Conversation

HELLO I KNOW WHAT YOU'RE THINKING "I could really use a 40 page treatise on uninitialized memory, move semantics, and destructors. Hopefully it's an absolute feverdream." DO NOT WORRY, I HAVE CREATED IT FOR YOU gankra.github.io/blah/deinitial YOU ARE WELCOME NOW I MUST COLLAPSE

Destroy All Values: Designing Deinitialization in Programming Languages - Faultlore

28

177

736

best non-fiction I've read all year! seriously tho, I really enjoy your style, this is a lot of fun :D ⁽ᶦ ᵐᶦᵍʰᵗ ᶠᵘʳᵗʰᵉʳ ʳᵉᵖˡʸ ʷᶦᵗʰ ᵃⁿʸᵗʰᶦⁿᵍ ᶦ ᵗʰᵒᵘᵍʰᵗ ᵒᶠ ʷʰᶦˡᵉ ʳᵉᵃᵈᶦⁿᵍ⁾

1

6

made me look at boost::tribool, saw boost.org/doc/libs/1_59_ and, entirely predictably, immediatelly thought of: BOOST_TRIBOOL_THIRD_STATE(FileNotFound)

2

7

> Rust absolutely hates intrusive pointers (although there’s plenty of Fun Tricks to deal with it, the most common just being “store array indices instead of pointers into the array”) aww was expecting a Pin honorable mention (tho idk how sound that usecase is yet, so maybe not)

2

5

oops forgot about the Swift COW thing if anyone wants it in Rust, you're in luck! Rc::make_mut is your friend, have fun COW-ing (mooing?) Lean4 liked that trick so much they built an entire functional programming language (well, a proof assistant) on it: leanprover.github.io/talks/IFL2019.

2

5

Brendan Zabarauskas

Doesn't Koka use a similar trick? Or am I mis-remembering?

1

sorry, sounds right: microsoft.com/en-us/research heh, they even mention Lean4: > In particular, we build on the pioneering reference counting work in the Lean theorem prover [42], I must've forgotten about it, since I've found logs where

was clearly telling me about it, lol

1

2

Brendan Zabarauskas

Replying to

and

Nice! Yeah was checking as I've not really taken the time to understand it in my perpetual distraction. Seems neat though. koka-lang.github.io/koka/doc/book.

9:41 PM · Jan 24, 2022Twitter Web App

Replying to

and

I'm not seeing anything like Lean4's Array type (as flat as a Rust Vec AFAIK), they still seem to be using cons lists? so maybe they're not as far as Lean4 in exploiting ref-counting? or maybe they're just doing different things with it overall idk

1