Using PHP substr() and strip_tags() while retaining formatting and without breaking HTML

Not amazing, but works.

function html_cut($text, $max_length){    $tags   = array();    $result = "";    $is_open   = false;    $grab_open = false;    $is_close  = false;    $in_double_quotes = false;    $in_single_quotes = false;    $tag = "";    $i = 0;    $stripped = 0;    $stripped_text = strip_tags($text);    while ($i < strlen($text) && $stripped < strlen($stripped_text) && $stripped < $max_length)    {        $symbol  = $text{$i};        $result .= $symbol;        switch ($symbol)        {           case '<':                $is_open   = true;                $grab_open = true;                break;           case '"':               if ($in_double_quotes)                   $in_double_quotes = false;               else                   $in_double_quotes = true;            break;            case "'":              if ($in_single_quotes)                  $in_single_quotes = false;              else                  $in_single_quotes = true;            break;            case '/':                if ($is_open && !$in_double_quotes && !$in_single_quotes)                {                    $is_close  = true;                    $is_open   = false;                    $grab_open = false;                }                break;            case ' ':                if ($is_open)                    $grab_open = false;                else                    $stripped++;                break;            case '>':                if ($is_open)                {                    $is_open   = false;                    $grab_open = false;                    array_push($tags, $tag);                    $tag = "";                }                else if ($is_close)                {                    $is_close = false;                    array_pop($tags);                    $tag = "";                }                break;            default:                if ($grab_open || $is_close)                    $tag .= $symbol;                if (!$is_open && !$is_close)                    $stripped++;        }        $i++;    }    while ($tags)        $result .= "</".array_pop($tags).">";    return $result;}

Usage example:

$content = html_cut($content, 100);

I'm not claiming to have invented this, but there is a very complete Text::truncate() method in CakePHP which does what you want:

function truncate($text, $length = 100, $ending = '...', $exact = true, $considerHtml = false) {    if (is_array($ending)) {        extract($ending);    }    if ($considerHtml) {        if (mb_strlen(preg_replace('/<.*?>/', '', $text)) <= $length) {            return $text;        }        $totalLength = mb_strlen($ending);        $openTags = array();        $truncate = '';        preg_match_all('/(<\/?([\w+]+)[^>]*>)?([^<>]*)/', $text, $tags, PREG_SET_ORDER);        foreach ($tags as $tag) {            if (!preg_match('/img|br|input|hr|area|base|basefont|col|frame|isindex|link|meta|param/s', $tag[2])) {                if (preg_match('/<[\w]+[^>]*>/s', $tag[0])) {                    array_unshift($openTags, $tag[2]);                } else if (preg_match('/<\/([\w]+)[^>]*>/s', $tag[0], $closeTag)) {                    $pos = array_search($closeTag[1], $openTags);                    if ($pos !== false) {                        array_splice($openTags, $pos, 1);                    }                }            }            $truncate .= $tag[1];            $contentLength = mb_strlen(preg_replace('/&[0-9a-z]{2,8};|&#[0-9]{1,7};|&#x[0-9a-f]{1,6};/i', ' ', $tag[3]));            if ($contentLength + $totalLength > $length) {                $left = $length - $totalLength;                $entitiesLength = 0;                if (preg_match_all('/&[0-9a-z]{2,8};|&#[0-9]{1,7};|&#x[0-9a-f]{1,6};/i', $tag[3], $entities, PREG_OFFSET_CAPTURE)) {                    foreach ($entities[0] as $entity) {                        if ($entity[1] + 1 - $entitiesLength <= $left) {                            $left--;                            $entitiesLength += mb_strlen($entity[0]);                        } else {                            break;                        }                    }                }                $truncate .= mb_substr($tag[3], 0 , $left + $entitiesLength);                break;            } else {                $truncate .= $tag[3];                $totalLength += $contentLength;            }            if ($totalLength >= $length) {                break;            }        }    } else {        if (mb_strlen($text) <= $length) {            return $text;        } else {            $truncate = mb_substr($text, 0, $length - strlen($ending));        }    }    if (!$exact) {        $spacepos = mb_strrpos($truncate, ' ');        if (isset($spacepos)) {            if ($considerHtml) {                $bits = mb_substr($truncate, $spacepos);                preg_match_all('/<\/([a-z]+)>/', $bits, $droppedTags, PREG_SET_ORDER);                if (!empty($droppedTags)) {                    foreach ($droppedTags as $closingTag) {                        if (!in_array($closingTag[1], $openTags)) {                            array_unshift($openTags, $closingTag[1]);                        }                    }                }            }            $truncate = mb_substr($truncate, 0, $spacepos);        }    }    $truncate .= $ending;    if ($considerHtml) {        foreach ($openTags as $tag) {            $truncate .= '</'.$tag.'>';        }    }    return $truncate;}

Use PHP's DOMDocument class to normalize an HTML fragment:

$dom= new DOMDocument();$dom->loadHTML('<div><p>Hello World');      $xpath = new DOMXPath($dom);$body = $xpath->query('/html/body');echo($dom->saveXml($body->item(0)));

This question is similar to an earlier question and I've copied and pasted one solution here. If the HTML is submitted by users you'll also need to filter out potential Javascript attack vectors like onmouseover="do_something_evil()" or <a href="javascript:more_evil();">...</a>. Tools like HTML Purifier were designed to catch and solve these problems and are far more comprehensive than any code that I could post.