Keeping line breaks when using PHP's DomDocument appendChild

Question

I'm trying to use the DOMDocument in PHP to add/parse things in an HTML document. From what I could read, setting the formOutput to true and preserveWhiteSpace to false should keep the tabs and newlines in order, but it doesn't seem like it is for newly created or appended nodes.

Here's the code:

$dom = new \DOMDocument;
$dom->formatOutput = true;
$dom->preserveWhiteSpace = false;
$dom->loadHTMLFile($htmlsource);
$tables = $dom->getElementsByTagName('table');
foreach($tables as $table)
{
    $table->setAttribute('class', 'tborder');
    $div = $dom->createElement('div');
    $div->setAttribute('class', 'm2x');
    $table->parentNode->insertBefore($div, $table);
    $div->appendChild($table);
}
$dom->saveHTMLFile($html)

Here's what the HTML looks like:

<table>
    <tr>
        <td></td>
    </tr>
</table>

Here's what I want:

<div class="m2x">
    <table class="tborder">
        <tr>
            <td></td>
        </tr>
    </table>
</div>

Here's what I get:

<div class="m2x"><table class="tborder"><tr>
<td></td>
        </tr></table></div>

Is there something I'm doing wrong? I've tried googling this as many different ways as I could thing of with no luck.

score 2 · Accepted Answer · answered Oct 20 '11 at 19:12

Unfortunately, you might need to write a function that indents the output how you want it. I made a little function you might find helpful.

function indentContent($content, $tab="\t")
{               

        // add marker linefeeds to aid the pretty-tokeniser (adds a linefeed between all tag-end boundaries)
        $content = preg_replace('/(>)(<)(\/*)/', "$1\n$2$3", $content);

        // now indent the tags
        $token = strtok($content, "\n");
        $result = ''; // holds formatted version as it is built
        $pad = 0; // initial indent
        $matches = array(); // returns from preg_matches()

        // scan each line and adjust indent based on opening/closing tags
        while ($token !== false) 
        {
                $token = trim($token);
                // test for the various tag states

                // 1. open and closing tags on same line - no change
                if (preg_match('/.+<\/\w[^>]*>$/', $token, $matches)) $indent=0;
                // 2. closing tag - outdent now
                elseif (preg_match('/^<\/\w/', $token, $matches))
                {
                        $pad--;
                        if($indent>0) $indent=0;
                }
                // 3. opening tag - don't pad this one, only subsequent tags
                elseif (preg_match('/^<\w[^>]*[^\/]>.*$/', $token, $matches)) $indent=1;
                // 4. no indentation needed
                else $indent = 0;

                // pad the line with the required number of leading spaces
                $line = str_pad($token, strlen($token)+$pad, $tab, STR_PAD_LEFT);
                $result .= $line."\n"; // add to the cumulative result, with linefeed
                $token = strtok("\n"); // get the next token
                $pad += $indent; // update the pad size for subsequent lines    
        }       

        return $result;
}

indentContent($dom->saveHTML()) will return:

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "http://www.w3.org/TR/REC-html40/loose.dtd">
<html>
    <body>
        <div class="m2x">
            <table class="tborder">
                <tr>
                    <td>
                    </td>
                </tr>
            </table>
        </div>
    </body>
</html>

I created this function starting with this one.

Great function! But unfortunately it indents too many spaces when using void elements. — Stan, Oct 30 '13 at 15:39
I've made changes to this code block, as to fit my needs, which is something you both have described. Link: https://stackoverflow.com/a/61990936/7646824 — Artis Zel, Sep 28 '22 at 06:55

Stan · Answer 2 · 2013-10-31T10:02:07.030

I modified the great function ghbarratt wrote, so it doesn't indent void elements.

function indentContent($content, $tab="\t")
{
    // add marker linefeeds to aid the pretty-tokeniser (adds a linefeed between all tag-end boundaries)
    $content = preg_replace('/(>)(<)(\/*)/', "$1\n$2$3", $content);

    // now indent the tags
    $token = strtok($content, "\n");
    $result = ''; // holds formatted version as it is built
    $pad = 0; // initial indent
    $matches = array(); // returns from preg_matches()

    // scan each line and adjust indent based on opening/closing tags
    while ($token !== false) 
    {
        $token = trim($token);
        // test for the various tag states

