How to saveHTML of DOMDocument without HTML wrapper?

Question

I'm the function below, I'm struggling to output the DOMDocument without it appending the XML, HTML, body and p tag wrappers before the output of the content. The suggested fix:

$postarray['post_content'] = $d->saveXML($d->getElementsByTagName('p')->item(0));

Only works when the content has no block level elements inside it. However, when it does, as in the example below with the h1 element, the resulting output from saveXML is truncated to...

<p>If you like</p>

I've been pointed to this post as a possible workaround, but I can't understand how to implement it into this solution (see commented out attempts below).

Any suggestions?

function rseo_decorate_keyword($postarray) {
    global $post;
    $keyword = "Jasmine Tea"
    $content = "If you like <h1>jasmine tea</h1> you will really like it with Jasmine Tea flavors. This is the last ocurrence of the phrase jasmine tea within the content. If there are other instances of the keyword jasmine tea within the text what happens to jasmine tea."
    $d = new DOMDocument();
    @$d->loadHTML($content);
    $x = new DOMXpath($d);
    $count = $x->evaluate("count(//text()[contains(translate(., 'ABCDEFGHJIKLMNOPQRSTUVWXYZ', 'abcdefghjiklmnopqrstuvwxyz'), '$keyword') and (ancestor::b or ancestor::strong)])");
    if ($count > 0) return $postarray;
    $nodes = $x->query("//text()[contains(translate(., 'ABCDEFGHJIKLMNOPQRSTUVWXYZ', 'abcdefghjiklmnopqrstuvwxyz'), '$keyword') and not(ancestor::h1) and not(ancestor::h2) and not(ancestor::h3) and not(ancestor::h4) and not(ancestor::h5) and not(ancestor::h6) and not(ancestor::b) and not(ancestor::strong)]");
    if ($nodes && $nodes->length) {
        $node = $nodes->item(0);
        // Split just before the keyword
        $keynode = $node->splitText(strpos($node->textContent, $keyword));
        // Split after the keyword
        $node->nextSibling->splitText(strlen($keyword));
        // Replace keyword with <b>keyword</b>
        $replacement = $d->createElement('strong', $keynode->textContent);
        $keynode->parentNode->replaceChild($replacement, $keynode);
    }
$postarray['post_content'] = $d->saveXML($d->getElementsByTagName('p')->item(0));
//  $postarray['post_content'] = $d->saveXML($d->getElementsByTagName('body')->item(1));
//  $postarray['post_content'] = $d->saveXML($d->getElementsByTagName('body')->childNodes);
return $postarray;
}

score 266 · Answer 1 · answered Mar 18 '14 at 20:58

266

All of these answers are now wrong, because as of PHP 5.4 and Libxml 2.6 loadHTML now has a $option parameter which instructs Libxml about how it should parse the content.

Therefore, if we load the HTML with these options

$html->loadHTML($content, LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD);

when doing saveHTML() there will be no doctype, no <html>, and no <body>.

LIBXML_HTML_NOIMPLIED turns off the automatic adding of implied html/body elements LIBXML_HTML_NODEFDTD prevents a default doctype being added when one is not found.

Full documentation about Libxml parameters is here

(Note that loadHTML docs say that Libxml 2.6 is needed, but LIBXML_HTML_NODEFDTD is only available in Libxml 2.7.8 and LIBXML_HTML_NOIMPLIED is available in Libxml 2.7.7)

answered Mar 18 '14 at 20:58

Alessandro Vendruscolo

14,493
4
32
41

10

This works like a charm. Should be the accepted answer. I just added one flag and all my headaches went away ;-) – Aug 23 '14 at 08:34
8

This doesn't work with PHP 5.4 and Libxml 2.9. loadHTML doesn't accept any options :( – Acyra Sep 10 '14 at 13:50
15

Note that this is not quite perfect. See http://stackoverflow.com/questions/29493678/loadhtml-libxml-html-noimplied-on-an-html-fragment-generates-incorrect-tags – Josh Levinson Sep 10 '15 at 16:37
Should be the answer. – kaYko Oct 07 '15 at 15:35
5

Sorry, but this doesn’t seem to be a good solution at all (at least not in practice). It really shouldn’t be the accepted answer. Besides the mentioned issues, there’s also a [nasty encoding issue](https://3v4l.org/LJjQF) with `DOMDocument` that also affects the code in this answer. Afaik, `DOMDocument` always interprets input data as latin-1 *unless the input specifies a different charset*. In other words: The `` tag seems to be needed for input data that is not latin-1. Otherwise the output will be broken for e. g. UTF-8 multibyte characters. – mermshaus Dec 05 '16 at 04:58
In a system that doesn't provide clean best practice solutions for common use cases the best solution is the one working best for most. The solution should work for elements with no siblings on the first level, shouldn't it? The encoding issue is yet another system flaw in my opinion and can be workarounded using `mb_convert_encoding` as described here: https://stackoverflow.com/a/8218649/2477364 I'd say we're in need of proper single-HTML-element handling. `echo`ing HTML elements together is so 2000. – Neonit May 23 '17 at 08:25
@Neon take a look at my answer down below tell me what you think. I want to implement it on a production environment. – Vixxs Aug 14 '17 at 18:59
Take a look to my answer tell my what you think @mermshaus – Vixxs Aug 14 '17 at 19:01
1

LIBXML_HTML_NOIMPLIED also messes up the HTML code by removing the tabs, indents and the line breaks – Zoltán Süle Jan 10 '18 at 11:12
4

This will cause issues: '
First Paragraph
Second Paragraph
' will be converted to '
First Paragraph
Second Paragraph Take not of the two innter paragraph tags. – Shawn Pivonka Oct 15 '19 at 16:15

score 78 · Answer 2 · edited Oct 30 '14 at 18:00

78

Just remove the nodes directly after loading the document with loadHTML():

