[fine] Gobs of work

- Assertion improvements
- Type check function calls
- Functions in the environment
- Compile function calls
This commit is contained in:
John Doty 2024-01-13 08:12:39 -08:00
parent 9d226b205d
commit 5ebede4a21
6 changed files with 539 additions and 142 deletions

View file

@ -57,7 +57,7 @@ impl fmt::Display for Error {
}
}
#[derive(Copy, Clone)]
#[derive(Clone)]
pub enum Type {
// Signals a type error. If you receive this then you know that an error
// has already been reported; if you produce this be sure to also note
@ -69,12 +69,17 @@ pub enum Type {
// everything's fine.
Unreachable,
// This is until generics are working
MagicPrintGarbage,
Nothing,
// TODO: Numeric literals should be implicitly convertable, unlike other
// types. Maybe just "numeric literal" type?
F64,
String,
Bool,
Function(Vec<Box<Type>>, Box<Type>),
}
impl Type {
@ -119,50 +124,98 @@ impl fmt::Display for Type {
F64 => write!(f, "f64"),
String => write!(f, "string"),
Bool => write!(f, "bool"),
MagicPrintGarbage => write!(f, "MagicPrintGarbage"),
Function(args, ret) => {
write!(f, "(")?;
let mut first = true;
for arg in args.iter() {
if !first {
write!(f, ", ")?;
}
write!(f, "{arg}")?;
first = false;
}
write!(f, ") -> {ret}")
}
}
}
}
#[derive(Clone, Copy)]
#[derive(Clone, Copy, Debug)]
pub enum Location {
Argument,
Local,
Module,
}
pub struct Declaration {
pub declaration_type: Type,
pub location: Location,
pub index: usize,
// TODO: Is `usize` what we want? Do we want e.g. dyn trait for invoke?
#[derive(Clone, Debug)]
pub struct ExternalFunctionId(usize);
impl ExternalFunctionId {
pub fn id(&self) -> usize {
self.0
}
}
pub enum Declaration {
Variable {
declaration_type: Type,
location: Location,
index: usize,
},
Function {
declaration_type: Type,
declaration: TreeRef, //?
},
ExternFunction {
declaration_type: Type,
id: ExternalFunctionId,
},
}
pub struct Environment {
pub parent: Option<EnvironmentRef>,
pub location: Location,
pub base_index: usize,
pub next_index: usize,
pub declarations: HashMap<Box<str>, Declaration>,
}
impl Environment {
pub fn new(parent: Option<EnvironmentRef>, location: Location, base_index: usize) -> Self {
pub fn new(parent: Option<EnvironmentRef>, location: Location) -> Self {
let parent_location = parent
.as_ref()
.map(|p| p.location)
.unwrap_or(Location::Module);
let base = parent.as_ref().map(|p| p.next_index).unwrap_or(0);
let next_index = match (parent_location, location) {
(_, Location::Argument) => 0,
(Location::Local, Location::Local) => base,
(_, Location::Local) => 0,
(Location::Module, Location::Module) => base,
(_, Location::Module) => panic!("What?"),
};
Environment {
parent,
location,
base_index,
next_index,
declarations: HashMap::new(),
}
}
pub fn insert(&mut self, token: &Token, t: Type) {
let index = self.base_index + self.declarations.len();
self.declarations.insert(
token.as_str().into(),
Declaration {
Declaration::Variable {
declaration_type: t,
location: self.location,
index,
index: self.next_index,
},
);
self.next_index += 1;
}
pub fn bind(&self, token: &Token) -> Option<&Declaration> {
@ -237,6 +290,24 @@ fn set_logical_parents(
}
}
}
TreeKind::FunctionDecl => {
// In a function declaration, the logical parent of the body is
// the parameter list.
