RSH 17: Rust for science

Video

Collaborative notes taken during the session

Warm-up question: what do you like in a programming language?

• tooling, documentation, community +1
• available packages/libraries
• clarity of code written in it

Material used during the stream

• https://github.com/dev-cafe/rust-demo
• https://github.com/annefou/rust-examples

Rough outline

• About RSH
• Why Rust
• History of Rust
• Memory model
• Examples
• Running Rust on Binder
• Interfacing with other languages

Questions/comments/resources

History:

• http://venge.net/graydon/talks/intro-talk-2.pdf
• The History of Rust: https://www.youtube.com/watch?v=79PSagCD_AY

Tools written in Rust:

• exa, ripgrep, bat, and fd written in Rust (you really want these tools!).
• Great overview of tools written in Rust: https://www.wezm.net/v2/posts/2020/100-rust-binaries/

Questions:

• How good is the multi-dimensional array and plotting functionality in Rust? (numpy, matplotlib equivalents)

  • there are plotting libraries but what I have done until now is to plot in Python/R and let Rust to compute the csv data :-) not because Rust was bad at it but because I knew the Python/R better.

• Can you comment on where Rust stands right now with respect to the packages available that offer functionalities for scientific computing ?, for example, numpy, scipy equivalents, packages for machine learning, etc. Do you think it is sufficient for general scientific computing with what it provides out of the box? Assuming that it is a relatively new language, the packages/library will certainly grow over time but would like to know what is the current scenario.

  • this is not really answering the question but I find the library ecosystem surprisingly rich. Still it's not yet on par with numpy and others but then a nice combination can be Python on top, orchestrating the input and output and libraries and Rust underneath for the "heavy lifting".
  • Thanks, that gives an insight, especially since Rust is primarily a systems language as I understand as of now. So, I can imagine that one would most often not need all the packages/libraries which can instead come from python as you mentioned, and Rust can make the whole application execute fast and offer a programming environment that is memory safe.