function getNextSibling($HTML_Anchor, $index)
{
if ($index == 1) {
//echo "<br/>__________________<br/>Before:".$HTML_Anchor." ";
//echo "<br/><br/>END :: getNextSibling(index:$index)<br/>".$HTML_Anchor->next_sibling()." ";
return $HTML_Anchor->next_sibling();
} else {
if ($index > 1) {
//echo "<br/>__________________<br/>getNextSibling(HTML_Anchor,index:$index)";
return getNextSibling($HTML_Anchor->next_sibling(), --$index);
} else {
//echo "<br/>__________________<br/>Error";
return null;
}
}
}
function parseAFDContent()
{
$GLOBALS['log'] .= "<br/> <span class='startCall'> ****************** Call ParseAFD->parseAFDContent() <a target='_blank' <a href='GetAFDListbyDebateDateListID.php?DebateDateListID=" . $this->afd->debateDateListID . "#" . $this->afd->AFDTitleID . "'>" . $this->afd->AFDTitle . "</a> </span>";
try {
$dom = new simple_html_dom();
$dom->load($this->givenHTML, false);
$i = 0;
//main section
foreach ($dom->find('div.afd') as $div_afd) {
$i++;
//get the result of discussion
$endResult_Html = $div_afd->find("p", 0);
$flag_error = "";
//echo $div_afd;
$result = ereg("^The", $endResult_Html->plaintext);
$result_2 = ereg("UTC)", $endResult_Html->plaintext);
$container_nourlexpansion = "";
$container_sibiling_index = 1;
$container_nourlexpansion_temp = getNextSibling($div_afd->find("p", 0), $container_sibiling_index);
//1
$checking = ereg("talk", $container_nourlexpansion_temp->plaintext);
$checking_2 = ereg("watch", $container_nourlexpansion_temp->plaintext);
$checking_3 = ereg("views", $container_nourlexpansion_temp->plaintext);
while (!($checking != 0 && $checking_2 != 0 && $checking_3 != 0)) {
++$container_sibiling_index;
$container_nourlexpansion_temp = getNextSibling($div_afd->find("p", 0), $container_sibiling_index);
//2..n
$checking = ereg("talk", $container_nourlexpansion_temp->plaintext);
$checking_2 = ereg("watch", $container_nourlexpansion_temp->plaintext);
$checking_3 = ereg("views", $container_nourlexpansion_temp->plaintext);
if ($container_sibiling_index > 7) {
$GLOBALS['log'] .= "<br/><span class='bad'>Error in MainParsing (container_sibiling_index={$container_sibiling_index})</span>";
break;
}
}
if ($checking != 0 && $checking_2 != 0 && $checking_3 != 0) {
$container_nourlexpansion = getNextSibling($div_afd->find("p", 0), $container_sibiling_index);
}
//1..n
//Add extraNote to endResult as well as fix the bug of endResult
$endResult_ExtraNote = "";
for ($endresult_i = 1; $endresult_i < $container_sibiling_index - 1; $endresult_i++) {
$endResult_Html .= getNextSibling($div_afd->find("p", 0), $endresult_i);
}
$endResult_SeprateCondition = preg_match_all("/\\((UTC)\\)/", $endResult_Html, $matches_endResultSeprate, PREG_OFFSET_CAPTURE);
$matches_endResultSeprate_0 = $matches_endResultSeprate[0];
if (count($matches_endResultSeprate_0) > 0) {
$result_2 = 1;
$endResult_Html_temp = substr($endResult_Html, 0, $matches_endResultSeprate_0[0][1] + 5);
$endResult_ExtraNote = substr($endResult_Html, $matches_endResultSeprate_0[0][1] + 5, strlen($endResult_Html) - $matches_endResultSeprate_0[0][1] + 5);
} else {
$endResult_Html_temp = $endResult_Html;
}
// such as . or '' might be in the end that would be removed.
