Interestingly, I find both ridiculous. I now just use discount and a Makefile
Why not use Pandoc I get the fuzzies knowing that my documents got turned into an AST by Haskell.
Hakyll is really great. I started with Jekyll, migrated to Octopress, didn’t like it, and eventually moved over to Hakyll. It strikes me as more of a library made for writing your own static blog generator, totally customized to your own blog, as opposed to a pre-built tool. However, this means it’s really great for customization - my website is run by a custom file format written in 80 lines using Parsec, and can accept four or five different formats for its blog entries (IHaskell or IPython notebooks, LaTeX, Markdown, HTML, plain text, or Asciidoc). All of this was trivial to add to the Hakyll blog over the years, at the price of a pretty steep upfront configuration cost.
Please tell me more, this sounds amazing. Is your parsec available?
Sure, you can take a look at the entire blog on Github.
It’s not that fancy, and it’s not that generalizable, but it works really well for me. I have a directory called posts. In that directory, I have a file called postlist, with contents like this:
title "Finger Trees";
categories haskell, ihaskell;
title "Linguistics and Syntax";
The posts directory also contains a subdirectory for each post; the source field specifies which directory corresponds to which post. Each of these directories must contain a post.ext file, where ext is tex, md, ipynb, and maybe a few more formats, I don’t remember. The directories can also contain img, data, and files directories, which the post.ext file may use. When you run the blog generator, it’ll parse the postlist, generate a post for each post.ext file, and create all the relevant HTML pages. It’s very nice, because it means I can write my blog in any format - sometimes I use HTML, sometimes latex, sometimes IPython/IHaskell notebook. Then I just create a new directory, add something to the postlist, regenerate the blog, and run a single command to deploy.
I haven’t seriously taken a look at pandoc recently. Perhaps I should?
It’s definitely worth a look. I try to not depend on Haskell stuff in general, because it’s a rather large dependency and I still haven’t forgiven it for portability issues years back, but I make an exception for pandoc because it’s just that good.
That’s my experience as well. Jekyll is much simpler than Octopress in every aspect.
I wrote my own, mainly because I wanted it to be dynamic - just dump and view. I should really go back to refine it from what I’ve learned.