# remove <!DOCTYPE 
$doc->removeChild($doc->doctype);           

# remove <html><body></body></html> 
$doc->replaceChild($doc->firstChild->firstChild->firstChild, $doc->firstChild);

edited Oct 30 '14 at 18:00

Community

1
1

answered Aug 05 '11 at 09:00

Alex

12,205
7
42
52

this is the cleaner answer to me. – KnF Jan 29 '13 at 19:53
41

should be noted that this works if has only one child node. – Yann Milin Dec 10 '13 at 23:15
Worked great. Thank you! Much cleaner and faster than the other preg answer. – akahunahi Feb 11 '14 at 22:56
Thank you for this! I just added another snip at the bottom to handle empty nodes. – Henry Ruhs Oct 01 '14 at 23:01
cleaner answer for me in php5.3 – stloc Feb 02 '15 at 14:24
Upvoted this but had to come again and get the @Alessandro Vendruscolo solution – Mariano Argañaraz Jan 10 '17 at 18:44
3

The code to remove ` ` has more than one child note. – Free Radical Feb 06 '18 at 09:26
In some cases it may fail - try `$doc->documentElement` instead of `$doc->firstChild` – l00k Oct 31 '19 at 14:02

Super Cat · Answer 3 · 2017-07-02T01:15:21.393

The issue with the top answer is that LIBXML_HTML_NOIMPLIED is unstable.

It can reorder elements (particularly, moving the top element's closing tag to the bottom of the document), add random p tags, and perhaps a variety of other issues[1]. It may remove the html and body tags for you, but at the cost of unstable behavior. In production, that's a red flag. In short:

Don't use LIBXML_HTML_NOIMPLIED. Instead, use substr.

Think about it. The lengths of <html><body> and </body></html> are fixed and at both ends of the document - their sizes never change, and neither do their positions. This allows us to use substr to cut them away:

$dom = new domDocument; 
$dom->loadHTML($html, LIBXML_HTML_NODEFDTD);

echo substr($dom->saveHTML(), 12, -15); // the star of this operation

(THIS IS NOT THE FINAL SOLUTION HOWEVER! See below for the complete answer, keep reading for context)

We cut 12 away from the start of the document because <html><body> = 12 characters (<<>>+html+body = 4+4+4), and we go backwards and cut 15 off the end because \n</body></html> = 15 characters (\n+//+<<>>+body+html = 1 + 2 + 4 + 4 + 4)

Notice that I still use LIBXML_HTML_NODEFDTD omit the !DOCTYPE from being included. First, this simplifies the substr removal of the HTML/BODY tags. Second, we don't remove the doctype with substr because we don't know if the 'default doctype' will always be something of a fixed length. But, most importantly, LIBXML_HTML_NODEFDTD stops the DOM parser from applying a non-HTML5 doctype to the document - which at least prevents the parser from treating elements it doesn't recognize as loose text.

We know for a fact that the HTML/BODY tags are of fixed lengths and positions, and we know that constants like LIBXML_HTML_NODEFDTD are never removed without some type of deprecation notice, so the above method should roll well into the future, BUT...

...the only caveat is that the DOM implementation could change the way in HTML/BODY tags are placed within the document - for instance, removing the newline at the end of the document, adding spaces between the tags, or adding newlines.

This can be remedied by searching for the positions of the opening and closing tags for body, and using those offsets as for our lengths to trim off. We use strpos and strrpos to find the offsets from the front and back, respectively:

$dom = new domDocument; 
$dom->loadHTML($html, LIBXML_HTML_NODEFDTD);

$trim_off_front = strpos($dom->saveHTML(),'<body>') + 6;
// PositionOf<body> + 6 = Cutoff offset after '<body>'
// 6 = Length of '<body>'

$trim_off_end = (strrpos($dom->saveHTML(),'</body>')) - strlen($dom->saveHTML());
// ^ PositionOf</body> - LengthOfDocument = Relative-negative cutoff offset before '</body>'

echo substr($dom->saveHTML(), $trim_off_front, $trim_off_end);

In closing, a repeat of the final, future-proof answer:

$dom = new domDocument; 
$dom->loadHTML($html, LIBXML_HTML_NODEFDTD);

$trim_off_front = strpos($dom->saveHTML(),'<body>') + 6;
$trim_off_end = (strrpos($dom->saveHTML(),'</body>')) - strlen($dom->saveHTML());

echo substr($dom->saveHTML(), $trim_off_front, $trim_off_end);

No doctype, no html tag, no body tag. We can only hope the DOM parser will receive a fresh coat of paint soon and we can more directly eliminate these unwanted tags.

Great answer, one small comment, why not `$html = $dom -> saveHTML();` instead of `$dom -> saveHTML();` repeatedly? — Steven, May 20 '19 at 09:36

Jonah · Answer 4 · 2011-02-03T03:51:18.737

21

Use saveXML() instead, and pass the documentElement as an argument to it.

$innerHTML = '';
foreach ($document->getElementsByTagName('p')->item(0)->childNodes as $child) {
    $innerHTML .= $document->saveXML($child);
}
echo $innerHTML;

http://php.net/domdocument.savexml

edited Feb 03 '11 at 03:51

answered Feb 02 '11 at 21:17

Jonah

9,991
5
45
79

That's better, but I'm still getting
wrapping the contents.
– Scott B Feb 02 '11 at 21:27
4

[As of PHP 5.3.6, you can use `saveHTML` as well](http://bugs.php.net/bug.php?id=39771) ([example](http://codepad.viper-7.com/obvdaS)) – Gordon Feb 02 '11 at 21:29
3

