Just released: github.com/tonbo-io/typ.... When working with arrow-rs, we noticed that schemas are declared at runtime. typed-arrow provides pure compile‑time Arrow schema declaration:
✅ No runtime errors
✅ No runtime overhead
Heavily using Rust macros and abstract type system.
Refactoring code with Cursor is akin to static program analysis followed by optimizations for performance or readability. It would be fantastic if agents could leverage runtime information in just-in-time automatic coding tasks.
LSM Tree pushes lots of works to background and improves data writing. However, it is not easy to create and manage background tasks in the browser, this limits the usage of LSM Tree based DB in the browser. We explored the practice of threads in WASM:
tonbo.io/blog/threads...
IMO tradisional Chinese is the real point-free style, let's define:
各(each): (a -> b) -> [a] -> [b]
加(+): int -> int -> int
一(one): int
于(at): .
相(fold0): (a -> b -> b) -> [a] -> b
相加于各加一 `fold0 (+) . map (+1)` is also a natural chinese sentence and a valid function, it shows the composability
Merry Christmas!!! and guess what's inside the box?
Simply listing the I/O requirements for building a log-structured DB on S3:
- Manifest: Stored as logs, requiring sequential writes.
- SSTable: Full-writes work fine with S3.
- WAL: Stored as logs, requiring sequential writes.
I really wish sequential writes were better supported.
A time-series database is essentially a HSAP database, as it needs to handle both real-time point queries for monitoring in observability use cases and offline range queries for tasks like historical analysis and AIOps training.
People always ignore that Box<dyn Trait> equals to Box<dyn Trait + 'static> in Rust, I ran into this issue again: github.com/apache/arrow...
🚀 In the next release, Tonbo will support browser deployment with WASM. As the first step, Fusio is taking a big step forward with OPFS support—perfect for those needing a virtual file system that bridges local disk and S3 access right in the browser. Dive in the preview here: t.co/KCM5jPS3jW
Fusio released 0.3.0, in which read and write semantics migrated to "exactly read at" and "truncate to 0 and append to". Compare with standard file op, random read and sequential write are more useful in DB scenarios and also suitable for io_uring. check it out: docs.rs/fusio/latest...
