hm-inference #6

Merged
sb10q merged 136 commits from hm-inference into master 2021-08-19 11:46:50 +08:00
3 changed files with 224 additions and 200 deletions
Showing only changes of commit 99276c8f31 - Show all commits

View File

@ -1,6 +1,6 @@
use crate::location::Location; use crate::location::Location;
use crate::typecheck::typedef::Type;
use crate::top_level::DefinitionId; use crate::top_level::DefinitionId;
use crate::typecheck::typedef::Type;
use rustpython_parser::ast::Expr; use rustpython_parser::ast::Expr;
#[derive(Clone, PartialEq)] #[derive(Clone, PartialEq)]

View File

@ -27,7 +27,7 @@ pub enum TopLevelDef {
// ancestor classes, including itself. // ancestor classes, including itself.
ancestors: Vec<DefinitionId>, ancestors: Vec<DefinitionId>,
// symbol resolver of the module defined the class, none if it is built-in type // symbol resolver of the module defined the class, none if it is built-in type
resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>> resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>>,
}, },
Function { Function {
// prefix for symbol, should be unique globally, and not ending with numbers // prefix for symbol, should be unique globally, and not ending with numbers
@ -47,7 +47,7 @@ pub enum TopLevelDef {
/// rigid type variables that would be substituted when the function is instantiated. /// rigid type variables that would be substituted when the function is instantiated.
instance_to_stmt: HashMap<String, (Stmt<Option<Type>>, usize)>, instance_to_stmt: HashMap<String, (Stmt<Option<Type>>, usize)>,
// symbol resolver of the module defined the class // symbol resolver of the module defined the class
resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>> resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>>,
}, },
Initializer { Initializer {
class_id: DefinitionId, class_id: DefinitionId,
@ -83,7 +83,6 @@ pub struct CodeGenContext<'ctx> {
pub loop_bb: Option<(BasicBlock<'ctx>, BasicBlock<'ctx>)>, pub loop_bb: Option<(BasicBlock<'ctx>, BasicBlock<'ctx>)>,
} }
pub fn name_mangling(mut class_name: String, method_name: &str) -> String { pub fn name_mangling(mut class_name: String, method_name: &str) -> String {
// need to further extend to more name mangling like instantiations of typevar // need to further extend to more name mangling like instantiations of typevar
class_name.push_str(method_name); class_name.push_str(method_name);
@ -171,24 +170,31 @@ impl TopLevelComposer {
} }
/// already include the definition_id of itself inside the ancestors vector /// already include the definition_id of itself inside the ancestors vector
pub fn make_top_level_class_def(index: usize, resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>>) -> TopLevelDef { pub fn make_top_level_class_def(
index: usize,
resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>>,
) -> TopLevelDef {
TopLevelDef::Class { TopLevelDef::Class {
object_id: DefinitionId(index), object_id: DefinitionId(index),
type_vars: Default::default(), type_vars: Default::default(),
fields: Default::default(), fields: Default::default(),
methods: Default::default(), methods: Default::default(),
ancestors: vec![DefinitionId(index)], ancestors: vec![DefinitionId(index)],
resolver resolver,
} }
} }
pub fn make_top_level_function_def(name: String, ty: Type, resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>>) -> TopLevelDef { pub fn make_top_level_function_def(
name: String,
ty: Type,
resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>>,
) -> TopLevelDef {
TopLevelDef::Function { TopLevelDef::Function {
name, name,
signature: ty, signature: ty,
instance_to_symbol: Default::default(), instance_to_symbol: Default::default(),
instance_to_stmt: Default::default(), instance_to_stmt: Default::default(),
resolver resolver,
} }
} }
@ -207,7 +213,7 @@ impl TopLevelComposer {
pub fn register_top_level( pub fn register_top_level(
&mut self, &mut self,
ast: ast::Stmt<()>, ast: ast::Stmt<()>,
resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>> resolver: Option<Arc<Mutex<dyn SymbolResolver + Send>>>,
) -> Result<Vec<(String, DefinitionId, Type)>, String> { ) -> Result<Vec<(String, DefinitionId, Type)>, String> {
match &ast.node { match &ast.node {
ast::StmtKind::ClassDef { name, body, .. } => { ast::StmtKind::ClassDef { name, body, .. } => {
@ -221,7 +227,8 @@ impl TopLevelComposer {
params: Default::default(), params: Default::default(),
}); });
let mut ret_vector: Vec<(String, DefinitionId, Type)> = vec![(class_name.clone(), DefinitionId(class_def_id), ty)]; let mut ret_vector: Vec<(String, DefinitionId, Type)> =
vec![(class_name.clone(), DefinitionId(class_def_id), ty)];
// parse class def body and register class methods into the def list // parse class def body and register class methods into the def list
// NOTE: module's symbol resolver would not know the name of the class methods, // NOTE: module's symbol resolver would not know the name of the class methods,
// thus cannot return their definition_id? so we have to manage it ourselves? // thus cannot return their definition_id? so we have to manage it ourselves?