It should be noted that saveXML() will save XHTML, not HTML. – alexantd Feb 02 '11 at 21:32
@Scott: that's really strange. It shows what you're trying to do right there in the examples section. Are you sure that you don't have that HTML in your DOM? Exactly what HTML is in your DOMDocument? It could be that we need to access a child node. – Jonah Feb 02 '11 at 21:39
@Jonah it's not strange. When you do `loadHTML` libxml uses the HTML parser module and that will insert the missing HTML skeleton. Consequently, `$dom->documentElement` will be the root HTML element. I have fixed your example code. It should now do what Scott is asking for. – Gordon Feb 02 '11 at 21:57
#Gordon - That's much better. Only have the p tags wrapper now, but I can live with that. Unless you know how to remove that as well :) – Scott B Feb 02 '11 at 22:01
@Scott that would be the innerHTML of the node. Unfortunately, there is no built in method for that. There is a Userland function for it at http://stackoverflow.com/questions/2087103/innerhtml-in-phps-domdocument – Gordon Feb 02 '11 at 22:06
@Gordon, I've updated the question with the suggestions. When the document contains block level elements, the 'p' tag suggestion fails. – Scott B Feb 02 '11 at 23:58
@Scott: I made a small change to the code, try it. saveXML's first argument must be a DOMNode. Fortunately, DOMText extends DOMNode. It just might work. – Jonah Feb 03 '11 at 01:31
@Gordon, I was really hoping that would work. It still returns only "if you like" and cuts off the rest. – Scott B Feb 03 '11 at 02:15
@Scott: tested, doesn't work. The reason is, when you load the HTML, it puts that first phrase inside of a paragraph. There rest is outside of the paragraph. `echo $document->saveHTML();` and see for yourself. Apparently the problem is (as @lonesomeday said), it's manipulating the markup on load. – Jonah Feb 03 '11 at 03:56
@Scott: you may have to go with @lonesomeday's answer. It seems that loadHTML is messing things up in an unavoidable sort of way. Either that or don't put block tags inside? It's invalid HTML anyway. – Jonah Feb 03 '11 at 04:01
@Jonah, see my last comment on @lonesomeday's answer. I'm not sure why, but that method is completely failing for me. – Scott B Feb 03 '11 at 14:49
@Scott: are you sure you did it right? Could you post the code you used? But I would really recommend that you just use valid HTML instead. – Jonah Feb 03 '11 at 16:35
@Jonah, It looks like you have updated your answer with the innerHTML bits. I have not tried that yet. Lemme check :) – Scott B Feb 03 '11 at 21:57

score 18 · Answer 5 · answered May 22 '14 at 17:44

18

use DOMDocumentFragment

$html = 'what you want';
$doc = new DomDocument();
$fragment = $doc->createDocumentFragment();
$fragment->appendXML($html);
$doc->appendChild($fragment);
echo $doc->saveHTML();

answered May 22 '14 at 17:44

jcp

181
1
2

3

The cleanest answer for pre php5.4. – Nick Johnson Sep 09 '14 at 22:27
This works for me, both older and newer than version Libxml 2.7.7. Why would this be for pre php5.4 solely? – RobbertT Apr 08 '15 at 14:21
This should have more votes. Great option for versions of libxml that don't support LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD. Thanks! – Marty Mulligan Aug 04 '15 at 13:54
Thanks, this is the best solutions – JpBaena13 Apr 26 '21 at 17:59

Vixxs · Answer 6 · 2017-12-05T22:45:13.583

It's 2017, and for this 2011 Question I don't like any of the answers. Lots of regex, big classes, loadXML etc...

Easy solution which solves the known problems:

$dom = new DOMDocument();
$dom->loadHTML( '<html><body>'.mb_convert_encoding($html, 'HTML-ENTITIES', 'UTF-8').'</body></html>' , LIBXML_HTML_NODEFDTD);
$html = substr(trim($dom->saveHTML()),12,-14);

Easy, Simple, Solid, Fast. This code will work regarding HTML tags and encoding like:

$html = '<p>äöü</p><p>ß</p>';

If anybody finds an error , please tell, I will use this myself.

Edit, Other valid options that work without errors (very similar to ones already given):

@$dom->loadHTML(mb_convert_encoding($html, 'HTML-ENTITIES', 'UTF-8'));
$saved_dom = trim($dom->saveHTML());
$start_dom = stripos($saved_dom,'<body>')+6;
$html = substr($saved_dom,$start_dom,strripos($saved_dom,'</body>') - $start_dom );

You could add body yourself to prevent any strange thing on the furure.

Thirt option:

 $mock = new DOMDocument;
 $body = $dom->getElementsByTagName('body')->item(0);
  foreach ($body->childNodes as $child){
     $mock->appendChild($mock->importNode($child, true));
  }
$html = trim($mock->saveHTML());

You should improve your answer by avoiding the more expensive `mb_convert_encoding` and instead adding `` and modifying `substr` accordingly. Btw, yours is the most elegant solution here. Upvoted. — Hlsg, Mar 31 '18 at 10:20

score 15 · Answer 7 · answered Feb 02 '11 at 21:45

15

A neat trick is to use loadXML and then saveHTML. The html and body tags are inserted at the load stage, not the save stage.

$dom = new DOMDocument;
$dom->loadXML('<p>My DOMDocument contents are here</p>');
echo $dom->saveHTML();

NB that this is a bit hacky and you should use Jonah's answer if you can get it to work.

answered Feb 02 '11 at 21:45

lonesomeday

233,373
50
316
318

5

This will fail for invalid HTML though. – Gordon Feb 02 '11 at 21:54
1

@Gordon Exactly why I put the disclaimer at the bottom! – lonesomeday Feb 02 '11 at 21:56
1

When I try this, and echo $dom->saveHTML(), it just returns an empty string. As if loadXML($content) is empty. When I do the same with $dom->loadHTML($content), then echo $dom->saveXML() I get the content as expected. – Scott B Feb 03 '11 at 02:44
Using loadXML when willing to load HTMl is thumb. Especially because LoadXML does no know how to handle HTML. – botenvouwer May 27 '13 at 16:23

score 11 · Answer 8 · edited May 23 '17 at 12:26

I'm a bit late in the club but didn't want to not share a method I've found out about. First of all I've got the right versions for loadHTML() to accept these nice options, but LIBXML_HTML_NOIMPLIED didn't work on my system. Also users report problems with the parser (for example here and here).

The solution I created actually is pretty simple.

HTML to be loaded is put in a <div> element so it has a container containing all nodes to be loaded.

Then this container element is removed from the document (but the DOMElement of it still exists).

Then all direct children from the document are removed. This includes any added <html>, <head> and <body> tags (effectively LIBXML_HTML_NOIMPLIED option) as well as the <!DOCTYPE html ... loose.dtd"> declaration (effectively LIBXML_HTML_NODEFDTD).