if (strlen(trim($endResult_ExtraNote)) < 6) {
$endResult_ExtraNote = "";
}
$endResult_Html = $endResult_Html_temp;
if ($result_2 == 0) {
$this->parsed_resultError++;
$GLOBALS['log'] .= "<div class='bad'> Result failed. (parsedError = True)</div> debateDateListID=<a target='_blank' href='getHtmlByID.php?id=" . $this->afd->debateDateListID . "'>" . $this->afd->debateDateListID . "</a>, AFDID=<a target='_blank' href='getAFDHtmlByID.php?id=" . $this->afd->AFDID . "'>" . $this->afd->AFDID . "</a>";
$parse_e_result_s = 0;
$parse_e_result_e = 0;
} else {
$GLOBALS['log'] .= "<div class='good'> Result Matched. </div> debateDateListID=<a target='_blank' href='getHtmlByID.php?id=" . $this->afd->debateDateListID . "'>" . $this->afd->debateDateListID . "</a>, AFDID=<a target='_blank' href='getAFDHtmlByID.php?id=" . $this->afd->AFDID . "'>" . $this->afd->AFDID . "</a>";
$parse_e_result_s = 1;
$parse_e_result_e = 1;
}
//check end result now
$GLOBALS['log'] .= "<div class='percentage'>After Result Sibiling index: {$container_sibiling_index} </div>";
$GLOBALS['log'] .= "<div class='percentage'> endResult_Html:</div> {$endResult_Html}";
$GLOBALS['log'] .= "<div class='percentage'> endResult_ExtraNote:</div> {$endResult_ExtraNote}";
//Main Comment Secetion
$sibling_index = 0;
$mainComment_Html = "";
$otherComment_Html = "";
++$sibling_index;
$mainComment_Html_temp = getNextSibling($container_nourlexpansion, $sibling_index);
//1
$mainComment_EndCondition = preg_match_all("/\\((UTC)\\)/", $mainComment_Html_temp, $matches_mainComment, PREG_OFFSET_CAPTURE);
$mainComment_Html_temp_2 = "";
//+++++ fix the bug when have multiple (UTC) in $mainComment_Html_temp and add the rest (2..n) to the other_comments
if ($mainComment_EndCondition > 1) {
$matches_mainComment_array = $matches_mainComment[0];
for ($substring_i = 0; $substring_i < count($matches_mainComment_array); $substring_i++) {
if ($substring_i == 0) {
$start_position = 0;
$end_position = $matches_mainComment_array[$substring_i][1] - $start_position + 5;
//for 1st occurence
$mainComment_Html_temp_2 = substr($mainComment_Html_temp, $start_position, $end_position);
} else {
$start_position = $matches_mainComment_array[$substring_i - 1][1] + 5;
$end_position = $matches_mainComment_array[$substring_i][1] - $start_position + 5;
//for 2nd to n occurence
$otherComment_Html .= substr($mainComment_Html_temp, $start_position, $end_position);
}
//echo "substring_i=$substring_i($start_position,".($matches_mainComment_array[$substring_i][1]+5).")<br/>". substr($mainComment_Html_temp,$start_position, $end_position )." <br/><hr/> ";
}
$flag_error = 'check';
}
//---- fix the bug when have multiple (UTC) in $mainComment_Html_temp and add the rest (2..n) to the other_comments
if (empty($mainComment_Html_temp_2)) {
$mainComment_Html = $mainComment_Html_temp;
} else {
$mainComment_Html .= $mainComment_Html_temp_2;
}
while ($mainComment_EndCondition == 0) {
++$sibling_index;
$mainComment_Html_temp = getNextSibling($container_nourlexpansion, $sibling_index);
//2..n
//echo "<div class='newFunction'> check sibling_index=$sibling_index<br/>$mainComment_Html_temp</div>";
//$mainComment_EndCondition = ereg("UTC)", trim(strip_tags($mainComment_Html)));
//New Way for calculation
