progress

progress.txt

@@ -10,8 +10,56 @@ status:
       2) experimental implementation of polymorphic types and type inferencing
          values currently treated as white-box, hardcoded generic types (e.g., list, function) in type inferencing algorithm
 
-todo:
+wip:
   - interfaces via typeclasses?
 
+  - revise the way types are encoded
+     we need only one 'type of type' (called 'kind' in Haskell), and we can call it 'Type'.
+     types explicitly contain their underlying types. the type inferencing algorithm needs this information.
+
+      Int
+       { symbol: Int, params: [] }
+
+      [Int]
+       { symbol: List, params: [{ symbol: Int, params: [] }] }
+
+      [[Int]]
+       { symbol: List,
+         params: [{ symbol: List, params: [{symbol: Int, params: []}]}]}
+
+      Int -> Bool
+       { symbol: Function, params: [
+            { symbol: Int, params: [] },
+            { symbol: Bool, params: [] },
+         ]
+       }
+
+      Int | Bool
+       { symbol: Sum, params: [
+            { symbol: Int, params: [] },
+            { symbol: Bool, params: [] },
+         ]
+       }
+
+      (Int, Bool)
+       { symbol: Product, params: [
+            { symbol: Int, params: [] },
+            { symbol: Bool, params: [] },
+         ]
+       }
+
+      Point2D   <-- custom nominal type! maybe it contains two Doubles, but we don't expose this
+       { symbol: Point2D, params: [] }
+
+      Type constructors are just functions that return a 'Type'.
+         For instance:
+         lsType :: Type -> Type
+         fnType :: Type -> Type -> Type
+
+      The sad(?) part about all of this, is that I'm converging with Haskell/Lean.
+
+  - treat all values as polymorphic? (non-polymorphic values simply have empty set of type variables)
+
+todo:
   - type inferencing can be reduced to finding a graph isomorphism?