I'm on a Bash spree lately, or as some might say, "going absolutely bonkers", "getting down with the sickness", "getting a few screws knocked loose"...whichever you prefer. I just really had a craving for, what I call, "the arts and crafts of programming", AKA gluing a bunch of Unix/CLI tools together until you have something that kind of works. Imagine sitting in an insane asylum all day and making art from macaroni, but you connect the macaroni (elbows) to each other so it forms a kind of pipe system. Or maybe you remember that screensaver from the 90's with the pipes -- that's basically writing Bash!

None of that makes any sense, but all you need to know is that I really wanted to read some RSS feeds and couldn't settle on a reader. They either seemed too complicated or just a pain to learn. I had used some of the suckless ones. I think elfeed is really nice and used that one for a while, even though I make fun of emacs all the time and don't even use it as an editor. I was thinking about it and I was able to break down what I needed into a few high level steps:

1. Download the feed files (XML)
2. Parse the XML
3. Store the parsed data
4. Display the necessary fields to the user
5. Have some minimal controls, like marking as read and opening in a browser

Immediately, I knew I wanted to use fzf for displaying and previewing the feed items, and providing some navigation and controls. XML is really annoying to me and an unintuitive format for querying data, in my opinion. This is why I decided quickly that I was going to write a small tool to parse XML into JSON, using the Rust rss. It uses minimal libraries and was only 34 lines to slap together. It was also really easy to support Atom alongside RSS and just move them into JSON as soon as possible. This will probably make the reader less accessible because now people will need to install this unpopular tool I made, which isn't widely available unless you have Rust installed.

With the feeds in a nice JSON format, I could use the powerful tool jq to not only extract what I needed, but to chunk up the feed into individual files per feed item. This made it easy to just store everything on the filesystem and avoid having to use a database. Each feed has it's own folder with read/ and unread/ subfolders containing JSON files per feed item. Upon startup jq will just parse all the .json files and feed the data to fzf.

That's pretty much the interesting part. The rest is just configuration and glue. It's better just to take a look at this demo:

After adding around 10 feeds with ~200 posts total, the performance is still decent, but you can tell that it isn't as snappy as other readers. This could probably be resolved with a sqlite database.

Another thing is the mechanism to load the posts. Right now, on startup the feeds are asynchronously downloaded and the user has to manual refresh for them to appear in the list. I'm not sure if there is a way to somehow "push" the posts in using a pipe or something. I don't think there is a way for them to maintain their order though, since fzf doesn't seem to have a way to keep things sorted.

frdr (feed reader)