Where parallels cross

Interesting bits of life

Moldable Emacs: learning syntax from your editor!

Too long; didn't read

How moldable-emacs highlights and (tries to) explain syntax of code and English.

The problem

If you read my blog, you must have tried to learn a language in your life. It is hard. I bet that even some syntax of your own native language can be arcane to you! For spoken languages you rely on teachers because rules be pretty ambiguous (take English past tenses as an example: verbs don't follow the same rule). But programming languages should not require a teacher. At least not for their syntax. Indeed, the computer itself needs a translator program that teaches it how to consume a language. It is funny that we don't use the same program to teach us that language as well!

So I wondered: can moldable-emacs teach me syntax?

It is a problem indeed

This is a focal point of my work on moldable-emacs. Following the steps of Moldable Development, I want to make complex information easy to grok. Receiving support in understanding the syntax of a programming language is a first step. The second step is recognizing that an if-else block in a programming language is a pattern for choice. The most important bit is pattern.

If moldable-emacs can highlight patterns to users and explain that to them, we open an interesting path. At the end of the day, isn't recursion a pattern as well? We can define the syntax needed to define a recursion pattern in code. So I imagine that we can go down the rabbit hole and describe complex patterns to our peers via molds! For example, a poem is full of patterns (as poetry.el shows us!).

Indeed, there are many steps in front of us, but first things first.

And there is a solution

I will rely on tree-sitter and elit for this demonstration. Tree-sitter parses programming languages with plug-and-play grammars, while elit parses natural language with machine learning. The lovely thing is that I could explain both syntax with the same code because my molds produce the same data format.

Let me show how moldable-emacs can explain syntax.

I have a piece of C code and I produce an Org Mode buffer that allows me to visualize the syntax used. Then I swap to a Java file and again I can do the same! Finally, I highlight a piece of text and you can see I get to a similar view, with nice syntax explanations!

The English syntax has better description because the type of nodes are standard: Elit uses the Stanford notation of Natural Language Processing. I extracted the description (with a Playground mold) and used in my LearnSyntax mold. The idea is that if you define syntax, you should provide an explanation and examples. If you have those, the LearnSyntax mold can look beautiful.

In the case of tree-sitter, the grammar developers don't use a common standard and the best fallback for now is to have an Org Mode link that browse for the type of node and the language used.

All in all this is a powerful teaching tool. If you want to teach a programming language this can help your pupil (or yourself) visualize the syntax you are teaching.

By the way, to run this mold you just need an emacs-lisp buffer with a list of plists looking like:

((:type my-type :begin 1 :end 2 :mode some-mode :buffer "my-buffer"))

This format makes it easy to define your own syntax types: you get the highlighting for free then! (Generalizing the nice looking syntax description is a TODO for next time!)


Software can become self explaining. Even natural language can. We just have to refine our tools. Feel free to try this by installing moldable-emacs and running the "LearnSyntax" mold after another like "CodeAsTree".

Happy learning!