Following up on last week’s post, today I’m gonna talk about the latest insights, hurdles and developments.

Current Status

Last week I left off after getting some basic rendering working and linking up the newly generated GalleryPage instances into the jekyll site. After posting the write-up I did investigate a bit further, as I was having troubles getting the paths between the source and destination sorted out. Eventually I found that jekyll has a separate Document class that takes care of entries in collections, which does handle path handling when the source is a underscore-directory, but duplicates/re-implements a lot of the remaining functionality, but differently.

To stay as close as possible to the original plan, I’ll substitute and Page for Document and see whether that fixes the path-translation issues I’ve been encountering when building the site. If that doesn’t work out, I’ll have to go back and modify the design and workflow of the generator. Maybe putting the entire _galleries folder content instead directly into the site and only go back and patch-up already instantiated Pages also works? This might even give more flexibility in the long-run, as any layout/page could access the new data for any images in that page/tree. Decision anxiety is a thing :-(, I’ve been on-and-off mulling over this for the last few days and both approaches have their merit. I’ll need to build (at least some of) both to test my assumptions.


I’ll start with replacing the Pages with Documents. This seems to be the smaller operation and will give me a better understanding of the feasibility of this and what else is there in Document.

Collections are configured by adding an entry in the _config.yml:

  - galleries

doing so already creates an instance of Jekyll::Collection in site.collections['galleries'] that has entries for every file in the directory tree:

[7] pry(#<Jekyll::Collection>):1> entries
=> [".",
[8] pry(#<Jekyll::Collection>):1>

but there is no rendering of anything happening, as the documentation helpfully points out. To change that we need to set output: true on the collection:

    output: true

and lo and behold, the index is rendered without any functionality in the cheesy-generator:

david@zion:~/Projects/cheesy-gallery/spec/fixtures/test_site$ cat _galleries/first/index.html
x_layout: gallery
title: "The first Gallery"
This is a test gallery. The first of its kind.
david@zion:~/Projects/cheesy-gallery/spec/fixtures/test_site$ cat _site/galleries/first/index.html
This is a test gallery. The first of its kind.

I’ve disabled layouting with the x_ prefix there to avoid a lot of irrelevant HTML, but enabling it does do the right thing and applies the layout tree through Liquid.

Image data, of course, is passed through unmodified.

The docs attribute on the Collection contains all Document instances. There is one for the first/index.html, but none for the image-only directories. The files attribute contains all StaticFile instances.

Another annoying quirk is that entries is a flat array. The documentation explains how to iterate over that in Liquid.

The documentation enumerates a few other attributes a collection entry (i.e. Document) can have. None of them specifically useful to the nested gallery usecase.

In this commit I implement all that I thought I had last weekend, but didn’t work, in a really nice and clean fashion:

def generate(site)
  @site = site
  collection = site.collections['galleries']

  # all galleries in the site
  galleries = Set[* { |e| File.dirname(e) }]

  # all galleries with an index.html
  galleries_with_index = Set[*collection.entries.find_all { |e| e.end_with?('/index.html') }.map { |e| File.dirname(e) }]

  # fill in Documents for galleries that don't have an index.html
  (galleries - galleries_with_index).each do |e|
    doc ='_galleries', e, 'index.html'), site: site, collection: collection) << doc if site.unpublished || doc.published?

  files_by_dirname = {}
  collection.files.each { |e| (files_by_dirname[File.dirname(e.relative_path)] ||= []) << e } do |doc|
    # attach images['images'] = files_by_dirname[File.dirname(doc.relative_path)]

First, galleries and galleries_with_index are calculated to be able to create new CheesyGallery::GalleryIndex instances for galleries that do not have a index.html. Then, the new Documents get created to fill in the blanks. Since jekyll maintains everything in flat lists, files_by_dirname is used to make the image linking easier. Finally, the images attribute is filled with all files belonging to a specific document.

In the gallery.html, I make some changes to show the images as an example how it can be used. For the next time I get around to this, I update the project on github with what I managed today and enhance the description of what to do next, as well as a couple of ideas for going forward.