$mainComment_EndCondition = preg_match_all("/\\((UTC)\\)/", $mainComment_Html_temp, $matches_mainComment, PREG_OFFSET_CAPTURE);
$mainComment_Html_temp_2 = "";
//+++++ fix the bug when have multiple (UTC) in $mainComment_Html_temp and add the rest (2..n) to the other_comments
if ($mainComment_EndCondition > 1) {
$matches_mainComment_array = $matches_mainComment[0];
//echo $mainComment_Html."<hr/>SSSS";
for ($substring_i = 0; $substring_i < count($matches_mainComment_array); $substring_i++) {
if ($substring_i == 0) {
$start_position = 0;
$end_position = $matches_mainComment_array[$substring_i][1] - $start_position + 5;
//for 1st occurence
$mainComment_Html_temp_2 = substr($mainComment_Html_temp, $start_position, $end_position);
} else {
$start_position = $matches_mainComment_array[$substring_i - 1][1] + 5;
$end_position = $matches_mainComment_array[$substring_i][1] - $start_position + 5;
//for 2nd to n occurence
$otherComment_Html .= substr($mainComment_Html_temp, $start_position, $end_position);
}
//echo "substring_i=$substring_i($start_position,".($matches_mainComment_array[$substring_i][1]+5).")<br/>". substr($mainComment_Html_temp,$start_position, $end_position )." <br/><hr/> ";
}
$flag_error = 'check';
}
//---- fix the bug when have multiple (UTC) in $mainComment_Html_temp and add the rest (2..n) to the other_comments
if (empty($mainComment_Html_temp_2)) {
$mainComment_Html = $mainComment_Html_temp;
} else {
$mainComment_Html .= $mainComment_Html_temp_2;
}
if ($sibling_index == 10) {
throw new Exception('There is record in DB!');
}
}
if ($sibling_index > 1) {
echo "<div class='newFunction'> from 2 to {$sibling_index} for (sibling_index) </div>";
for ($mainComment_i = 2; $mainComment_i < $sibling_index; $mainComment_i++) {
$mainComment_Html .= getNextSibling($container_nourlexpansion, $mainComment_i);
}
}
//Add extraNote to endResult as well as fix the bug of endResult
$mainComment_ExtraNote = "";
/*$mainComment_ExtraSeprateCondition = preg_match_all ("/\((UTC)\)/", $mainComment_Html, $matches_mainCommentSeprate, PREG_OFFSET_CAPTURE);
$matches_mainCommentSeprate_0 = $matches_mainCommentSeprate[0];
$mainComment_Html_temp = substr($mainComment_Html,0, $matches_mainCommentSeprate_0[0][1]+5 );
$mainComment_ExtraNote = substr($mainComment_Html, $matches_mainCommentSeprate_0[0][1]+5 , strlen($mainComment_Html) - $matches_mainCommentSeprate_0[0][1]+5 );
$mainComment_Html = $mainComment_Html_temp;
*/
$otherComment_Html_pre = $otherComment_Html;
//other Comment Secetion
$otherComment_Html = "";
$sibling_index_otherComment = $sibling_index;
++$sibling_index_otherComment;
$otherComment_Html .= getNextSibling($container_nourlexpansion, $sibling_index_otherComment);
//1
$otherComment_EndCondition = ereg("No further edits should be made to this page", trim(strip_tags($otherComment_Html)));
$extra_i = 0;
// other comment is not empty
if ($otherComment_EndCondition == 0) {
//$GLOBALS['log'] .= "<br/><span class='percentage'> otherComment_EndCondition=$otherComment_EndCondition, sibling_index_otherComment=$sibling_index_otherComment</span>";
$flag_otherComment_empty = 0;
$otherComment_Html_next = getNextSibling($container_nourlexpansion, $sibling_index_otherComment + 1);
//2
