People are allowed to have preferences. I have more fun writing Nim than writing Rust ¯\\_(ツ)_/¯.
Now let's address the rest: > Rust is safer Rust is less safe than Ada Sparks or Formally Verified C. There is a reason why on audit Rustls was recommended to switch to formally verified C primitives for cryptography: <https://github.com/rustls/rustls/blob/master/audit/TLS-01-report.pdf> Nim also has a borrow checker via `lent` and `owned ref` semantics. A borrow checker however only catches memory issues and data races. It doesn't catch design bugs or race conditions, you need formal verification for that (like Ada Spark). Hopefully in the future Nim gets the Z3 theorem prover support (see [DrNim](https://nim-lang.org/docs/drnim.html)) and we can have primitives with mathematical proofs of correctness: <https://github.com/nim-lang/RFCs/issues/222> By the way regarding safety, I don't understand the choice of default unsigned integer when those are a recipe for underflow bugs everytime you do a substraction, say `for a in 0..(vec.len - 2) {...}` > and faster I'm still waiting for: * Matrix multiplication in Rust faster than mine which achieve 97% to 102% throughput of almost pure Assembly OpenBLAS: <https://github.com/numforge/laser/blob/e23b5d6/benchmarks/gemm/gemm_bench_float32.nim#L393-L407> * state-of-the-art cryptography: <https://hackmd.io/@gnark/eccbench#Pairing> (This was before solving a Nim bug on parameter passing large objects) * Multithreading runtime for high-performance computing: <https://github.com/mratsim/weave> Nim is probably the best language to develop domain-specific language and new languages in. It is excellent at rapid development and time to market with it's scripting feel. Rust doesn't run on all hardware ISA as LLVM doesn't support all ISAs. All ISAs have a C compiler. > Write a Rust wrapper should be much easier than develop a pure Nim library So it should be easy for you to write an equivalent to my cryptography library? <https://github.com/mratsim/constantine>, pro tips you need: * a compile-time assembler because GCC and LLVM don't know how to generate add-with-carry and are 70% slower than assembly on big int multiplication: <https://github.com/mratsim/constantine#compiler-caveats> * Working "integer generics" and "enum generics". Rust still doesn't have this key feature working reliably. * Higher-Kinded Types I would also be interested to see a Rust high-performance computing and image manipulation library similar to <https://github.com/numforge/laser/tree/master/laser/lux_compiler/core> with codegen for parallel CPUs and GPUs and the possibility to setup vectorization (SSE, AVX, Neon), parallelization and other HPC goodies that can generate code at runtime or compile-time. Pro tip: you need AST manipulation at compile-time, <https://github.com/numforge/laser/blob/master/laser/lux_compiler/core/lux_types.nim> I would also be very interesting in a JIT compiler that doesn't do text interpolation for generating opcodes: * javascript in asmjit (<https://github.com/asmjit/asmjit/blob/2a706fd/tools/tablegen-x86.js>) * C++ in xbyak (<https://github.com/herumi/xbyak/blob/c79311a/gen/gen_code.cpp>) Instead in Nim you can do everything in a clean macro, with proper checks of types and arguments: <https://github.com/mratsim/photon-jit/blob/747ae2d/photon_jit/x86_64/x86_64_ops.nim>