Then all direct children of the container are added to the document again and it can be output.

$str = '<p>Lorem ipsum dolor sit amet.</p><p>Nunc vel vehicula ante.</p>';

$doc = new DOMDocument();

$doc->loadHTML("<div>$str</div>");

$container = $doc->getElementsByTagName('div')->item(0);

$container = $container->parentNode->removeChild($container);

while ($doc->firstChild) {
    $doc->removeChild($doc->firstChild);
}

while ($container->firstChild ) {
    $doc->appendChild($container->firstChild);
}

$htmlFragment = $doc->saveHTML();

XPath works as usual, just take care that there are multiple document elements now, so not a single root node:

$xpath = new DOMXPath($doc);
foreach ($xpath->query('/p') as $element)
{   #                   ^- note the single slash "/"
    # ... each of the two <p> element

PHP 5.4.36-1+deb.sury.org~precise+2 (cli) (built: Dec 21 2014 20:28:53)

it didn't work for me with more complex HTML source. It also removed given part of the HTML. — Zoltán Süle, Jan 10 '18 at 13:50

score 5 · Answer 9 · answered Jul 27 '12 at 13:28

Okay I found a more elegant solution, but it's just tedious:

$d = new DOMDocument();
@$d->loadHTML($yourcontent);
...
// do your manipulation, processing, etc of it blah blah blah
...
// then to save, do this
$x = new DOMXPath($d);
$everything = $x->query("body/*"); // retrieves all elements inside body tag
if ($everything->length > 0) { // check if it retrieved anything in there
      $output = '';
      foreach ($everything as $thing) {
           $output .= $d->saveXML($thing);
      }
      echo $output; // voila, no more annoying html wrappers or body tag
}

Alright, hopefully this does not omit anything and helps somebody?

Doesn't handle the case when loadHTML load a string without markup — copndz, May 16 '13 at 10:44

score 4 · Answer 10 · answered Jun 12 '12 at 21:30

None of the other solutions at the time of this writing (June, 2012) were able to completely meet my needs, so I wrote one which handles the following cases:

Accepts plain-text content which has no tags, as well as HTML content.
Does not append any tags (including <doctype>, <xml>, <html>, <body>, and <p> tags)
Leaves anything wrapped in <p> alone.
Leaves empty text alone.

So here is a solution which fixes those issues:

class DOMDocumentWorkaround
{
    /**
     * Convert a string which may have HTML components into a DOMDocument instance.
     *
     * @param string $html - The HTML text to turn into a string.
     * @return \DOMDocument - A DOMDocument created from the given html.
     */
    public static function getDomDocumentFromHtml($html)
    {
        $domDocument = new DOMDocument();

        // Wrap the HTML in <div> tags because loadXML expects everything to be within some kind of tag.
        // LIBXML_NOERROR and LIBXML_NOWARNING mean this will fail silently and return an empty DOMDocument if it fails.
        $domDocument->loadXML('<div>' . $html . '</div>', LIBXML_NOERROR | LIBXML_NOWARNING);

        return $domDocument;
    }

    /**
     * Convert a DOMDocument back into an HTML string, which is reasonably close to what we started with.
     *
     * @param \DOMDocument $domDocument
     * @return string - The resulting HTML string
     */
    public static function getHtmlFromDomDocument($domDocument)
    {
        // Convert the DOMDocument back to a string.
        $xml = $domDocument->saveXML();

        // Strip out the XML declaration, if one exists
        $xmlDeclaration = "<?xml version=\"1.0\"?>\n";
        if (substr($xml, 0, strlen($xmlDeclaration)) == $xmlDeclaration) {
            $xml = substr($xml, strlen($xmlDeclaration));
        }

        // If the original HTML was empty, loadXML collapses our <div></div> into <div/>. Remove it.
        if ($xml == "<div/>\n") {
            $xml = '';
        }
        else {
            // Remove the opening <div> tag we previously added, if it exists.
            $openDivTag = "<div>";
            if (substr($xml, 0, strlen($openDivTag)) == $openDivTag) {
                $xml = substr($xml, strlen($openDivTag));
            }

            // Remove the closing </div> tag we previously added, if it exists.
            $closeDivTag = "</div>\n";
            $closeChunk = substr($xml, -strlen($closeDivTag));
            if ($closeChunk == $closeDivTag) {
                $xml = substr($xml, 0, -strlen($closeDivTag));
            }
        }

        return $xml;
    }
}

I also wrote some tests which would live in that same class:

public static function testHtmlToDomConversions($content)
{
    // test that converting the $content to a DOMDocument and back does not change the HTML
    if ($content !== self::getHtmlFromDomDocument(self::getDomDocumentFromHtml($content))) {
        echo "Failed\n";
    }
    else {
        echo "Succeeded\n";
    }
}

public static function testAll()
{
    self::testHtmlToDomConversions('<p>Here is some sample text</p>');
    self::testHtmlToDomConversions('<div>Lots of <div>nested <div>divs</div></div></div>');
    self::testHtmlToDomConversions('Normal Text');
    self::testHtmlToDomConversions(''); //empty
}

You can check that it works for yourself. DomDocumentWorkaround::testAll() returns this:

    Succeeded
    Succeeded
    Succeeded
    Succeeded

HTML =/= XML, you should use the HTML loader for HTML. – hakre Apr 07 '15 at 19:16 — hakre, Apr 07 '15 at 19:16

Free Radical · Answer 11 · 2018-02-07T09:06:18.257

I am struggling with this on RHEL7 running PHP 5.6.25 and LibXML 2.9. (Old stuff in 2018, I know, but that is Red Hat for you.)

I have found that the much upvoted solution suggested by Alessandro Vendruscolo breaks the HTML by rearranging tags. I.e.:

<p>First.</p><p>Second.</p>'

becomes:

<p>First.<p>Second.</p></p>'

This goes for both the options he suggests you use: LIBXML_HTML_NOIMPLIED and LIBXML_HTML_NODEFDTD.

