private void add_rule(string key, rule_information _ri)
{
if (HTML.stripHtml(_ri._htmlText) != "")
{
if (!_ht_rules.ContainsKey(key))
{
_ht_rules.Add(key, _ri); //new rule
}
else //rule ok
{
rule_information _ri_temp = (rule_information)_ht_rules[key];
if (_ri_temp._Classname == _ri._Classname)
{//prediction ok
_ri_temp._count++;
if (HTML.stripHtml(_ri._htmlText).Trim() != "")
{
if (_ri_temp._htmlText == _ri._htmlText)
{
_ri_temp._repetive = true;
_ht_rules[key] = _ri_temp;
}
else
{
_ri_temp._repetive = true;
_ht_rules[key] = _ri_temp;
}
}
else
{
_ri_temp._htmlText = _ri._htmlText;
_ht_rules[key] = _ri_temp;
}
}
else
{ //prediction error
if (_ri_temp._count < 3) //maybe mistake so delete
{
_ht_rules.Remove(key);
}
}
} //rule ok else
}
}
//prepare information for a given hmtl
public element AnalyzeGivenHTML_AE(string html_content, string inner_text, ref element _element)
{
_element.BagofWords_AE = inner_text;
_element.wordCount_AE = HTML.WordsCountGivenText(_element.BagofWords_AE);
string pattern = "href=.*?>(.*?)</a";
Regex exp = new Regex(pattern, RegexOptions.IgnoreCase | RegexOptions.Multiline | RegexOptions.Singleline | RegexOptions.IgnorePatternWhitespace);
MatchCollection matchList = exp.Matches(html_content);
string[] _list = new string[matchList.Count];
string URL_INNER = "";
for (int i = 0; i < matchList.Count; i++)
{
Match match = matchList[i];
if (match.Value.Length > 0)
{
URL_INNER = URL_INNER + " " + HTML.stripHtml(match.Groups[1].Value);
}
}
_element.LinkCount_AE = matchList.Count;
_element.wordCountinLink_AE = HTML.WordsCountGivenText(URL_INNER);
if (_element.LinkCount_AE != 0)
{
_element.meanofWordinLinks_AE = (double)_element.wordCountinLink_AE / _element.LinkCount_AE;
}
else
{
_element.meanofWordinLinks_AE = 0;
}
if (_element.wordCount_AE != 0)
{
_element.meanofWordinLinksAllWords_AE = (double)_element.wordCountinLink_AE / _element.wordCount_AE;
}
else
{
_element.meanofWordinLinksAllWords = 0;
}
return(_element);
}
//for test finding operation
public static string[] Contents_of_givenLayout_Tags_TESTER(string html_content, string pattern, bool cut_sub_blocks)
{
Hashtable _tags_in_HTML = filtergivenHTMLtag_TESTER(html_content, pattern);
string[] _content = null;
string s_tag = pattern.Replace(".", " ");
string e_tag = find_EndTag(s_tag);
int elementsize = 0;
foreach (DictionaryEntry d in _tags_in_HTML)
{
string _tag = (string)d.Key;
int _cnt = (int)d.Value;
_content = GrabbingofHTMLTags(html_content, _tag, _cnt);
string temp = "";
for (int i = 0; i < _content.Length; i++)
{
string t_content = _content[i];
if (cut_sub_blocks)
{
t_content = HTML.trimDIV(t_content);
t_content = HTML.trimTD(t_content);
_content[i] = t_content;
}
temp = temp + t_content;
//başlagıç ve bitiş etiketi tekrar yazılıyor.
