• 1 Post
  • 11 Comments
Joined 1 year ago
cake
Cake day: June 18th, 2023

help-circle




  • Whatever works for you. Just do it. It is convenient as f when you are just starting. You can always improve incrementally later on when (if) you encounter a problem.

    Too much noise/power costs to run a small thing - get a pi and run it there. Too much impct on your desktop performance - okay, buy a dedicated monster. Want to deep dive into isolating things (and VMs are too much of a hassle) - get multiple devices.

    No need to spend money (maybe sponsoring more e-waste) and time until it's justified for your usecases.


  • Better dependency control. I strongly prefer software that only depends on the stuff I can get from the package manager. This lowers the chance of supply chain attacks. Doesn’t prevent them, but I expect repo maintiners to do a better job looking at packages, than a developer who just puts another pip/gem/npm install in a dockerfile.

    Also if something is only available in a container, it sort of screams “this code is such a mess, we don’t even know a simple way to run it” to me.





  • Something like that, where I just write a function that spits out a numpy array or something like that and it gets plotted, would be great, but there is one thing Grafana can do and vega-altair, plotly and even matplotlib (*): a UI that allows to select a time interval to view.

    So I can freely pan/zoom in/out in time, and only the required part of the data will be loaded (with something like select ... where time between X and Y under the hood). So if I look at a single day, it will only load that day, and only if I dare to zoom out too much it will spend some time loading everything from the last year.

    (*) yes, you can do interactive things with matplotlib, but you don’t really want to, unless you must…