using System.IO.Compression;
using System.Text;
using System.Text.RegularExpressions;

namespace Rmsys.Infrastructure;

/// <summary>
/// 任免文件 ↔ 上会干部 的姓名匹配。
///
/// 做法：不从文件里"猜"姓名（中文姓名边界不可靠，会出现"去刘志强""于陈晓东"这类误抽），
/// 而是拿库里已知的姓名去文件中查找——姓名是确定的，命中即关联，不存在边界歧义。
///
/// 支持 .docx（直接读 word/document.xml，不依赖任何外部组件）；
/// .pdf / .doc 需服务器有 pdftotext / antiword，没有时返回 NotSupported，由界面提示改为手动设置人员。
/// </summary>
public static class DocNameMatcher
{
    public class Result
    {
        /// <summary>文件中出现、且能在库里唯一确定的干部（按在文件中出现的位置排序）</summary>
        public List<(int CadreId, string Name, int Pos)> InFile { get; set; } = new();
        /// <summary>文件里出现但库内有多个同名人 → 需要人工确认</summary>
        public List<(string Name, List<int> CadreIds)> Ambiguous { get; set; } = new();
        /// <summary>文件里出现、但本场会议上会名单里没有的姓名（疑似写错或未纳入名单）</summary>
        public List<string> UnknownNames { get; set; } = new();
        /// <summary>能否抽取文本</summary>
        public bool TextOk { get; set; }
        /// <summary>抽取失败原因</summary>
        public string? Error { get; set; }
        public int TextLength { get; set; }
    }

    public static Result Match(string fileFullPath, IEnumerable<(int Id, string Name)> cadres)
    {
        var r = new Result();
        var text = ExtractText(fileFullPath, out var err);
        if (text == null)
        {
            r.TextOk = false;
            r.Error = err;
            return r;
        }
        r.TextOk = true;
        r.TextLength = text.Length;

        var list = cadres.Where(x => !string.IsNullOrWhiteSpace(x.Name)).ToList();

        // 1) 先按姓名长度倒序找：避免"王明"抢走"王小明"的位置
        foreach (var g in list.GroupBy(x => x.Name).OrderByDescending(g => g.Key.Length))
        {
            var name = g.Key;
            var poses = new List<int>();
            var idx = text.IndexOf(name, StringComparison.Ordinal);
            while (idx >= 0 && poses.Count < 20)
            {
                poses.Add(idx);
                idx = text.IndexOf(name, idx + name.Length, StringComparison.Ordinal);
            }
            if (poses.Count == 0) continue;

            if (g.Count() > 1)
            {
                r.Ambiguous.Add((name, g.Select(x => x.Id).ToList()));
                // 同名时不做自动关联，交给人工确认，避免把两个人的材料挂错
                continue;
            }
            r.InFile.Add((g.First().Id, name, poses[0]));
        }

        r.InFile = r.InFile.OrderBy(x => x.Pos).ToList();

        // 2) 找"库外姓名"：把已匹配姓名抹掉后，剩余的「XX同志」才算可疑
        var masked = text;
        foreach (var g in list.GroupBy(x => x.Name).OrderByDescending(g => g.Key.Length))
            masked = masked.Replace(g.Key, new string('○', g.Key.Length));

        var known = new HashSet<string>(list.Select(x => x.Name));
        foreach (Match m in Regex.Matches(masked, @"([\u4e00-\u9fa5]{2,4})(?=同志)"))
        {
            var cand = TrimLeading(m.Groups[1].Value);
            if (cand.Length < 2) continue;
            if (known.Contains(cand)) continue;
            if (IsStopWord(cand)) continue;
            if (!r.UnknownNames.Contains(cand)) r.UnknownNames.Add(cand);
        }
        return r;
    }

    /// <summary>
    /// 从正文里提取公文标题（"关于……的通知/决定/批复"）。
    /// 用途：后台列表显示文件自己的标题，而不是系统拼的"关于XX等N名同志任免职的通知"。
    /// </summary>
    public static string? ExtractTitle(string text)
    {
        foreach (var raw in text.Split('\n'))
        {
            var line = raw.Trim();
            if (line.Length < 6 || line.Length > 60) continue;
            if (line.Contains("号") && line.Contains("〔")) continue;   // 文号行
            var m = Regex.Match(line, @"关于.{2,40}?(通知|决定|批复|意见|任免)");
            if (m.Success) return m.Value;
        }
        return null;
    }

    /// <summary>剥掉动词/介词前缀（"免去刘志强"→"刘志强" 这类没有边界的残留）</summary>
    private static string TrimLeading(string s)
    {
        string[] junk = { "免去", "任命", "同意", "决定", "关于", "提名", "拟任", "拟免", "提请", "同志", "其中", "以及", "和", "及", "与", "为", "任", "免", "去", "于" };
        foreach (var j in junk)
        {
            if (s.Length - j.Length >= 2 && s.StartsWith(j, StringComparison.Ordinal))
                return s[j.Length..];
        }
        return s;
    }

    private static bool IsStopWord(string s)
    {
        string[] stop = { "全体", "各位", "有关", "上述", "以上", "各位同志", "领导班子", "干部群众", "会议", "研究", "决定", "通知", "文件" };
        return stop.Contains(s);
    }

