public void OnSpiderInformation(SpiderArgs args)
{
if (SpiderInformation != null)
{
SpiderInformation(this, args);
}
}
public void OnSpiderScreenConsole(SpiderArgs args)
{
if (SpiderScreenConsole != null)
{
SpiderScreenConsole(this, args);
}
}
public void OnSpiderReloadForUpdate(SpiderArgs args)
{
if (SpiderReloadForUpdate != null)
{
SpiderReloadForUpdate(this, args);
}
}
public void OnSpiderProcessing(SpiderArgs args)
{
if (SpiderProcessing != null)
{
SpiderProcessing(this, args);
}
}
public void OnSpiderStatusChanged(SpiderArgs args)
{
if (SpiderStatusChanged != null)
{
SpiderStatusChanged(this, args);
}
}
private void GetSetArgs()
{
SpiderArgs args = new SpiderArgs
{
Hour = ConnectionStrings.GetArgsValue("Hour"),
Minute = ConnectionStrings.GetArgsValue("Minute"),
SpiderType = Convert.ToInt32(ConnectionStrings.GetArgsValue("SpiderType"))
};
bool IsSpl;
if (Convert.ToBoolean(ConnectionStrings.GetArgsValue("IsSql")))
{
// args = ArgsDal.LoadEntities(b => true).First();
IsSpl = true;
}
else
{
IsSpl = false;
}
this.BeginInvoke(new MethodInvoker(() =>
{
checkIsSql.Checked = IsSpl;
if (args.SpiderType == 1)
{
rbtnDay.Checked = true;
numHour.Value = Convert.ToDecimal(args.Hour.Trim());
numMinute.Value = Convert.ToDecimal(args.Minute.Trim());
}
else
{
rbtnHour.Checked = true;
numMinute2.Value = Convert.ToDecimal(args.Minute.Trim());
}
SetSetControlEnable(false);
}));
}
private void btnSetSave_Click(object sender, EventArgs e)
{
//ArgsDal.LoadEntities(null).First();
//SpiderArgs args = ArgsDal.LoadEntities(b=>true).First();
SpiderArgs args = new SpiderArgs();
args.SpiderArgsId = 1;
if (rbtnDay.Checked)
{
args.SpiderType = 1;
args.Hour = numHour.Value.ToString();
args.Minute = numMinute.Value.ToString();
}
else
{
args.SpiderType = 2;
args.Hour = String.Empty;
args.Minute = numMinute2.Value.ToString();
}
if (checkIsSql.Checked)
{
//var ar = ArgsDal.LoadEntities(b => true).First();
//ar.SpiderType = args.SpiderType;
//ar.Hour = args.Hour;
//ar.Minute = args.Minute;
//ArgsDal.EditEntity(ar);
//ArgsDal.SaveChange();
}
else
{
ConnectionStrings.UpdateArgsValue("SpiderType", args.SpiderType.ToString());
ConnectionStrings.UpdateArgsValue("Hour", args.Hour.ToString());
ConnectionStrings.UpdateArgsValue("Minute", args.Minute.ToString());
}
ConnectionStrings.UpdateArgsValue("IsSql", checkIsSql.Checked.ToString());
}
protected override void ProcessParseLink()
{
try
{
while (!IS_STOP)
{
int number = CounterManager.GetNumber();
_removelinks_detail = this.RemoveLinks.ToString();
string url = string.Empty;
if (QueueDetailURL.Count > 0)
{
ConcreteLink objUrl;
lock (QueueDetailURL)
{
try
{
objUrl = QueueDetailURL.Dequeue();
}
catch
{
continue;
}
}
string html = string.Empty;
try
{
if (string.IsNullOrEmpty(objUrl.href))
{
continue;
}
}
catch { continue; }
url = objUrl.href;
if (!objUrl.href.Contains("https://") && !objUrl.href.Contains("http://"))
{
url = this.BaseURL + objUrl.href;
}
try
{
html = DownloadExpress.getResponseString(url, "");
}
catch (Exception _error)
{
//-------------------------------------------------------------------------------------------------------
string _errorString = string.Format("{0}-ERROR:{1}-URLDownload:{2}", this.ThreadName, _error.ToString(), url);
SpiderArgs args = new SpiderArgs()
{
Message = _errorString
};
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(args);
}
if (string.IsNullOrEmpty(html))
{
continue;
}
#region FIND DETAIL LINK
/*foreach (var item in base.ConfigLinks.configlinks)
* {
* FindLinkBase findLink = null;
* if (item.pattern_type == "REGEX")
* {
* //if(item.link_type == "REMOVE_LINK")
*
* if (_removelinks_detail != string.Empty)
* {
* findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, _removelinks_detail);
* }
* else
* {
* findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, "");
* }
* }
*
*
* if (findLink == null)
* continue;
*
* if (item.link_type == "DETAIL_LINK")
* {
* List<string> linkDetail = findLink.FindLink(html);
* if (linkDetail == null || linkDetail.Count == 0)
* continue;
*
*
* foreach (string href in linkDetail)
* {
* string _url = href;
* if (IsNewUrl(ref _url))
* {
* ConcreteLink _concreteLink = new ConcreteLink();
* _concreteLink.link_type = "DETAIL_LINK";
* _concreteLink.href = href;
* QueueDetailURL.Enqueue(_concreteLink);
* }
* }
* }
* }*/
#endregion
#region PARSE FIELD
string url_md5 = StringHelpers.MD5Hash(url);
//check exist
try
{
string isExist = DownloadExpress.getResponseString(string.Format("http://localhost:9999/ContentHub/CheckExist?url_md5={0}", url_md5), "");
bool _isExist = Convert.ToBoolean(isExist);
if (_isExist)
{
CounterManager.InCreaseTryingCount();
continue;
}
}
catch (Exception _error)
{
}
//################################################################################
var dict_post = new Dictionary <string, object>();
dict_post.Add("UrlMd5", url_md5);
dict_post.Add("Url", url);
Dictionary <string, string> _dataString = new Dictionary <string, string>();
string the_last_config = string.Empty;
string content_filter = string.Empty;
string content_khongdau = string.Empty;
#region process parsing
try
{
string title = StringHelpers.GetStringBetween(html, "<h1><p class=\"title font_28\">", "</p></h1>");
if (string.IsNullOrEmpty(title))
{
continue;
}
dict_post.Add("Title", title);
string Price = StringHelpers.GetStringBetween(html, "<li class=\"li_100 clear icon_bds font_14 roboto_regular\">Giá : <span class=\"font_14 roboto_bold\">", "</span></li>");
dict_post.Add("Price", Price);
string Acreage = StringHelpers.GetStringBetween(html, "<li class=\"li_100 clear icon_bds font_14 roboto_regular\">Diện tích : <span class=\"font_14 roboto_bold\">", "<sup>2</sup></span></li>");
