Published on 4 July 2021.
This is what I’ve been up to in June 2021.
I switched my main laptop and also upgraded it to use the latest version of Fedora.
That led me to revisited and tweak my dotfiles for the new setup.
One tool that I rely on heavily is rlselect, and it only works with Python 2. The new version of Fedora doesn’t ship Python 2 with all the libraries that rlselect needs. And Python 2 is deprecated anyway. That gave me the urge to port rlselect to Python 3. But I also wanted to write the port using a literate programming tool. I had previously thought about trying to use Smart Notes as a literate programming tool, but it lacked support now.
Because of my laptop upgrade, I had a new top priority for Smart Notes: implement some kind of literate programming support.
I started by adding a script to convert RLiterate (my previous attempt at a literate programming tool) documents to Smart Notes documents. That way I could convert my projects that use RLiterate to Smart Notes so that I could continue editing them from within Smart Notes.
I converted the Smart Notes RLiterate document to a Smart Notes document. Smart Notes was now written in itself. But there was no way to edit the code notes created by the convert script from within Smart Notes. The only way to edit the code was by editing the generated files. But if I did that, the code in the Smart Notes document and the generated files would be out of sync. So I decided to write a script that would examine changes in the external files and merge them back into the Smart Notes document. I got the idea that this was possible by reading about a similar approach used in Leo.
I got the merge script working well enough. I was able to edit smartnotes.py and merge the changes back into smartnotes.notes. So that when smartnotes.py was generated from smartnotes.notes it would include the changes.
I used this workflow to add support to Smart Notes to create and edit code notes. At this point I was no longer in need of the merge script. I could do all the editing from within Smart Notes. However, as I talked about in the previous update, such a merge script would be crucial for working with others not using Smart Notes. And I believe it will also come in handy to make editing external files from within Smart Notes more convenient. For example, it would allow a whole class to be edited even though that class is split into multiple code notes. When the whole class is saved, the merge script will merge changes back into the corresponding code notes.
At this point I could use Smart Notes as a literate programming tool to create a new version of rlselect that worked with Python 3.
I started by importing all the existing code into a new Smart Notes document. I did that by creating a minimal set of code notes. At this point, the code notes were large and didn’t really benefit from the literate approach.
Then I started making changes to make rlselect work with Python 3 and got it working quite easily.
The next step was to split the code notes into smaller chunks to be able to annotate and document smaller portions of the code. For that, Smart Notes needed some additional features. At least to make the process convenient.
I added support in Smart Notes for splitting code notes and tried it on the Smart Notes document itself. I didn’t have time to use this feature to improve the rlselect document, but I plan to.
I also added virtual links so that code notes that are related are automatically linked.
This process of dogfooding is really useful and fun. It answers the question very clearly what to work on next.
To see a demo of the literate programming features in Smart Notes, have a look at 5:14.
I thought about how the literate programming in Smart Notes feels different from the literate programming in RLiterate. In Smart Notes you don’t create a single linear document. You connect pieces of information that are related to each other. It will therefore also be difficult to extract a single linear document from Smart Notes. So this kind of literate programming will not result in “a book” that is the whole application with source code explained. It will rather be annotated source code. Linear documents might help with understanding sometimes. So I’m not quite sure how it fits into the Smart Notes literate programming approach. One possibility is to generate documentation files using regular code notes. For example, a README.md could be generated by assembling code notes from Smart Notes. But that README.md would not contain the whole program with documentation. Just a small part. Perhaps that is ok. By using Smart Notes as my main literate programming tool, I will experiment and see what it feels like.
I am really happy with the progress I made on the literate programming support in Smart Notes. The script to convert RLiterate documents to Smart Notes documents and the script to merge changes from external files back into Smart Notes documents were killer features that allowed me to progress very fast.
Aside from programming, I also continued to learn about my new hobby of videography:
I learned more about how to properly use my camera when it comes to white balance and exposure. I also ordered white balance cards to help me with that.
I tried to figure out how to use different frame rates and shutter speeds to eliminate flicker from light sources. I found a good tool from RED to list good combinations: flicker free video. I don’t fully understand the logic behind it, but at least I can now get rid of some flicker in practice.
I tested what happens to sound clips when cut in Kdenlive vs. in Ardour. Ardour adds short fade ins and fade outs, but Kdenlive doesn’t. So probably cuts in Kdenlive can create unwanted sounds if they don’t match up.
I thought that my Rode Vide Mic NTG sounded noisy. I tried a firmware update, but it didn’t help. I tried various methods of recording, and all the results were similar. Perhaps I just need to be closer to the microphone so that the main audio is much louder than the background noise. The other option would be to reduce background noise in post.
This time last year I was working on projectional editing support for RLiterate. I am no longer convinced that projectional editing is a good idea. It might just be that I am very comfortable with my current tools (Vim + command line tools) that I feel frustrated when I can’t use them. Or it might be that I haven’t used any good enough projectional editing system. My focus has at least shifted to build tools that are more practical and comfortable for me to use.
This time last year I also read about compilers. It is a topic that I am still interested in, and I hope to work more and write more about compiler related stuff in the future.
What is Rickard working on and thinking about right now?
Every month I write a newsletter about just that. You will get updates about my current projects and thoughts about programming, and also get a chance to hit reply and interact with me. Subscribe to it below.