olydis

So without anything else, we'd have to talk about programs in terms of "(t t) t ..." or binary trees, which gets unwieldy quickly. The first natural step, for practical purposes, is to allow definitions "foo = ...", then some syntactic sugar for lists, functions, etc. Ooops and now we have a "language". If you open the "Playground" page there's a summary of what exactly is TC and what is syntactic sugar (and really nothing more!) on top of it.

I like to think that the line is so blurry precisely because TC needs nothing but a bit of syntactic sugar to feel like a usable PL haha.

Comment · 1 year ago

Super cool! I think a (potentially animated) version of this would be an excellent addition for the website's "Specification" page :)

Comment · 1 year ago

Note that it refers to his book on the "Specification" page :)

> Seems to be a pattern for them sadly, not sure why

Can you elaborate? Agreed I could (and will) attribute more explicitly on the website, but the intention is in no way to grab credit. I just posted this reply for more background on everything: https://news.ycombinator.com/item?id=42375914

Comment · 1 year ago

Thanks for the feedback! The target audience I had in mind was certainly developers (like me), not "all people". And the wording was indeed inspired by PL talks and blog posts I consumed over the years.

Here is a (slightly provocative) thought: We developers were promised "first class functions" with functional programming languages. And it's true, you can pass them around like any other value! Cool. But first: What about inspecting those values? I can "look inside of" a number or string or array any day. But functions, I can only call. Huh, so an intensional view (not that anyone thinks that out loud) for all kinds of values, except functions. Yes sure, many languages do allow you do dig into functions. But it is all not the same as or as powerful as supporting it right down at the logic level! TC is also not first to do that. But IMO the most compact and practical calculus/language to do so, yet. Second, a practical example: We had "first class functions" for decades now. But where is our main stream config language (JSON etc) that has first class support for them? Of course the answer is: Because it remains tricky. In industry I've seen all sorts of work arounds, usually involving strings that get compiled/interpreted on the fly. Which usually means some amount of scary security, and no static guarantees whatsoever. With TC, a parser/runtime for arbitrary (but pure) functions is a few dozens lines of code. And thanks to being intensional, one can imagine type checking those functions at the time of loading the config, not only when executing! Concrete demo/blog post for exactly this is in the works.

So anyways, I do belive this enables truly fundamental shifts, hence "democratizing".

Comment · 1 year ago

Hi all, author of the website here (I'm https://johannes-bader.com/). Wow, thanks for the reactions and many good suggestions! I thought I'd add a bit of context here.

As has correctly been pointed out, Tree Calculus was developed by Barry Jay! The "Specification" page links to his book. And a preview of his PEPM 2025 paper (hi chewxy!) can now be found here: https://github.com/barry-jay-personal/typed_tree_calculus/bl...

Compared to how long Barry has been professionally researching this, I entered the picture yesterday and joined the effort to help. Potential mistakes on the website are mine, not Barry's, but I do firmly believe in some of the "ambitious" wording there. Blog posts and more concrete demos or details to come!

Just for the curious, here is my angle in all this:

I happened to (hobbyist!) research tiny, yet practical, languages for over a decade (see e.g. https://lambada.pages.dev/). In that work I started noticing that the Iota combinator (https://en.wikipedia.org/wiki/Iota_and_Jot) is not a combinator in the most traditional sense: It is usually defined as behaving like "\x x S K", but like, how can we just refer to a lambda in the definition of a logic? One could write reduction rules such as "Iota Iota x -> x", but now Iota appears on the left hand side, in argument position! Doesn't that allow reflecting on arguments? Horror! I just started realizing that, while SK is perfectly Turing complete and sane, forcing down the number of combinators from 2 to 1 magically forces some amount of intensionality! Curious, isn't it?

And that's when I came across Barry's work and book! A calculus that embraces intensionality. So I reached out to see if I can help. How can I be most useful? Barry has been figuring out theory, professionally, before I could think. So I felt like helping spread the word, maybe to a less academic developer crowd, would be a more effective way to contribute. I spent my entire career building developer tools, so I have some ideas what could be useful and what might be necessary to convince various kinds of developers. That's how the website came to be, and it is very much a work in progress. We sync regularly, for instance Barry had excellent ideas for demos and I suggested slightly tweaked reduction rules at some point (not more powerful in theory, just more convenient in practice), see "triage calculus" in typed_program_analysis.pdf