_content[i] = s_tag + _content[i] + e_tag;
}
elementsize = elementsize + temp.Length;
}
//başlangıç etiketini tekrar yapıştır.
return(_content);
}
public ArrayList prepareDOM(string htmlContent)
{
htmlContent = HTML.trimOptions(htmlContent);
htmlContent = HTML.trimScript(htmlContent);
IHTMLDocument2 htmlDocument = new mshtml.HTMLDocumentClass();
htmlDocument.write(htmlContent);
IHTMLElementCollection allElements = htmlDocument.all;
_ht = new Hashtable();
_list = new ArrayList();
_xmllist = new ArrayList();
string _tempinner_text = "";
if (htmlDocument.body != null)
{
if (htmlDocument.body.innerText != null)
{
_tempinner_text = htmlDocument.body.innerText.Replace("\r\n", "");
}
}
element _firstelement = AnalyzeGivenHTML(htmlContent, _tempinner_text);
int i = 0;
foreach (IHTMLElement htmlelement in allElements)
{
if (htmlelement.outerHTML != null)
{
element _element = new element();
_element.id = i;
_element.outerHTML = htmlelement.outerHTML;
_element.outerHTML = _element.outerHTML.Replace("\r\n", "");
if (htmlelement.innerHTML != null)
{
_element.innerHTML = htmlelement.innerHTML;
_element.innerHTML = _element.innerHTML.Replace("\r\n", "");
}
else
{
_element.innerHTML = "";
}
if (_element.id == 0)
{
_element.elementlinked_id = -1;//root
savehtmlContent = _element.outerHTML;
resulthmtlContent = _element.outerHTML;
}
else
{
_element.elementlinked_id = 0;
}
if (htmlelement.tagName == "HTML")
{//html bazen geç geliyor...
savehtmlContent = _element.outerHTML;
resulthmtlContent = _element.outerHTML;
}
string _str = _element.outerHTML;
int _start = _str.IndexOf('<');
int _end = _str.IndexOf('>');
_element.elementName = _str.Substring(_start, _end - _start + 1);
//<!--className::(.*?)--> ???garanti başlangıçtaki olmalı diğerlerine kaymamalı????
_start = _str.IndexOf("<!--className::");
_end = _str.IndexOf("-->");
if (_start >= 0)
{
if (_start == _element.elementName.Length)
{
_start = 15 + _element.elementName.Length;
if (_end - _start > 0)
{
_element.className = _str.Substring(_start, _end - _start);
}
}
}
_element.tagName = htmlelement.tagName;
if (htmlelement.id != null)
{
_element.tag_id = 1;
}
if (htmlelement.className != null)
{
_element.tag_class = 1;
}
_element.tag_idORclass = _element.tag_id + _element.tag_idORclass;
if (_element.tag_idORclass == 2)
{
_element.tag_idORclass = 1;
}
string tempinner_text = htmlelement.innerText;
if (tempinner_text != null)
{
tempinner_text = tempinner_text.Replace("\r\n", " ");
tempinner_text = tempinner_text.Trim();
}
else
{
tempinner_text = "";
}
element _tempelement = AnalyzeGivenHTML(htmlelement.outerHTML, tempinner_text);
_element.BagofWords = _tempelement.BagofWords;
_element.wordCount = _tempelement.wordCount;
_element.DensityinHTML = (double)_element.wordCount / _firstelement.wordCount;
_element.LinkCount = _tempelement.LinkCount;
_element.wordCountinLink = _tempelement.wordCountinLink;
_element.meanofWordinLinks = _tempelement.meanofWordinLinks;
_element.meanofWordinLinksAllWords = _tempelement.meanofWordinLinksAllWords;
_element.similarity_with_other_web_page = 1;
_element.relevant = false;
_element.parent_elementName = "";
_list.Add(_element);
int key = (int)htmlelement.sourceIndex;//for fast searching
_ht.Add(key, i);
i++;
}
}
foreach (IHTMLElement htmlelement in allElements)
{
if (htmlelement.outerHTML != null)
{
string[] _sonuclar = ExtractionofSubLayouts(htmlelement);
string tempinner_text = _sonuclar[1];
string tempOuterHTML = _sonuclar[2];
string tempinnerHTML = _sonuclar[3];
string str_i = _sonuclar[0];
i = Convert.ToInt32(str_i);
//After Extraction
element _element = (element)_list[i];
element _tempelement = AnalyzeGivenHTML_AE(tempOuterHTML, tempinner_text);
_element.outerHTML_AE = tempOuterHTML;
_element.innerHTML_AE = tempinnerHTML;
_element.BagofWords_AE = _tempelement.BagofWords_AE;
_element.wordCount_AE = _tempelement.wordCount_AE;
_element.DensityinHTML_AE = (double)_element.wordCount_AE / _firstelement.wordCount;
_element.LinkCount_AE = _tempelement.LinkCount_AE;
_element.wordCountinLink_AE = _tempelement.wordCountinLink_AE;
_element.meanofWordinLinks_AE = _tempelement.meanofWordinLinks_AE;
_element.meanofWordinLinksAllWords_AE = _tempelement.meanofWordinLinksAllWords_AE;
//dot endofcontent
if (htmlelement.innerText != null)
{
if (htmlelement.innerText.Trim() != "")
{
if (htmlelement.innerText[htmlelement.innerText.Length - 1] == '.')
