I have serious doubts about Python

My most recent side project is an Arduino Programming Language AST interpreter fully written in Python.

The only way I could get around finishing my Bachelor’s Thesis was framing it as a side project.

During this time using Python for a real world-ish medium sized project I have gone through multiple phases. All of them converge to being skeptical about modern computers’ lingua franca. Or it could just probably be a skill issue on my side.

My background

Around a year and a half ago my functional bro ¹ arc began, as I changed jobs and Elixir became the most used tool under my programmer belt. I fell in love with pattern matching, pipes and everything-is-an-expression. Currently I am fully sold on functional programming and I see it as a superior way of punching modern day cards.

I think of myself as a computer guy. I want to be able to build systems by telling computers what to do. I should be able to tickle transistors with a broad set of symbols. So around the time I was wondering if I should try to touch grass outside of the superior programming paradigm, I came across the oportunity of this project.

Reality check

Ok! So let’s build a scanner ²! I see that Python has a match statement and I also learn to not look for stuff outside of the official Python docs. Seriously, the top two pages for almost any search related to Python in DuckDuckGo and Google are full of SEO driven pages. Not cool.

Going back to the match statement. It’s a statement 😓. I have to rely on mutability if I want to do something with the result of my operations inside the plethora of cases. Oh wait, now I’m scared about the typing being dynamic. This did not happen with Elixir.

Type check

I guess I’ll have to learn type hints. I’ve seen Python code without them and I got lost pretty quick, I want to avoid that.

Bottom line: it’s cool but I feel like it’s missing a lot of inference power. The last time I’ve been doing types (not typespecs) it was with Rust and it felt effortless. With Python type hinting it feels like I’m typing just to hug myself and whisper to my ear that everything will be fine.

Metaprogramming

Writing an interpreter means you have to conform to a language specification. The Arduino Language Reference is lackluster at best. It does not state anywhere that the Arduino Language is a subset of C++. It does not explain what is a sketch, or what happens with code that your write outside top level functions. It does not explain if it supports classes. ³

Anyways, I have to implement many tests, a lot of them look like an input file and an expected output file:

Input file

int a = 2;
int b = a + 3;

Expected output file

In this case, we want to check that the parser does actually parse code correctly.

[
  VariableStmtSpec(
    var_type=TokenSpec(token=TokenType.INT),
    name=TokenSpec(token=TokenType.IDENTIFIER, lexeme='a'),
    initializer=LiteralExprSpec(
      value=TokenSpec(token=TokenType.INT_LITERAL, literal=2)
    )
  ),
  VariableStmtSpec(
    var_type=TokenSpec(token=TokenType.INT),
    name=TokenSpec(token=TokenType.IDENTIFIER, lexeme='b'),
    initializer=BinaryExprSpec(
      lhs=VariableExprSpec(
        vname=TokenSpec(token=TokenType.IDENTIFIER, lexeme='a')
      ),
      op=TokenSpec(token=TokenType.PLUS),
      rhs=LiteralExprSpec(
        value=TokenSpec(token=TokenType.INT_LITERAL, literal=3)
      )
    )
  )
]

Well, to run these tests with unittest you need to create dynamic tests. I did this creating a unittest.TestCase subclass and reimplement the magic runTest method. I still don’t fully understand it. I know it would have been much easier doing some metaprogramming wizardry in Elixir. ⁴

Closing thoughts

I feel like doing this project is teaching me a lot about programming languages. Not only I’m implementing one, but the host language is also showing me a different way of doing things. I still feel like Python is overrated for serious programming™.