I’ve been thinking about programming languages a lot recently. A question I asked myself was: why do we work on, refine, and create new programming languages?
I thought of several reasons, but they seemed to boil down into two broader reasons:
Better abstractions and more automation: some languages automate and ease some tedious tasks like memory management, concurrency, or type annotations. Almost all languages give you some ways of creating abstractions that let you reason with concepts in your problem domain, but different languages do this in different ways.
I’m running macOS Catalina (10.15.5). Originally I tried using the pre-built packages via brew (brew cask install emacs) and those available at Emacs for Mac OS X. However, all these pre-built binaries crashed on Catalina. I guess it’s a problem with Catalina. 🙄
An analogy occurred to me this evening as I was thinking about programming language design:
Choosing good keywords and function names is like picking a good font; the ideas conveyed may be the same, but a change can drastically impact legibility and enjoyment of use.
PHP does a spectacular job of providing a bad example. It’s like the Comic Sans of programming languages. Now there are many reasons why PHP is not a good language—I’d like to investigate this particular aspect of its design here briefly.
Computers are funny things. At the lowest level they’re just a pile of ones and zeros that we assign meaning to. It’s something you can easily take for granted, but there’s a disconnect with how we talk about how things operate at the hardware level and then again at the software level.
Since writing a compiler, I’ve been able to bridge that gap in part. The fundamental idea is that we represent some meaning in a concrete, though still high-level form. Example: we might represent an entry in a contact book with a struct.
I’ve been building a compiler for a small lambda calculus that compiles to x86. It’s pretty broken, and I decided to start from scratch. I checked out a new branch in Git, and then deleted the entirety of my compiler before I had a chance to do anything else.
It hurt. But it was a good kind of hurt.
I don’t usually just blow everything away like that. Even this time, I’m keeping many of my auxiliary functions. I’m not rewriting the parser or some of my x86-helper routines. Those can stay.
I’m a FreeBSD guy. My first computer was a FreeBSD machine that my dad had running in a closet. I learned how to use Emacs as well as the command line on that black-screen white-text no-mouse interface. That’s how real programmers spend their childhood! 😎 😜
I’ve only heard good things about FreeBSD. While not known as particularly desktop-friendly (various Linux distros win here) I’ve heard tales of its rock-solid stability. I wanted to try running on FreeBSD again, just to see what all the fuss was about.
I’d like to say right off the bat that Helm is a great tool. I used it for several months and enjoyed it. Once thing that I love about helm is how discoverable it makes commands and functions. helm also got me into using bookmarks. I don’t keep many bookmarks; I tend to collect a few when working on a multi-file project long-term. The bookmark that I use most consistently is to my .emacs file; these days I’m fiddling constantly with my settings.
Today I figured out how to get Pandoc to automatically generate MLA citations for me!
I used Pandoc and the Biblatex bibliography format. What’s nice about this is that you can enter in all the information you know about the source, keep it nice and organized in a file, and then change the citation style on the fly. Imagine if you thought you had to use MLA, but then realized you needed to switch to APA citation styles. You can do that instantly with Pandoc and Biblatex.
Marked Man (mm) is a little program I wrote to view Markdown files like UNIX man pages. (Because who wants to leave their terminal just to open a file?)
It uses Pandoc to convert between Markdown and the groff format. As a happy side-effect, this program can read basically anything as a man page: HTML, LaTeX, Word files (seriously), ePub, etc. Anything that Pandoc can read, Marked Man can handle.