New week, new Rust! What are you folks up to?
@llogiq I think itās week 15, not 14, at least according to ISO.
EDIT: I see itās obviously scripted and not a real account. Nevermind.
Iām testing some early LALR integration that Iād like to bring to the lexer/parser generator I announced here last week, which is currently LL(1). I toyed with the idea of LL(k) for a while, but Iām more interested in a more complementary option right now, though it probably wonāt offer as much flexibility.
Honestly still reading the rust book . I worked in java 9 to 5s for 7 years still doing one.
Just learning the language on weekends and nights.
I finished (I thinkā¦maybeā¦.) for now my thread-local allocators. The doc now starts:
Thread-local allocators.
Temp is for heap allocations that last a short time.
Local is for heap allocations that last a longer time ( but not longer than the thread ).
Temp and Local are not Send or Sync, so values allocated from them cannot be accssed by other threads.
Temp uses ābumpā allocation, deallocate just decreases a count of outstanding allocations. This means there is no minimum allocation internally. Allocations larger than 64K bytes, or having more than 128 byte alignment, are routed to Global.
Local has an array of free lists, one for each size class 16,32,64ā¦64K. The minimum internal allocation (on a 64-bit system) is 16 bytes, which is also the maximum alignment. Allocations larger than 64K bytes, or having more than 16 byte alignment, are routed to Global.
Implementing Sanqi (Chinese äøę£, SÄnqĆ, āThree Chessā), a two-player strategy game in which coordination between pieces plays a central role.
I published a typenum replacement called gnat this week, which can do everything that typenum can, but without operator trait bounds.
Polishing this article: https://medium.com/@amid.ukr/can-rust-have-zero-cost-dependency-injection-fc0c9ae6abd3
- Submitted to dzone for review.
- Converting to md format, to publish in dev.to
And some code to support idea with examples:

Working on Anima, a native macOS desktop app that extends Claude Code's /buddy companion system. Built with Tauri v2 and Rust.
Every project gets its own ASCII familiar, token usage earns in-app currency for re-rolls, and a Rust daemon polls all active sessions to surface observations about your work. 4MB binary, 43 tests, MIT licensed. First app I've ever shipped!
I have added UCI HAR and Isolet datasets as classification examples to the Hypervector crate I have been working on.
HAR is activity recognition from mobile phone sensors (Walking, sitting, ...) and Isolet is speech recognition (letters).
Both datasets are relatively small by today's standards - but still often used on kaggle and similar sites.
Hypervectors are fast and compact - useful for edge devices and for brain-inspired computing.