The solution suggested by Alex goes half way to solve it, but it does not work if <body> has more than one child node.

The solution that works for me is the follwing:

First, to load the DOMDocument, I use:

$doc = new DOMDocument()
$doc->loadHTML($content);

To save the document after massaging the DOMDocument, I use:

// remove <!DOCTYPE 
$doc->removeChild($doc->doctype);  
$content = $doc->saveHTML();
// remove <html><body></body></html> 
$content = str_replace('<html><body>', '', $content);
$content = str_replace('</body></html>', '', $content);

I am the first to agree that this this is not a very elegant solution - but it works.

This isn't going to work - if the $content includes information then it fails. — user1432181, Jan 05 '22 at 16:06

score 3 · Answer 12 · answered Jul 24 '12 at 15:42

3

Use this function

$layout = preg_replace('~<(?:!DOCTYPE|/?(?:html|head|body))[^>]*>\s*~i', '', $layout);

answered Jul 24 '12 at 15:42

boksiora

1,168
13
16

13

There may be some readers who have stumbled across this post via [this post](http://stackoverflow.com/questions/1732348/regex-match-open-tags-except-xhtml-self-contained-tags), have decided not to use regex to parse their HTML and use a DOM parser instead, and end up potentially needing a regex answer to achieve a complete solution... ironic – scrowler Jul 24 '14 at 05:03
I do't understand why noboy just returns the content of BODY. Is that tag not suppossed to always be present when the parser adds the whole document header/doctype? The regex above would even be shorter. – sergio Jul 06 '15 at 02:37
@boksiora "it does the job" - then why are we using DOM parser methods in the first place? – Mulan Sep 30 '15 at 11:50
@naomik i haven't said not to use a DOM parser, there are of course many different way to achieve the same result, its up to you, at the time i used this function i had a problem with the built-in php dom parser, which was not parsing html5 correctly. – boksiora Oct 01 '15 at 13:31
1

I had to use `preg_replace` because using DOMDocument-based methods of removing the html and body tags were not preserving the UTF-8 encoding :( – wizonesolutions Aug 21 '17 at 21:37

score 3 · Answer 13 · answered Oct 21 '14 at 16:51

Much like other members, I first revelled in the simplicity and awesome power of @Alessandro Vendruscolo answer. The ability to simply pass in some flagged constants to the constructor seemed too good to be true. For me it was. I have the correct versions of both LibXML as well as PHP however no matter what it still would add the HTML tag to the node structure of the Document object.

My solution worked way better than using the...

$html->loadHTML($content, LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD);

Flags or....

# remove <!DOCTYPE 
$doc->removeChild($doc->firstChild);            

# remove <html><body></body></html>
$doc->replaceChild($doc->firstChild->firstChild->firstChild, $doc->firstChild);

Node Removal, which gets messy without a structured order in the DOM. Again code fragments have no way to predetermine DOM structure.

I started this journey wanting a simple way to do DOM traversal how JQuery does it or at least in some fashion that had a structured data set either singly linked, doubly linked or tree'd node traversal. I didn't care how as long as I could parse a string the way HTML does and also have the amazing power of the node entity class properties to use along the way.

So far DOMDocument Object has left me wanting... As with many other programmers it seems... I know I have seen a lot of frustration in this question so since I FINALLY.... (after roughly 30 hours of try and fail type testing) I have found a way to get it all. I hope this helps someone...

First off, I am cynical of EVERYTHING... lol...

I would have went a lifetime before agreeing with anyone that a third party class is in anyway needed in this use case. I very much was and am NOT a fan of using any third party class structure however I stumbled onto a great parser. (about 30 times in Google before I gave in so don't feel alone if you avoided it because it looked lame of unofficial in any way...)

If you are using code fragments and need the, code clean and unaffected by the parser in any way, without extra tags being used then use simplePHPParser.

It's amazing and acts a lot like JQuery. I not often impressed but this class makes use of a lot of good tools and I have had no parsing errors as of yet. I am a huge fan of being able to do what this class does.

You can find its files to download here, its startup instructions here, and its API here. I highly recommend using this class with its simple methods that can do a .find(".className") the same way a JQuery find method would be used or even familiar methods such as getElementByTagName() or getElementById()...

When you save out a node tree in this class it doesn't add anything at all. You can simply say $doc->save(); and it outputs the entire tree to a string without any fuss.

I will now be using this parser for all, non-capped-bandwidth, projects in the future.

score 3 · Answer 14 · answered May 09 '16 at 21:14

If the flags solution answered by Alessandro Vendruscolo doesn't works, you may try this:

$dom = new DOMDocument();
$dom->loadHTML($content);

//do your stuff..

$finalHtml = '';
$bodyTag = $dom->documentElement->getElementsByTagName('body')->item(0);
foreach ($bodyTag->childNodes as $rootLevelTag) {
    $finalHtml .= $dom->saveHTML($rootLevelTag);
}
echo $finalHtml;

$bodyTag will contain your full processed HTML code without all those HTML wraps, except for the <body> tag, which is the root of your content. Then you can use a regex or a trim function to remove it from the final string (after saveHTML) or, like in the case above, iterate over all of its childen, saving their content into a temporary variable $finalHtml and return it (what i believe being safer).

Panagiotis Koursaris · Answer 15 · 2016-12-08T06:54:12.973

I came across this topic to find a way to remove HTML wrapper. Using LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD works great, but I have a problem with utf-8. After much effort I found a solution. I post it bellow for anyone has the same problem.

