Performance of atx-style header regex whit lot of whitespace

[aprendendofelipe]

The problem reported here #35 and here #52 still persists.

Any string with a lot of whitespace makes the atx-style header regex very slow.

The proposed regex, in addition to being faster in any situation, does not encounter the same performance problem with whitespace.

[blaw2422]

any chance a bug was introduced with this pr? https://github.com/stiang/remove-markdown/commit/fd5e8034eae0dc1b61ecdda09ab11d18a9a43f2d#diff-e727e4bdf3657fd1d798edcd6b099d6e092f8573cba266154583a746bba0f346L74

I'm not sure if your change fixes it but I can't get this to work with markdown thats got a table with more than 5 or 6 columns.

[aprendendofelipe]

I'm not sure if your change fixes it but I can't get this to work with markdown thats got a table with more than 5 or 6 columns.

Please test with this function and tell us the result

function removeMarkdown(md, options) {
  options = options || {};
  options.listUnicodeChar = options.hasOwnProperty('listUnicodeChar') ? options.listUnicodeChar : false;
  options.stripListLeaders = options.hasOwnProperty('stripListLeaders') ? options.stripListLeaders : true;
  options.gfm = options.hasOwnProperty('gfm') ? options.gfm : true;
  options.useImgAltText = options.hasOwnProperty('useImgAltText') ? options.useImgAltText : true;
  options.abbr = options.hasOwnProperty('abbr') ? options.abbr : false;
  options.replaceLinksWithURL = options.hasOwnProperty('replaceLinksWithURL') ? options.replaceLinksWithURL : false;
  options.htmlTagsToSkip = options.hasOwnProperty('htmlTagsToSkip') ? options.htmlTagsToSkip : [];

  var output = md || '';

  // Remove horizontal rules (stripListHeaders conflict with this rule, which is why it has been moved to the top)
  output = output.replace(/^(-\s*?|\*\s*?|_\s*?){3,}\s*/gm, '');

  try {
    if (options.stripListLeaders) {
      if (options.listUnicodeChar)
        output = output.replace(/^([\s\t]*)([\*\-\+]|\d+\.)\s+/gm, options.listUnicodeChar + ' $1');
      else output = output.replace(/^([\s\t]*)([\*\-\+]|\d+\.)\s+/gm, '$1');
    }
    if (options.gfm) {
      output = output
        // Header
        .replace(/\n={2,}/g, '\n')
        // Fenced codeblocks
        .replace(/~{3}.*\n/g, '')
        // Strikethrough
        .replace(/~~/g, '')
        // Fenced codeblocks
        .replace(/`{3}.*\n/g, '');
    }
    if (options.abbr) {
      // Remove abbreviations
      output = output.replace(/\*\[.*\]:.*\n/, '');
    }
    output = output
      // Remove HTML tags
      .replace(/<[^>]*>/g, '');

    var htmlReplaceRegex = new RegExp('<[^>]*>', 'g');
    if (options.htmlTagsToSkip.length > 0) {
      // Using negative lookahead. Eg. (?!sup|sub) will not match 'sup' and 'sub' tags.
      var joinedHtmlTagsToSkip = '(?!' + options.htmlTagsToSkip.join('|') + ')';

      // Adding the lookahead literal with the default regex for html. Eg./<(?!sup|sub)[^>]*>/ig
      htmlReplaceRegex = new RegExp('<' + joinedHtmlTagsToSkip + '[^>]*>', 'ig');
    }

    output = output
      // Remove HTML tags
      .replace(htmlReplaceRegex, '')
      // Remove setext-style headers
      .replace(/^[=\-]{2,}\s*$/g, '')
      // Remove footnotes?
      .replace(/\[\^.+?\](\: .*?$)?/g, '')
      .replace(/\s{0,2}\[.*?\]: .*?$/g, '')
      // Remove images
      .replace(/\!\[(.*?)\][\[\(].*?[\]\)]/g, options.useImgAltText ? '$1' : '')
      // Remove inline links
      .replace(/\[([^\]]*?)\][\[\(].*?[\]\)]/g, options.replaceLinksWithURL ? '$2' : '$1')
      // Remove blockquotes
      .replace(/^(\n)?\s{0,3}>\s?/gm, '$1')
      // .replace(/(^|\n)\s{0,3}>\s?/g, '\n\n')
      // Remove reference-style links?
      .replace(/^\s{1,2}\[(.*?)\]: (\S+)( ".*?")?\s*$/g, '')
      // Remove atx-style headers
      .replace(/^(\n)?\s{0,}#{1,6}\s*( (.+))? +#+$|^(\n)?\s{0,}#{1,6}\s*( (.+))?$/gm, '$1$3$4$6')
      // Remove * emphasis
      .replace(/([\*]+)(\S)(.*?\S)??\1/g, '$2$3')
      // Remove _ emphasis. Unlike *, _ emphasis gets rendered only if
      //   1. Either there is a whitespace character before opening _ and after closing _.
      //   2. Or _ is at the start/end of the string.
      .replace(/(^|\W)([_]+)(\S)(.*?\S)??\2($|\W)/g, '$1$3$4$5')
      // Remove code blocks
      .replace(/(`{3,})(.*?)\1/gm, '$2')
      // Remove inline code
      .replace(/`(.+?)`/g, '$1')
      // // Replace two or more newlines with exactly two? Not entirely sure this belongs here...
      // .replace(/\n{2,}/g, '\n\n')
      // // Remove newlines in a paragraph
      // .replace(/(\S+)\n\s*(\S+)/g, '$1 $2')
      // Replace strike through
      .replace(/~(.*?)~/g, '$1');
  } catch (e) {
    console.error(e);
    return md;
  }
  return output;
}

[stiang]

Thanks for the PR. I would love to merge this but I’m struggling to find the time to maintain the package. Please see issue #61.

[zuchka]

@aprendendofelipe @stiang I just tested this PR on node 16 and it worked well. I'm going to merge it, with the understanding that we can roll it back as needed. 👍

great work @aprendendofelipe

[zuchka]

also appears to fix #58