Huh. This has not been my experience, rather the other way around.
I do see a whole heap of benchmarks that contradict it, so I probably should put in some effort to find out why, or if I'm just plain wrong... But at a guess? Guile doesn't have Racket's startup time penalty, and most of what I do is IO bound - and Guile's IO story is better than Racket's.
For the onlookers, this bunch of microbenchmarks suggests Racket is faster than Guile, but the fastest is Chez (which Racket uses as a substrate).
Not bad at comparing compilers and optimisation. Great for checking R7RS compliance. Not what I'd trust to make decisions with.
I'm a complete outsider, cannot find that blog post any more, and also just not invested in the language at all, so I'm hesitant draw conclusions about the validity of any of the accusations that were thrown back and forth at the time, but it seems pretty obvious to me that it's inevitable that the community will end up being split after an event like that.
i was coming from the CL side of the isle, i hadn't known his work prior to that.
* The recent Guile work on WASM is promising. (Note also Jens Axel Soegaard's recent work on WASM with a Racket-related compiler.)
* Racket's rehosting atop Chez seems like a good idea, and I'd guess that the Racket internals are now easier to work with than Guile's.
* Racket has done a lot of great work, and is a nice platform for people who can choose their tools without worrying about employability keywords for their resume. It made some missteps for broader adoption when it had a chance, and several of the most prominent industry practitioner contributors left.
* Racket still has the best metaprogramming facilities, AFAIK. But even more important than `syntax-parse` and `#lang`, one thing I'd really like from Guile and other Schemes is to support Racket's module system.
(I really like the ability to define Racket submodules inline, in fragments, for things like embedded unit tests <https://docs.racket-lang.org/overeasy/> and embedded API docs <https://docs.racket-lang.org/mcfly/>.)
(I also wanted to play with Racket's module system for PL research compilers: having early compiler implementation for a new language first expand into Scheme code, and then later (with submodules) also do native/VM code generation, while keeping the option to still expand to Scheme code (for better development tools, or for when changing the language). For example, imagine targeting a microcontroller or a GPU.)
* Right now, any Scheme is for people who don't have to do techbro/brogrammer interviews. The field has been in a bad place for awhile, professionalism-wise, and the troubled economy (and post-ZIRP disruption of the familiar VC growth investment scams) and the push to "AI" robo-plagiarism (albeit with attendant new investment scams) are suddenly making the field worse for ICs.
Maybe, I couldn't say, but I find Chez's source code very cryptic and hard to read. More so than any other Scheme implementation.
I wrote that Guile "greatly lags" Racket.
This is a different implementation of Guile, though. Has Hoot (on, say, V8) been benchmarked?
A few problems remain though. A good debugger, a good macro expander (geiser in Emacs is able to expand somehow), and solving the issues with R6RS library syntax and standard library bindings, are what comes to mind.
Racket's multi-core abilities for a long time were mostly heavy weight (places, starting whole new Racket VMs), except for their implementation of futures, but that one was not always useful. I think recently the situation in Racket has improved, but I don't know, whether it is as good as Guile fibers and futures (which are different from futures in Racket).
Except that ecosystem Guix provides is not on Windows nor MacOS, making a serious limitation to anyone who wants to develop guile on those platforms.
I support GNU's mission in general, but I find it ironic that when they push freedom of choice you're forced to make the "right" choice or you're left twisting in the wind.
I am asking, because I might want to get more familiar with it, if it is great for something like GUI or web app. I know there is Guile Golf, but to me it feels unfinished, and I experienced segfaults and crashes, when not doing things exactly as shown in examples and using standard GTK procedures, so that one is kind of out for me, and guile-gi is still changing and seemed kinda complicated to use. On the web front, there is Artanis, but it is too opinionated for me, not a minimalistic framework. Still can build things on top of guile fibers concurrent web server, but then have to build lots of things myself.
Recently, I used Python and tkinter, which works really well, and I would love to have good bindings for tk in Guile.
On my search for Scheme and GUI, I also found stklos, but I have no idea how ready it is and what kind of libraries I would have available, if I built an application with it.
Racket of course also has a GUI framework, that I have tried to use in the past, but it does not have a treeview widget. Has listbox though, with multiple columns.
What does Gauche have to offer, specifically for application development? (GUI, or web, or other)
It got so bad I moved to racket as my daily driver.
It has a graphical experience similar to the survivors from Lisp is Great days, like LispWorks and Allegro Common Lisp, or Clojure with Cursive.
I have my doubts, given its age.
Overall understanding of the code, code generation capabilities and ability to explain are all pretty good.
As agents become the dominant code writers, the top concerns for a “working class” programming language would become reducing errors and improving clarity. I think that will lead to languages becoming more explicit and less fun for humans to write, but great for producing code that has a clear intent and can be easily modified without breaking. Rust in its rawest form with lifetimes and the rigmarole will IMO top the charts.
The big question that I still ponder over: will languages like Hoot have a place in the professional world? Or will they be relegated to hobbyists, who still hand-type code for the love of the craft. It could be the difference between having a kitchen gardening hobby vs modern farming…
I'm focusing on very high-quality feedback from the compiler, and sandboxing via WASM to be able to safely iterate without human intervention - which Hoot has as well.
- Very small methods that function as standalone compilation units, enabling extremely fast compilation.
- Built-in, fast, and effective code browsing capabilities (e.g., listing senders, implementors, and instance variable users...). This makes it easy for the agent to extract only the required context from the system.
- Powerful runtime reflectivity and easily accessible debugging capabilities.
- A simple grammar with a more natural, language-like feel compared to Lisp.
- Natural sandboxing
Some links to start off with:
https://github.com/feenkcom/gt4llm
https://omarabedelkader.github.io/ChatPharo/
Edit: I suppose the next step would be to teach an LLM about "moldable exceptions", https://arxiv.org/pdf/2409.00465 (PDF), have it create its own debuggers.
Seems a bit much. How large would the todo example be?
Hoot is primarily an ahead-of-time compiler, not an interpreter, so this REPL demo has a lot of stuff that a production binary wouldn't have like a macro expander, runtime module system, and an interpreter. A significant chunk of extra code. We have a todo list example [0] in a separate repo, btw. On my machine, todo.wasm is 566K uncompressed, 143K gzipped. And that includes a virtual DOM diffing algorithm implemented in Scheme.
That said, there is quite some room to optimize binary size. For example, we know we're generating too many local variables in functions [1], many of which could be optimized away with an additional compilation pass. We expect that to have a significant impact on binary size. There's also certain workarounds we're doing due to features missing from Wasm at the time of implementation that add to total binary size. Lack of native stack switching for continuations is one example. Adopting the stack switching proposal should shave off some bytes.