News for hackers. Stuff that matters.
Re:Implementing a simple object system from scratch in Ruby by block_dagger on Sunday December 15, 2024 @08:57AM (#42423394)
Interesting. Why the inconsistent use of hash rockets?
Re:Implementing a simple object system from scratch in Ruby by metacircu1ar on Sunday December 15, 2024 @09:09AM (#42423451)
Thanks for spotting this. The parts of code were written at different times. I'll fix this to standardize on :key => value form, as I want symbols to be clearly distinguishable both on call sites and in the method table.
Re:Implementing a simple object system from scratch in Ruby by alehlopeh on Sunday December 15, 2024 @09:15AM (#42423482)
Not that it necessarily applies here but any time I see hash rockets in a PR it just screams “an llm wrote this”
Re:Implementing a simple object system from scratch in Ruby by metacircu1ar on Sunday December 15, 2024 @09:22AM (#42423527)
I write some ruby code for my job to automate some API testing stuff, where I use hash rockets exclusively.
Re:Implementing a simple object system from scratch in Ruby by caseyohara on Sunday December 15, 2024 @10:01AM (#42423761)
You should know that using the hash rocket syntax has not been idiomatic Ruby since the new hash syntax was introduced in 2012. Unless your keys are not symbols, there’s no reason to use hash rockets.
Re:Implementing a simple object system from scratch in Ruby by VeejayRampay on Sunday December 15, 2024 @09:24AM (#42423539)
Ruby really is a powerful language

it might not be the fastest or the strictest, but man is it expressive

Re:Implementing a simple object system from scratch in Ruby by uticus on Sunday December 15, 2024 @09:38AM (#42423613)
You are correct technically, but the implication that you must sacrifice speed or static typing is practically incorrect: faster on startup than .NET [0]. You can get benefits similar to static typing in for a while now [1].

[0] tested locally on a Linux environment

[1] https://github.com/ruby/rbs#readme

Re:Implementing a simple object system from scratch in Ruby by neonsunset on Sunday December 15, 2024 @10:56AM (#42424084)
> faster on startup than .NET [0]

This is unlikely. When you invoke 'dotnet run' you're essentially going through the entire build system, it doesn't "just" launch the application - it also builds it or verifies that it doesn't need to.

The actual startup time for compiled applications is <100ms for JIT (sometimes even <50ms) and ~10ms for AOT.

Feel free to verify this with:

  dotnet new console --aot
  dotnet publish -o .
  hyperfine ./{name of the program}
(make sure to use up-to-date version of the SDK, can be installed with sudo apt install dotnet9 / brew install dotnet)

Of course performance of the code itself is going to have more than an order of magnitude of difference: https://benchmarksgame-team.pages.debian.net/benchmarksgame/...

Re:Implementing a simple object system from scratch in Ruby by igouy on Sunday December 15, 2024 @08:05PM (#42427124)
> [Ruby] ... faster on startup than .NET

Please share.

fwiw

C# aot in-process timing (0.029) about 0.005 faster than process timing.

Ruby yjit in-process timing (1.565) about 0.083 faster than process timing.

https://benchmarksgame-team.pages.debian.net/benchmarksgame/...