🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Agree with this assessment. Rust frameworks are far away from the convenience of something like springs ExceptionHandler advice (docs.spring.io/spring-framewo…), requiring everyone to reinvent the wheels.

How do y'all test your #rust #axum IntoResponse implementations? If I create a Response from serde_json::Value or Bytes I should be able to get the body and assert it in a unit test, without any async shenanigans, right?

Another pro-tip is that `[cfg(miri)]` implies a nightly rust version, so you can use unstable strict_provenance apis to silence warnings.

I will be talking at today's Rust Berlin Meetup about hash tables and hash codes: meetup.com/rust-berlin/ev…

If we don't keep raising awareness of low-level programming skills, the skillset will disppear forever under a sea of SDKs, binary blobs and proprietary chips. C programming matters, assembly language matters.

There is no such thing as an apolitical company. "No politics at work" means "we support the status quo and will enforce it by firing people who disagree".

Columnar file formats like Parquet/ORC are ubiquitous. Our VLDB paper with Xinyu Zeng + Huanchen Zhang + Wes McKinney studies their internals. TLDR: They're not optimized for modern hardware. Something new is needed. Paper: vldb.org/pvldb/vol17/p1… Code: github.com/XinyuZeng/Eval…

Branchless decoding of uleb128 encoded integers, as used in thrift and parquet, using bmi2 instructions: rust.godbolt.org/z/4cdPEGc41

It wouldn't be fun implementing a specification, if the existing implementations agreed with that spec. Case in point: Collections of boolean values in #apache #thrift github.com/jhorstmann/com…

Parsing parquet metadata is inherently sequential because of the thrift encoding, but this writeup shows that significant performance improvements are possible. Metadata should not be a limiting factory, compared to http/object store requests and loading the actual data.

#Zen5 desktop reviews are out! Time for another technical #AVX512 teardown. numberworld.org/blogs/2024_8_7…

Fascinated by how some people think boolean expressions are somehow special and not just returning a different data type. The example of `if x==y then true else false` is well known, but I recently talked with someone who did not believe ` SELECT x==y, COUNT(*)` to be valid sql.

Radix sort implementation using AVX-512: github.com/jhorstmann/rad… I've been sitting on this code for some years, since I could never get it to outperform the scalar implementation, on either tigerlake or zen4. A tweet by Robert Clausecker reminded me to clean it up and publish.

This looks super useful for writing complex declarative rust macros:

Experimented with different ways of reading from files in #rust, without having to initialize buffers. Benchmarks show that initialization (memset) adds significant overhead. AFAIK current stable rust has no sound way to do this for arbitrary readers. github.com/jhorstmann/rus…

And just in time for my last post, there is a blog describing the issue with initializing buffers and also problems with nightly BorrowedBuf in much more detail: mina86.com/2025/rusts-wor…

Small #rust quiz: For which `T` does `Arc::<T>::default()` NOT allocate? Assuming Rust 1.84

Answer: [T] if the alignment of T is <= 16, str and CStr. All these are backed by a shared slice and only require incrementing the reference count. github.com/rust-lang/rust…

GraphQL...not even once

Up to 45% faster #parquet writing by staring at a profiler for a while and changing ~100 lines of code. Previous title was overly optimistic, the main branch was missing one commit to make it a fair comparison.