let param_list = tree.child_of_kind(syntax_tree, TreeKind::ParamList);
let body = tree.child_of_kind(syntax_tree, TreeKind::Block);
for child in &tree.children {
match child {
Child::Token(_) => (),
Child::Tree(ct) => {
if Some(*ct) == body {
set_logical_parents(parents, syntax_tree, *ct, param_list);
} else {
set_logical_parents(parents, syntax_tree, *ct, Some(t));
}
}
}
}
}
_ => {
// By default, the parent for each child is current tree.
for child in &tree.children {
@ -270,7 +341,7 @@ pub struct Semantics<'a> {
errors: RefCell<Vec<Error>>,
types: RefCell<Vec<Incremental<Type>>>,
environments: RefCell<Vec<Incremental<EnvironmentRef>>>,
empty_environment: EnvironmentRef,
root_environment: EnvironmentRef,
}
impl<'a> Semantics<'a> {
@ -280,6 +351,15 @@ impl<'a> Semantics<'a> {
set_logical_parents(&mut logical_parents, tree, root, None);
}
let mut root_environment = Environment::new(None, Location::Module);
root_environment.declarations.insert(
"print".into(),
Declaration::ExternFunction {
declaration_type: Type::MagicPrintGarbage,
id: ExternalFunctionId(0),
},
);
let mut semantics = Semantics {
syntax_tree: tree,
lines,
@ -287,7 +367,7 @@ impl<'a> Semantics<'a> {
errors: RefCell::new(vec![]),
types: RefCell::new(vec![Incremental::None; tree.len()]),
environments: RefCell::new(vec![Incremental::None; tree.len()]),
empty_environment: EnvironmentRef::new(Environment::new(None, Location::Module, 0)),
root_environment: EnvironmentRef::new(root_environment),
};
// NOTE: We ensure all the known errors are reported before we move
@ -379,7 +459,7 @@ impl<'a> Semantics<'a> {
Incremental::Complete(e) => return e.clone(),
Incremental::InProgress => {
// NOTE: Set the state so the ICE doesn't loop on itself.
*state = Incremental::Complete(self.empty_environment.clone());
*state = Incremental::Complete(self.root_environment.clone());
drop(borrow);
//eprintln!("environment_of circular => {t:?}");
@ -394,12 +474,14 @@ impl<'a> Semantics<'a> {
let parent = match self.logical_parents[t.index()] {
Some(t) => self.environment_of(t),
None => self.empty_environment.clone(),
None => self.root_environment.clone(),
};
let result = match tree.kind {
TreeKind::LetStatement => self.environment_of_let(parent, tree),
TreeKind::FunctionDecl => self.environment_of_func(parent, tree),
TreeKind::ParamList => self.environment_of_paramlist(parent, tree),
TreeKind::File => self.environment_of_file(parent, tree),
TreeKind::Block => self.environment_of_block(parent, tree),
// TODO: Blocks should introduce a local environment if required.
// Test with a type error in a block statement and a
@ -426,6 +508,62 @@ impl<'a> Semantics<'a> {
result
}
fn environment_of_block(&self, parent: EnvironmentRef, tree: &Tree) -> EnvironmentRef {
let mut environment = Environment::new(Some(parent), Location::Local);
for child in tree.children.iter() {
match child {
Child::Tree(t) => {
let ct = &self.syntax_tree[*t];
if ct.kind == TreeKind::FunctionDecl {
// TODO: Should I have accessors for function decls?
let Some(name) = ct.nth_token(1) else {
continue;
};
environment.declarations.insert(
name.as_str().into(),
Declaration::Function {
declaration_type: self.type_of(*t),
declaration: *t,
},
);
}
}
_ => {}
}
}
EnvironmentRef::new(environment)
}
fn environment_of_file(&self, parent: EnvironmentRef, tree: &Tree) -> EnvironmentRef {
let mut environment = Environment::new(Some(parent), Location::Module);
for child in tree.children.iter() {
match child {
Child::Tree(t) => {
let ct = &self.syntax_tree[*t];
if ct.kind == TreeKind::FunctionDecl {
// TODO: Should I have accessors for function decls?
let Some(name) = ct.nth_token(1) else {
continue;
};
environment.declarations.insert(
name.as_str().into(),
Declaration::Function {
declaration_type: self.type_of(*t),
declaration: *t,
},
);
}
}
_ => {}
}
}
EnvironmentRef::new(environment)
}
fn environment_of_let(&self, parent: EnvironmentRef, tree: &Tree) -> EnvironmentRef {
let Some(name) = tree.nth_token(1) else {
return parent; // Error is already reported?
