diff options
| author | Christian Weiske <cweiske@cweiske.de> | 2016-02-10 14:56:20 +0100 |
|---|---|---|
| committer | Christian Weiske <cweiske@cweiske.de> | 2016-02-10 14:56:20 +0100 |
| commit | cd02bac646f42a0cb402ff2dc8240aa01f1f0fb8 (patch) | |
| tree | 8cc7ee5d841f868e38ccc0b54d8cc6d33a852ed7 /src/phinde/LinkExtractor/Html.php | |
| parent | f67e8f0bc3f51f2d280a86a8c7cffa68d812efe1 (diff) | |
| download | phinde-cd02bac646f42a0cb402ff2dc8240aa01f1f0fb8.tar.gz phinde-cd02bac646f42a0cb402ff2dc8240aa01f1f0fb8.zip | |
rework crawler; add atom link extraction
Diffstat (limited to 'src/phinde/LinkExtractor/Html.php')
| -rw-r--r-- | src/phinde/LinkExtractor/Html.php | 67 |
1 files changed, 67 insertions, 0 deletions
diff --git a/src/phinde/LinkExtractor/Html.php b/src/phinde/LinkExtractor/Html.php new file mode 100644 index 0000000..538d6c4 --- /dev/null +++ b/src/phinde/LinkExtractor/Html.php @@ -0,0 +1,67 @@ +<?php +namespace phinde\LinkExtractor; + +use phinde\LinkInfo; + +class Html +{ + public function extract(\HTTP_Request2_Response $res) + { + $url = $res->getEffectiveUrl(); + + $linkInfos = array(); + + //FIXME: mime type switch for cdata + $doc = new \DOMDocument(); + //@ to hide parse warning messages in invalid html + @$doc->loadHTML($res->getBody()); + + //FIXME: extract base url from html + $base = new \Net_URL2($url); + + $xpath = new \DOMXPath($doc); + $links = $xpath->evaluate('//a'); + //FIXME: link rel, img, video + + $alreadySeen = array(); + + foreach ($links as $link) { + $linkTitle = $link->textContent; + $href = ''; + foreach ($link->attributes as $attribute) { + if ($attribute->name == 'href') { + $href = $attribute->textContent; + } + } + if ($href == '' || $href{0} == '#') { + //link on this page + continue; + } + + $linkUrlObj = $base->resolve($href); + $linkUrlObj->setFragment(false); + $linkUrl = (string) $linkUrlObj; + if (isset($alreadySeen[$linkUrl])) { + continue; + } + + switch ($linkUrlObj->getScheme()) { + case 'http': + case 'https': + break; + default: + continue 2; + } + + //FIXME: check target type + //FIXME: check nofollow + $linkInfos[] = new LinkInfo( + $linkUrl, $linkTitle, $url + ); + $alreadySeen[$linkUrl] = true; + } + + return $linkInfos; + } +} +?> |