@ -229,52 +236,49 @@ impl TopLevelComposer {
// application to be used to build symbol resolver? <- current implementation // application to be used to build symbol resolver? <- current implementation
// FIXME: better do not return and let symbol resolver to manage the mangled name // FIXME: better do not return and let symbol resolver to manage the mangled name
for b in body { for b in body {
if let ast::StmtKind::FunctionDef {name, ..} = &b.node { if let ast::StmtKind::FunctionDef { name, .. } = &b.node {
let fun_name = name_mangling(class_name.clone(), name); let fun_name = name_mangling(class_name.clone(), name);
let def_id = self.definition_list.len(); let def_id = self.definition_list.len();
// add to unifier // add to unifier
let ty = self.unifier.add_ty(TypeEnum::TFunc(crate::typecheck::typedef::FunSignature { let ty = self.unifier.add_ty(TypeEnum::TFunc(
crate::typecheck::typedef::FunSignature {
args: Default::default(), args: Default::default(),
ret: self.primitives.none, ret: self.primitives.none,
vars: Default::default() vars: Default::default(),
})); },
));
// add to the definition list // add to the definition list
self.definition_list.push( self.definition_list.push(TopLevelDefInfo {
TopLevelDefInfo {
def: Self::make_top_level_function_def(fun_name.clone(), ty, None), // FIXME: def: Self::make_top_level_function_def(fun_name.clone(), ty, None), // FIXME:
ty, ty,
ast: None // since it is inside the class def body statments ast: None, // since it is inside the class def body statments
} });
);
ret_vector.push((fun_name, DefinitionId(def_id), ty)); ret_vector.push((fun_name, DefinitionId(def_id), ty));
// if it is the contructor, special handling is needed. In the above // if it is the contructor, special handling is needed. In the above
// handling, we still add __init__ function to the class method // handling, we still add __init__ function to the class method
if name == "__init__" { if name == "__init__" {
self.definition_list.push( self.definition_list.push(TopLevelDefInfo {
TopLevelDefInfo {
def: TopLevelDef::Initializer { def: TopLevelDef::Initializer {
class_id: DefinitionId(class_def_id) class_id: DefinitionId(class_def_id),
}, },
ty: self.primitives.none, // arbitary picked one ty: self.primitives.none, // arbitary picked one
ast: None, // it is inside the class def body statments ast: None, // it is inside the class def body statments
} })
)
// FIXME: should we return this to the symbol resolver?, should be yes // FIXME: should we return this to the symbol resolver?, should be yes
} }
} else { } // else do nothing } else {
} // else do nothing
} }
// add to the definition list // add to the definition list
self.definition_list.push( self.definition_list.push(TopLevelDefInfo {
TopLevelDefInfo {
def: Self::make_top_level_class_def(class_def_id, resolver), def: Self::make_top_level_class_def(class_def_id, resolver),
ast: Some(ast), ast: Some(ast),
ty, ty,
} });
);
Ok(ret_vector) Ok(ret_vector)
}, }
ast::StmtKind::FunctionDef { name, .. } => { ast::StmtKind::FunctionDef { name, .. } => {
let fun_name = name.to_string(); let fun_name = name.to_string();
@ -284,14 +288,14 @@ impl TopLevelComposer {
self.unifier.add_ty(TypeEnum::TFunc(crate::typecheck::typedef::FunSignature { self.unifier.add_ty(TypeEnum::TFunc(crate::typecheck::typedef::FunSignature {
args: Default::default(), args: Default::default(),
ret: self.primitives.none, ret: self.primitives.none,
vars: Default::default() vars: Default::default(),
})); }));
// add to the definition list // add to the definition list
self.definition_list.push(TopLevelDefInfo { self.definition_list.push(TopLevelDefInfo {
