to-markdown.php

<?php

/*
 * to-markdown - an HTML to Markdown converter for PHP
 *
 * Copyright 2013, Blake Turner
 * Licenced under the MIT licence
 *
 */

function toMarkdown($string, $housekeeping = TRUE) {
  $markdown = $string;

  // Get rid of attributes!
  // Particularly useful when converting garbage wysiwyg code to markdown
  if ($housekeeping) {
    $els = array('p', 'span', 'ul', 'ol', 'li', 'strong', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6');
    foreach ($els as $el) {
      $markdown = preg_replace("/<$el.*?>/", "<$el>", $markdown);
    }
  }

  $regexMap = array(
    'p'         => "",
    '\/p'       => "\n",
    'span'      => '',
    'span.*?'   => '',
    '\/span'    => '',
    'h1'        => '#',
    'h2'        => '##',
    'h3'        => '###',
    'h4'        => '####',
    'h5'        => '#####',
    'h6'        => '######',
    '\/h\d'     => "\n",
    'br'        => "\n",
    'br\s\/'    => "\n",
    'strong'    => "__",
    '\/strong'  => "__",
    'em'        => "_",
    '\/em'      => "_",
  );

  // Replace simple tags w/ markdown equivalent
  foreach ($regexMap as $el => $replacement) {
    $markdown = preg_replace("/<$el>/", $replacement, $markdown);
  }

  // Images: Capture <img> tags, isolate src
  if (preg_match_all('/<img.*src="([^\s"]*?)".*>?/', $markdown, $matches)) {

    // Swap markdown replacement for <img> markup for each match
    foreach ($matches[0] as $i => $img_markup) {
      $url = $matches[1][$i];

      // Alt tags
      $alt = NULL;
      if (preg_match('/alt="([^\s"]*?)"/', $img_markup, $alt_match)) {
        $alt = $alt[1];
      }

      // Build markdown and swap
      $img_markdown = "![$alt]($url)";
      $markdown = str_replace($img_markup, $img_markdown, $markdown);
    }
  }

  // Links: Capture <a> tags, isolate href
  if (preg_match_all('/<a.*?href="([^\s"]*?)".*?>(.*?)?<\/a>/', $markdown, $matches)) {

    // Swap markdown replacement for <a> markup for each match
    foreach ($matches[0] as $i => $a_markup) {
      $href = $matches[1][$i];
      $text = $matches[2][$i];

      // Build markdown and swap
      $a_markdown = "[$text]($href)";
      $markdown = str_replace($a_markup, $a_markdown, $markdown);
    }
  }

  // Unordered lists
  if (preg_match_all('/<ul>(.*?)<\/ul>/s', $markdown, $matches)) {
    $markdown = preg_replace('/<ul>|<\/ul>/', '', $markdown);

    foreach ($matches[0] as $ul) {
      if (preg_match_all('/<li>(.*?)<\/li>/s', $ul, $list_items)) {
        foreach ($list_items[0] as $i => $li_markup) {
          $li_inner = $list_items[1][$i];
          $li_markdown = '- ' . $li_inner;
          $markdown = str_replace($li_markup, $li_markdown, $markdown);
        }
      }
    }
  }

  // Ordered lists
  if (preg_match_all('/<ol>(.*?)<\/ol>/s', $markdown, $matches)) {
    $markdown = preg_replace('/<ol>|<\/ol>/', '', $markdown);

    foreach ($matches[0] as $ol) {
      if (preg_match_all('/<li>(.*?)<\/li>/s', $ol, $list_items)) {
        foreach ($list_items[0] as $i => $li_markup) {
          $li_inner = $list_items[1][$i];
          $li_markdown = $i+1 . '. ' . $li_inner;
          $markdown = str_replace($li_markup, $li_markdown, $markdown);
        }
      }
    }
  }

  // Blockquotes
  if (preg_match_all('/<blockquote>(.*?)<\/blockquote>/s', $markdown, $matches)) {

    foreach ($matches[1] as $i => $inner_html) {
      $blockquote_markup = $matches[$i];
      $blockquote_markdown = '';

      $lines = explode("\n", $inner_html);
      foreach ($lines as $line) {
        $blockquote_markdown .= '> ' . $line . "\n";
      }

      $markdown = str_replace($blockquote_markup, $blockquote_markdown, $markdown);
    }
  }

  return $markdown;
}