__"; }, System.Text.RegularExpressions.RegexOptions.Singleline);
            html = System.Text.RegularExpressions.Regex.Replace(html, "<[^>]*>", "");
            for (int i = 0; i < figureBlocks.Count; i++)
                html = html.Replace($"__FIGURE_{i}__", figureBlocks[i]);
            return html;
        }

        private static string ConvertList(string listHtml, string bullet)
        {
            listHtml = System.Text.RegularExpressions.Regex.Replace(listHtml, @"<li[^>]*>(.*?)</li>", m => bullet + " " + CleanInnerHtml(m.Groups[1].Value) + Environment.NewLine, System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);
            return StripTags(listHtml).Trim() + Environment.NewLine;
        }

        private static string ConvertOrderedList(string listHtml)
        {
            int counter = 0;
            listHtml = System.Text.RegularExpressions.Regex.Replace(listHtml, @"<li[^>]*>(.*?)</li>", m => { counter++; return counter + ". " + CleanInnerHtml(m.Groups[1].Value) + Environment.NewLine; }, System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);
            return StripTags(listHtml).Trim() + Environment.NewLine;
        }

        private static string ConvertTables(string html)
        {
            return System.Text.RegularExpressions.Regex.Replace(html, @"<table[^>]*>(.*?)</table>", m =>
            {
                string tableContent = m.Groups[1].Value;
                var headerMatch = System.Text.RegularExpressions.Regex.Match(tableContent, @"<tr[^>]*>(.*?)</tr>", System.Text.RegularExpressions.RegexOptions.Singleline);
                if (!headerMatch.Success) return StripTags(tableContent) + Environment.NewLine;

                string headers = CleanInnerHtml(headerMatch.Groups[1].Value);
                headers = headers.Replace("<th", "<td").Replace("</th>", "</td>");
                string headerLine = "| " + string.Join(" | ", StripTags(headers).Split(new[] { "</td>", "<td" }, StringSplitOptions.RemoveEmptyEntries).Select(c => c.Trim())) + " |";
                string separator = "|" + string.Join("|", StripTags(headers).Split(new[] { "</td>", "<td" }, StringSplitOptions.RemoveEmptyEntries).Select(c => "---")) + "|";

                var rows = System.Text.RegularExpressions.Regex.Matches(tableContent, @"<tr[^>]*>(.*?)</tr>", System.Text.RegularExpressions.RegexOptions.Singleline);
                var dataRows = new List<string>();
                for (int i = 1; i < rows.Count; i++)
                {
                    string rowText = CleanInnerHtml(rows[i].Groups[1].Value);
                    string[] cells = rowText.Split(new[] { "</td>", "<td" }, StringSplitOptions.RemoveEmptyEntries);
                    dataRows.Add("| " + string.Join(" | ", cells.Select(c => c.Trim().Replace("<b>", "").Replace("</b>", "").Replace("<strong>", "").Replace("</strong>", ""))) + " |");
                }

                return headerLine + Environment.NewLine + separator + Environment.NewLine + string.Join(Environment.NewLine, dataRows) + Environment.NewLine + Environment.NewLine;
            }, System.Text.RegularExpressions.RegexOptions.Singleline | System.Text.RegularExpressions.RegexOptions.IgnoreCase);
        }
    }
}