I used to learn a lot of programming languages, just out of curiosity. It was something very fun to take have a new idea about a project I’d like to build, and use this as an opportunity to try a new language, a new setup, potentially a new paradigm. For instance, I learned Elixir this way, and I remember what impact it had on me. It was a brand new way of thinking about software development.
So I’m curious to hear about similar experiences. Which languages did you pick? Why? And how did it changed the way you see programming?
Golang because despite its flaws, that’s what we’re gonna write microservices in as an industry I guess. It’s fine, honestly it’s a good work language. A bit joyless to write. Feels designed to let people start working quickly but once you get it you don’t feel like you ever get better.
Same here - every time you want to try something fancy it slaps your hand and you end up writing it the boring simple way. Any time I’m back in typescript land I try something fancy… I arrive in hell and end up writing it the boring simple way anyway.
Not quite sure what to take away from this.
Agree with this. Unlike some other languages, there aren’t a lot of penny-dropping concepts - it does what it says on the tin
I have learned Golang like in 2012 I think, and never touched it again. So I was wondering if I would “learn Go a second time” at some point.
I was using Go around that point, and came back to it. I had to learn the module system, and the Context type being used everywhere, but otherwise it was entirely unchanged.
Generics are a big recent change. They did them well; they don’t make the language feel much bigger but it is much more expressive, and there are a lot of things I wouldn’t have considered using Go for before where I’d be open to it now.
I feel like I still get better at designing good APIs and interfaces in Go. And I keep making new tools and putting them into my toolbox for later.
I’ve been working on Kotlin lately. I always avoided the JVM so this has been eye opening to me. I came along this route by getting introduced to Intellij. Between the conveniences of IJ and the succinctness of Kotlin, i feel like kind of a sucker for avoiding this combo for so long.
Last week I learned some of the rudiments of Pure Data, which is a visual reactive programming language for creating music. Reason: curiousity (it is extremely different from other languages I used), and I want to learn more about reactive programming in the context of a pure reactive language (as opposed to a reactive library API in a conventional language).
In Pure Data, you write code by editing a running program, and you have immediate audio / visual feedback about your changes as you edit the program. This live coding experience is quite compelling.
Pure Data is a visual language in which you connect together nodes using wires. Data flows from one component to another over these wires. Based on screen shots alone, I thought I would hate this, compared to a textual language. The actual experience of programming using direct manipulation was a lot more compelling than I expected. The experience of dragging boxes and wires around made me feel like I was building things with my hands, which feels very different in my head from reading and reasoning about text, and I liked it.
This kinda sounds like my ideal environment for solving Advent of Code puzzles. I’ll check this out maybe this year.
That was almost two years ago, though. I should probably pick up another language.
I learned a new – style? paradigm? – of programming this year that basically feels like writing a new language. It’s a library for building stateful self-adjusting computation graphs with an API the likes of which I’d never seen before (a “split arrow” or “relative monad” interface). It’s still OCaml, but it was really eye-opening to see UIs modeled as graphs instead of trees. It’s so much nicer!
Bonsai seems extremely cool; I seriously considered using it for the frontend in my latest big project (tempest), But the compiled size of even hello-world app executables was enormous (like >10MiB), which was a dealbreaker :( I wonder if this just wasn’t an issue for Jane Street because they’re using it for internal tools and not pages randos on the internet are loading over possibly-slow connections.
Pony, because of the Actor approach to concurrency. (ponylang.io)
Still amazed that there a so few languages where concurrency is either not built in at all or just some form of async/await.
I semi-recently learned scheme and then C. Scheme because I was curious and it looked like “simpler lisp” which I had tried a bit with eLisp and it’d been fun. I liked how scheme felt “right” for my brain. I was able to rite very short and to the point functions that I could then compose into bigger functions to accomplish what I wanted, so if kind of matched the natural thinking “to do C and I to do B and to do be I need A”.
I decided to learn C not too long ago because frankly, I’ve just been avoiding it. It feels like I’ve tried everything around it just to not try it. Rust, Zig, LuaJIT FFI… but then when I decided to actually take a look I was just blown away!! C in itself is quite simple and easy to have the whole thing in your brain (C89/99 at least) but what blew me away was the ecosystem! EVERYTHING is written in C! And I could write and do absolutely everything and a bunch of ways! If I wanted to watch a file for changes, I could do anything from polling it myself, using a library, or just straight up talking to the kernel to do it for me. I’ve never felt so empowered and in control before with any other language.
pkg-config —list-allbecame a dear friend and a Pandora’s box to explore :)
So yeah, C has really changed how I code inasmuch and I now feel like o can write anything to run anywhere (I did some embedded dev, so much fun!)
