Demonstrate reputation calculation over Holochain data #4
Assignees
Comments
mhuesch
added a commit
that referenced
this issue
Apr 12, 2021
* add types module * beginnings of type inference! * add `env` module for typing environment sdiehl has a module which presents a subset of (the wrapped) `Data.Map` API. this is my take on that. it doesn't seems strictly necessary - we can probably just use `HashMap` directly, and perhaps it would be nicer to just do that, because it implements so many handy traits. I thought this would teach me a bit about API design and also about Rust trait stuff. we can rewrite later if we feel it's not worth its maintenance cost. * more on `infer` finish implementing the analogue of `Substitutable`. still seems like this approach is pretty good. I am getting schooled on Rust a bit which is fun. * cargo fmt this was long overdue because I was working on my Mac and it doesn't have `rustfmt` through the nix install for some reason... * eschew (zero cost) indirection: turn `Subst` into a type synonym seems cleaner - we don't get much from the newtype. * cargo fmt * wip: start on infer * wip 2: more `infer`, before realizing we need a diff approach pr#89 on the repo demonstrates that inference was broken. the new approach should work, which is excellent. * pivot: return `Vec<Constraint>` from `infer` as documented in pr#89 sdiehl/write-you-a-haskell#89 the original approach of the tutorial had an issue with `let` inference. the PR implemented a new approach which resolves the issue. this commit adapts the code to take that approach. * flesh out the body of `infer` for all match cases seems to be going smoothly ⛵ * implement `generalize` * implement most of the solver we are now only missing `unify_many`. * implement `unify_many` * finish type inference?? I have now implemented all the functions of `poly`'s `Infer` module. * example of inference - it fails to propagate constraints * rename `Decl` to `Defn` (lisp/scheme style) & define a parser for it had to lift out a bunch of common sub-parsers. turns out the type sigs are verbose but almost entirely the same from one to the next - only the arguments and return type really differ. * make a repl which parses & typechecks `Expr`s * setup for demo asciinema * fix bug in `App` type inferece I forgot to perform the substitution from the first unification. d'oh 🤦♂️ * repl: pretty print the scheme * fix import in test broken by adding `util::pretty` module.
|
this is subsumed by neighbour-hoods/social_sensemaker#8, although that will only demonstrate calculations over simple types. however, we are already tracking user-defined types in #54, and implementing that, plus neighbour-hoods/social_sensemaker#8 above, should subsume the entire issue here. closing. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Integrate the reputation DSL with some REA event data coming from a Holo-REA module and make calculation execution workable
The text was updated successfully, but these errors were encountered: