X-Git-Url: https://git.cweiske.de/phinde.git/blobdiff_plain/915b66fe6ca517610a41acec0a71597e7cee0807..87670eb07d1b0e82e0a8b4c1f9b9d20e3cafdb42:/bin/index.php diff --git a/bin/index.php b/bin/index.php index c6de5a9..7550ad3 100755 --- a/bin/index.php +++ b/bin/index.php @@ -14,6 +14,16 @@ if ($argc < 2) { exit(1); } +function removeTags($doc, $tag) { + $elems = array(); + foreach ($doc->getElementsbyTagName($tag) as $elem) { + $elems[] = $elem; + } + foreach ($elems as $elem) { + $elem->parentNode->removeChild($elem); + } +} + $es = new Elasticsearch($GLOBALS['phinde']['elasticsearch']); $url = $argv[1]; @@ -22,6 +32,7 @@ if ($existingDoc && $existingDoc->status == 'indexed') { echo "URL already indexed: $url\n"; exit(0); } +//FIXME: size limit //FIXME: sourcetitle, sourcelink $req = new \HTTP_Request2($url); @@ -49,6 +60,7 @@ if (!in_array($mimetype, $supportedIndexTypes)) { //FIXME: update index only if changed since last index time //FIXME: extract base url from html +//FIXME: check if effective url needs updating $url = $res->getEffectiveUrl(); $base = new \Net_URL2($url); @@ -61,16 +73,14 @@ $doc = new \DOMDocument(); $dx = new \DOMXPath($doc); //remove script tags -$elems = array(); -foreach ($doc->getElementsbyTagName('script') as $elem) { - $elems[] = $elem; -} -foreach ($elems as $elem) { - $elem->parentNode->removeChild($elem); -} +removeTags($doc, 'script'); +removeTags($doc, 'style'); +removeTags($doc, 'nav'); //default content: $xpContext = $doc->getElementsByTagName('body')->item(0); +//FIXME: follow meta refresh, no body +// example: https://www.gnu.org/software/coreutils/ //use microformats content if it exists $xpElems = $dx->query( @@ -95,13 +105,13 @@ $indexDoc->domain = parse_url($url, PHP_URL_HOST); $indexDoc->author = new \stdClass(); -$arXpElems = $dx->query('/html/head/meta[@name="author"]'); +$arXpElems = $dx->query('/html/head/meta[@name="author" and @content]'); if ($arXpElems->length) { $indexDoc->author->name = trim( $arXpElems->item(0)->attributes->getNamedItem('content')->textContent ); } -$arXpElems = $dx->query('/html/head/link[@rel="author"]'); +$arXpElems = $dx->query('/html/head/link[@rel="author" and @href]'); if ($arXpElems->length) { $indexDoc->author->url = trim( $base->resolve( @@ -128,7 +138,7 @@ foreach (array('h1', 'h2', 'h3', 'h4', 'h5', 'h6') as $headlinetype) { } } -//FIXME: limit to h-entry e-content +//FIXME: split paragraphs //FIXME: insert space after br $indexDoc->text = array(); $indexDoc->text[] = trim( @@ -141,7 +151,7 @@ $indexDoc->text[] = trim( //tags $tags = array(); -foreach ($dx->query('/html/head/meta[@name="keywords"]') as $xkeywords) { +foreach ($dx->query('/html/head/meta[@name="keywords" and @content]') as $xkeywords) { $keywords = $xkeywords->attributes->getNamedItem('content')->textContent; foreach (explode(',', $keywords) as $keyword) { $tags[trim($keyword)] = true; @@ -150,7 +160,7 @@ foreach ($dx->query('/html/head/meta[@name="keywords"]') as $xkeywords) { $indexDoc->tags = array_keys($tags); //dates -$arXpdates = $dx->query('/html/head/meta[@name="DC.date.created"]'); +$arXpdates = $dx->query('/html/head/meta[@name="DC.date.created" and @content]'); if ($arXpdates->length) { $indexDoc->crdate = date( 'c', @@ -162,7 +172,7 @@ if ($arXpdates->length) { //FIXME: keep creation date from database, or use modified date if we // do not have it there -$arXpdates = $dx->query('/html/head/meta[@name="DC.date.modified"]'); +$arXpdates = $dx->query('/html/head/meta[@name="DC.date.modified" and @content]'); if ($arXpdates->length) { $indexDoc->modate = date( 'c', @@ -182,12 +192,10 @@ if ($arXpdates->length) { //language //there may be "en-US" and "de-DE" -$indexDoc->language = strtolower( - substr( - $doc->documentElement->attributes->getNamedItem('lang')->textContent, - 0, 2 - ) -); +$xlang = $doc->documentElement->attributes->getNamedItem('lang'); +if ($xlang) { + $indexDoc->language = strtolower(substr($xlang->textContent, 0, 2)); +} //FIXME: fallback, autodetection //FIXME: check noindex