That’s a pretty nice feedback, thanks for sharing! Out of curiosity, what editor setup did you use to hack in C? I wanted to try C lsp server at some point.
I dusted vscode. The C/C++ extension is pretty good, it uses the makefile to determine where to find the headers and thus it autocompletes from them, which was really all that I needed. So, not really lsp, but it was enough for me :)
For last year’s Advent of Code, I picked up Gleam. Having a lot of experience in Elixir, I wanted to see what a statically typed language for the BEAM felt like.
The ML-ish syntax was very easy for me to get used to, but ultimately I did not fall in love with the language. There were lots of opportunities to violate the type system and cause a runtime crash, which felt like the main advantage of Gleam evaporating before my eyes. And on the other side, I felt a lot of friction integrating with other BEAM code (like OTP). I didn’t try any Gleam in the browser.
I couldn’t help but wonder if Gleam were written in Gleam, if it would have fewer rough edges.
Perl. I used to use it for trivial glue and CGI; I’m right now learning more of the language in order to write a non-trivial program with it.
Why? Because I didn’t want to use Go or C, and I want to eventually support Plan 9.
TypeScript. I needed to write something that could run in a browser, but I was sick of JS from my last experience of it and I’m so over dynamically-typed languages in general.
I actually really like TypeScript. The only frustrating part is the tooling — NPM and browserify and all that crap. It makes CMake and C++ link errors look like child’s play.
Can I ask why you were using browserify? It has been generally considered subpar compared to other bundling tools for many years at this point.
Because I had just discovered that stuff that runs in node.js doesn’t automatically run in a browser, and googled about for advice and found way too much stuff and picked one.
If you’re open to looking again, vite is quite good & low config.
Ah. That makes sense. The persistence of outdated technical advice on Google is a bit of a problem. For what it’s worth, I concur with @marcellerusu that Vite is an excellent option.
Zig is the only one I have learned in-depth lately.
It’s the “better C” for me. It has taught me that dependent(ish, limited to compile-time values) types aren’t that scary, but very useful, and now I miss them elsewhere.
I learned a new language every year for Advent of Code (Haskell, Clojure, Rust) until I got sick of it and didn’t feel there was a point in racking up languages just for the hell of it. If I had wanted to go on, I’d probably learn APL, Racket, Forth, Julia.
The last language I ‘learned’, just very recently and reluctantly, has been Nixlang which is a disgusting amalgam of Haskell and Bash. There’s absolutely no point of learning this if not for the ecosystem behind it.
I learned LISP! (well, scheme) Because I have fallen in love with functional and it’s something I have always heard is powerful and beautiful, and it is! Probably one of my new favorites.
Hmm this post made me realize that I haven’t learned a language in a long time (by my standards, which include building something)
FWIW I no longer plan to use OCaml. I was investigating it as a language for implementing languages, but settled out on the “middle out” / language-oriented programming / DSL style instead
I was about to say I haven’t really learned a language since 2010 or 2014. (Python continues to be my go-to, jack of all trades tool)
But I forgot that I did learn PHP !!!
I always “hated” PHP because it was ugly and inconsistent compared to Python. But I like how PHP has a “web framework” builtin – the .php files just correspond to pages. And you don’t have to import libraries to parse web requests, or accept file uploads. What I can’t really get over is that string escaping seems to still be problematic without a template engine in PHP. If I’m going to use a template engine, it feels like I might just go back to Python.
I also wrote a web app in Python Flask around the same time.
So even though I appreciated many things about PHP, I don’t see myself using it again. It was more like a language experiment, to see how much “whipupitude” it has :)
Funny thing is that the next language I will learn may be Elixir. I bought an Erlang book in 2008, and thought it was cool, but I didn’t really have anything to build with it.
I’ve gone through some Elixir and Phoenix docs, and it looks cool, and I want to build some dynamic web sites, including some that have client-side progress indicators and so forth.
For someone interested in languages, I also find not knowing Haskell to be a bit of barrier, since it comes up a lot in papers and demos of languages. Ironically I think ChatGPT could help with the Haskell syntax
Here are some others I’ve looked at:
Go - I had to write some Go code for work, and I have read a lot about its language design. I definitely respect the design, but it doesn’t seem fun. It seems low on “whipupitude” :)
Rust - I’ve read a lot about it, but don’t have a project to justify really learning it. And I already know C++, which takes up a lot of head space. I’m impressed at anyone who “knows” BOTH C++ and Rust :)
SQL - I wrote SQL by hand in a toy project awhile ago. I still have mixed feelings – it’s powerful, but also really flawed.
