In particular, it tries to maintain the following features:
*
* - Links are maintained, with the 'href' copied over
*
- Information in the <head> is lost
*
*
* @param string $html the input HTML
* @param boolean $ignore_error Ignore xml parsing errors
* @return string the HTML converted, as best as possible, to text
* @throws Html2TextException if the HTML could not be loaded as a {@link DOMDocument}
*/
public static function convert($html, $ignore_error = false) {
// replace with spaces
$html = str_replace(" ", " ", $html);
$html = str_replace("\xc2\xa0", " ", $html);
$is_office_document = static::isOfficeDocument($html);
if ($is_office_document) {
// remove office namespace
$html = str_replace(array("", ""), "", $html);
}
$html = static::fixNewlines($html);
if (mb_detect_encoding($html, "UTF-8", true)) {
$html = mb_convert_encoding($html, "HTML-ENTITIES", "UTF-8");
}
$doc = static::getDocument($html, $ignore_error);
$output = static::iterateOverNode($doc, null, false, $is_office_document);
// remove leading and trailing spaces on each line
$output = preg_replace("/[ \t]*\n[ \t]*/im", "\n", $output);
$output = preg_replace("/ *\t */im", "\t", $output);
// unarmor pre blocks
$output = str_replace("\r", "\n", $output);
// remove unnecessary empty lines
$output = preg_replace("/\n\n\n*/im", "\n\n", $output);
// remove leading and trailing whitespace
$output = trim($output);
return $output;
}
/**
* Unify newlines; in particular, \r\n becomes \n, and
* then \r becomes \n. This means that all newlines (Unix, Windows, Mac)
* all become \ns.
*
* @param string $text text with any number of \r, \r\n and \n combinations
* @return string the fixed text
*/
static function fixNewlines($text) {
// replace \r\n to \n
$text = str_replace("\r\n", "\n", $text);
// remove \rs
$text = str_replace("\r", "\n", $text);
return $text;
}
/**
* Parse HTML into a DOMDocument
*
* @param string $html the input HTML
* @param boolean $ignore_error Ignore xml parsing errors
* @return DOMDocument the parsed document tree
*/
static function getDocument($html, $ignore_error = false) {
$doc = new \DOMDocument();
$html = trim($html);
if (!$html) {
// DOMDocument doesn't support empty value and throws an error
// Return empty document instead
return $doc;
}
if ($html[0] !== '<') {
// If HTML does not begin with a tag, we put a body tag around it.
// If we do not do this, PHP will insert a paragraph tag around
// the first block of text for some reason which can mess up
// the newlines. See pre.html test for an example.
$html = '' . $html . '';
}
if ($ignore_error) {
$doc->strictErrorChecking = false;
$doc->recover = true;
$doc->xmlStandalone = true;
$old_internal_errors = libxml_use_internal_errors(true);
$load_result = $doc->loadHTML($html, LIBXML_NOWARNING | LIBXML_NOERROR | LIBXML_NONET);
libxml_use_internal_errors($old_internal_errors);
}
else {
$load_result = $doc->loadHTML($html);
}
if (!$load_result) {
throw new Html2TextException("Could not load HTML - badly formed?", $html);
}
return $doc;
}
/**
* Can we guess that this HTML is generated by Microsoft Office?
*/
static function isOfficeDocument($html) {
return strpos($html, "urn:schemas-microsoft-com:office") !== false;
}
static function isWhitespace($text) {
return strlen(trim($text, "\n\r\t ")) === 0;
}
static function nextChildName($node) {
// get the next child
$nextNode = $node->nextSibling;
while ($nextNode != null) {
if ($nextNode instanceof \DOMText) {
if (!static::isWhitespace($nextNode->wholeText)) {
break;
}
}
if ($nextNode instanceof \DOMElement) {
break;
}
$nextNode = $nextNode->nextSibling;
}
$nextName = null;
if (($nextNode instanceof \DOMElement || $nextNode instanceof \DOMText) && $nextNode != null) {
$nextName = strtolower($nextNode->nodeName);
}
return $nextName;
}
static function iterateOverNode($node, $prevName = null, $in_pre = false, $is_office_document = false) {
if ($node instanceof \DOMText) {
// Replace whitespace characters with a space (equivilant to \s)
if ($in_pre) {
$text = "\n" . trim($node->wholeText, "\n\r\t ") . "\n";
// Remove trailing whitespace only
$text = preg_replace("/[ \t]*\n/im", "\n", $text);
// armor newlines with \r.