{
_element.dot_endofstence = 1;
}
else
{
_element.dot_endofstence = 0;
}
}
}
if (_element.wordCount_AE > _element.wordCount)
{
_element.wordCount_AE = _element.wordCount; //istisnayi durum scriptler sorun olduğu için nadir bir durum...
}
_list[i] = _element;
if (htmlelement.tagName == "DIV" || htmlelement.tagName == "TD")
{
element _e = (element)_list[i];
if (_e.elementName.Contains("vAlign=bot"))
{
_e.relevant = true;
}
bool _decision = HTMLMarkerClass.desicionClass.determineIrrevelantLayout(_element);
if (_decision == false)
{
_e.relevant = true;
_list[i] = _e;
//Update child elements
for (int m = 0; m < _list.Count; m++)
{
element _et = (element)_list[m];
if (_et.elementlinked_id == _e.id)
{
if (_et.tagName != "DIV")//div'ler için karar verme
{
if (_et.tagName != "TD")
{
_et.relevant = true;
_et.parent_elementName = _e.elementName;
_list[m] = _et;
}
}
}
} //for m
} //decision = true
} //if div or td
} // if not null
} //for each
for (int m = 0; m < _list.Count; m++)
{
element _element = (element)_list[m];
if (_element.relevant == true)
{
if (_element.tagName == "DIV" || _element.tagName == "TD")
{
bool _mainlayout = HTMLMarkerClass.desicionClass.determineLayout(_element);
if (_mainlayout)
{
_element.predicted_className = "MAIN";
}
else
{
_element.predicted_className = HTMLMarkerClass.desicionClass.determineHEADLINE_INFORMATION(_element);
}
}
else
{
_element.predicted_className = HTMLMarkerClass.desicionClass.determineHEADLINE_INFORMATION(_element);
if (_element.predicted_className == "MAIN")
{
_element.predicted_className = "IRRELEVANT";
}
}
/*if (_element.predicted_className == "IRRELEVANT")
* _element.predicted_className = "INFORMATIONABOUTARTICLE";*/
if (clear_illegal_characters_for_XML(_element.BagofWords_AE.Trim()).Trim() != "")
{
_list[m] = _element;
xml_elemet _xml = new xml_elemet();
_xml.elementName = _element.elementName;
_xml.content = _element.BagofWords_AE;
_xml.predicted_className = _element.predicted_className;
_xml.parent_elementName = _element.parent_elementName;
//equal content in _xml_list
bool find = false;
for (int v = 0; v < _xmllist.Count; v++)
{
xml_elemet item = (xml_elemet)_xmllist[v];
if (item.content == _xml.content)
{
item.elementName = item.elementName + ", " + _element.elementName;
_xmllist[v] = item;
find = true;
}
}
if (!find)
{
_xmllist.Add(_xml);
}
}
}
}
return(_list);
}
public ArrayList prepareDOM(string htmlContent2)
{
string htmlContent = htmlContent2;
htmlContent = HTML.trim_commenttags(htmlContent);
htmlContent = HTML.trimOptions(htmlContent);
htmlContent = HTML.trimScript(htmlContent);
htmlContent = HTML.trim_HREF_SCR(htmlContent);
htmlContent = HTML.trim_some_cases(htmlContent);
//for fast processing otherwise image, link, javascript loading...