HTML, CSS - I still write these by hand, and improved my knowledge, which I’m grateful will last
I like to read about other people’s nascent languages, e.g. research languages like Virgil
Hm I actually forgot that I did deeper dives into Awk and Make around 2016-2017, writing a few bigger Awk scripts and nontrivial Makefiles. I even read this whole GNU Make book.
Like the PHP experiment, it was kind of “research” for https://www.oilshell.org/
But as far as my personal toolkit, I don’t plan to use them much more (same with OCaml):
So I have been learning languages, but I tend to “eject” them from my toolkit once I see what’s good and bad. Make is DEFINITELY ejected – in favor of shell for “task files”, and Ninja for incremental builds.
In the last year I’ve really been focusing on R and Zig. R for classes & my grad research, hacking on my advisor’s R/C++ packages. I don’t particularly enjoy R, but I’ve become comfortable enough in it to (somewhat) read the insane incantations of veteran R statisticians. I’ve been learning Zig in an attempt to use it as a replacement for where I end up using C, and have recently started a compiler that uses Zig for small runtime components (similar to Roc). I find myself having a lot more fun in Zig than in C or C++, but Rust is still my go-to systems/low-level language.
Neither of those were my “latest” which goes to Mojo. I wrote a neural network from scratch a last year in Python, and I wanted to recreate it in Mojo to compare the languages. There were some issues I ran into, but I was able to get a small model working. Mojo caries over the mutable value semantics of Swift which I’m still trying to wrap my head around (the implicit copies were killing me before I added debug to every constructor I had).
As for “how did it changed the way you see programming”, I feel like every language has a bit to teach you. The compiler I started is for a language I’ve been designing for a while that takes inspiration from some of these and more. Zig’s comptime is incredibly eye opening in terms of what a language could provide for type manipulation. Mojo shows us a relatively simple language can achieve insane performance compared to standard C/C++ by utilizing better/different hardware and compiler architecture. R taught me…. well, I’m not sure what, but at least I can use ggplot well.
Clojure, for work - I wanted to work with some folks from the CHICKEN Scheme community I always enjoyed working with, and in the not-so secret hope that we’d be able to do a project in CHICKEN later (still on the original project and no new one in sight so far). I wrote up my thoughts on the language. I don’t have any language on my radar that I’d like to learn right now. Honestly I’m a bit burnt out on software stuff right now and don’t really have the energy to even think about learning something new.
Rust and typescript, around the same time, though that was a few years ago now.
Typescript fills the hole in my heart left by the death of coffeescript, and while the underlying JS is full of design flaws, the ES specs keep making incremental quality-of-life improvements that keep me happy. It’s high-level enough and fast enough for rapid iteration on things like servers (a MUD, an activitypub node, …) and toy web apps.
My day job and some side projects are firmware or other extremely low-level stuff, which is where rust comes in. TBH it has a lot of the same “too many features” problems of C++ and scala, but the benefits far outweigh that pain, and it’s straightforward to write code that’s absurdly fast and tiny.
I’m glad other people are enjoying golang, but I’m too opinionated and like modern affordances too much to really get into it. I can see that it fills the gap between high-level TS and low-level rust, but both languages are “good enough” for me, and I’m writing code mostly for myself.
Coq, though it was a bit ago; I’d mucked with it and other dependently typed languages, but the better part of a year ago was inspired to sit down and actually wortk all the way through software foundations volume 1 & a bit of volume 2, and finally felt like I grokked dependently typed programming & using proof assistants. I then also went a bit deeper into Idris than I had before.
I spend a lot less time learning languages than I once did, and mostly dive into it these days if I think a language is going to teach me something deeper.
My last one was Elixir, for work. My team already did Erlang and we needed something rest-api-like and Elixir+Phoenix just sounded a lot more fitting than doing the hard work in Erlang proper. I like it but I wouldn’t call it revolutionary, guess it depends what you worked with before.
Coming from Elrang, I guess Elixir is not that esoteric indeed, but I learned Elixir without any prior knowledge of Erlang. At the end of the day, I guess I was more impress by OTP than the language, though the macro system is pretty cool.
A while back, Elixir. It was a bit difficult to get hang of it as I didn’t have any experience with the paradigm and functional aspects. But oh boy, it did really change the way I look at things. Pattern matching along changed a lot of aspects on how I look at data.
