|
|
|
|
|
|
by c-cube
1113 days ago
|
|
|
You're moving the goalposts here. Rust is being used, in significant projects (eg: proxies at Cloudflare, a company where http is somewhat of a big deal: https://blog.cloudflare.com/introducing-oxy/ ; Dropbox's new sync algorithm, a company where file syncing id kind of a big deal : https://dropbox.tech/infrastructure/rewriting-the-heart-of-o...). Equating the existence of unsafe with C is laughable imho (it'd be barely comparable even if 100% of the rust code was in unsafe blocs, which never happens). Not even then it doesn't matter for the original point : rust is used in production for business critical functions, in large companies. |
|
|
Parts of systems written in a language doesn't really mean anything for its adoption into mainstream. For example, Amazon uses Ruby heavily for a bunch of deployment stuff, but Ruby (sans Ruby on Rails that is in decline), is not really a mainstream language any more.
>Equating the existence of unsafe with C is laughable imho
Im not comparing it. The point is to demonstrate that unsafe exists for the sole reason of performance. In fast code you often want to directly access x[y] where x and y are variables, without having to run extra code around it. Its a well known computer science thing, as most of the code challenges given in interviews rely on this access pattern for optimal solutions.
And because of Rice theorem, a compiler cannot determine whether x[y] is always safe, as determining all the values y could take would involve running the program.
So as such, for all the advantages that Rust offers, you can have the same advantages with C with macros and LLVM extensions, albeit with less concise syntax.
https://www.microsoft.com/en-us/research/project/checked-c/
Similar arguments were used to justify Haskel about 6-7 years ago, and Haskel is pretty much dead in the water at this point.
The modern way to make a memory safe language is to focus on a high level language that doesn't require programmer to deal with memory directly, and then work on the compiler to make the resultant code optimal.