Cake
  • Log In
  • Sign Up
    • I saw that Medium article a few weeks ago. I’m currently learning the programming language R for data science and it feels like in the past few years the trend has been to make things “plug and play” throughout the whole data science process: there are now utilities out there for building pipelines without a heavy dose of coding required. I know @ChrisJenkins is involved in this space so I’m tagging him in case he’d care to chime in. “Drag and drop coding” sounds like an intriguing idea: I remember writing SQL scripts in Oracle where you would do table joins by literally clicking on a visual display of their common keys, and the related code would auto-generate. But I think that you would still need a strong Computer Science background to “no code” effectively.

    • @StephenL nails the thing on the head. Visual, drag-n-drop programming is fine as long as you understand what is happening, and what will be happening with your program. As soon as the situation moves to a stage where you just connect some blocks of stuff and the resulting code is a black box, what you get is not a software product, but an unmaintainable piece of something. Which is usually not what you want to have, especially if you are going to deploy it and run it and maintain it.

      Unless we get benevolent general AIs doing all that maintenance in the background, it hardly flies for me.

    • The thing with all those point'n'click systems is that it works, but only if your use-case is exactly what the creators envisioned. As soon as you want to make a single step outside the lines, you're out of luck. It may seem like a minor or a trivial thing (move this out of this box and into that box), but if the support for that is not built in, forget it.

      And for any non-trivial system, you'll bump up against such a limitation sooner or later. That's why such systems are good for fooling around, trying stuff out and quick mockups. For anything else, you'll need a proper development platform.