IHTMLDocument2 htmlDocument = new mshtml.HTMLDocumentClass();
htmlDocument.write(htmlContent);
IHTMLElementCollection allElements = htmlDocument.all;
_ht = new Hashtable();
_list = new ArrayList();
_xmllist = new ArrayList();
string _tempinner_text = "";
if (htmlDocument.body != null)
{
if (htmlDocument.body.innerText != null)
{
_tempinner_text = htmlDocument.body.innerText.Replace("\r\n", "");
domhtmlContent = htmlDocument.body.outerHTML.Replace("\r\n", "");
}
}
element _firstelement = AnalyzeGivenHTML(htmlDocument.body.innerHTML, _tempinner_text);
all_words = _firstelement.BagofWords;
int i = 0;
foreach (IHTMLElement htmlelement in allElements)
{
if (htmlelement.outerHTML != null)
{
element _element = new element();
_element.id = i;
_element.outerHTML = htmlelement.outerHTML;
_element.outerHTML = _element.outerHTML.Replace("\r\n", "");
if (htmlelement.innerHTML != null)
{
_element.innerHTML = htmlelement.innerHTML;
_element.innerHTML = _element.innerHTML.Replace("\r\n", "");
}
else
{
_element.innerHTML = "";
}
if (_element.id == 0)
{
_element.elementlinked_id = -1;//root
savehtmlContent = _element.outerHTML;
resulthmtlContent = _element.outerHTML;
}
else
{
_element.elementlinked_id = 0;
}
if (htmlelement.tagName == "HTML")
{//html bazen geç geliyor...
savehtmlContent = _element.outerHTML;
resulthmtlContent = _element.outerHTML;
}
string _str = _element.outerHTML;
int _start = _str.IndexOf('<');
int _end = _str.IndexOf('>');
_element.elementName = _str.Substring(_start, _end - _start + 1);
_element.tagName = htmlelement.tagName;
_element.tag_id_Name = "";
_element.tag_class_Name = "";
if (htmlelement.id != null)
{
_element.tag_id = 1;
_element.tag_id_Name = htmlelement.id;
}
if (htmlelement.className != null)
{
_element.tag_class = 1;
_element.tag_class_Name = htmlelement.className;
}
if (_element.tag_id != 1 || _element.tag_class != 1)
{
_element.tag_idORclass = 1;
}
string tempinner_text = htmlelement.innerText;
if (tempinner_text != null)
{
tempinner_text = tempinner_text.Replace("\r\n", " ");
tempinner_text = tempinner_text.Trim();
}
else
{
tempinner_text = "";
}
element _tempelement = AnalyzeGivenHTML(htmlelement.outerHTML, tempinner_text);
_element.BagofWords = _tempelement.BagofWords;
_element.wordCount = _tempelement.wordCount;
_element.DensityinHTML = (double)_element.wordCount / _firstelement.wordCount;
_element.LinkCount = _tempelement.LinkCount;
_element.wordCountinLink = _tempelement.wordCountinLink;
_element.meanofWordinLinks = _tempelement.meanofWordinLinks;
_element.meanofWordinLinksAllWords = _tempelement.meanofWordinLinksAllWords;
string temp_innerhtml_ = _element.innerHTML.ToUpper(new CultureInfo("en-US", false)); //for english words thus html tags
_element.dot_count = webfilter.CountStringOccurrences(temp_innerhtml_, ".");
_element.h1_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<H1");
_element.h2_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<H2");
_element.h3_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<H3");
_element.h4_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<H4");
_element.h5_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<H5");
_element.h6_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<H6");
_element.img_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<IMG");
_element.p_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<P");
_element.br_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<BR");
_element.span_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<SPAN");
_element.object_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<OBJECT");
_element.ul_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<UL");
_element.li_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<LI");
_element.input_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<INPUT")
+ webfilter.CountStringOccurrences(temp_innerhtml_, "<BUTTON")
+ webfilter.CountStringOccurrences(temp_innerhtml_, "<LABEL");
_element.div_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<DIV");