dict_post.Add("Acreage", Acreage);
string Address = StringHelpers.GetStringBetween(html, "<span class=\"roboto_bold font_14 cl_333 dc_new icon_bds\">Địa chỉ: </span>", "<span");
dict_post.Add("Address", Address);
string Room = StringHelpers.GetStringBetween(html, "<li class=\"li_100 clear icon_bds font_14 roboto_regular\">Có <span class=\"font_14 roboto_bold\">", "</span></li>");
dict_post.Add("Room", Room);
string Phone = StringHelpers.GetStringBetween(html, "<p id=\"mobile_show\" class=\"mobile_hide show_mobile\" style=\"*display: none; \"><span style=\"*padding-left: 10px\">", "</span></p>");
dict_post.Add("Phone", Phone);
string Content = StringHelpers.GetStringBetween(html, "<div class=\"info_text\">", "<div class=\"info_box\">");
Content = HtmlUtility.RemoveAllTag(Content);
dict_post.Add("Content", Content);
dict_post.Add("Location", "Ha Noi");
dict_post.Add("Source", "Rongbay");
string DatePost = StringHelpers.GetStringBetween(html, "Thời gian đăng : <span class=\"color cl_888 font_13\">", " -");
dict_post.Add("DatePost", DatePost);
string CategoryName = StringHelpers.GetStringBetween(html, "<span class=\"nameScate\">", "</span>");
CategoryName = HtmlUtility.RemoveAllTag(CategoryName);
dict_post.Add("CategoryName", CategoryName);
dict_post.Add("Metadata", html);
string jsonContent = JsonConvert.SerializeObject(dict_post, Formatting.Indented);
if (!string.IsNullOrWhiteSpace(jsonContent))
{
string results = string.Empty;
if (this.Mode != "TEST") // if not test then insert to api db
{
string _url_post = String.Join("&", dict_post.Select(i => $"{i.Key}={i.Value}"));
results = DownloadExpress.DownloadPost(this.PostURL, jsonContent);
if (!string.IsNullOrEmpty(results)) //check data return from api if not null or empty
{
try
{
var _info = JsonConvert.DeserializeObject <ContentHub>(results);
if (_info.Id > 0)
{
this._BaseManagement.TotalInsertLink = this._BaseManagement.TotalInsertLink + 1;
CounterManager.ResetTryingCount();
}
else
{
}
}
catch (Exception error)
{
string message = string.Format("[{0}] ERROR: {1} , {2}, {3}", ThreadName, error.Message, DateTime.Now, url);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
CounterManager.InCreaseTryingCount();
continue;
}
}
else //empty or error
{
CounterManager.InCreaseTryingCount();
}
}
try
{
SpiderArgs args = new SpiderArgs()
{
SourceName = this.ThreadName,
Index = number,
Href = url,
Title = (string)dict_post["Title"],
Message = results
};
SpiderSingletonEvent.Instance.OnSpiderProcessing(args);
}
catch { }
}
}
catch (Exception _errorParsing)
{
string message = string.Format("[{0}][PROCESS-PARSING] ERROR: {1} , {2}, {3}", ThreadName, _errorParsing.Message, DateTime.Now, url);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
CounterManager.InCreaseTryingCount();
continue;
}
#endregion
#endregion
}
else
{
CounterManager.InCreaseTryingCount();
}
if (this.Mode == "UPDATE")
{
if (number >= 2000)
{
CounterManager.CurrentTryingCount = 100000;
}
if (CounterManager.CurrentTryingCount >= this.MaxTryingCount)
{
string message = string.Format("[{0}] SLEEP FOR UPDATE AFTER {1} seconds, {2}", ThreadName, (this.UpdateSleep / 1000), DateTime.Now);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
Thread.Sleep(this.UpdateSleep);
try
{
//this._BaseManagement.TotalUpdateLink = 0;
//this._BaseManagement.TotalInsertLink = 0;
//UrlStoragClear();
//CounterManager.ResetTryingCount();
CounterManager.Reset(0, 100);
}
catch
{
//this._BaseManagement.TotalUpdateLink = 0;
//this._BaseManagement.TotalInsertLink = 0;
//UrlStoragClear();
//CounterManager.ResetTryingCount();
CounterManager.Reset(0, 100);
}
}
}
Thread.Sleep(this.THREAD_SLEEP);
}
}
catch (Exception ex)
{
string message = string.Format("[{0}] ERROR OUT OF THREAD: {1} , {2}", ThreadName, ex.Message.ToString(), DateTime.Now);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
}
}
protected override void ProcessFindLink()
{
try
{
while (!IS_STOP)
{
//int number = CounterManager.GetNumber();
_removelinks = this.RemoveLinks.ToString();
string url = string.Empty;
if (FrontierURL.Count > 0)
{
ConcreteLink objUrl;
lock (FrontierURL)
{
try
{
objUrl = FrontierURL.Dequeue();
}
catch { continue; }
}
string html = string.Empty;
if (string.IsNullOrEmpty(objUrl.href))
{
continue;
}
url = objUrl.href;
if (!objUrl.href.Contains("https://") && !objUrl.href.Contains("http://"))
{
url = this.BaseURL + objUrl.href;
}
string baseURL = string.Empty;
try
{
//html = DownloadExpress.Download(url, Utility.GetEncoding(EncodingType.UTF8));
html = DownloadExpress.getResponseString(url, "", out baseURL);
}
catch (Exception _error)
{
//-------------------------------------------------------------------------------------------------------
string _errorString = string.Format("{0}-ERROR:{1}-URLDownload:{2}", this.ThreadName, _error.ToString(), url);
SpiderArgs args = new SpiderArgs()
{
Message = _errorString
};
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(args);
//--------------------------------------------------------------------------------------------------------
try
{
html = DownloadExpress.getResponseString(url, "", out baseURL);
}
catch (Exception ex)
{
_errorString = string.Format("{0}-ERROR:{1}-URLDownloadBySocket:{2}", this.ThreadName, ex.ToString(), url);
args = new SpiderArgs()
{
Message = _errorString
};
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(args);
//--------------------------------------------------------------------------------------------------------
continue;
}
}
if (string.IsNullOrEmpty(html))
{
continue;
}
var _configLinkDetailLink = base.ConfigLinks.configlinks.Where(c => c.link_type == "DETAIL_LINK").ToList();
foreach (var item in base.ConfigLinks.configlinks)
{
FindLinkBase findLink = null;