The problem caused because of <meta http-equiv="Content-Type" content="text/html; charset=utf-8">

The problem:

$dom = new DOMDocument();
$dom->loadHTML('<meta http-equiv="Content-Type" content="text/html; charset=utf-8">' . $document, LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD);
$dom->saveHTML();

Solution 1:

$dom->loadHTML(mb_convert_encoding($document, 'HTML-ENTITIES', 'UTF-8'), LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD);
    $dom->saveHTML($dom->documentElement));

Solution 2:

$dom->loadHTML($document, LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD);
utf8_decode($dom->saveHTML($dom->documentElement));

I find it nice that you share your findings, but Solution 2 is already present with this exact questions here and Solution 1 is elsewhere. Also for the Problem of Solution 1 the answer given is unclear. I honour your good intentions, nbut please be aware that it can create a lot of noise as well hindering others to find the solutions they are looking for which I guess is kinda the opposite of what you want to achieve with your answer. Stackoverflow works best if you handle one question at a time. Just a hint. — hakre, Dec 04 '16 at 09:26

score 2 · Answer 16 · edited May 23 '17 at 11:55

Adding the <meta> tag will trigger the fixing behavior of DOMDocument. The good part is that you don't need to add that tag at all. If you wan't to use an encoding of your choosing just pass it as a constructor argument.

http://php.net/manual/en/domdocument.construct.php

$doc = new DOMDocument('1.0', 'UTF-8');
$node = $doc->createElement('div', 'Hello World');
$doc->appendChild($node);
echo $doc->saveHTML();

Output

<div>Hello World</div>

Thanks to @Bart

score 2 · Answer 17 · answered Aug 27 '13 at 12:13

I had this requirement, too, and liked the solution posted by Alex above. There are a couple of issues, though - if the <body> element contains more than one child element, the resulting document will only contain only the first child element of <body>, not all of them. Also, I needed the stripping to handle things conditionally - only when you had document with the HTML headings. So I refined it as follows. Instead of removing <body>, I transformed it to a <div>, and stripped out the XML declaration and <html>.

function strip_html_headings($html_doc)
{
    if (is_null($html_doc))
    {
        // might be better to issue an exception, but we silently return
        return;
    }

    // remove <!DOCTYPE 
    if (!is_null($html_doc->firstChild) &&
        $html_doc->firstChild->nodeType == XML_DOCUMENT_TYPE_NODE)
    {
        $html_doc->removeChild($html_doc->firstChild);     
    }

    if (!is_null($html_doc->firstChild) &&
        strtolower($html_doc->firstChild->tagName) == 'html' &&
        !is_null($html_doc->firstChild->firstChild) &&
        strtolower($html_doc->firstChild->firstChild->tagName) == 'body')
    {
        // we have 'html/body' - replace both nodes with a single "div"        
        $div_node = $html_doc->createElement('div');

        // copy all the child nodes of 'body' to 'div'
        foreach ($html_doc->firstChild->firstChild->childNodes as $child)
        {
            // deep copies each child node, with attributes
            $child = $html_doc->importNode($child, true);
            // adds node to 'div''
            $div_node->appendChild($child);
        }

        // replace 'html/body' with 'div'
        $html_doc->removeChild($html_doc->firstChild);
        $html_doc->appendChild($div_node);
    }
}

score 2 · Answer 18 · answered May 31 '16 at 10:47

I have PHP 5.3 and the answers here did not work for me.

$doc->replaceChild($doc->firstChild->firstChild->firstChild, $doc->firstChild); replaced all the document with only the first child, I had many paragraphs and only the first was being saved, but the solution gave me a good starting point to write something without regex I left some comments and I am pretty sure this can be improved but if someone has the same problem as me it can be a good starting point.

function extractDOMContent($doc){
    # remove <!DOCTYPE
    $doc->removeChild($doc->doctype);

    // lets get all children inside the body tag
    foreach ($doc->firstChild->firstChild->childNodes as $k => $v) {
        if($k !== 0){ // don't store the first element since that one will be used to replace the html tag
            $doc->appendChild( clone($v) ); // appending element to the root so we can remove the first element and still have all the others
        }
    }
    // replace the body tag with the first children
    $doc->replaceChild($doc->firstChild->firstChild->firstChild, $doc->firstChild);
    return $doc;
}

Then we could use it like this:

$doc = new DOMDocument();
$doc->encoding = 'UTF-8';
$doc->loadHTML('<p>Some html here</p><p>And more html</p><p>and some html</p>');
$doc = extractDOMContent($doc);

Note that appendChild accepts a DOMNode so we do not need to create new elements, we can just reuse existing ones that implement DOMNodesuch as DOMElement this can be important to keep code "sane" when manipulating multiple HTML/XML documents

This won't work for fragments, only for a single child-element that you want to make the first-child of the document. This is pretty limited and effectively not doing the job of the `LIBXML_HTML_NOIMPLIED` as it does that only partially. Removing the doctype is effectively `LIBXML_HTML_NODEFDTD`. — hakre, Dec 04 '16 at 09:29

smarteist · Answer 19 · 2020-01-01T13:42:45.240

I face 3 problems with DOMDocument class.

1- This class loads html with ISO encoding and utf-8 characters not showing in output.

2- Even if we give ‍‍‍LIBXML_HTML_NOIMPLIED flag to loadHtml method, until our input html does not contain a root tag, it will not be parse correctly.

3- This class considers the HTML5 tags invalid.

So I've override this class to solve these problems and I changed some of the methods.

class DOMEditor extends DOMDocument
{
    /**
     * Temporary wrapper tag , It should be an unusual tag to avoid problems
     */
    protected $tempRoot = 'temproot';

    public function __construct($version = '1.0', $encoding = 'UTF-8')
    {
        //turn off html5 errors
        libxml_use_internal_errors(true);
        parent::__construct($version, $encoding);
    }

    public function loadHTML($source, $options = LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD)
    {
        // this is a bitwise check if LIBXML_HTML_NOIMPLIED is set
        if ($options & LIBXML_HTML_NOIMPLIED) {
            // it loads the content with a temporary wrapper tag and utf-8 encoding
            parent::loadHTML("<{$this->tempRoot}>" . mb_convert_encoding($source, 'HTML', 'UTF-8') . "</{$this->tempRoot}>", $options);
        } else {
            // it loads the content with utf-8 encoding and default options
            parent::loadHTML(mb_convert_encoding($source, 'HTML', 'UTF-8'), $options);
        }
    }

    private function unwrapTempRoot($output)
    {
        if ($this->firstChild->nodeName === $this->tempRoot) {
            return substr($output, strlen($this->tempRoot) + 2, -strlen($this->tempRoot) - 4);
        }
        return $output;
    }

    public function saveHTML(DOMNode $node = null)
    {
        $html = html_entity_decode(parent::saveHTML($node));
        if (is_null($node)) {
            $html = $this->unwrapTempRoot($html);
        }
        return $html;
    }

    public function saveXML(DOMNode $node = null, $options = null)
    {
        if (is_null($node)) {
            return '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>' . PHP_EOL . $this->saveHTML();
        }
        return parent::saveXML($node);
    }

