Newsletter September 2019

Published on 1 October 2019.

This is what I’ve been up to in September 2019:

• I finished the article on parsing left associative operators in RLMeta. I’m pleased with how it turned out.

• I played with Cython to see if RLMeta VM could be made faster if converted to a C extension module. (It could!) I might write a blog post about it.

• I learned more about Python C extension modules from What are (c)python extension modules?.

• I watched A Look at the Design of Lua. I was inspired by the setting in which it seems to have been created. Small office. Whiteboard. Collaboration. I want to implement a minimal OOP-language in order to learn how to do it.

• I finished porting Timeline to Python 3.

  • The most difficult part was to verify that division operations were still working the same. (Python 3 integer division gives float instead of integer as in Python 2.)

  • I concluded that maintaining software is hard work. Even if your program does not change, the external world changes, and you need to change your program accordingly.

  • I finished the article on the segfault failure that I found during porting.

• I learned about a trick help me get started on a task faster: leave work undone. Leave a test failing. Leave a sentence half written.

• I learned that your number one priority is not what you write on top of the TODO list, but what you actually do.

• I finished a post with my notes on Alan Kay. He is a big inspiration to me.

• I read most parts of Common Systems Programming Optimizations & Tricks. Low level optimizations are interesting. It reminded me of Bob Nystrom’s game programming book that discusses some similar topics.

• I continued with the article about memoizing failures in RLMeta. It is almost finished, but because I got distracted by other work, it is not. I hope to finish it next month.

• I started working on compiling expressions to x86 machine code:

  • I don’t really know assembly, so I found A fundamental introduction to x86 assembly programming useful.

  • I got segafults because of 32/64-bit mixup. It just shows that I don’t really know what I’m doing yet.

  • I learned that add and imul instructions do not work the same. imul can not store result back in memory but must go via register.

  • I learned that I subtracted operands in the wrong order.

  • I read Writing a basic x86-64 JIT compiler from scratch in stock Python.

  • I read JIT compiling a subset of Python to x86-64.