@ -440,35 +578,23 @@ impl<'a> Semantics<'a> {
None => Type::Error,
};
let (location, base_index) = match parent.location {
Location::Local => (
Location::Local,
parent.base_index + parent.declarations.len(),
),
Location::Module => (
Location::Module,
parent.base_index + parent.declarations.len(),
),
Location::Argument => (Location::Local, 0),
let location = match parent.location {
Location::Local => Location::Local,
Location::Module => Location::Module,
Location::Argument => Location::Local,
};
let mut environment = Environment::new(Some(parent), location, base_index);
let mut environment = Environment::new(Some(parent), location);
environment.insert(name, declaration_type);
EnvironmentRef::new(environment)
}
fn environment_of_func(&self, parent: EnvironmentRef, tree: &Tree) -> EnvironmentRef {
let Some(param_list) = tree.nth_tree(2) else {
return parent; // SE
};
let param_list = &self.syntax_tree[param_list];
if param_list.kind != TreeKind::ParamList {
return parent; // SE
}
fn environment_of_paramlist(&self, parent: EnvironmentRef, tree: &Tree) -> EnvironmentRef {
assert!(tree.kind == TreeKind::ParamList);
let mut environment = Environment::new(Some(parent), Location::Argument, 0);
for child in param_list.children.iter() {
let mut environment = Environment::new(Some(parent), Location::Argument);
for child in tree.children.iter() {
let Child::Tree(ct) = child else {
continue;
};
@ -529,12 +655,18 @@ impl<'a> Semantics<'a> {
TreeKind::ExpressionStatement => self.type_of_expression_statement(tree),
TreeKind::Identifier => self.type_of_identifier(t, tree),
TreeKind::FunctionDecl => self.type_of_function_decl(tree),
_ => self.internal_compiler_error(Some(t), "asking for a nonsense type"),
};
// NOTE: These return `None` if they encounter some problem.
let result = result.unwrap_or(Type::Error);
if result.is_error() {
eprintln!("OH NO AN ERROR AT {}: {:?}", t.index(), tree);
}
self.types.borrow_mut()[t.index()] = Incremental::Complete(result.clone());
result
}
@ -635,7 +767,10 @@ impl<'a> Semantics<'a> {
"string" => Some(Type::String),
"bool" => Some(Type::Bool),
"()" => Some(Type::Nothing),
_ => Some(Type::Error),
_ => {
self.report_error_tree(tree, "Unrecognized type");
Some(Type::Error)
}
}
}
@ -748,12 +883,64 @@ impl<'a> Semantics<'a> {
fn type_of_call(&self, tree: &Tree) -> Option<Type> {
assert_eq!(tree.kind, TreeKind::CallExpression);
Some(Type::Error)
let f_ref = tree.nth_tree(0)?;
let f = self.type_of(f_ref);
let arg_list = &self.syntax_tree[tree.nth_tree(1)?];
let arg_types: Vec<_> = arg_list
.children
.iter()
.filter_map(|c| match c {
Child::Tree(t) => Some((*t, self.type_of(*t))),
_ => None,
})
.collect();
if f.is_error() || arg_types.iter().any(|(_, t)| t.is_error()) {
return Some(Type::Error);
}
match f {
Type::Function(params, ret) => {
let mut any_errors = false;
if params.len() != arg_types.len() {
// TODO: Augment with function name if known
self.report_error_tree(tree, format!("expected {} parameters", params.len()));
any_errors = true;
}
for (i, ((t, a), p)) in arg_types.iter().zip(params.iter()).enumerate() {