if (item.pattern_type == "REGEX")
{
//if(item.link_type == "REMOVE_LINK")
if (_removelinks != string.Empty)
{
findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, _removelinks);
}
else
{
findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, "");
}
}
else if (item.pattern_type == "XPATH")
{
if (_removelinks != string.Empty)
{
findLink = new FindLinkByXPath(this.BaseURL, item.url_pattern, _removelinks);
}
else
{
findLink = new FindLinkByXPath(this.BaseURL, item.url_pattern, "");
}
}
if (findLink == null)
{
continue;
}
#region SUB_LINK
if (item.link_type == "SUB_LINK")
{
List <string> listLinkSub = findLink.FindLink(html);
if (listLinkSub == null || listLinkSub.Count == 0)
{
continue;
}
List <string> listSubLinkRemove = new List <string>();
//##################### filter sub link ##############################
//#####################################################################
if (listSubLinkRemove.Count > 0)
{
listLinkSub = listLinkSub.Where(s => !listSubLinkRemove.Contains(s)).ToList();
}
foreach (string linkSub in listLinkSub)
{
string _link = linkSub;
if (!_link.Contains("http://") && !_link.Contains("https://"))
{
_link = baseURL + linkSub;
}
string _url = _link;
if (IsNewUrl(ref _url))
{
ConcreteLink _concreteLink = new ConcreteLink();
_concreteLink.link_type = "SUB_LINK";
_concreteLink.href = _url;
FrontierURL.Enqueue(_concreteLink);
SpiderArgs args = new SpiderArgs()
{
Message = string.Format("{0}-SUB-{1}", _url, this.ThreadName)
};
SpiderSingletonEvent.Instance.OnSpiderInformation(args);
}
}
}
#endregion
#region DETAIL_LINK
else if (item.link_type == "DETAIL_LINK")
{
List <string> linkDetail = findLink.FindLink(html);
if (linkDetail == null || linkDetail.Count == 0)
{
continue;
}
foreach (string href in linkDetail)
{
string _link = href;
if (!_link.Contains("http://") && !_link.Contains("https://"))
{
_link = baseURL + href;
}
string _url = _link;
if (IsNewUrl(ref _url))
{
ConcreteLink _concreteLink = new ConcreteLink();
_concreteLink.link_type = "DETAIL_LINK";
_concreteLink.href = _url;
QueueDetailURL.Enqueue(_concreteLink);
}
}
}
#endregion
}
}
if (this.Mode == "UPDATE")
{
if (CounterManager.CurrentTryingCount >= this.MaxTryingCount)
{
string message = string.Format("[{0}] SLEEP FOR UPDATE AFTER {1} seconds, {2}", ThreadName, (this.UpdateSleep / 1000), DateTime.Now);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
SpiderSingletonEvent.Instance.OnSpiderReloadForUpdate(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = string.Format("{0} RELOAD FOR UPDATE \r\n", ThreadName)
});
Thread.Sleep(this.UpdateSleep);
try
{
this._BaseManagement.TotalUpdateLink = 0;
this._BaseManagement.TotalInsertLink = 0;
//CounterManager.ResetTryingCount();
//CounterManager.CurrentNumber = 0;
UrlStoragClear();
CounterManager.Reset(0, 100);
}
catch
{
this._BaseManagement.TotalUpdateLink = 0;
this._BaseManagement.TotalInsertLink = 0;
//CounterManager.ResetTryingCount();
//CounterManager.CurrentNumber = 0;
UrlStoragClear();
CounterManager.Reset(0, 100);
}
}
}
Thread.Sleep(this.THREAD_SLEEP);
}
}
catch (Exception ex)
{
string test = ex.Message;
}
}
protected override void ProcessParseLink()
{
try
{
while (!IS_STOP)
{
int number = CounterManager.GetNumber();
_removelinks_detail = this.RemoveLinks.ToString();
string url = string.Empty;
if (QueueDetailURL.Count > 0)
{
ConcreteLink objUrl;
lock (QueueDetailURL)
{
try {
objUrl = QueueDetailURL.Dequeue();
}
catch
{
continue;
}
}
string html = string.Empty;
try
{
if (string.IsNullOrEmpty(objUrl.href))
{
continue;
}
}
catch { continue; }
url = objUrl.href;
if (!objUrl.href.Contains("https://") && !objUrl.href.Contains("http://"))
{
url = this.BaseURL + objUrl.href;
}
try
{
html = DownloadExpress.getResponseString(url, "");
}
catch (Exception _error)
{
//-------------------------------------------------------------------------------------------------------
string _errorString = string.Format("{0}-ERROR:{1}-URLDownload:{2}", this.ThreadName, _error.ToString(), url);
SpiderArgs args = new SpiderArgs()
{
Message = _errorString
};
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(args);
}
if (string.IsNullOrEmpty(html))
{
continue;
}
//#region FIND DETAIL LINK
//foreach (var item in base.ConfigLinks.configlinks)
//{
// FindLinkBase findLink = null;
// if (item.pattern_type == "REGEX")
// {
// //if(item.link_type == "REMOVE_LINK")
// if (_removelinks_detail != string.Empty)
// {
// findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, _removelinks_detail);
// }
// else
// {
// findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, "");
// }
// }
// if(item.pattern_type == "XPATH")
// {
// if (_removelinks_detail != string.Empty)
// {
// findLink = new FindLinkByXPath(this.BaseURL, item.url_pattern, _removelinks_detail);
// }
// else
// {
// findLink = new FindLinkByXPath(this.BaseURL, item.url_pattern, "");
// }
// }
// if (findLink == null)
// continue;
// if (item.link_type == "DETAIL_LINK")
// {
// List<string> linkDetail = findLink.FindLink(html);
// if (linkDetail == null || linkDetail.Count == 0)
// continue;
// foreach (string href in linkDetail)
// {
// string _url = href;
// if (IsNewUrl(ref _url))
// {
// ConcreteLink _concreteLink = new ConcreteLink();
// _concreteLink.link_type = "DETAIL_LINK";
// _concreteLink.href = href;
// QueueDetailURL.Enqueue(_concreteLink);
// }
// }
// }
//}
//#endregion
#region PARSE FIELD
var dict_post = new Dictionary <string, object>();
dict_post.Add("source_id", this.SourceID);
dict_post.Add("href", url);
Dictionary <string, string> _dataString = new Dictionary <string, string>();
string content_filter = string.Empty;
string content_khongdau = string.Empty;
