A Series on Generating this Static Website

The generation of this website is far from being trivial, and requires the combination of —probably too— many tools. For instance, my write-ups about Coq are actually Coq files, and I use coqdoc to generate the HTML pages you read. The theme is not written in CSS, but in SASS that needs to be compiled. Even more, the whole website is postprocessed using soupault.

At some point, I felt like the whole process was interesting enough so that it would deserve a write-up of its own, but the risk was that such a piece of text would quickly become out-dated. This is reminescent of documenting any software project, and I was aware at that time of a dedicated paradigm to prevent these kind of issues: literate programming. I spent quite some time turning my custom toolchain into a literate program, so that its actual code source would actually be the write-ups I wanted to add to my website. This was an interesting challenge, since it meant cleopatra would have to generate itself before it could build my website. In other words, cleopatra achieves the bootstsrapping challenge!

I really enjoyed this first experiment with literate programming, and I started using cleopatra for other projects of mine where literate programming felt like an interesting choice. In doing so, it quickly became clear cleopatra was cumbersome to set-up for a new project. At the end, I ended up rewriting it to overcome the specific issues posed by its initial design¹. But the so-called generation processes I had written for cleopatra the first basically “just worked” with cleopatra the second.

So, coming back to this series, it is just the very reason why I started using cleopatra in the first place: the generation processes used by this website, written as literate programs.