def: Self::make_top_level_function_def( def: Self::make_top_level_function_def(
name.into(), name.into(),
self.primitives.none, self.primitives.none,
resolver resolver,
), ),
ast: Some(ast), ast: Some(ast),
ty, ty,
@ -462,18 +466,23 @@ impl TopLevelComposer {
} }
} }
pub fn parse_type_var<T>(
input: &ast::Expr<T>,
pub fn parse_type_var<T>(input: &ast::Expr<T>, resolver: &dyn SymbolResolver) -> Result<Type, String> { resolver: &dyn SymbolResolver,
) -> Result<Type, String> {
match &input.node { match &input.node {
ast::ExprKind::Name {id, ..} => { ast::ExprKind::Name { id, .. } => resolver
resolver.get_symbol_type(id).ok_or_else(|| "unknown type variable identifer".to_string()) .get_symbol_type(id)
}, .ok_or_else(|| "unknown type variable identifer".to_string()),
ast::ExprKind::Attribute {value, attr, ..} => { ast::ExprKind::Attribute { value, attr, .. } => {
if let ast::ExprKind::Name {id, ..} = &value.node { if let ast::ExprKind::Name { id, .. } = &value.node {
let next_resolver = resolver.get_module_resolver(id).ok_or_else(|| "unknown imported module".to_string())?; let next_resolver = resolver
next_resolver.get_symbol_type(attr).ok_or_else(|| "unknown type variable identifer".to_string()) .get_module_resolver(id)
.ok_or_else(|| "unknown imported module".to_string())?;
next_resolver
.get_symbol_type(attr)
.ok_or_else(|| "unknown type variable identifer".to_string())
} else { } else {
unimplemented!() unimplemented!()
// recursively resolve attr thing, FIXME: new problem: how do we handle this? // recursively resolve attr thing, FIXME: new problem: how do we handle this?
@ -486,8 +495,8 @@ pub fn parse_type_var<T>(input: &ast::Expr<T>, resolver: &dyn SymbolResolver) ->
// class B(Generic[A.A.T]): // class B(Generic[A.A.T]):
// pass // pass
} }
}, }
_ => Err("not supported".into()) _ => Err("not supported".into()),
} }
} }

View File

@ -1,8 +1,11 @@
use crate::typecheck::{
type_inferencer::*,
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
};
use rustpython_parser::ast;
use rustpython_parser::ast::{Cmpop, Operator, Unaryop};
use std::borrow::Borrow; use std::borrow::Borrow;
use std::collections::HashMap; use std::collections::HashMap;
use rustpython_parser::ast::{Cmpop, Operator, Unaryop};
use crate::typecheck::{type_inferencer::*, typedef::{FunSignature, FuncArg, TypeEnum, Unifier, Type}};
use rustpython_parser::ast;
pub fn binop_name(op: &Operator) -> &'static str { pub fn binop_name(op: &Operator) -> &'static str {
match op { match op {
@ -61,12 +64,17 @@ pub fn comparison_name(op: &Cmpop) -> Option<&'static str> {
} }
} }
pub fn impl_binop(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, other_ty: &[Type], ret_ty: Type, ops: &[ast::Operator]) { pub fn impl_binop(
if let TypeEnum::TObj {fields, ..} = unifier.get_ty(ty).borrow() { unifier: &mut Unifier,
_store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
ops: &[ast::Operator],
) {
if let TypeEnum::TObj { fields, .. } = unifier.get_ty(ty).borrow() {
for op in ops { for op in ops {
fields.borrow_mut().insert( fields.borrow_mut().insert(binop_name(op).into(), {
binop_name(op).into(),
{
let other = if other_ty.len() == 1 { let other = if other_ty.len() == 1 {
other_ty[0] other_ty[0]
} else { } else {
@ -75,18 +83,11 @@ pub fn impl_binop(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, othe
unifier.add_ty(TypeEnum::TFunc(FunSignature { unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty, ret: ret_ty,
vars: HashMap::new(), vars: HashMap::new(),
args: vec![FuncArg { args: vec![FuncArg { ty: other, default_value: None, name: "other".into() }],
ty: other,
default_value: None,
name: "other".into()
}]
})) }))
} });
);