#region process parsing
try
{
//------------------------------- video ------------------------------------------------------------
try
{
//process parsing video
foreach (var itemvideo in base.ConfigVideos.configvideos)
{
string field_name = itemvideo.name;
string pattern_type = itemvideo.config.pattern_type;
if (pattern_type == PatternType.XPATH.ToString()) //pattern type xpath
{
foreach (var config in itemvideo.config.step)
{
string video_url = XPATHHelpers.ParsingHTML(html, config.Value.field_value);
int is_url_video_cache = 0;
int.TryParse(config.Value.is_url_video_cache, out is_url_video_cache);
if (!string.IsNullOrEmpty(video_url))
{
if (is_url_video_cache == 1) //if video url is cache
{
var dict_video_post = new Dictionary <string, object>();
dict_video_post.Add("content_id", "");
dict_video_post.Add("source_id", this.SourceID);
dict_video_post.Add("video_url", video_url);
dict_video_post.Add("video_url_cache", video_url);
dict_video_post.Add("status", "NEED_CACHE");
dict_video_post.Add("video_url_path", "empty");
dict_video_post.Add("status_download", "PREPARING");
string json_video_Content = JsonConvert.SerializeObject(dict_video_post, Formatting.Indented);
var response_video = DownloadExpress.DownloadPost(this.VIDEO_POST_URL, json_video_Content);
}
else
{
var dict_video_post = new Dictionary <string, object>();
dict_video_post.Add("content_id", "");
dict_video_post.Add("source_id", this.SourceID);
dict_video_post.Add("video_url", video_url);
dict_video_post.Add("video_url_cache", "empty");
dict_video_post.Add("status", "NEED_PARSING");
dict_video_post.Add("video_url_path", "empty");
dict_video_post.Add("status_download", "PREPARING");
string json_video_Content = JsonConvert.SerializeObject(dict_video_post, Formatting.Indented);
var response_video = DownloadExpress.DownloadPost(this.VIDEO_POST_URL, json_video_Content);
}
}
}
}
}
}
catch { }
//---------------------------------------------------------------------------------------------------
foreach (var itemfield in base.ConfigFields.configfields)
{
string field_name = itemfield.name;
string htmlParser = html;
string content_text = "";
string the_last_config = string.Empty;
string pattern_type = itemfield.config.pattern_type;
if (pattern_type == PatternType.XPATH.ToString()) //pattern type xpath
{
foreach (var config in itemfield.config.step)
{
//################# check break_parsing point ##############################
bool break_parsing = false;
try
{
if (!string.IsNullOrWhiteSpace(config.Value.break_parsing))
{
if (Convert.ToInt32(config.Value.break_parsing) == 0)
{
break_parsing = false;
}
else
{
break_parsing = true;
}
}
}
catch { break_parsing = false; }
//###########################################################################
if (break_parsing) //breaking parsing
{
htmlParser = XPATHHelpers.ParsingHTML(html, config.Value.xpath);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
if (field_name == "content")
{
the_last_config = config.Value.xpath;
}
}
}
else
{
htmlParser = XPATHHelpers.ParsingHTML(htmlParser, config.Value.xpath);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
if (field_name == "content")
{
the_last_config = config.Value.xpath;
}
}
}
//################## process remove html tag ###############################
try
{
if (!string.IsNullOrWhiteSpace(config.Value.remove_html))
{
if (Convert.ToInt32(config.Value.remove_html) == 1)
{
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = HtmlUtility.RemoveAllTag(htmlParser);
}
}
}
}
catch { }
//##########################################################################
}
}
else // STRING BETWEEN
{
foreach (var config in itemfield.config.step)
{
//################# check break_parsing point ##############################
bool break_parsing = false;
try
{
if (!string.IsNullOrWhiteSpace(config.Value.break_parsing))
{
if (Convert.ToInt32(config.Value.break_parsing) == 0)
{
break_parsing = false;
}
else
{
break_parsing = true;
}
}
}
catch { break_parsing = false; }
//###########################################################################
if (break_parsing) //breaking parsing
{
if (string.IsNullOrWhiteSpace(content_text))
{
htmlParser = StringHelpers.GetStringBetween(html, config.Value.start_pattern, config.Value.end_pattern);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
the_last_config = config.Value.start_pattern;
}
}
else
{
htmlParser = StringHelpers.GetStringBetween(content_text, config.Value.start_pattern, config.Value.end_pattern);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
the_last_config = config.Value.start_pattern;
}
}
}
else
{
htmlParser = StringHelpers.GetStringBetween(htmlParser, config.Value.start_pattern, config.Value.end_pattern);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
the_last_config = config.Value.start_pattern;
}
}
//################## process remove html tag ###############################
try
{
if (!string.IsNullOrWhiteSpace(config.Value.remove_html))
{
if (Convert.ToInt32(config.Value.remove_html) == 1)
{
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = HtmlUtility.RemoveAllTag(htmlParser);
}
}
}
}
catch { }
//##########################################################################
} //endfor
}
if (!string.IsNullOrWhiteSpace(content_text))
{
if (field_name.ToLower() == "title")
{
dict_post.Add("title", content_text);
content_khongdau += StringHelpers.ConvertToKD(content_text) + "\n\n";
}
if (field_name.ToLower() == "tag_name")
{
if (string.IsNullOrWhiteSpace(content_text))
{
content_text = "Không xác định";
}
dict_post.Add("tag_name", content_text.Trim().Replace(" ", ""));
}
if (field_name.ToLower() == "published_at")
{
bool checkDateTime = false;
if (content_text.Contains("."))