}

Now im using DOMEditor instead of DOMDocument and it has worked well for me so far

        $editor = new DOMEditor();
        $editor->loadHTML($html, LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD);
        // works like a charm!
        echo $editor->saveHTML();

Your point 1. is solved by using mb_convert_encoding($string, 'HTML-ENTITIES', 'UTF-8'); before using loadHTML() and the 2.nd by having a DIV tag around in your helper function, around mb_convert_encoding() you use for example. Worked out for me good enough. Indeed if no DIV is present, then it automatically adds a paragraph in my case which is inconvenient since usually they have some margin applied (bootstrap..) — trainoasis, Jan 27 '20 at 08:54

score 2 · Answer 20 · answered Nov 18 '21 at 09:19

After reading lots of code about this topic, I ended up with the following solution that works very well for me and is easy to understand.

It fixes unwanted Doctype and <html> and <body> as well as encoding issues.

This code assumes that $htmlContent is encoded in utf-8.

$htmlContent = "<h1>This is a heading</h1><p>This is a paragraph</p>";

// 1.) Load the html
$dom = new DOMDocument();
$dom->loadHTML("<meta http-equiv='Content-Type' content='charset=utf-8' /><div>$htmlContent</div>");

// 2.) Do you logic
$dom->getElementsByTagName('h1')[0]->setAttribute('class', 'happy');

// 3.) Render the html
$wrapperNode = $dom->getElementsByTagName('div')[0];
$renderedHtml = $dom->saveHTML($wrapperNode);
// If you want to keep the wrapper div
echo $renderedHtml;
// Or remove the wrapper <div>
echo substr(trim($renderedHtml), 5, -6);

The key take aways are:

loadHTML assumes content to be iso-8859-1, if this is not the case, you need to add encoding information.
Wrap your html code in a div and render just this div, you can remove it with substring if you don’t want to keep it.

score 1 · Answer 21 · answered Jan 29 '21 at 12:16

My universal solution independent of how the HTML was loaded:

function getNodeHtml(DOMNode $node, $outer = true) {
    $doc = new DOMDocument();
    $node = $node instanceof DOMDocument ? $node->documentElement : $node;
    foreach(($outer ? array($node) : $node->childNodes) as $n) {
        $doc->appendChild($doc->importNode($n->cloneNode(true), true));
    }
    return $doc->saveHTML();
}

Sample results:

<p>foo bar </p> ━▶ <p>foo bar </p>
<p>foo</p><p>bar</p> ━▶ <p>foo</p><p>bar</p>
<p>foo </p> <p> bar</p> ━▶ <p>foo </p> <p> bar</p>
Hello! ━▶ Hello!
<html><body><b>foo</b></body></html> ━▶ <html><body><b>foo</b></body></html>

This does absolutely nothing... Why not just use "$doc->saveHTML()" ? — Martin Zvarík, Apr 27 '21 at 21:23

score 0 · Answer 22 · answered May 13 '13 at 14:30

0

I came upon this issue as well.

Unfortunately, I did not felt comfortably using any of the solutions provided in this thread, so I went to check one that would satisfy me.

Here's what I made up and it works without issues:

$domxpath = new \DOMXPath($domDocument);

/** @var \DOMNodeList $subset */
$subset = $domxpath->query('descendant-or-self::body/*');

$html = '';
foreach ($subset as $domElement) {
    /** @var $domElement \DOMElement */
    $html .= $domDocument->saveHTML($domElement);
}

In essense it works in similar way to most of the solutions provided here, but instead of doing manual labor it uses xpath selector to select all the elements within the body and concatenates their html code.

answered May 13 '13 at 14:30

Nikola Petkanski

4,724
1
33
41

Like all solutions here, it doesn't work for every case : if the loaded string didn't start with markup,
has been added, then your code doesn't work, since it'll add the
markup in the saved content – copndz May 15 '13 at 15:26
To be fair, I haven't tested it with raw text, but in theory should work. For your specific case you may need to change the xpath to something like ```descendant-or-self::body/p/*```. – Nikola Petkanski May 16 '13 at 08:57

Tomer Ofer · Answer 23 · 2014-05-29T08:31:56.377

my server got php 5.3 and can't upgrade so those options

LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD

are not for me.

To solve this i tell to the SaveXML Function to print the Body element and then just replace the "body" with "div"

here is my code, hope it's helping someone:

<? 
$html = "your html here";
$tabContentDomDoc = new DOMDocument();
$tabContentDomDoc->loadHTML('<?xml encoding="UTF-8">'.$html);
$tabContentDomDoc->encoding = 'UTF-8';
$tabContentDomDocBody = $tabContentDomDoc->getElementsByTagName('body')->item(0);
if(is_object($tabContentDomDocBody)){
    echo (str_replace("body","div",$tabContentDomDoc->saveXML($tabContentDomDocBody)));
}
?>

the utf-8 is for Hebrew support.

score 0 · Answer 24 · answered Oct 01 '14 at 22:59

Alex answer is correct, but might cause following error on empty nodes:

Argument 1 passed to DOMNode::removeChild() must be an instance of DOMNode

Here comes my little mod:

    $output = '';
    $doc = new DOMDocument();
    $doc->loadHTML($htmlString); //feed with html here

    if (isset($doc->firstChild)) {

        /* remove doctype */

        $doc->removeChild($doc->firstChild);

        /* remove html and body */

        if (isset($doc->firstChild->firstChild->firstChild)) {
            $doc->replaceChild($doc->firstChild->firstChild->firstChild, $doc->firstChild);
            $output = trim($doc->saveHTML());
        }
    }
    return $output;

Adding the trim() is also a good idea to remove whitespace.