fields.borrow_mut().insert( fields.borrow_mut().insert(binop_assign_name(op).into(), {
binop_assign_name(op).into(),
{
let other = if other_ty.len() == 1 { let other = if other_ty.len() == 1 {
other_ty[0] other_ty[0]
} else { } else {
@ -95,153 +96,167 @@ pub fn impl_binop(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, othe
unifier.add_ty(TypeEnum::TFunc(FunSignature { unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty, ret: ret_ty,
vars: HashMap::new(), vars: HashMap::new(),
args: vec![FuncArg { args: vec![FuncArg { ty: other, default_value: None, name: "other".into() }],
ty: other,
default_value: None,
name: "other".into()
}]
})) }))
});
} }
); } else {
unreachable!("")
} }
} else { unreachable!("") }
} }
pub fn impl_unaryop(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, ret_ty: Type, ops: &[ast::Unaryop]) { pub fn impl_unaryop(
if let TypeEnum::TObj {fields, ..} = unifier.get_ty(ty).borrow() { unifier: &mut Unifier,
_store: &PrimitiveStore,
ty: Type,
ret_ty: Type,
ops: &[ast::Unaryop],
) {
if let TypeEnum::TObj { fields, .. } = unifier.get_ty(ty).borrow() {
for op in ops { for op in ops {
fields.borrow_mut().insert( fields.borrow_mut().insert(
unaryop_name(op).into(), unaryop_name(op).into(),
unifier.add_ty(TypeEnum::TFunc(FunSignature { unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty, ret: ret_ty,
vars: HashMap::new(), vars: HashMap::new(),
args: vec![] args: vec![],
})) })),
); );
} }
} else { unreachable!() } } else {
unreachable!()
}
} }
pub fn impl_cmpop(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: Type, ops: &[ast::Cmpop]) { pub fn impl_cmpop(
if let TypeEnum::TObj {fields, ..} = unifier.get_ty(ty).borrow() { unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: Type,
ops: &[ast::Cmpop],
) {
if let TypeEnum::TObj { fields, .. } = unifier.get_ty(ty).borrow() {
for op in ops { for op in ops {
fields.borrow_mut().insert( fields.borrow_mut().insert(
comparison_name(op).unwrap().into(), comparison_name(op).unwrap().into(),
unifier.add_ty(TypeEnum::TFunc(FunSignature { unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: store.bool, ret: store.bool,
vars: HashMap::new(), vars: HashMap::new(),
args: vec![FuncArg { args: vec![FuncArg { ty: other_ty, default_value: None, name: "other".into() }],
ty: other_ty, })),
default_value: None,
name: "other".into()
}]
}))
); );
} }
} else { unreachable!() } } else {
unreachable!()
}
} }
/// Add, Sub, Mult, Pow /// Add, Sub, Mult, Pow
pub fn impl_basic_arithmetic(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: &[Type], ret_ty: Type) { pub fn impl_basic_arithmetic(
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ unifier: &mut Unifier,
ast::Operator::Add, store: &PrimitiveStore,
ast::Operator::Sub, ty: Type,
ast::Operator::Mult, other_ty: &[Type],
]) ret_ty: Type,
) {
impl_binop(
unifier,
store,
ty,
other_ty,
ret_ty,
&[ast::Operator::Add, ast::Operator::Sub, ast::Operator::Mult],
)
} }
pub fn impl_pow(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: &[Type], ret_ty: Type) { pub fn impl_pow(
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ unifier: &mut Unifier,
ast::Operator::Pow, store: &PrimitiveStore,
]) ty: Type,
other_ty: &[Type],
ret_ty: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::Pow])
} }
/// BitOr, BitXor, BitAnd /// BitOr, BitXor, BitAnd
pub fn impl_bitwise_arithmetic(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) { pub fn impl_bitwise_arithmetic(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_binop(unifier, store, ty, &[ty], ty, &[ impl_binop(
ast::Operator::BitAnd, unifier,
ast::Operator::BitOr, store,
ast::Operator::BitXor, ty,
]) &[ty],
ty,