$otherComment_EndCondition = ereg("No further edits should be made to this page", trim(strip_tags($otherComment_Html_next)));
while ($otherComment_EndCondition == 0) {
++$sibling_index_otherComment;
$otherComment_Html .= getNextSibling($container_nourlexpansion, $sibling_index_otherComment);
//1
$otherComment_Html_next = getNextSibling($container_nourlexpansion, $sibling_index_otherComment + 1);
//2
$otherComment_EndCondition = ereg("No further edits should be made to this page", trim(strip_tags($otherComment_Html_next)));
if ($sibling_index == 30) {
throw new Exception('There is record in DB!');
}
}
//add extra note before the very end of other comment ( before "No further edit should be made to this page")
$otherComment_Html_next_dd = find_innder_dd($otherComment_Html_next);
$otherComment_EndCondition_dd = ereg("No further edits should be made to this page", trim(strip_tags($otherComment_Html_next_dd)));
while ($otherComment_EndCondition_dd == 0) {
$extra_i++;
$otherComment_Html .= $otherComment_Html_next_dd;
$otherComment_Html_next = str_replace($otherComment_Html_next_dd, "", $otherComment_Html_next);
$otherComment_Html_next_dd = find_innder_dd($otherComment_Html_next);
$otherComment_EndCondition_dd = ereg("No further edits should be made to this page", trim(strip_tags($otherComment_Html_next_dd)));
}
} else {
//$GLOBALS['log'] .= "<br/><span class='percentage'> (empty) otherComment_EndCondition=$otherComment_EndCondition, sibling_index_otherComment=$sibling_index_otherComment</span>";
$otherComment_Html_next = $otherComment_Html;
$otherComment_Html = "";
//add extra note before the very end of other comment ( before "No further edit should be made to this page")
$otherComment_Html_next_dd = find_innder_dd($otherComment_Html_next);
$otherComment_EndCondition_dd = ereg("No further edits should be made to this page", trim(strip_tags($otherComment_Html_next_dd)));
while ($otherComment_EndCondition_dd == 0) {
$extra_i++;
$otherComment_Html .= $otherComment_Html_next_dd;
$otherComment_Html_next = str_replace($otherComment_Html_next_dd, "", $otherComment_Html_next);
$otherComment_Html_next_dd = find_innder_dd($otherComment_Html_next);
$otherComment_EndCondition_dd = ereg("No further edits should be made to this page", trim(strip_tags($otherComment_Html_next_dd)));
}
if ($extra_i == 0) {
$flag_otherComment_empty = 1;
} else {
$flag_otherComment_empty = 0;
}
}
$otherComment_Html = $otherComment_Html_pre . $otherComment_Html;
$GLOBALS['log'] .= "<div class='percentage'> mainComment_Html(sibling_index={$sibling_index}):</div>{$mainComment_Html}";
$GLOBALS['log'] .= "<div class='percentage'> mainComment_ExtraNote:</div>{$mainComment_ExtraNote}";
$GLOBALS['log'] .= str_replace("background-color: #F3F9FF;", " ", $this->givenHTML);
$GLOBALS['log'] .= "<div class='percentage'>extra_i:{$extra_i} for AFDID:" . $this->givenAFDID . "</div>, <a target='_blank' href='https://en.wikipedia.org{$this->afdURL}'>https://en.wikipedia.org{$this->afdURL}</a>";
$GLOBALS['log'] .= "<div class='percentage'>otherComment_Html:</div><br/>(flag_otherComment_empty={$flag_otherComment_empty})<br/> {$otherComment_Html} <hr/>";
$GLOBALS['log'] .= "<div class='percentage'> otherComment_Html_next:</div><br/> {$otherComment_Html_next} ";