_element.td_count = webfilter.CountStringOccurrences(temp_innerhtml_, "<TD");
_element.parent_elementName = "";
//sim control
//-1 : not available for sim control
//0 : similar
//0..1: similarity degree
//1 : not similar
_element.sim_bagofword = -1;
_element.sim_bagofword_AE = -1;
_element.sim_innerHTML = -1;
_element.sim_innerHTML_AE = -1;
_list.Add(_element);
int key = (int)htmlelement.sourceIndex;//for fast searching
_ht.Add(key, i);
i++;
}
}
foreach (IHTMLElement htmlelement in allElements)
{
if (htmlelement.outerHTML != null)
{
string[] _sonuclar = ExtractionofSubLayouts(htmlelement);
string tempinner_text = _sonuclar[1];
string tempOuterHTML = _sonuclar[2];
string tempinnerHTML = _sonuclar[3];
string str_i = _sonuclar[0];
i = Convert.ToInt32(str_i);
//After Extraction
element _element = (element)_list[i];
element _tempelement = AnalyzeGivenHTML_AE(tempOuterHTML, tempinner_text);
if (_element.elementlinked_id > 0)
{
element _p_element = (element)_list[_element.elementlinked_id];
_element.parent_elementName = _p_element.elementName;
}
if (_element.tagName == "DIV" || _element.tagName == "TD" || _element.tagName == "UL" ||
_element.tagName == "H1" || _element.tagName == "H2" || _element.tagName == "H3" ||
_element.tagName == "H4" || _element.tagName == "H5" || _element.tagName == "H6" ||
_element.tagName == "SPAN" || _element.tagName == "B" || _element.tagName == "STRONG" ||
_element.tagName == "P")
{
_element.outerHTML_AE = tempOuterHTML;
_element.innerHTML_AE = tempinnerHTML;
_element.BagofWords_AE = _tempelement.BagofWords_AE;
_element.wordCount_AE = _tempelement.wordCount_AE;
_element.DensityinHTML_AE = (double)_element.wordCount_AE / _firstelement.wordCount;
_element.LinkCount_AE = _tempelement.LinkCount_AE;
_element.wordCountinLink_AE = _tempelement.wordCountinLink_AE;
_element.meanofWordinLinks_AE = _tempelement.meanofWordinLinks_AE;
_element.meanofWordinLinksAllWords_AE = _tempelement.meanofWordinLinksAllWords_AE;
string temp_innerhtml_AE = _element.innerHTML_AE.ToUpper(new CultureInfo("en-US", false));//for english words thus html tags
_element.dot_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, ".");
_element.h1_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<H1");
_element.h2_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<H2");
_element.h3_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<H3");
_element.h4_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<H4");
_element.h5_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<H5");
_element.h6_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<H6");
_element.img_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<IMG");
_element.p_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<P");
_element.br_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<BR");
_element.span_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<SPAN");
_element.object_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<OBJECT");
_element.ul_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<UL");
_element.li_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<LI");
_element.input_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<INPUT")
+ webfilter.CountStringOccurrences(temp_innerhtml_AE, "<BUTTON")
+ webfilter.CountStringOccurrences(temp_innerhtml_AE, "<LABEL");
_element.div_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<DIV");
_element.td_count_AE = webfilter.CountStringOccurrences(temp_innerhtml_AE, "<TD");
if (_element.wordCount_AE > _element.wordCount)
{
_element.wordCount_AE = _element.wordCount; //istisnayi durum scriptler sorun olduğu için nadir bir durum...
}
//etiketin tekrar sayısı
//_element.repeat_tag_count = webfilter.CountStringOccurrences(htmlDocument.body.innerHTML, _element.elementName);
int benzertagsayisi = 0;
for (int k = 0; k < _list.Count; k++)
{
element _e1 = (element)_list[k];
if (_element.elementName == _e1.elementName)
{
benzertagsayisi++;
}
}
_element.repeat_tag_count = benzertagsayisi;
}
_list[i] = _element;
} // if not null
} //for each
return(_list);
}
