彻底解决Html工具HtmlAgilityPack 中文乱码问题

作者:小鱼的互联网观察 发布时间:November 23, 2009 分类:互联网观察

对于HtmlAgilityPack,修改HtmlWeb.cs的如下方法(红色为修改代码):

        private HttpStatusCode Get(Uri uri, string method, string path, HtmlDocument doc)
        {
            string cachePath = null;
            HttpWebRequest req;
            bool oldFile = false;

            req = WebRequest.Create(uri) as HttpWebRequest;
            req.Method = method;

            _fromCache = false;
            _requestDuration = 0;
            int tc = Environment.TickCount;
            if (UsingCache)
            {
                cachePath = GetCachePath(req.RequestUri);
                if (File.Exists(cachePath))
                {
                    req.IfModifiedSince = File.GetLastAccessTime(cachePath);
                    oldFile = true;
                }
            }

            if (_cacheOnly)
            {
                if (!File.Exists(cachePath))
                {
                    throw new HtmlWebException("File was not found at cache path: '" + cachePath + "'");
                }

                if (path != null)
                {
                    IOLibrary.CopyAlways(cachePath, path);
                    // touch the file
                    File.SetLastWriteTime(path, File.GetLastWriteTime(cachePath));
                }
                _fromCache = true;
                return HttpStatusCode.NotModified;
            }

            if (_useCookies)
            {
                req.CookieContainer = new CookieContainer();
            }

            if (PreRequest != null)
            {
                // allow our user to change the request at will
                if (!PreRequest(req))
                {
                    return HttpStatusCode.ResetContent;
                }

                // dump cookie
//                if (_useCookies)
//                {
//                    foreach(Cookie cookie in req.CookieContainer.GetCookies(req.RequestUri))
//                    {
//                        HtmlLibrary.Trace("Cookie " + cookie.Name + "=" + cookie.Value + " path=" + cookie.Path + " domain=" + cookie.Domain);
//                    }
//                }
            }

            HttpWebResponse resp;

            try
            {
                resp = req.GetResponse() as HttpWebResponse;
            }
            catch (WebException we)
            {
                _requestDuration = Environment.TickCount - tc;
                resp = (HttpWebResponse)we.Response;
                if (resp == null)
                {
                    if (oldFile)
                    {
                        if (path != null)
                        {
                            IOLibrary.CopyAlways(cachePath, path);
                            // touch the file
                            File.SetLastWriteTime(path, File.GetLastWriteTime(cachePath));
                        }
                        return HttpStatusCode.NotModified;
                    }
                    throw;
                }
            }
            catch(Exception)
            {
                _requestDuration = Environment.TickCount - tc;
                throw;
            }

            // allow our user to get some info from the response
            if (PostResponse != null)
            {
                PostResponse(req, resp);
            }

            _requestDuration = Environment.TickCount - tc;
            _responseUri = resp.ResponseUri;
            bool html = IsHtmlContent(resp.ContentType);
            System.Text.Encoding respenc;

            if ((resp.ContentEncoding != null) && (resp.ContentEncoding.Length>0))
            {
                respenc = System.Text.Encoding.GetEncoding(resp.ContentEncoding);
            }
            else if ((resp.CharacterSet != null) && (resp.CharacterSet.Length > 0))//根据Content-Type中获取的charset
            {
                if (string.Compare(resp.CharacterSet, "ISO-8859-1", true, System.Globalization.CultureInfo.InvariantCulture) == 0)
                    respenc = System.Text.Encoding.GetEncoding("GB2312");
                else
                    respenc = System.Text.Encoding.GetEncoding(resp.CharacterSet);
            }
            else
            {
                respenc = System.Text.Encoding.GetEncoding("GB2312");
            }

            if (resp.StatusCode == HttpStatusCode.NotModified)
            {
                if (UsingCache)
                {
                    _fromCache = true;
                    if (path != null)
                    {
                        IOLibrary.CopyAlways(cachePath, path);
                        // touch the file
                        File.SetLastWriteTime(path, File.GetLastWriteTime(cachePath));
                    }
                    return resp.StatusCode;
                }
                else
                {
                    // this should *never* happen...
                    throw new HtmlWebException("Server has send a NotModifed code, without cache enabled.");
                }
            }
            Stream s = resp.GetResponseStream();
            if (s != null)
            {
                if (UsingCache)
                {
                    // NOTE: LastModified does not contain milliseconds, so we remove them to the file
                    SaveStream(s, cachePath, RemoveMilliseconds(resp.LastModified), _streamBufferSize);

                    // save headers
                    SaveCacheHeaders(req.RequestUri, resp);

                    if (path != null)
                    {
                        // copy and touch the file
                        IOLibrary.CopyAlways(cachePath, path);
                        File.SetLastWriteTime(path, File.GetLastWriteTime(cachePath));
                    }
                }
                else
                {
                    // try to work in-memory
                    if ((doc != null) && (html))
                    {
                        if (respenc != null)
                        {
                            doc.Load(s, respenc);
                        }
                        else
                        {
                            doc.Load(s, true);
                        }
                    }
                }
                resp.Close();
            }
            return resp.StatusCode;
        }

标签: 彻底解决Html工具HtmlAgilityPack, 中文乱码问题

互联网观察
彻底解决Html工具HtmlAgilityPack 中文乱码问题
本文地址:http://www.tianmeng.org/archives/173/

相关文章

  • 无相关文章
文档信息

当前暂无评论 »

网站地图 京ICP证030173号