//storeInformation
$this->afd->flag_error = $flag_error;
$this->afd->flag_otherComment_empty = $flag_otherComment_empty;
$this->afd->endResult_Html = $endResult_Html;
$this->afd->endResult_ExtraNote = $endResult_ExtraNote;
$this->afd->parse_e_result_s = $result;
$this->afd->parse_e_result_e = $result_2;
$this->afd->mainComment_Html = $mainComment_Html;
$this->afd->mainComment_ExtraNote = $mainComment_ExtraNote;
$this->afd->otherComment_Html = $otherComment_Html;
$this->afd->plainlinks_Html = $container_nourlexpansion;
$this->afd->parse_e_result_s = $parse_e_result_s;
$this->afd->parse_e_result_e = $parse_e_result_e;
$this->afd->updateAFD_withoutAFDHTML_byAFDID();
}
} catch (Exception $e) {
echo '<br/>Caught exception: ', $e->getMessage(), "\n";
}
$GLOBALS['log'] .= "<br/><span class='endCall'>**** End Called ParseAFD->parseAFDContent()*******************</span>";
echo $GLOBALS['log'];
$GLOBALS['log'] = "";
flush();
}
function parseContent($totalCalculated)
{
$GLOBALS['log'] .= "<br/> <span class='startCall'> ****************** Call ParseDebateDate->parseContent() <a target='_blank' <a href='getHtmlByID.php?id=" . $this->debateDate->crawlerID . "'>" . $this->debateDate->url . "</a> </span>";
try {
$dom = new simple_html_dom();
$dom->load($this->givenHTML, false);
$i = 0;
//main section
foreach ($dom->find('div.afd') as $div_afd) {
$i++;
//get the result of discussion
$endResult_Html = $div_afd->find("p", 0);
$GLOBALS['log'] .= "<hr/>{$i} out of {$totalCalculated} (<span class='good'>" . round($i / $totalCalculated, 2) . "%</span>) For ";
$GLOBALS['log'] .= "<a target='_blank' <a href='getHtmlByID.php?id=" . $this->debateDate->crawlerID . "'>" . $this->debateDate->url . "</a>";
$GLOBALS['log'] .= ", crawlerID =" . $this->debateDate->crawlerID . ", and debateDateListID =" . $this->debateDate->debateDateListID . "<br/> {$endResult_Html}";
$flag_error = "";
$result = ereg("^The", $div_afd->find("p", 0)->plaintext);
$result_2 = ereg("UTC)\$", $div_afd->find("p", 0)->plaintext);
if ($result == 0) {
$GLOBALS['log'] .= "<br/><span class='bad'>{$i}. Result did Not match. (parsedError = {$this->parsedError})</span>";
$this->parsedError++;
$flag_error = "endResult_Start";
$parse_e_result_s = 0;
if ($result_2 == 0) {
$GLOBALS['log'] .= "<br/><span class='bad'>{$i}. Result_end failed. (parsedError = {$this->parsedError})</span>";
$this->parsedError++;
$flag_error = $flag_error . "; endResult_End";
$parse_e_result_e = 0;
} else {
$GLOBALS['log'] .= "<br/><span class='good'>{$i}. Result_end Matched.</span>";
$parse_e_result_e = 1;
}
} else {
$GLOBALS['log'] .= "<br/><span class='good'>{$i}. Result Matched.</span>";
$parse_e_result_s = 1;
if ($result_2 == 0) {
$GLOBALS['log'] .= "<br/><span class='bad'>{$i}. Result_end failed. (parsedError = {$this->parsedError})</span>";
$this->parsedError++;
$flag_error = $flag_error . "endResult_End";
$parse_e_result_e = 0;
} else {
$GLOBALS['log'] .= "<br/><span class='good'>{$i}. Result_end Matched.</span>";
$parse_e_result_e = 1;
}
}
$container_nourlexpansion = "";
$container_sibiling_index = 1;
$container_nourlexpansion_temp = getNextSibling($div_afd->find("p", 0), $container_sibiling_index);
//1
$checking = ereg("talk", $container_nourlexpansion_temp->plaintext);
