Hacker News Reader

Comment by tiffanyh | Hacker News Reader

What is the dominant flavor of Lisp these days?

My guess is Common Lisp has the most usage in industry, its just quiet compared to the latest dog food sharing app built in Clojure. Racket seems the most enthusiastic and fun community, and has some great learning tools. And Emacs Lisp secretly runs all of it.

I vaguely recall a story about some enterprise critical app that some madman hacked out in elisp because time was tight.

Steve Yegge’s customer support app for Amazon possibly

Might it be this article?

https://news.ycombinator.com/item?id=26531660

This one about german ATC running in elisp

https://news.ycombinator.com/item?id=26202517

More than disturbing that ATC software was written quickly because of tight constraints.

Yes that's the one. Thanks!

As measured by what? Clojure is pretty popular, as is Racket in other circles. Common Lisp is robust, standardized, and "decentralized" with both free and commercial implementations. It and Python power a few quantum computing technologies.

In addition to this, Guile is big in the GNU space, used for Guix and other things.

There was also sustained effort over the past two decades to move Emacs from Elisp to Guile but that has stalled:

https://emacsninja.com/posts/state-of-emacs-lisp-on-guile.ht...

And now thanks to the JIT patch elisp outperforms Guile considerably. Emacs dev is very much motivated by pragmatism, jokes about Emacs being a tool created to write Emacs config notwithstanding.

I think of emacs as a lisp machine with an editor attached.

You sure about that? Guile also has a JIT compiler these days and I don't know of any benchmarks comparing the two in a meaningful way.

No, Guile got a JIT too and elisp is not that fast.

I mean, I'm sure I could write a fast enough GB emulator with Guile.

Not with Elisp, even with the XCB or XLIB bindings for Elisp/Emacs.

My impression is that part of the problem is that Emacs Lisp is a large language (lots of data types and functions implemented in C) and one without any formal specification, which makes duplicating it correctly in another system very difficult. By contrast, Scheme is a smaller language and has a reasonably clear spec.

Which makes me wonder – instead of trying to port GNU Emacs to run on top of Scheme, might it be easier to produce a Scheme implementation that ran on top of Emacs Lisp? Possibly Emacs Lisp is missing some of the primitives required to implement Scheme efficiently, but the amount of code required to add those primitives will likely turn out to be less than that required to reimplement the whole of GNU Emacs in Scheme – and adding those new primitives may make Emacs Lisp a better language, thus benefitting even those users who aren't interested in Scheme in itself.

Guile has a generic VM and extensible compiler tower, so a bug for bug port of elisp was proving to be an effective method. Unfortunately, the project did not interest the Emacs devs so there just wasn't the buy-in necessary to work through the finer details and make it happen for real. You can open a Guile REPL today and switch the language from Scheme to Elisp and hack away. What makes the whole thing so appealing is that the Elisp compiler simply lowers the code to Guile's "tree intermediate language" (Tree-IL) and gets a free ride the rest of the way through optimization and bytecode emission. The biggest impact Elisp support had on core Guile was that a new immediate type had to be added to represent 'nil'. Soon it won't even be much of a stretch to compile Elisp to WebAssembly (would be funny!) once that backend is ready for Scheme. Guile is undervalued as a platform for language implementation and Emacs missed out on a cool opportunity.

I'm a user of one of those QC technologies (ARTIQ), but what's the CL one?

https://github.com/quil-lang/quilc

Didn't realize Rigetti was a CL shop. That's cool!

At least HRL Laboratories, Rigetti Computing, and D-Wave all use/deploy Common Lisp.

One measure is git pushes on GitHub. By that measure[0], in Q1 2023, we have Emacs Lisp (2995 pushes) > Clojure (2135) > Scheme (1350) > Common Lisp (236) > Racket (below detection; latest in Q1 2022: 102).

[0]: https://madnight.github.io/githut/

Github is not that popular among Lisp users.

It seems pretty popular nowadays in CL and Clojure. Scheme less so.

SBCL is on sourceforge, CLISP is on Gitlab, McCLIM is on Codeberg, Common-Lisp.net with its projects is on Gitlab, ...

Example: Microsoft owns Github and uses the hosted code to train its AI offerings.

https://www.theverge.com/2022/11/8/23446821/microsoft-openai...

https://www.osnews.com/story/136052/microsofts-github-copilo...

https://www.theregister.com/2023/06/09/github_copilot_lawsui...

It's not clear what "dominant" means in an already-niche language, but while Guile isn't very popular you still find little snippets of it everywhere from back when it was GNU's official extension language. Even GNU make embeds a guile interpreter, though I think most packagers turn that off at this point, sadly.

> while Guile isn't very popular you still find little snippets of it everywhere from back when it was GNU's official extension language

Your use of past tense here seems to imply it is no longer GNU's official extension language. GNU's own website still says [0] "Guile is the GNU Ubiquitous Intelligent Language for Extensions, and the official extension language of the GNU project."

[0] https://www.gnu.org/software/guile/

And it's all of the infrastructure in the official GNU Linux distribution, so I guess you're right, but it was supposed to be the scripting language for the GNU desktop, and to the extent that's still Gnome, that didn't happen. Also the Guile team itself keeps toying with writing a (shudder) Javascript front-end.

The ecmascript frontend is more of a toy than something you could write production code in, but it's a fun example of Guile's extensible compiler tower. No one has seriously worked on it in a long time.

I really wish Gauche was more popular. It's everything I wanted out of Guile and Chicken but never felt like they achieved as well.

Typescript?

Reply on news.ycombinator.com