{
content_text = content_text.Replace(".", "/");
}
//content_text = "Thứ hai, 15/5/2017 | 09:48 GMT+7";
DateTime _published_date = Utility.ConvertDateTimeStringForArchived(content_text, out checkDateTime).Date;
if (checkDateTime)
{
dict_post.Add("published_at", string.Format("{0}-{1}-{2}", _published_date.Year, _published_date.Month, _published_date.Day));
}
else
{
DateTime _published_date_ = Utility.ConvertDateTimeStringForArchived(content_text, out checkDateTime).Date;
string message = string.Format("[{0}] ERROR: {1} , {2}, {3}", ThreadName, "NOT PARSING DATATIME", content_text, url);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
continue;
}
content_text = content_text.Trim().Replace(" ", "");
dict_post.Add("published_time", content_text);
}
if (field_name.ToLower() == "content")
{
content_filter = HtmlUtility.RemoveAllTag(content_text);
content_khongdau += StringHelpers.ConvertToKD(content_text) + "\n\n";
dict_post.Add("content", content_text);
}
//_dataString.Add(field_name, content_text);
}
}
//_dataString.Add("html_data", html);
//string jsonContent = JsonConvert.SerializeObject(_dataString, Formatting.Indented);
//dict_post.Add("data", jsonContent);
dict_post.Add("content_filter", content_filter);
dict_post.Add("content_khongdau", content_khongdau);
dict_post.Add("full_html", WebUtility.HtmlEncode(html));
string jsonContent = JsonConvert.SerializeObject(dict_post, Formatting.Indented);
if (!string.IsNullOrWhiteSpace(jsonContent))
{
string results = string.Empty;
if (this.Mode != "TEST") // if not test then insert to api db
{
results = DownloadExpress.DownloadPost(this.PostURL, jsonContent);
if (!string.IsNullOrEmpty(results)) //check data return from api if not null or empty
{
try
{
var _info = JsonConvert.DeserializeObject <Dictionary <string, string> >(results);
string type = _info["type"];
switch (type)
{
case "UPDATE_VERSION_SUCCESS":
string content_id = _info["content_id"];
this._BaseManagement.TotalUpdateLink = this._BaseManagement.TotalUpdateLink + 1;
CounterManager.ResetTryingCount();
//--------------BEGIN capture content to images -------------------------------------------------------------------
CaptureArchivied.PDFConverterServiceClient _captureService = new CaptureArchivied.PDFConverterServiceClient();
CaptureArchivied.ConvertPDFRequest _pdfRequest = new CaptureArchivied.ConvertPDFRequest();
_pdfRequest.contentId = content_id;
_pdfRequest.pageURL = url;
_pdfRequest.filterBy = this.chk_unique_css == 1 ? 0 : 1;
_pdfRequest.filterText = this.filter_pdf;
_pdfRequest.deleteItems = this.remove_filter_pdf;
var return_data = _captureService.ConvertToPDF(_pdfRequest);
//--------------END capture content to images ---------------------------------------------------------------------
break;
case "INSERT_SUCCESS":
string content_insert_id = _info["content_id"];
this._BaseManagement.TotalInsertLink = this._BaseManagement.TotalInsertLink + 1;
CounterManager.ResetTryingCount();
//--------------BEGIN capture content to images -------------------------------------------------------------------
CaptureArchivied.PDFConverterServiceClient insert_captureService = new CaptureArchivied.PDFConverterServiceClient();
CaptureArchivied.ConvertPDFRequest insert_pdfRequest = new CaptureArchivied.ConvertPDFRequest();
insert_pdfRequest.contentId = content_insert_id;
insert_pdfRequest.pageURL = url;
insert_pdfRequest.filterBy = this.chk_unique_css == 1 ? 0 : 1;
insert_pdfRequest.filterText = this.filter_pdf;
insert_pdfRequest.deleteItems = this.remove_filter_pdf;
var insert_return_data = insert_captureService.ConvertToPDF(insert_pdfRequest);
//--------------END capture content to images ---------------------------------------------------------------------
break;
default:
CounterManager.InCreaseTryingCount();
break;
}
}
catch (Exception error)
{
string message = string.Format("[{0}] ERROR: {1} , {2}, {3}", ThreadName, error.Message, DateTime.Now, url);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
CounterManager.InCreaseTryingCount();
continue;
}
}
else //empty or error
{
CounterManager.InCreaseTryingCount();
}
}
try
{
SpiderArgs args = new SpiderArgs()
{
SourceName = this.ThreadName,
Index = number,
Href = url,
Title = (string)dict_post["title"],
Message = results
};
SpiderSingletonEvent.Instance.OnSpiderProcessing(args);
}
catch { }
}
}
catch (Exception _errorParsing) {
string message = string.Format("[{0}][PROCESS-PARSING] ERROR: {1} , {2}, {3}", ThreadName, _errorParsing.Message, DateTime.Now, url);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
CounterManager.InCreaseTryingCount();
continue;
}
#endregion
#endregion
}
else
{
CounterManager.InCreaseTryingCount();
}
if (this.Mode == "UPDATE")
{
if (number >= 2000)
{
CounterManager.CurrentTryingCount = 100000;
}
if (CounterManager.CurrentTryingCount >= this.MaxTryingCount)
{
string message = string.Format("[{0}] SLEEP FOR UPDATE AFTER {1} seconds, {2}", ThreadName, (this.UpdateSleep / 1000), DateTime.Now);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
Thread.Sleep(this.UpdateSleep);
try
{
//this._BaseManagement.TotalUpdateLink = 0;
//this._BaseManagement.TotalInsertLink = 0;
//UrlStoragClear();
//CounterManager.ResetTryingCount();
CounterManager.Reset(0, 100);
}
catch
{
//this._BaseManagement.TotalUpdateLink = 0;
//this._BaseManagement.TotalInsertLink = 0;
//UrlStoragClear();
//CounterManager.ResetTryingCount();
CounterManager.Reset(0, 100);
}
}
}
Thread.Sleep(this.THREAD_SLEEP);
}
}
catch (Exception ex)
{
string message = string.Format("[{0}] ERROR OUT OF THREAD: {1} , {2}", ThreadName, ex.Message.ToString(), DateTime.Now);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
}
}
protected override void ProcessParseLink()
{
try
{
while (!IS_STOP)
{
int number = CounterManager.GetNumber();
_removelinks_detail = this.RemoveLinks.ToString();
string url = string.Empty;
if (QueueDetailURL.Count > 0)
{
ConcreteLink objUrl;
lock (QueueDetailURL)
{
try
{
objUrl = QueueDetailURL.Dequeue();
}
catch
{
continue;
}
}