Joe · Answer 25 · 2016-04-28T18:23:36.523

0

I maybe too late. But maybe somebody (like me) still has this issue.
So, none of the above worked for me. Because $dom->loadHTML also close open tags as well, not only add html and body tags.
So add a < div > element is not working for me, because I have sometimes like 3-4 unclosed div in the html piece.
My solution:

1.) Add marker to cut, then load the html piece

$html_piece = "[MARK]".$html_piece."[/MARK]";
$dom->loadHTML($html_piece);

2.) do whatever you want with the document
3.) save html

$new_html_piece = $dom->saveHTML();

4.) before you return it, remove < p >< /p > tags from marker, strangely it is only appear on [MARK] but not on [/MARK]...!?

$new_html_piece = preg_replace( "/<p[^>]*?>(\[MARK\]|\s)*?<\/p>/", "[MARK]" , $new_html_piece );

5.) remove everything before and after marker

$pattern_contents = '{\[MARK\](.*?)\[\/MARK\]}is';
if (preg_match($pattern_contents, $new_html_piece, $matches)) {
    $new_html_piece = $matches[1];
}

6.) return it

return $new_html_piece;

It would be a lot easier if LIBXML_HTML_NOIMPLIED worked for me. It schould, but it is not. PHP 5.4.17, libxml Version 2.7.8.
I find really strange, I use the HTML DOM parser and then, to fix this "thing" I have to use regex... The whole point was, not to use regex ;)

edited Apr 28 '16 at 18:23

answered Apr 27 '16 at 20:19

Joe

129
3
6

Looks dangerous what you do here, http://stackoverflow.com/a/29499718/367456 should do the job for you. – hakre Dec 04 '16 at 09:31
Unfortunately this (https://stackoverflow.com/questions/4879946/how-to-savehtml-of-domdocument-without-html-wrapper/29499718#29499718) will not work for me. As I said: "So add a < div > element is not working for me, because I have sometimes like 3-4 unclosed div in the html piece" For some reason, the DOMDocument want to close all "not closed" elements. In may case, I will get a fregment within a shortcode or other marker, remove the fregment and I want to manipulate the other piece of the document, when I'm done with that, I will insert the fregment back. – Joe Jun 01 '17 at 07:27
It should be possible to leave the div element out and operate on the body element after loading your own content instead. The body element should be added implicitly when you load a fragment. – hakre Jun 01 '17 at 10:08
My problem is, my fregment containt unclosed tag. It should be remain unclosed and DOMDocument will close those elements. Fregment like: `< div >< div > ... < /div >`. I still looking for solutions. – Joe Jun 02 '17 at 10:10
Hmm, I think div tags always have a closing pair. Perhaps Tidy can handle that, it can work with fragments, too. – hakre Jun 02 '17 at 17:37
It should and in the end it will. This "markers" is from the enduser. I use this markers eg. for a lazy loading plugin to ignore a part of the page. I do not know where the user will add the marks. It shouldn't metter. What my code need to do, replace all < img src to 1px blank placeholder but ignore the fregment between markers. – Joe Jul 04 '17 at 06:00
Sound more like a custom filtering of the underlying nodes. The markers will be within text-nodes. This is a bit more advanced, text node processing (but in the meaning of editing these and inserting element nodes there in I have shown in https://stackoverflow.com/a/8193700/367456 - just referencing, this is *not* exactly what you're looking for but it *might* give you some ideas. – hakre Jul 04 '17 at 17:41

score -1 · Answer 26 · answered Sep 06 '16 at 15:27

For anyone using Drupal, there's a built in function to do this:

https://api.drupal.org/api/drupal/modules!filter!filter.module/function/filter_dom_serialize/7.x

Code for reference:

function filter_dom_serialize($dom_document) {
  $body_node = $dom_document->getElementsByTagName('body')->item(0);
  $body_content = '';

  if ($body_node !== NULL) {
    foreach ($body_node->getElementsByTagName('script') as $node) {
      filter_dom_serialize_escape_cdata_element($dom_document, $node);
    }

    foreach ($body_node->getElementsByTagName('style') as $node) {
      filter_dom_serialize_escape_cdata_element($dom_document, $node, '/*', '*/');
    }

    foreach ($body_node->childNodes as $child_node) {
      $body_content .= $dom_document->saveXML($child_node);
    }
    return preg_replace('|<([^> ]*)/>|i', '<$1 />', $body_content);
  }
  else {
    return $body_content;
  }
}

Upvoted. Usig this function from the Drupal API works fine on my Drupal 7 site. I guess those not using Drupal can just copy the function into their own site - as there is nothing Drupal-specific about this. — Free Radical, Feb 07 '18 at 09:23

score -1 · Answer 27 · answered Jun 15 '19 at 23:32

-1

#remove doctype tag
$doc->removeChild($doc->doctype); 

#remove html & body tags
$html = $doc->getElementsByTagName('html')[0];
$body = $html->getElementsByTagName('body')[0];
foreach($body->childNodes as $child) {
    $doc->appendChild($child);
}
$doc->removeChild($html);

answered Jun 15 '19 at 23:32

Dylan Maxey

976
10
9

Care to share why the -1? – Dylan Maxey Apr 07 '20 at 21:46

score -1 · Answer 28 · answered Jun 05 '20 at 13:37

You can use tidy with show-body-only:

$tidy = new tidy();
$htmlBody = $tidy->repairString($html, [
  'indent' =>  true,
  'output-xhtml' => true,
  'show-body-only' => true
], 'utf8');

But, remeber: tidy remove some tags like Font Awesome icons: Problems Indenting HTML(5) with PHP

score -1 · Answer 29 · answered Nov 18 '20 at 14:22

-1

This is the solution that helped me:

$content = str_replace(array('<html>','</html>') , '' , $doc->saveHTML());

answered Nov 18 '20 at 14:22

Sven Johanson

1

How to saveHTML of DOMDocument without HTML wrapper?

29 Answers29

Linked

Related