Issue with meta description links in Discourse page source

Today, I have been trying to fix the following issue with the Discourse source code:

Screen Shot 2020-12-05 at 12.03.02 PM1195×106 42.1 KB

As you can see Discourse is not stripping newlines and extra white space in their meta description tags.

In addition, there is also the oddly injected HTML entities, in this case: "…"

I tried adding this code to one of my plugins for Discourse, but it did not work as expected, and in fact, this code crashed the topics view:

TopicsController.class_eval do
    after_action :strip_newlines_and_more_from_meta_description

    private

    def strip_newlines_and_more_from_meta_description
      @description_meta = @description_meta&.gsub("…", "")
      @description_meta = @description_meta&.gsub(/\n+/, " ")
      @description_meta = @description_meta&.gsub(/\s+/, " ")&.strip
    end
  end

Not sure why. The code works fine in the rails console when not used as an action controller filter.

I tried both:

TopicsController.class_eval do

and

TopicsController.instance_eval do

No joy yet.

Anyway, I'm going to take a break since this is not working as expected; but I might keep working on this when the site is quiet with few members logged in

If anyone has Discourse running in a development environment, if you can get this code working so we can fix our meta data, that would be helpful.

Thanks.

Hmm.

I added some debug to log statements and the filter works:

>>>>NEONEO_BEGIN>>>>> @description_meta: Today, I have been trying to fix the following issue with the Discourse source code: 
 
As you can see Discourse is not stripping newlines and  extra white space in their meta description tags. 
In addition, there is als…  uri: /t/issue-with-meta-description-links-in-discourse-page-source/381000
>>>>NEONEO_END>>>>> @description_meta: Today, I have been trying to fix the following issue with the Discourse source code: As you can see Discourse is not stripping newlines and extra white space in their meta description tags. In addition, there is als…  uri: /t/issue-with-meta-description-links-in-discourse-page-source/381000

But for some reason, the Rails filter after_action appears to not work as expected.

My expectation was that before_action would filter before the page is rendered; but maybe not?

Will try moving the logic "up the chain", as follows and debug:

TopicsController.class_eval do
    before_action :strip_newlines_and_more_from_meta_description, on: [:perform_show_response]

    private

    def strip_newlines_and_more_from_meta_description
      Discourse.warn(">>>>NEONEO_BEGIN>>>>> @topic_view.topic.excerpt: #{@topic_view.topic.excerpt}", uri: request.env["REQUEST_URI"])
      tmp_description = @topic_view.topic.excerpt&.dup
      tmp_description = tmp_description&.gsub("…", "")
      tmp_description = tmp_description&.gsub("…", "")
      tmp_description = tmp_description&.gsub(/\n+/, " ")
      tmp_description = tmp_description&.gsub(/\s+/, " ")&.strip
      @topic_view.topic.excerpt = tmp_description
      Discourse.warn(">>>>NEONEO_END>>>>> @topic_view.topic.excerpt: #{@topic_view.topic.excerpt}", uri: request.env["REQUEST_URI"])
    end
  end

Hmm. None of those techniques worked as expected, so I am going to try this:

  # overwrite TopicsController private method to clean up @description_meta text
  class TopicsController
    def perform_show_response
      if request.head?
        head :ok
        return
      end

      topic_view_serializer = TopicViewSerializer.new(
        @topic_view,
        scope: guardian,
        root: false,
        include_raw: !!params[:include_raw],
      )

      respond_to do |format|
        format.html do
          @tags = SiteSetting.tagging_enabled ? @topic_view.topic.tags : []
          @breadcrumbs = helpers.categories_breadcrumb(@topic_view.topic) || []
          @description_meta = @topic_view.topic.excerpt.present? ? @topic_view.topic.excerpt : @topic_view.summary
          @description_meta = @description_meta&.gsub("…", "")
          @description_meta = @description_meta&.gsub("…", "")
          @description_meta = @description_meta&.gsub(/\n+/, " ")
          @description_meta = @description_meta&.gsub(/\s+/, " ")&.strip
          store_preloaded("topic_#{@topic_view.topic.id}", MultiJson.dump(topic_view_serializer))
          render :show
        end

        format.json do
          render_json_dump(topic_view_serializer)
        end
      end
    end
  end
end

EDIT: Which also did not work as expected

Success!

This monkey patch worked as expected:

  # overwrite TopicsController private method to clean up @description_meta text
 TopicsController.class_eval do
  def perform_show_response
    if request.head?
      head :ok
      return
    end

    topic_view_serializer = TopicViewSerializer.new(
      @topic_view,
      scope: guardian,
      root: false,
      include_raw: !!params[:include_raw],
    )

    respond_to do |format|
      format.html do
        @tags = SiteSetting.tagging_enabled ? @topic_view.topic.tags : []
        @breadcrumbs = helpers.categories_breadcrumb(@topic_view.topic) || []
        @description_meta = @topic_view.topic.excerpt.present? ? @topic_view.topic.excerpt : @topic_view.summary
        @description_meta = @description_meta&.gsub("…", " ")
        @description_meta = @description_meta&.gsub("…", " ")
        @description_meta = @description_meta&.gsub(/\n+/, " ")
        @description_meta = @description_meta&.gsub(/\s+/, " ")&.strip
        store_preloaded("topic_#{@topic_view.topic.id}", MultiJson.dump(topic_view_serializer))
        render :show
      end

      format.json do
        render_json_dump(topic_view_serializer)
      end
    end
  end

Took me the better part of three hours to work it out; but it's worked out and now, and the meta descriptions important for SEO are not mangled now:

Screen Shot 2020-12-05 at 3.52.49 PM1875×115 52.3 KB

Soon, I will release this new plugin which fixes this and other issues, which I call:

Plugin: discourse-selfish-seo-layout-changes

Anyway, popular SEO wisdom says the meta description is not of much SEO value and that Google does not use this meta tag; and if they did, they just omit the newlines. However, I prefer to get rid of the extra white space and newlines. So, this code is done for now; until I see other little issues that need to be addressed in the Discourse source code.