Recently, Go. Seeing the need for it all job posts and people constantly asking about it in interviews decided to learn it. I think Go does well in many aspects where you just want something done quickly and don’t want to spend too much time on getting setup with the language. I can see now why people prefer it for business. Though, my main idea to pick this up is the job market.
I did try Zig, Odin, Rust and OCaml. I really want to keep doing OCaml, but I absolutely have no time or incentive on doing so. It’s still in pending.
Learning about databases on the side, so probably SQL also comes in at times.
I think most of us have been here. Especially when pattern matching feels like a first-class feature with exhaustiveness and not a janky
switchit really does change how you program and you don’t want to use languages without it.
Yes, exactly. And I have been struggling with this now
The languages that I seem to be having to use for work doesn’t have the great support for it. Ruby, recently added it so trying to get that working but nothing like in Elixir.
Go, because I felt stuck with supporting old apps written in Perl
The last language that I learned is Cython for work. It didn’t feel like learning a language from scratch, because it’s sorta Python, sorta C with Python syntax. There is not a lot of documentation, so sometimes I have to derive the semantics from reading the generated C++ code (we mostly target C++, because that gives us access to STL which is handy for containers).
I used Rust + PyO3 in the past, which I like a lot more, but Cython undeniably gives deeper integration with Python.
Manatee (https://cs.lmu.edu/~ray/notes/manatee/) because I like writing compilers and I hadn’t written one in a while and it looked like a fun one to write. When/if it’s finished I’ll post it here.
Golang, because I like https://charm.sh and decided to build a similar looking TUI.
Initially was not very fond of learning Golang. But even with downsides, it’s a pretty cool language to have under a belt.
Starlark, an embeddable dialect of Python, specifically the Golang implementation starlark-go.
I’m at the RIPE DNS Hackathon in Rotterdam. Our team is prototyping a mechanism for scriptable network measurements using the RIPE Atlas distributed probing system. At the moment Atlas supports a fixed collection of measurement types; scripting should make the measurements extensible, and allow data reduction on the probes.
For this purpose we wanted a scripting language that’s reasonably friendly for data scientists, and which can be sandboxed so that the Atlas probes don’t become a programmable botnet, and which has a host language that’s quick for prototyping.
Before that I learned Jai, in a more iterative learning process, because the language comes with examples and tutorials that go over the basics and less basics of the language. I managed to create a naive HTTP parser that I plugged into a naive socket based server.
Last languages learned and used:
brew); I love Homebrew, but Ruby not so much.
Currently kind-of learning (async mode):
Languages I’d like to get time to learn:
Just a note that Idris 2 is under active development and has a lively Discord community. https://github.com/idris-lang/Idris2
I had to really think about this because I start to learn something and then, you know, don’t (at least not in any meaningful way), but working my way back I think Lua was the last language I took the time to learn and actually use.
It’s used to write tools for Renoise, the DAW I use (along with Reaper) to create music Over the last few years I’ve written a number tools for myself to aid my workflow.
I’m using now to play around with Solar2d, a game engine.
The “indexing starts at 1” thing was a bit weird at first but you get used to it. :)
Can’t say it’s changed how I think about programming, though.
Last one I had to learn was last year with TypeScript. I hated it as a massive step down from the PureScript and Elm I had previously done a bulk of my work in. I was tasked to write in a functional style with the libraries that helped, but overall the ergonomics were just awful for FP, taking 4–10× the amount of code to express the same thing with another language. The cherry on top was a junior had to use some of the code I wrote & didn’t understand
Optionand just called
o.valuewhich rendered the whole exercise moot. If folks can slap Haskell on the server-side and realize they want FP on the client-side, they should recognize there are a wealth of options where the ergonomics and readability are better suited for a team that thinks within the FP reference.
Lately I’ve been reading and doing slow introductory work with OCaml now that multicore is here (while libraries catch up). I don’t fully grok it, but the goal is to get less lost in the ‘nice’ parts of the Haskell-realm type system and more grounded with modules (which Elm, despite it’s clear limitations, clicked better with me even if the boilerplate sucked). It helps that the compiler is super fast and the output binaries are efficient.
C++. It started when I applied to a random job offer, which listed the language as a plus. During the first interview, I was asked how much I knew about the language, to which I answered truthfully: not much; I believe it is mostly C, with classes. Boy, was I wrong! Thankfully, I didn’t get that job, but it made me curious to learn how C++ actually works, and what “Modern C++” means. I did get a different job since then, programming in C++20. I don’t know if it is the language, the topic, or even the team, but I am having a lot of fun with it! While it’s not perfect, I am not in a rush to move onto something else.