        // 1. open and closing tags on same line - no change
        if (preg_match('/.+<\/\w[^>]*>$/', $token, $matches)) $indent=0;
        // 2. closing tag - outdent now
        elseif (preg_match('/^<\/\w/', $token, $matches))
        {
            $pad--;
            if($indent>0) $indent=0;
        }
        // 3. opening tag - don't pad this one, only subsequent tags (only if it isn't a void tag)
        elseif (preg_match('/^<\w[^>]*[^\/]>.*$/', $token, $matches))
        {
            $voidTag = false;
            foreach ($matches as $m)
            {
                // Void elements according to http://www.htmlandcsswebdesign.com/articles/voidel.php
                if (preg_match('/^<(area|base|br|col|command|embed|hr|img|input|keygen|link|meta|param|source|track|wbr)/im', $m))
                {
                    $voidTag = true;
                    break;
                }
            }

            if (!$voidTag) $indent=1;
        }
        // 4. no indentation needed
        else $indent = 0;

        // pad the line with the required number of leading spaces
        $line = str_pad($token, strlen($token)+$pad, $tab, STR_PAD_LEFT);
        $result .= $line."\n"; // add to the cumulative result, with linefeed
        $token = strtok("\n"); // get the next token
        $pad += $indent; // update the pad size for subsequent lines    
    }    

    return $result;
}

All credits go to ghbarratt.

+1 Way to make it better. And I hardly deserve much credit for it since I mostly swiped it from http://recursive-design.com/blog/2007/04/05/format-xml-with-php/ — ghbarratt, Jan 14 '14 at 21:03

Artis Zel · Answer 3 · 2020-04-28T09:01:54.337

Both @Stan and @ghbarrat does not go well with <!DOCTYPE html> html5 declaration. It kind of passes indentation to <head> element.

Expected:

<!DOCTYPE html>
<html>
  <head>
    <meta charset="UTF-8">
  </head>
  <body>
    <!-- all good -->
  </body>
</html>

Result:

<!DOCTYPE html>
<html>
  <head>
    <meta charset="UTF-8">
    </head>
    <body>
      <!-- all good -->
    </body>
  </html>

A little bit of testing, reveals a partial fix when I add <html> element to Void element list, however that does not solve the problem with head and it also flattens children (namely head and body).

Edit #1 It appears <meta charset="UTF-8"> after all is responsible for incorrect indentation.

Edit #2 - Solution

After little troubleshooting I discovered that <meta> as a self-closing tag would impact next closing tag, which is solved by adding a flag. The flag defines if we found self-closing tag, then next instance of a closing tag would have an extra negative indent.

function indentContent($content, $tab="\t"){
    // add marker linefeeds to aid the pretty-tokeniser (adds a linefeed between all tag-end boundaries)
    $content = preg_replace('/(>)(<)(\/*)/', "$1\n$2$3", $content);

    // now indent the tags
    $token = strtok($content, "\n");
    $result = ''; // holds formatted version as it is built
    $pad = 0; // initial indent
    $matches = array(); // returns from preg_matches()

    // scan each line and adjust indent based on opening/closing tags
    while ($token !== false && strlen($token)>0)
    {
        $token = trim($token);
        // test for the various tag states

        // 1. open and closing tags on same line - no change
        if (preg_match('/.+<\/\w[^>]*>$/', $token, $matches)) $indent=0;
        // 2. closing tag - outdent now
        elseif (preg_match('/^<\/\w/', $token, $matches))
        {
            $pad--;
            if($indent>0) $indent=0;
            if($nextTagNegative){
                $pad--;$nextTagNegative=false;
            }
        }
        // 3. opening tag - don't pad this one, only subsequent tags (only if it isn't a void tag)
        elseif (preg_match('/^<\w[^>]*[^\/]>.*$/', $token, $matches))
        {
            $voidTag = false;
            foreach ($matches as $m)
            {
                // Void elements according to http://www.htmlandcsswebdesign.com/articles/voidel.php
                if (preg_match('/^<(area|base|br|col|command|embed|hr|img|input|keygen|link|meta|param|source|track|wbr)/im', $m))
                {
                    $voidTag = true;
                    break;
                }
            }

            if (!$voidTag) $indent=1;$nextTagNegative=true;
        }
        // 4. no indentation needed
        else $indent = 0;

        // pad the line with the required number of leading spaces
        $line = str_pad($token, strlen($token)+$pad, $tab, STR_PAD_LEFT);
        $result .= $line."\n"; // add to the cumulative result, with linefeed
        $token = strtok("\n"); // get the next token
        $pad += $indent; // update the pad size for subsequent lines
    }

    return $result;
}

There is a problem for LI tag with my solution. Have not found a solution, yet. — Artis Zel, May 06 '20 at 09:00

Keeping line breaks when using PHP's DomDocument appendChild

3 Answers3

Linked