• kibiz0r@midwest.social
    link
    fedilink
    English
    arrow-up
    111
    arrow-down
    1
    ·
    1 year ago

    One of the most frustrating things about null is that it has so many possible meanings:

    • We don’t plan to provide a value here, so use a default instead
    • We plan to provide a value, but memory for this value hasn’t been allocated yet
    • The memory has been allocated, but we haven’t attempted to compute/retrieve the proper value yet
    • We are in the process of computing/retrieving the value
    • There was a code-level problem computing/retrieving the value
    • We successfully got the value, and the value is “the abstract concept of nothingness”
    • or the value is “please use the default”
    • or the value is “please try again”

    And so on. “Null” probably has more different meanings based on context than the word “fuck”.

    • Badland9085@lemm.ee
      link
      fedilink
      arrow-up
      19
      ·
      1 year ago

      Many of these meanings seem to be captured in some modern solutions already:

      • We plan to provide a value, but memory for this value hasn’t been allocated yet.
      • The memory has been allocated, but we haven’t attempted to compute/retrieve the proper value yet
      • We are in the process of computing/retrieving the value

      Futures?

      • There was a code-level problem computing/retrieving the value

      Exception? Result monads? (Okay, yea, we try to avoid the m word, but bear with me there)

      • We successfully got the value, and the value is “the abstract concept of nothingness”

      An Option or Maybe monad?

      • or the value is “please use the default”
      • or the value is “please try again”

      An enumeration of return types would seem to solve this problem. I can picture doing this in Rust.

      • david@feddit.uk
        link
        fedilink
        arrow-up
        7
        arrow-down
        1
        ·
        1 year ago

        Don’t call it a monad, call it a structured data type or something, that’s what it is! Calling it a monad is like saying that you’re using a curve of constant normal intersection point. Why not just say it’s a wheel?

        Yes, it’s mathematically true that you’re having a smooth ride precisely because the normals have a constant intersection point, but it’s also true to say that it’s a wheel and it goes round and isn’t bumpy and doesn’t scrape, and people can get a handle on that.

        So yeah, use a Result or Option or Maybe structured data type because it keeps explicit track of whether there’s a value or not, and yeah, you can change or combine them and preserve the tracking, but there’s no point calling it a monad unless you’re trying to make people believe that avoiding the $1bn mistake of allowing/using null requires category theory. It doesn’t, it’s just a structured data type. It’s simpler than an array! Stop calling it a monad.

        • Spzi@lemm.ee
          link
          fedilink
          arrow-up
          3
          ·
          1 year ago

          “Monad” is a shorter term though. “Structured data type” reads almost as bulky as “Curve of constant normal intersection points”.

          • david@feddit.uk
            link
            fedilink
            arrow-up
            2
            ·
            1 year ago

            True. But the word Monad has done more harm to the accessibility, popularity and reputation of pure functional programming than pretty much anything else.

            Yeah, I could have said circle rather than curve of constant normal intersection points, but that word is very commonly understood, so it’s not that same as unnecessarily calling something a Monad. Maybe it’s the equivalent of calling it a 2-manifold instead of a wheel.

            Perhaps just ditch the generalisation, then, and just call them Result or Maybe. After all, circle is a short word, but we just call them wheels.

      • Camilo@discuss.tchncs.de
        link
        fedilink
        arrow-up
        5
        ·
        edit-2
        1 year ago

        Yeah, for this reason null shouldn’t be part of any production code. If there’s the possibility of having a null value, you need to check every variable or returned value to be safe.

        These monads tell the consumer of your functions to do something (a check for emptiness or wait for it to be ready, or iterate it) to access the value inside. In a safe language, if the value is not wrapped by a monad, then you should expect to access it without issues.

        I kind of get why people don’t want to call them monads, since it sounds like a heavy term and more things to learn that are not strictly “necessary”, but the earlier you learn about their importance, the earlier you can use any of their benefits in your codebase.

    • unalivejoy@lemm.ee
      link
      fedilink
      English
      arrow-up
      8
      ·
      1 year ago

      Null was added to JavaScript because Java had it. Null is unnatural. Undefined is the canon “no value” value.