string html = string.Empty;
try
{
if (string.IsNullOrEmpty(objUrl.href))
{
continue;
}
}
catch { continue; }
url = objUrl.href;
if (!objUrl.href.Contains("https://") && !objUrl.href.Contains("http://"))
{
url = this.BaseURL + objUrl.href;
}
try
{
html = DownloadExpress.getResponseString(url, "");
}
catch (Exception _error)
{
//-------------------------------------------------------------------------------------------------------
string _errorString = string.Format("{0}-ERROR:{1}-URLDownload:{2}", this.ThreadName, _error.ToString(), url);
SpiderArgs args = new SpiderArgs()
{
Message = _errorString
};
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(args);
}
if (string.IsNullOrEmpty(html))
{
continue;
}
#region FIND DETAIL LINK
/*foreach (var item in base.ConfigLinks.configlinks)
* {
* FindLinkBase findLink = null;
* if (item.pattern_type == "REGEX")
* {
* //if(item.link_type == "REMOVE_LINK")
*
* if (_removelinks_detail != string.Empty)
* {
* findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, _removelinks_detail);
* }
* else
* {
* findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, "");
* }
* }
*
*
* if (findLink == null)
* continue;
*
* if (item.link_type == "DETAIL_LINK")
* {
* List<string> linkDetail = findLink.FindLink(html);
* if (linkDetail == null || linkDetail.Count == 0)
* continue;
*
*
* foreach (string href in linkDetail)
* {
* string _url = href;
* if (IsNewUrl(ref _url))
* {
* ConcreteLink _concreteLink = new ConcreteLink();
* _concreteLink.link_type = "DETAIL_LINK";
* _concreteLink.href = href;
* QueueDetailURL.Enqueue(_concreteLink);
* }
* }
* }
* }*/
#endregion
#region PARSE FIELD
var dict_post = new Dictionary <string, object>();
dict_post.Add("source_id", this.SourceID);
dict_post.Add("href", url);
Dictionary <string, string> _dataString = new Dictionary <string, string>();
string the_last_config = string.Empty;
string content_filter = string.Empty;
string content_khongdau = string.Empty;
#region process parsing
try
{
#region Parsing Field
foreach (var itemfield in base.ConfigFields.configfields)
{
string field_name = itemfield.name;
string htmlParser = html;
string content_text = "";
string pattern_type = itemfield.config.pattern_type;
if (pattern_type == PatternType.XPATH.ToString()) //pattern type xpath
{
foreach (var config in itemfield.config.step)
{
//################# check break_parsing point ##############################
bool break_parsing = false;
try
{
if (!string.IsNullOrWhiteSpace(config.Value.break_parsing))
{
if (Convert.ToInt32(config.Value.break_parsing) == 0)
{
break_parsing = false;
}
else
{
break_parsing = true;
}
}
}
catch { break_parsing = false; }
//###########################################################################
if (break_parsing) //breaking parsing
{
htmlParser = XPATHHelpers.ParsingHTML(html, config.Value.xpath);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
if (field_name == "content")
{
the_last_config = config.Value.xpath;
}
}
}
else
{
htmlParser = XPATHHelpers.ParsingHTML(htmlParser, config.Value.xpath);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
if (field_name == "content")
{
the_last_config = config.Value.xpath;
}
}
}
//################## process remove html tag ###############################
try
{
if (!string.IsNullOrWhiteSpace(config.Value.remove_html))
{
if (Convert.ToInt32(config.Value.remove_html) == 1)
{
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = HtmlUtility.RemoveAllTag(htmlParser);
}
}
}
}
catch { }
//##########################################################################
}
}
else //pattern type string_between
{
foreach (var config in itemfield.config.step)
{
//################# check break_parsing point ##############################
bool break_parsing = false;
try
{
if (!string.IsNullOrWhiteSpace(config.Value.break_parsing))
{
if (Convert.ToInt32(config.Value.break_parsing) == 0)
{
break_parsing = false;
}
else
{
break_parsing = true;
}
}
}
catch { break_parsing = false; }
//###########################################################################
if (break_parsing) //breaking parsing
{
if (string.IsNullOrWhiteSpace(content_text))
{
htmlParser = StringHelpers.GetStringBetween(html, config.Value.start_pattern, config.Value.end_pattern);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
}
}
else
{
htmlParser = StringHelpers.GetStringBetween(content_text, config.Value.start_pattern, config.Value.end_pattern);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
}
}
}
else
{
htmlParser = StringHelpers.GetStringBetween(htmlParser, config.Value.start_pattern, config.Value.end_pattern);
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = htmlParser;
}
}
//################## process remove html tag ###############################
try
{
if (!string.IsNullOrWhiteSpace(config.Value.remove_html))
{
if (Convert.ToInt32(config.Value.remove_html) == 1)
{
if (!string.IsNullOrWhiteSpace(htmlParser))
{
content_text = HtmlUtility.RemoveAllTag(htmlParser);
}
}
}
}
catch { }
//##########################################################################
}
}
if (!string.IsNullOrWhiteSpace(content_text))
{
if (field_name.ToLower() == "title")
{
dict_post.Add("title", content_text);
content_khongdau += StringHelpers.ConvertToKD(content_text) + "\n\n";
}
if (field_name.ToLower() == "tag_name")
{
if (string.IsNullOrWhiteSpace(content_text))
{
content_text = "Không xác định";
}
dict_post.Add("tag_name", content_text.Trim().Replace(" ", ""));
}
if (field_name.ToLower() == "published_at")
{
bool checkDateTime = false;
//content_text = "Thứ hai, 15/5/2017 | 09:48 GMT+7";
DateTime _published_date = Utility.ConvertDateTimeStringForArchived(content_text, out checkDateTime).Date;
if (checkDateTime)
{
dict_post.Add("published_at", string.Format("{0}-{1}-{2}", _published_date.Year, _published_date.Month, _published_date.Day));
}
else
{
DateTime _published_date_ = Utility.ConvertDateTimeStringForArchived(content_text, out checkDateTime).Date;