if !a.compatible_with(p) {
self.report_error_tree_ref(
*t,
format!(
"parameter {i} has an incompatible type: expected {} but got {}",
p, a
),
);
any_errors = true;
}
}
if any_errors {
return Some(Type::Error);
}
Some(*ret.clone())
}
_ => {
self.report_error_tree_ref(f_ref, format!("expected a function type, got: {f}"));
Some(Type::Error)
}
}
}
fn type_of_argument(&self, tree: &Tree) -> Option<Type> {
assert_eq!(tree.kind, TreeKind::Argument);
Some(Type::Error)
let result = self.type_of(tree.nth_tree(0)?);
Some(result)
}
fn type_of_expression_statement(&self, tree: &Tree) -> Option<Type> {
@ -785,13 +972,42 @@ impl<'a> Semantics<'a> {
let id = tree.nth_token(0)?;
let environment = self.environment_of(t);
if let Some(declaration) = environment.bind(id) {
return Some(declaration.declaration_type);
return Some(match declaration {
Declaration::Variable {
declaration_type, ..
} => declaration_type.clone(),
Declaration::Function {
declaration_type, ..
} => declaration_type.clone(),
Declaration::ExternFunction {
declaration_type, ..
} => declaration_type.clone(),
});
}
self.report_error_tree(tree, format!("cannot find value {id} here"));
Some(Type::Error)
}
fn type_of_function_decl(&self, tree: &Tree) -> Option<Type> {
let param_list = tree.child_tree_of_kind(self.syntax_tree, TreeKind::ParamList)?;
let mut parameter_types = Vec::new();
for p in param_list.child_trees() {
let param = &self.syntax_tree[p];
// TODO: Missing type expression means it's a generic function.
let parameter_type = param.child_of_kind(self.syntax_tree, TreeKind::TypeExpression)?;
parameter_types.push(Box::new(self.type_of(parameter_type)));
}
let return_type = match tree.child_tree_of_kind(self.syntax_tree, TreeKind::ReturnType) {
Some(t) => self.type_of(t.child_of_kind(self.syntax_tree, TreeKind::TypeExpression)?),
None => Type::Nothing,
};
let return_type = Box::new(return_type);
Some(Type::Function(parameter_types, return_type))
}
pub fn dump_compiler_state(&self, tr: Option<TreeRef>) {
eprintln!("Parsed the tree as:");
eprintln!("\n{}", self.syntax_tree.dump(true));
@ -826,7 +1042,28 @@ impl<'a> Semantics<'a> {
let mut environment = Some(self.environment_of(tr));
while let Some(env) = environment {
for (k, v) in env.declarations.iter() {
eprintln!(" {k}: {:?}", v.declaration_type);
eprint!(" {k}: ");
match v {
Declaration::Variable {
declaration_type,
location,
index,
} => {
eprintln!("{declaration_type:?} (variable {location:?} {index})");
}
Declaration::Function {
declaration_type,
declaration,
} => {
eprintln!("{declaration_type:?} (function {declaration:?})");
}
Declaration::ExternFunction {
declaration_type,
id,
} => {
eprintln!("{declaration_type:?} (extern {id:?})");
}
};
}
environment = env.parent.clone();
}
@ -872,9 +1109,14 @@ pub fn check(s: &Semantics) {
let _ = s.type_of(t);
}
TreeKind::ArgumentList => {}
TreeKind::Argument => {}
TreeKind::Argument => {
let _ = s.type_of(t);
}
TreeKind::IfStatement => {}
TreeKind::Identifier => {}
TreeKind::Identifier => {
let _ = s.type_of(t);
}
TreeKind::ReturnType => {}
}
}
}