public function testDomDocument()
{
$html = $this->loadDat('dom_document.html');
// $html = mb_convert_encoding($html, 'HTML-ENTITIES', 'UTF-8');
$extractor = new ContentExtractor();
$extractor->exec($html);
$xpath = $extractor->calculateXpath();
$text = $extractor->scan($extractor->getExtractedNode());
d($extractor->getExtractedNode()->nodeName);
d($text);
d($xpath);
d($extractor->params);
d($extractor->text);
d($extractor->title);
d('pancutuationCountAll:' . $extractor->pancutuationCountAll);
d('domCountAll:' . $extractor->domCountAll);
d('textLengthAll:' . $extractor->textLengthAll);
d('textAll:' . $extractor->textAll);
d(mb_strlen('あ', 'utf-8'));
d(mb_strlen('ほげ
'));
}
public function testXpath()
{
$datPath = implode('/', [PATH_TEST, 'dat', 'ContentExtractor']);
foreach ($this->getData() as $name => $testData) {
// ファイル名が正解のxpathのキー
$path = implode('/', [$datPath, $name]);
if (is_dir($path)) {
continue;
}
$html = file_get_contents($path);
$extractor = new ContentExtractor();
$extractor->exec($html);
$xpath = $extractor->calculateXpath();
$text = $extractor->scan($extractor->getExtractedNode());
$hit = false;
foreach ($testData->xpathCandidates as $xpathExpected) {
if ($xpathExpected === $xpath) {
$hit = true;
}
}
if (!$hit) {
d($text);
file_put_contents('./test2.txt', $text);
d($extractor->getExtractedNode()->nodeName);
d($xpath);
d($testData->url);
d($extractor->params);
d('pancutuationCountAll:' . $extractor->pancutuationCountAll);
d('domCountAll:' . $extractor->domCountAll);
d('textLengthAll:' . $extractor->textLengthAll);
d('textAll:' . $extractor->textAll);
// d('preProcessedInput:' . $extractor->preProcessedInput);
}
$this->assertEquals(true, $hit, $xpath . ' ' . $testData->url);
// . PHP_EOL . var_export($extractor->params, true) . PHP_EOL);
}
}
