Since the type checker sucks less now, we can think about maybe
adding some features to the language.
...At some point I'd like to have the type checker clean up
the index map.
- Renamed "intrinsic" -> "primitive"
- I thought it was funny, but it's just annoying.
- Rename "Uninferred" -> "Inferred"
- It's a concrete type, but an inferred one.
- Add lifetimes to the Entry API
- Categorize does not care about recursive types. You can't have a recursive AST.
- Added helpful constructors to the inference engine
- Added some overloadable operators to the inference engine
- These are a MAJOR work in progress
- Reorganized the Inference implementation file by functionality.
stdlib:
- Updated standard library.
- Removed all unreferenced files
- Reimplemented missing/nonfunctional behavior
- Added module documentation for most things
- TODO: item-level docs on Entry(Mut)
- Reparented the stages of Table population into the `stage` module.
- TODO: rewrite type inference to use only the tools provided by Table.
- These implementations, like the types themselves, are/will be compiler-intrinsic. This allows the definitions of these functions to be (apparently) infinitely recursive.
- TODO: Implement these.
WARNING: The type checker is still a MAJOR work in progress. You'll be unable to ignore the stringly-typed error handling, effort duplication, and general poor code quality all around. However, this makes leaps and bounds toward a functional, if primitive, type checker.
Definitions now borrow their data from the AST, reducing needless copies.
- This unfortunately means the REPL has to keep around old ASTs, but... Eh. Probably have to keep those around anyway.
The "char" primitive type has been added. Semantics TBD.
Modules definition, type_kind, and value_kind have been consolidated into one.
Project now keeps track of the set of unnamed (anonymous) types (i.e. tuples, function pointers, references) and will yield them freely.
- Project can now also evaluate arbitrary type expressions via the EvaluableTypeExpression trait.
The NameCollector has been replaced with trait NameCollectable.
- This pass visits each node in the AST which can have an item declaration inside it, and constructs an unevaluated module tree.
The TypeResolver has been replaced with trait TypeResolvable and the function resolve()
- This pass visits each *Def* in the project, and attempts to derive all subtypes.
- It's important to note that for Items, unlike EvaluableTypeExpression, the ID passed into resolve_type is the ID of `self`, not of the parent!
typeck.rs:
- The Conlang "standard library" is included in the binary
- There's a main menu now! Type "help" for options.
- Queries have been upgraded from paths to full type expressions!
- Querying doesn't currently trigger resolution, but it could!
