33 ms·

I don’t see Go ever making headway in data science. For one, it’s not fast enough to implement algorithms directly but lacks a decent FFI; it would be burdensom

I don’t see Go ever making headway in data science. For one, it’s not fast enough to implement algorithms directly but lacks a decent FFI; it would be burdensome to create the metric ton of new libraries that would have to be put into place. And there are Python libraries for basically everything already, from numeric computing to Bayesian time series to Spark to Torch and TensorFlow. Plus Go’s kind of weird, relative to the languages that many data scientists have experience with. Maybe I’m being unfair, though.

IshKebab 5y ago
Go is definitely fast enough to do data science.
- mrtranscendence 5y ago
  Sure, it’s much faster than Python, agreed. But is it fast enough to implement highly optimized numerical algorithms in a way that can compete with Fortran or C? That I’m not so sure of, and if not you’re stuck with the crappy FFI.
  - IshKebab 5y ago
    Unless you need the last 10% of performance, yes it is. You can also use assembly directly from Go if you need bleeding edge speed.
    - azth 5y ago
      It's too verbose to do data science stuff.