Growing a Language [pdf] (1998)
langev.com

・

45 points

・

ColinWright

・

5 days ago

12 comments

ColinWright ・ 5 days ago

This is the PDF of the talk given here:

https://www.youtube.com/watch?v=_ahvzDzKdB0

It's Guy Steele[0]'s keynote at the 1998 ACM[1] OOPSLA[2] conference.

[0] https://en.wikipedia.org/wiki/Guy_L._Steele_Jr.

[1] Association for Computing Machinery

[2] The 13th annual ACM SIGPLAN[3] Conference on Object-Oriented Programming Systems, Languages, and Applications.

[3] SIGPLAN is the Association for Computing Machinery's Special Interest Group (SIG) on programming languages.

  • kreetx ・ a day ago

    This is one of the talks I watch over and over - though it has computer science value then its entertainment value is much much higher still.

  • gjvc ・ a day ago

    The OOPSLA 1997 keynote was a banger as well. That was truly an exciting time.

senkora ・ a day ago

> An operator can be overloaded in C plus plus, but right now operators in the Java pro- gramming language can not be overloaded by the programmer, though names of methods may be overloaded. I would like to change that

It’s a shame that he lost this battle. Operator overloading really does make it a lot nicer to design certain kinds of libraries.

  • chubot ・ a day ago

    FWIW another thing that Guy Steele advocated was value types, which as I understand it is ongoing work:

    https://en.wikipedia.org/wiki/Project_Valhalla_(Java_languag...

    https://openjdk.org/projects/valhalla/

    I have said in the past, and will say now, that I think it would be a good thing for the Java programming language to add generic types and to let the user define overloaded operators.

    He did get the generic types though, I think that was well over a decade ago now

    • steveklabnik ・ a day ago

      > I think that was well over a decade ago now

      This was in Java 5, in 2004. So two decades!

    • roetlich ・ a day ago

      Or you could use C# and get all of those features.

  • swyx ・ a day ago

    i mean okay but also it makes it impossible to have an open source ecosystem with shared libraries because suddenly things dont mean the same thing in one system vs another. i was interviewing at Jane St and realized their OCaml completely doesnt work with the rest of the world - and its fine if you're Jane St i guess but that really sucks for the "health" of a language ecosystem.

    • Someone ・ a day ago

      Why would

        a + b
      suffer more from this than

        plus(a,b)
      ? In both cases, libraries will only clash if both define a function with the same name taking an A and a B. The only difference is that it is called + in the former case, and plus in the latter.
      • mojifwisi ・ 6 hours ago

        It's syntactically more straightforward to resolve the issue of clashing definitions through namespacing with functions compared to operators.

        The following is pretty standard:

            foo::plus(a, b) // or foo.plus(a, b)
    bar::plus(a, b) // or bar.plus(a, b)
        Whereas this is more awkward:

            a foo::+ b // or a foo.+ b
    a bar::+ b // or a bar.+ b