return str_replace("\n", "\r", $text);
} else {
$text = preg_replace("/[\\t\\n\\f\\r ]+/im", " ", $node->wholeText);
if (!static::isWhitespace($text) && ($prevName == 'p' || $prevName == 'div')) {
return "\n" . $text;
}
return $text;
}
}
if ($node instanceof \DOMDocumentType) {
// ignore
return "";
}
if ($node instanceof \DOMProcessingInstruction) {
// ignore
return "";
}
$name = strtolower($node->nodeName);
$nextName = static::nextChildName($node);
// start whitespace
switch ($name) {
case "hr":
$prefix = '';
if ($prevName != null) {
$prefix = "\n";
}
return $prefix . "---------------------------------------------------------------\n";
case "style":
case "head":
case "title":
case "meta":
case "script":
// ignore these tags
return "";
case "h1":
case "h2":
case "h3":
case "h4":
case "h5":
case "h6":
case "ol":
case "ul":
// add two newlines, second line is added below
$output = "\n";
break;
case "td":
case "th":
// add tab char to separate table fields
$output = "\t";
break;
case "p":
// Microsoft exchange emails often include HTML which, when passed through
// html2text, results in lots of double line returns everywhere.
//
// To fix this, for any p element with a className of `MsoNormal` (the standard
// classname in any Microsoft export or outlook for a paragraph that behaves
// like a line return) we skip the first line returns and set the name to br.
if ($is_office_document && $node->getAttribute('class') == 'MsoNormal') {
$output = "";
$name = 'br';
break;
}
// add two lines
$output = "\n\n";
break;
case "pre":
case "tr":
case "div":
// add one line
$output = "\n";
break;
case "li":
$output = "- ";
break;
default:
// print out contents of unknown tags
$output = "";
break;
}
// debug
//$output .= "[$name,$nextName]";
if (isset($node->childNodes)) {
$n = $node->childNodes->item(0);
$previousSiblingName = null;
while($n != null) {
$text = static::iterateOverNode($n, $previousSiblingName, $in_pre || $name == 'pre', $is_office_document);
// Pass current node name to next child, as previousSibling does not appear to get populated
if ($n instanceof \DOMDocumentType
|| $n instanceof \DOMProcessingInstruction
|| ($n instanceof \DOMText && static::isWhitespace($text))) {
// Keep current previousSiblingName, these are invisible
}
else {
$previousSiblingName = strtolower($n->nodeName);
}
$node->removeChild($n);
$n = $node->childNodes->item(0);
// suppress last br tag inside a node list
if ($n != null || $previousSiblingName != 'br') {
$output .= $text;
}
}
}
// end whitespace
switch ($name) {
case "h1":
case "h2":
case "h3":
case "h4":
case "h5":
case "h6":
$output .= "\n";
break;
case "p":
// add two lines
$output .= "\n\n";
break;
case "pre":
case "br":
// add one line
$output .= "\n";
break;
case "div":
break;
case "a":
// links are returned in [text](link) format
$href = $node->getAttribute("href");
$output = trim($output);
// remove double [[ ]] s from linking images
if (substr($output, 0, 1) == "[" && substr($output, -1) == "]") {
$output = substr($output, 1, strlen($output) - 2);
// for linking images, the title of the overrides the title of the ![]()
if ($node->getAttribute("title")) {
$output = $node->getAttribute("title");
}
}
// if there is no link text, but a title attr
if (!$output && $node->getAttribute("title")) {
$output = $node->getAttribute("title");
}
if ($href == null) {
// it doesn't link anywhere
if ($node->getAttribute("name") != null) {
$output = "[$output]";
}
} else {
if ($href == $output || $href == "mailto:$output" || $href == "http://$output" || $href == "https://$output") {
// link to the same address: just use link
$output;
} else {
// replace it
if ($output) {
$output = "[$output]($href)";
} else {
// empty string
$output = $href;
}
}
}
// does the next node require additional whitespace?
switch ($nextName) {
case "h1": case "h2": case "h3": case "h4": case "h5": case "h6":
$output .= "\n";
break;
}
break;
case "img":
if ($node->getAttribute("title")) {
$output = "[" . $node->getAttribute("title") . "]";
} elseif ($node->getAttribute("alt")) {
$output = "[" . $node->getAttribute("alt") . "]";
} else {
$output = "";
}
break;
case "li":
$output .= "\n";
break;
default:
// do nothing
}
return $output;
}
}