# _plugins/filter.rb
module Filter
  def self.process(site, payload)
    # The link to the notes is named _linked which creates `linked` collection in Jekyll
    # This takes all the files from there and keeps only the ones that have the `publish_on`
    # attribute in the front matter set.
    # This allows me to keep track of drafts or schedule publishing.
    site.collections['linked'].docs.select!{|x| x.data['publish_on']}
    now = Time.now

    # Next, I loop through the collection in order inject each file into the `posts`
    # collection. By default, Jekyll expects the filename to contain the date.
    # Because my files do not have that their dates default  to the file
    # creation date. Since that isn't the order that I want to publish them
    # in I've added a `publish_on` attribute to the front matter
    # that is parsed and posts are then ordered by that instead.
    site.collections['linked'].docs.sort_by{|x| Time.parse(x.data['publish_on'])}.each do |x|
      t = Time.parse(x.data['publish_on'])

      # For normal builds, I want to include only the posts that should have
      # already been published. But, if the --future flag is set when running
      # Jekyll, then all scheduled posts will be rendered.
      if t <= now or site.config['future']

        # Here, a new Jekyll::Document is created that is a copy of the linked
        # document. It's then tied to the posts collection.
        new_doc = Jekyll::Document.new(
          x.path,
          {site: site, collection: site.collections['posts']}
        )

        # Jekyll triggers this under the hood, although I haven't found exactly
        # where. Instead, I trigger it manually. This reads the contents of the
        # file and sets the front matter into the data attribute of the document.
        new_doc.read

        # The date is set to the parsed `publish_on` attribute.
        new_doc.data['date'] = t

        # I do not use the draft attribute explicitly. I use only the `publish_on`.
        # This line can be excluded if you intend to have a separate attribute in
        # the front matter to explicitly manage the draft status.
        new_doc.data['draft'] = false

        # Copy the categories and the description from the front matter.
        # I'm not sure why this is not picked up immediately.
        new_doc.data['categories'] = x.data['categories']

        # I use `desc` in the notes but Jekyll uses `description`.
        new_doc.data['description'] = x.data['desc']

        # Remove the duplicate description
        x.data.delete 'desc'

        # Set the layout.
        new_doc.data['layout'] = 'post'

        # Remove any dots from the slug. Note that this isn't removing the extension
        # but because some of my earlier posts used to have files with the format
        # blog.category.slug
        new_doc.data['slug'] = x.data['slug'].split('.').last

        # Set the __coll attribute.
        # While looking into this I noticed this attribute set, although I'm not sure
        # if it's strictly necessary or where it's used.
        new_doc.data['__coll'] = "posts"

        # This line removes any content in between < !-- exclude --> and < !-- include -->
        # This lets me have additional content in blog posts that I do not want
        # to be included in the generated site
        new_doc.content = new_doc.content.split("< !-- exclude -->")
        new_doc.content = new_doc.content.reject{|x| x.include? "< !-- include -->"}.join("")
        # Note that here there is a space between < and !. The actual file does
        # not have the space. It's needed here because the syntax highligter will
        # render each term within a span which will cause the delimiters to render
        # as comments since they're no longer direct descendants of the code block.

        # Secondly, this has a limitation in that it does not support having an
        # exclude block at the beginning of the file.

        # Fix the links to images because my blog notes are in notes/blog/<category>
        # and the images are in notes/assets/images while in Jekyll images are /images
        new_doc.content.gsub!("", "")

        # I do maintain separate copies of images for the blog. This is mainly
        # so that I can keep higher resolution images in my notes and serve
        # smaller more efficient versions in the blog.

        # Fix links between posts because in the blog notes it'd be
        # [Link](<category>/slug.md). The category and the .md need
        # to be removed so that only the slug remains.
        new_doc.content.gsub!(/\[(.+)\]\(..\/(.+)\/(.+)\.md\)/, '[\1](/\3)')

        # Finally, include the new document into the posts collection
        site.collections['posts'].docs << new_doc
      end
    end
  end
end

# Register the hook to run whenever files are read.
Jekyll::Hooks.register :site, :post_read do |site, payload|
  Filter.process(site, payload)
end