    /// <summary>抽取纯文本；失败返回 null 并给出原因</summary>
    public static string? ExtractText(string fullPath, out string? error)
    {
        error = null;
        var ext = Path.GetExtension(fullPath).ToLowerInvariant();
        try
        {
            if (ext == ".docx") return FromDocx(fullPath);
            if (ext == ".txt") return File.ReadAllText(fullPath, Encoding.UTF8);

            if (ext == ".pdf")
            {
                var txt = RunTool("pdftotext", $"-enc UTF-8 -q \"{fullPath}\" -");
                if (txt != null) return txt;
                error = "该文件是 PDF，服务器缺少 PDF 文本抽取组件，无法自动匹配姓名；可改用 Word 上传，或在「人员」里手动勾选。";
                return null;
            }
            if (ext == ".doc")
            {
                // 1) 内置纯托管解析：不依赖任何外部程序，Windows 客户机上同样可用
                var txt = WordDocText.Extract(fullPath, out var docErr);
                if (!string.IsNullOrWhiteSpace(txt)) return txt;

                // 2) 兜底：服务器装了 LibreOffice 就转成文本再读（标准包自带）
                txt = ViaSoffice(fullPath);
                if (!string.IsNullOrWhiteSpace(txt)) return txt;

                // 3) 兜底：Linux 上的 antiword
                txt = RunTool("antiword", $"\"{fullPath}\"");
                if (!string.IsNullOrWhiteSpace(txt)) return txt;

                error = "该文件是老版 .doc，未能提取到文字"
                        + (string.IsNullOrWhiteSpace(docErr) ? "" : $"（{docErr}）")
                        + "；请另存为 .docx 后重新上传，或在「人员」里手动勾选。";
                return null;
            }
            error = $"暂不支持从 {ext} 文件自动识别姓名，请在「人员」里手动勾选。";
            return null;
        }
        catch (Exception ex)
        {
            error = "读取文件内容失败：" + ex.Message;
            return null;
        }
    }

    /// <summary>.docx = zip，正文在 word/document.xml；段落转成换行，去掉所有标签</summary>
    private static string? FromDocx(string path)
    {
        using var zip = ZipFile.OpenRead(path);
        var entry = zip.GetEntry("word/document.xml");
        if (entry == null) return null;
        using var sr = new StreamReader(entry.Open(), Encoding.UTF8);
        var xml = sr.ReadToEnd();

        var sb = new StringBuilder(xml.Length / 4);
        // 段落、换行、制表 → 分隔符，避免相邻段落文字粘连
        xml = Regex.Replace(xml, @"</w:p>", "\n");
        xml = Regex.Replace(xml, @"<w:br[^>]*/>", "\n");
        xml = Regex.Replace(xml, @"<w:tab[^>]*/>", " ");
        // 去掉所有标签
        var plain = Regex.Replace(xml, @"<[^>]+>", "");
        sb.Append(System.Net.WebUtility.HtmlDecode(plain));
        return sb.ToString();
    }

    /// <summary>用 LibreOffice 把 .doc 转成 UTF-8 纯文本（服务器装了 LO 时作为兜底）</summary>
    private static string? ViaSoffice(string fullPath)
    {
        var exe = PdfConverter.ExePath;
        if (string.IsNullOrWhiteSpace(exe)) return null;
        var tmp = Path.Combine(Path.GetTempPath(), "rmsys_doc2txt_" + Guid.NewGuid().ToString("N"));
        try
        {
            Directory.CreateDirectory(tmp);
            RunTool(exe!, $"--headless --norestore --convert-to \"txt:Text (encoded):UTF8\" --outdir \"{tmp}\" \"{fullPath}\"");
            foreach (var f in Directory.GetFiles(tmp, "*.txt"))
            {
                var t = File.ReadAllText(f, Encoding.UTF8);
                if (!string.IsNullOrWhiteSpace(t)) return t;
            }
            return null;
        }
        catch { return null; }
        finally
        {
            try { Directory.Delete(tmp, true); } catch { /* 临时目录清理失败不影响结果 */ }
        }
    }

    private static string? RunTool(string exe, string args)
    {
        try
        {
            var psi = new System.Diagnostics.ProcessStartInfo(exe, args)
            {
                RedirectStandardOutput = true,
                RedirectStandardError = true,
                UseShellExecute = false,
                CreateNoWindow = true
            };
            using var p = System.Diagnostics.Process.Start(psi);
            if (p == null) return null;
            var outp = p.StandardOutput.ReadToEnd();
            p.WaitForExit(20000);
            return p.ExitCode == 0 && outp.Length > 0 ? outp : null;
        }
        catch { return null; }
    }
}