totally-safe-transmute VS tinyvec

For example, there is this (pure safe Rust) code: https://github.com/ben0x539/totally-safe-transmute/blob/main... which accesses external resources (/proc/self/mem) in order to violate the safety guarantees.

The first is the totally_safe_transmute crate. I mean, who wouldn't love library code that has .expect("welp") and .expect("oof") as its error handling? But that's not even the really scary part. Issue #2 ("i hate this") remains open to this day, but for obvious reasons there's no chance of resolution. This post has some context and a line-by-line explanation of how it works.

You mean like this?

You can also do that in rust on linux: https://github.com/ben0x539/totally-safe-transmute/blob/master/src/lib.rs

I want to correct this statement: Rust can be safer, but not if a library you use contains unsound code. Unsoundness is most often caused by unsafe code, but not always (totally_safe_transmute, anyone?). There is a misconception that unsafe code blocks are always unsound and should be avoided at all costs, but they're completely fine if the safety contracts are upheld. In fact, unsafe blocks isolate the potential issues to make it easier to identify where undefined behavior may be occurring. unsafe code blocks are a feature of the language, and their usage should not be viewed as opting out of any safety the language provides, imo.

Just reimplement totally_safe_transmute in Zig. No need for unsafe.

Why transmute() when you can totally_safe_transmute()?

Well, there is always the totally-safe-transmute.

They should use https://github.com/ben0x539/totally-safe-transmute

funny indeed. i changed all my projects to use tinyvec

I've seen the cost of zeroing memory be measurable, or maybe significant enough to care about, but I've never seen 90% of cycles spent on it, the only case I know of that gets close is creating an empty TinyVec versus an empty SmallVec with an inline buffer of 256 bytes. In my opinion that's an unreasonably large inline buffer. At inline buffers of 128 bytes and below, the overhead is less than 50%, and that's on a microbenchmark of the Default impl; the effect is rapidly diluted in a real program.

My point is that "implementation that doesn't use unsafe" is not necessarily always slower than "implementation that does use unsafe". Often people assume that this is the case, and it isn't. tinyvec currently beats smallvec in more than a few benchmarks. Not all, but some. And this sometimes visible to users. The point is that if you want speed, you don't necessarily need to give up any safety at all. Most differences in performance are due to the amount of effort or expertise that has been spent on the codebase, not the amount of unsafe in it.

https://github.com/Lokathor/tinyvec will definitely benefit, although not as much as something currently relying on typenum.