Friday, March 8, 2019

State of play

I haven't made much progress since the start of Dec 18. Hopefully that will change.

I used to put minor stuff in wombatlang google+. That is going away and I'm not sure how to replace it. Maybe I'll post a "small stuff" post at times and put stuff in the comments. For the moment I'll use this post.

Because there is some discomfit with things having multiple types, I intend to use the word "Mode" instead (in deference to Algol68). For a given value, its Type will be the lowest Mode it can have in the type hierarchy (now "mode hierarchy"), without going into subset modes. Note that if you have 2 Modes with an isA relationship between them, but the lower one has no additional properties that aren't inherited from the higher one, then the lower one is regarded as a subset mode, since it is indistinguishable from one.

Nobody (except me) likes the Wombat name, so I'm going to change it. I rather like Emmy, in honour of one of the 20th century's greatest Mathematicians: Emmy Noether. Or I could look for a name that highlights some of the features of the language, such as: closure orientation, type/mode lattice, bidirectional typing and bidirectional execution driven by the type/mode lattice: BiCLat. Suggestions are welcome to robert.smart@wombatlang.org.

In https://golem.ph.utexas.edu/category/2019/03/left_adjoints_between_categori.html, John Baez makes a push to call "bounded lattice" just "lattice". I like this: there doesn't seem any reason for lattices that don't have a top and bottom, and these can be easily added if missing. However I probably will want sup/inf of more than finite sets: at most "computable sets". So somewhere between lattice and complete lattice. Anyway I'll just say "lattice" from now on.

4 comments:

  1. I tweet as @rks987. I'll put #wombatlang (or #emmylang after the name change) if its about the language.

    ReplyDelete
  2. They're discussing stuff relevant to my programming language ideas on the n-category blog. So I put my oar in, in case I get some useful ideas from the experts:
    https://golem.ph.utexas.edu/category/2019/03/left_adjoints_between_categori.html#c055534

    ReplyDelete
    Replies
    1. I'm very excited about this. Blog post coming. We will have Types, with values having only one type. Types form a partial order from the isA relationship. Then we have Modes which are up-sets of types. This is where subtyping happens (except that it is now submoding). A magic reconciliation between the desire for subtyping and the desire to have uniquely typed values.

      Delete
  3. A new document on the improved type system "Behaviours, Bases, Modes and Properties" (https://docs.google.com/document/d/1wJoLRP9bV70d7bL_JNF2pk7U0cvHQ1lNevDOI09FDG8/edit?usp=sharing)

    ReplyDelete