string message = string.Format("[{0}] ERROR: {1} , {2}, {3}", ThreadName, "NOT PARSING DATATIME", content_text, url);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
continue;
}
content_text = content_text.Trim().Replace(" ", "");
dict_post.Add("published_time", content_text);
}
if (field_name.ToLower() == "content")
{
content_filter = HtmlUtility.RemoveAllTag(content_text);
content_khongdau += StringHelpers.ConvertToKD(content_text) + "\n\n";
dict_post.Add("content", content_text);
}
_dataString.Add(field_name, content_text);
}
}
#endregion
dict_post.Add("content_filter", content_filter);
dict_post.Add("content_khongdau", content_khongdau);
dict_post.Add("full_html", html);
string jsonContent = JsonConvert.SerializeObject(dict_post, Formatting.Indented);
if (!string.IsNullOrWhiteSpace(jsonContent))
{
string results = string.Empty;
if (this.Mode != "TEST") // if not test then insert to api db
{
results = DownloadExpress.DownloadPost(this.PostURL, jsonContent);
if (!string.IsNullOrEmpty(results)) //check data return from api if not null or empty
{
try
{
var _info = JsonConvert.DeserializeObject <Dictionary <string, string> >(results);
string type = _info["type"];
switch (type)
{
case "UPDATE_VERSION_SUCCESS":
this._BaseManagement.TotalUpdateLink = this._BaseManagement.TotalUpdateLink + 1;
CounterManager.ResetTryingCount();
string content_id = _info["content_id"];
//--------------BEGIN capture content to images -------------------------------------------------------------------
if (this.SourceID == "58d0ee601d41c849a893a49d") //vnexpress
{
CaptureArchivied.PDFConverterServiceClient _captureService = new CaptureArchivied.PDFConverterServiceClient();
if (url.Contains("sohoa.vnexpress.net"))
{
CaptureArchivied.ConvertPDFRequest _pdfRequest = new CaptureArchivied.ConvertPDFRequest();
_pdfRequest.contentId = content_id;
_pdfRequest.pageURL = url;
_pdfRequest.filterBy = 1;
_pdfRequest.filterText = "main_content_detail";
_pdfRequest.deleteItems = "#box_tinlienquan,#box_comment,.banner_468,#social_like,.xemthem_new_ver,.banner_468x90,.block_goithutoasoan";
//capture by id
var return_data = _captureService.ConvertToPDF(_pdfRequest);
}
else
{
CaptureArchivied.ConvertPDFRequest _pdfRequest = new CaptureArchivied.ConvertPDFRequest();
_pdfRequest.contentId = content_id;
_pdfRequest.pageURL = url;
_pdfRequest.filterBy = 1;
_pdfRequest.filterText = "block_col_480";
_pdfRequest.deleteItems = "#box_tinlienquan,#box_comment,.banner_468,#social_like,.xemthem_new_ver,.banner_468x90,.block_goithutoasoan";
var return_data = _captureService.ConvertToPDF(_pdfRequest);
}
//capture by id
/*if (the_last_config.Contains("fck_detail"))
* {
* CaptureArchivied.ReturnData return_data = _captureService.Capture(url, 3, "fck_detail", null, content_id);
* }
* else if(the_last_config.Contains("article_content"))
* {
* CaptureArchivied.ReturnData return_data = _captureService.Capture(url, 2, "article_content", null, content_id);
* }
* else if (the_last_config.Contains("container_tab_live"))
* {
* CaptureArchivied.ReturnData return_data = _captureService.Capture(url, 2, "container_tab_live", null, content_id);
* }*/
}
//--------------END capture content to images ---------------------------------------------------------------------
break;
case "INSERT_SUCCESS":
string content_insert_id = _info["content_id"];
this._BaseManagement.TotalInsertLink = this._BaseManagement.TotalInsertLink + 1;
CounterManager.ResetTryingCount();
//--------------BEGIN capture content to images -------------------------------------------------------------------
if (this.SourceID == "58d0ee601d41c849a893a49d") //vnexpress
{
CaptureArchivied.PDFConverterServiceClient _captureService = new CaptureArchivied.PDFConverterServiceClient();
if (url.Contains("sohoa.vnexpress.net"))
{
CaptureArchivied.ConvertPDFRequest _pdfRequest = new CaptureArchivied.ConvertPDFRequest();
_pdfRequest.contentId = content_insert_id;
_pdfRequest.pageURL = url;
_pdfRequest.filterBy = 1;
_pdfRequest.filterText = "main_content_detail";
_pdfRequest.deleteItems = "#box_tinlienquan,#box_comment,.banner_468,#social_like,.xemthem_new_ver,.banner_468x90,.block_goithutoasoan";
//capture by id
var return_data = _captureService.ConvertToPDF(_pdfRequest);
}
else
{
CaptureArchivied.ConvertPDFRequest _pdfRequest = new CaptureArchivied.ConvertPDFRequest();
_pdfRequest.contentId = content_insert_id;
_pdfRequest.pageURL = url;
_pdfRequest.filterBy = 1;
_pdfRequest.filterText = "block_col_480";
_pdfRequest.deleteItems = "#box_tinlienquan,#box_comment,.banner_468,#social_like,.xemthem_new_ver,.banner_468x90,.block_goithutoasoan";
var return_data = _captureService.ConvertToPDF(_pdfRequest);
}
//capture by id
/*if (the_last_config.Contains("fck_detail"))
* {
* CaptureArchivied.ReturnData return_data = _captureService.Capture(url, 3, "fck_detail", null, content_insert_id);
* }
* else if (the_last_config.Contains("article_content"))
* {
* CaptureArchivied.ReturnData return_data = _captureService.Capture(url, 2, "article_content", null, content_insert_id);
* }
* else if (the_last_config.Contains("container_tab_live"))
* {
* CaptureArchivied.ReturnData return_data = _captureService.Capture(url, 2, "container_tab_live", null, content_insert_id);
* }*/
}
//--------------END capture content to images ---------------------------------------------------------------------
break;
default:
CounterManager.InCreaseTryingCount();
break;
}
}
catch (Exception error)
{
string message = string.Format("[{0}] ERROR: {1} , {2}, {3}", ThreadName, error.Message, DateTime.Now, url);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
CounterManager.InCreaseTryingCount();
continue;
}
}
else //empty or error
{
CounterManager.InCreaseTryingCount();
}
}
try
{
SpiderArgs args = new SpiderArgs()
{
SourceName = this.ThreadName,
Index = number,
Href = url,
Title = (string)dict_post["title"],
Message = results
};
SpiderSingletonEvent.Instance.OnSpiderProcessing(args);
}
catch { }
}
}
catch (Exception _errorParsing)
{