$checking_2 = ereg("watch", $container_nourlexpansion_temp->plaintext);
$checking_3 = ereg("views", $container_nourlexpansion_temp->plaintext);
while (!($checking != 0 && $checking_2 != 0 && $checking_3 != 0)) {
++$container_sibiling_index;
$container_nourlexpansion_temp = getNextSibling($div_afd->find("p", 0), $container_sibiling_index);
//2..n
$checking = ereg("talk", $container_nourlexpansion_temp->plaintext);
$checking_2 = ereg("watch", $container_nourlexpansion_temp->plaintext);
$checking_3 = ereg("views", $container_nourlexpansion_temp->plaintext);
if ($container_sibiling_index > 7) {
$GLOBALS['log'] .= "<br/><span class='bad'>Error in MainParsing (container_sibiling_index={$container_sibiling_index})</span>";
break;
}
}
if ($checking != 0 && $checking_2 != 0 && $checking_3 != 0) {
$container_nourlexpansion = getNextSibling($div_afd->find("p", 0), $container_sibiling_index);
//1..n
$GLOBALS['log'] .= "<div class='percentage'>After Result Sibiling index: {$container_sibiling_index} </div>";
}
//if($container_sibiling_index>2)
for ($result_i = 2; $result_i < $container_sibiling_index; $result_i++) {
$endResult_Html .= getNextSibling($div_afd->find("p", 0), $result_i);
}
$AFDTitleID = $container_nourlexpansion->find('span.nourlexpansion', 0)->prev_sibling()->id;
//echo $container_nourlexpansion ->find('span.nourlexpansion', 0);
//echo "<br/>";
$articleURL = $container_nourlexpansion->find('span.nourlexpansion', 0)->find('a', 0)->href;
$AFDTitle = $container_nourlexpansion->find('span.nourlexpansion', 0)->find('a', 0)->plaintext;
$flag_deletedArticle = "";
$flag_articleURL_Working = 0;
if (!empty($container_nourlexpansion->find('span.nourlexpansion', 0)->find('a', 0)->class)) {
$flag_deletedArticle = $container_nourlexpansion->find('span.nourlexpansion', 0)->find('a', 0)->class;
$GLOBALS['log'] .= "{$flag_deletedArticle}<br/>";
$flag_articleURL_Working = 0;
} else {
$flag_articleURL_Working = 1;
$flag_deletedArticle = "";
}
//echo $container_nourlexpansion ->find('span.nourlexpansion', 0)->next_sibling();
//echo "<br/>";
$AFDURL = $container_nourlexpansion->find('span.nourlexpansion', 0)->next_sibling()->href;
//storeInformation
$afd = new AFD($AFDTitle, $this->debateDate->debateDateListID, $this->debateDate->conn);
$afd->AFDTitleID = $AFDTitleID;
$afd->articleURL = $articleURL;
$afd->flag_deletedArticle = $flag_deletedArticle;
$afd->flag_articleURL_Working = $flag_articleURL_Working;
$afd->AFDURL = $AFDURL;
$afd->AFDHTML = $div_afd;
$afd->endResult_Html = $endResult_Html;
$afd->parse_e_result_s = $parse_e_result_s;
$afd->parse_e_result_e = $parse_e_result_e;
$this->debateDate->addNewAFDByTitle($afd);
$this->totalParsed++;
//echo $container_nourlexpansion;
$GLOBALS['log'] .= "\"<span class='percentage'>{$AFDTitle}</span>\" crawlerID=" . $this->debateDate->crawlerID . ", afd->debateDateListID={$afd->debateDateListID} , afd->AFDID= <a target='_blank' href='getAFDHtmlByID.php?id=" . $afd->AFDID . "'>" . $afd->AFDID . "</a> <br/>";
$GLOBALS['log'] .= "<br/>" . $div_afd->plaintext;
echo $GLOBALS['log'];
$GLOBALS['log'] = "";
flush();
}
$this->debateDate->updateTotalAFDContent($i);
} catch (Exception $e) {
echo '<br/>Caught exception: ', $e->getMessage(), "\n";
}
}
