The benefit of being a father again (Freya my 3rd child, was born last week) is that while on paternity leave & between two baby bottles, I can hack on fun stuff.
A few months ago, I've built for my running club a Pelican-based website, check it out at : http://acr-dijon.org. Nothing's special about it, except that I am not the one feeding it. The content is added by people from the club that have zero knowledge about softwares, let alone stuff like vim or command line tools.
I set up a github-based flow for them, where they add content through the github UI and its minimal reStructuredText preview feature - and then a few of my crons update the website on the server I host. For images and other media, they are uploading them via FTP using FireSSH in Firefox.
For the comments, I've switched from Disqus to ISSO after I got annoyed by the fact that it was impossible to display a simple Disqus UI for people to comment without having to log in.
I had to make my club friends go through a minimal reStructuredText syntax training, and things are more of less working now.
The system has a few caveats though:
- it's dependent on Github. I'd rather have everything hosted on my server.
- the github restTRucturedText preview will not display syntax errors and warnings and very often, articles get broken
- the resulting reST is ugly, and it's a bit hard to force my editors to be stricter about details like empty lines, not using tabs etc.
- adding folders or organizing articles from Github is a pain
- editing the metadata tags is prone to many mistakes
So I've decided to build my own web editing tool with the following features:
- resTructuredText cleanup
- content browsing
- resTructuredText web editor with live preview that shows warnings & errors
- a little bit of wsgi glue and a few forms to create articles without having to worry about metadata syntax.
resTructuredText cleanup
The first step was to build a reStructuredText parser that would read some reStructuredText and render it back into a cleaner version.
We've imported almost 2000 articles in Pelican from the old blog, so I had a lot of samples to make my parser work well.
I first tried rst2rst but that parser was built for a very specific use case (text wrapping) and was incomplete. It was not parsing all of the reStructuredText syntax.
Inspired by it, I wrote my own little parser using docutils.
Understanding docutils is not a small task. This project is very powerfull but quite complex. One thing that cruelly misses in docutils parser tools is the ability to get the source text from any node, including its children, so you can render back the same source.
That's roughly what I had to add in my code. It's ugly but it does the job: it will parse rst files and render the same content, minus all the extraneous empty lines, spaces, tabs etc.
Content browsing
Content browsing is pretty straightforward: my admin tool let you browse the Pelican content directory and lists all articles, organized by categories.
In our case, each category has a top directory in content. The browser parses the articles using my parser and displays paginated lists.
I had to add a cache system for the parser, because one of the directory contains over 1000 articles -- and browsing was kind of slow :)
resTructuredText web editor
The last big bit was the live editor. I've stumbled on a neat little tool called rsted, that provides a live preview of the reStructuredText as you are typing it. And it includes warnings !
Check it out: http://rst.ninjs.org/
I've stripped it and kept what I needed, and included it in my app.
I am quite happy with the result so far. I need to add real tests and a bit of documentation, and I will start to train my club friends on it.
The next features I'd like to add are:
- comments management, to replace Isso (working on it now)
- smart Pelican builds. e.g. if a comment is added I don't want to rebuild the whole blog (~1500 articles)
- media management
- spell checker
The project lives here: https://github.com/AcrDijon/henet
I am not going to release it, but if someone finds it useful, I could.
It's built with Bottle & Bootstrap as well.