string message = string.Format("[{0}][PROCESS-PARSING] ERROR: {1} , {2}, {3}", ThreadName, _errorParsing.Message, DateTime.Now, url);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
CounterManager.InCreaseTryingCount();
continue;
}
#endregion
#endregion
}
else
{
CounterManager.InCreaseTryingCount();
}
if (this.Mode == "UPDATE")
{
if (number >= 2000)
{
CounterManager.CurrentTryingCount = 100000;
}
if (CounterManager.CurrentTryingCount >= this.MaxTryingCount)
{
string message = string.Format("[{0}] SLEEP FOR UPDATE AFTER {1} seconds, {2}", ThreadName, (this.UpdateSleep / 1000), DateTime.Now);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
Thread.Sleep(this.UpdateSleep);
try
{
//this._BaseManagement.TotalUpdateLink = 0;
//this._BaseManagement.TotalInsertLink = 0;
//UrlStoragClear();
//CounterManager.ResetTryingCount();
CounterManager.Reset(0, 100);
}
catch
{
//this._BaseManagement.TotalUpdateLink = 0;
//this._BaseManagement.TotalInsertLink = 0;
//UrlStoragClear();
//CounterManager.ResetTryingCount();
CounterManager.Reset(0, 100);
}
}
}
Thread.Sleep(this.THREAD_SLEEP);
}
}
catch (Exception ex)
{
string message = string.Format("[{0}] ERROR OUT OF THREAD: {1} , {2}", ThreadName, ex.Message.ToString(), DateTime.Now);
SpiderSingletonEvent.Instance.OnSpiderScreenConsole(new BlankSpider.Spider.Events.SpiderArgs()
{
Message = message
});
}
}
protected override void ProcessParseLink()
{
try
{
while (!IS_STOP)
{
int number = CounterManager.GetNumber();
string url = string.Empty;
if (QueueDetailURL.Count > 0)
{
ConcreteLink objUrl = QueueDetailURL.Dequeue();
string html = string.Empty;
try {
if (string.IsNullOrEmpty(objUrl.href))
{
continue;
}
}
catch { continue; }
url = objUrl.href;
if (!objUrl.href.Contains("https://") && !objUrl.href.Contains("http://"))
{
url = this.BaseURL + objUrl.href;
}
try
{
html = DownloadExpress.Download(url, Utility.GetEncoding(EncodingType.UTF8));
}
catch
{
try
{
html = DownloadExpress.DownloadBySocket(url, Utility.GetEncoding(EncodingType.UTF8));
}
catch { continue; }
}
if (string.IsNullOrEmpty(html))
{
continue;
}
#region PARSE FIELD
var dict_post = new Dictionary <string, object>();
dict_post.Add("source_id", this.SourceID);
dict_post.Add("href", url);
Dictionary <string, string> _dataString = new Dictionary <string, string>();
foreach (var itemfield in base.ConfigFields.configfields)
{
string field_name = itemfield.name;
string htmlParser = html;
string content_text = "";
foreach (var config in itemfield.config.step)
{
htmlParser = StringHelpers.GetStringBetween(htmlParser, config.Value.start_pattern, config.Value.end_pattern);
content_text = htmlParser;
}
if (!string.IsNullOrWhiteSpace(content_text))
{
if (field_name.ToLower() == "title")
{
SpiderArgs args = new SpiderArgs()
{
SourceName = this.ThreadName,
Index = number,
Href = url,
Title = content_text,
Message = string.Empty
};
SpiderSingletonEvent.Instance.OnSpiderProcessing(args);
}
if (field_name.ToLower() == "tag_name")
{
dict_post.Add("tag_name", content_text);
}
_dataString.Add(field_name, content_text);
}
}
_dataString.Add("html_data", html);
string jsonContent = JsonConvert.SerializeObject(_dataString, Formatting.Indented);
dict_post.Add("data", jsonContent);
jsonContent = JsonConvert.SerializeObject(dict_post, Formatting.Indented);
if (!string.IsNullOrWhiteSpace(jsonContent))
{
string results = DownloadExpress.DownloadPost(this.PostURL, jsonContent);
}
#endregion
}
Thread.Sleep(this.THREAD_SLEEP);
}
}
catch (Exception ex)
{
string test = ex.Message;
}
}
protected override void ProcessFindLink()
{
try
{
while (!IS_STOP)
{
int number = CounterManager.GetNumber();
//_removelinks = this.RemoveLinks.ToString();
string url = string.Empty;
if (FrontierURL.Count > 0)
{
ConcreteLink objUrl = FrontierURL.Dequeue();
string html = string.Empty;
if (string.IsNullOrEmpty(objUrl.href))
{
continue;
}
url = objUrl.href;
if (!objUrl.href.Contains("https://") && !objUrl.href.Contains("http://"))
{
url = this.BaseURL + objUrl.href;
}
try {
html = DownloadExpress.Download(url, Utility.GetEncoding(EncodingType.UTF8));
}
catch {
try
{
html = DownloadExpress.DownloadBySocket(url, Utility.GetEncoding(EncodingType.UTF8));
}
catch (Exception ex)
{
continue;
}
}
if (string.IsNullOrEmpty(html))
{
continue;
}
foreach (var item in base.ConfigLinks.configlinks)
{
FindLinkBase findLink = null;
if (item.pattern_type == "REGEX")
{
//if(item.link_type == "REMOVE_LINK")
if (_removelinks != string.Empty)
{
findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, _removelinks);
}
else
{
findLink = new FindLinkByRegex(this.BaseURL, item.url_pattern, "");
}
}
if (findLink == null)
{
continue;
}
#region SUB_LINK
if (item.link_type == "SUB_LINK")
{
List <string> listLinkSub = findLink.FindLink(html);
if (listLinkSub == null || listLinkSub.Count == 0)
{
continue;
}
foreach (string linkSub in listLinkSub)
{
string _url = linkSub;
if (IsNewUrl(ref _url)) //check exist link
{
if (!_url.Contains(".html#") && !_url.Contains(".htm#"))
{
ConcreteLink _concreteLink = new ConcreteLink();
_concreteLink.link_type = "SUB_LINK";
_concreteLink.href = linkSub;
FrontierURL.Enqueue(_concreteLink);
SpiderArgs args = new SpiderArgs()
{
Message = linkSub + "\r\n"
};
SpiderSingletonEvent.Instance.OnSpiderInformation(args);
}
}
}
}
#endregion
#region DETAIL_LINK
else if (item.link_type == "DETAIL_LINK")
{
List <string> linkDetail = findLink.FindLink(html);
if (linkDetail == null || linkDetail.Count == 0)
{
continue;
}
foreach (string href in linkDetail)
{
string _url = href;
if (IsNewUrl(ref _url)) //check exist link
{
if (!_url.Contains(".html#") && !_url.Contains(".htm#"))
{
ConcreteLink _concreteLink = new ConcreteLink();
_concreteLink.link_type = "DETAIL_LINK";
_concreteLink.href = href;
QueueDetailURL.Enqueue(_concreteLink);
SpiderArgs args = new SpiderArgs()
{
Message = _url + " - DETAIL" + "\r\n"
};
SpiderSingletonEvent.Instance.OnSpiderInformation(args);
}
}
}
}
#endregion
}
}
Thread.Sleep(this.THREAD_SLEEP);
}
}
catch (Exception ex)
{
string test = ex.Message;
}
}