&[ast::Operator::BitAnd, ast::Operator::BitOr, ast::Operator::BitXor],
)
} }
/// LShift, RShift /// LShift, RShift
pub fn impl_bitwise_shift(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) { pub fn impl_bitwise_shift(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_binop(unifier, store, ty, &[ty], ty, &[ impl_binop(unifier, store, ty, &[ty], ty, &[ast::Operator::LShift, ast::Operator::RShift])
ast::Operator::LShift,
ast::Operator::RShift,
])
} }
/// Div /// Div
pub fn impl_div(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: &[Type]) { pub fn impl_div(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: &[Type]) {
impl_binop(unifier, store, ty, other_ty, store.float, &[ impl_binop(unifier, store, ty, other_ty, store.float, &[ast::Operator::Div])
ast::Operator::Div,
])
} }
/// FloorDiv /// FloorDiv
pub fn impl_floordiv(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: &[Type], ret_ty: Type) { pub fn impl_floordiv(
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ unifier: &mut Unifier,
ast::Operator::FloorDiv, store: &PrimitiveStore,
]) ty: Type,
other_ty: &[Type],
ret_ty: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::FloorDiv])
} }
/// Mod /// Mod
pub fn impl_mod(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: &[Type], ret_ty: Type) { pub fn impl_mod(
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ unifier: &mut Unifier,
ast::Operator::Mod, store: &PrimitiveStore,
]) ty: Type,
other_ty: &[Type],
ret_ty: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::Mod])
} }
/// UAdd, USub /// UAdd, USub
pub fn impl_sign(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) { pub fn impl_sign(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, store, ty, ty, &[ impl_unaryop(unifier, store, ty, ty, &[ast::Unaryop::UAdd, ast::Unaryop::USub])
ast::Unaryop::UAdd,
ast::Unaryop::USub,
])
} }
/// Invert /// Invert
pub fn impl_invert(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) { pub fn impl_invert(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, store, ty, ty, &[ impl_unaryop(unifier, store, ty, ty, &[ast::Unaryop::Invert])
ast::Unaryop::Invert,
])
} }
/// Not /// Not
pub fn impl_not(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) { pub fn impl_not(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, store, ty, store.bool, &[ impl_unaryop(unifier, store, ty, store.bool, &[ast::Unaryop::Not])
ast::Unaryop::Not,
])
} }
/// Lt, LtE, Gt, GtE /// Lt, LtE, Gt, GtE
pub fn impl_comparison(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: Type) { pub fn impl_comparison(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: Type) {
impl_cmpop(unifier, store, ty, other_ty, &[ impl_cmpop(
ast::Cmpop::Lt, unifier,
ast::Cmpop::Gt, store,
ast::Cmpop::LtE, ty,
ast::Cmpop::GtE, other_ty,
]) &[ast::Cmpop::Lt, ast::Cmpop::Gt, ast::Cmpop::LtE, ast::Cmpop::GtE],
)
} }
/// Eq, NotEq /// Eq, NotEq
pub fn impl_eq(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) { pub fn impl_eq(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_cmpop(unifier, store, ty, ty, &[ impl_cmpop(unifier, store, ty, ty, &[ast::Cmpop::Eq, ast::Cmpop::NotEq])
ast::Cmpop::Eq,
ast::Cmpop::NotEq,
])
} }
pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifier) { pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifier) {
let PrimitiveStore { let PrimitiveStore { int32: int32_t, int64: int64_t, float: float_t, bool: bool_t, .. } =
int32: int32_t, *store;
int64: int64_t,
float: float_t,
bool: bool_t,
..
} = *store;
/* int32 ======== */ /* int32 ======== */
impl_basic_arithmetic(unifier, store, int32_t, &[int32_t], int32_t); impl_basic_arithmetic(unifier, store, int32_t, &[int32_t], int32_t);
impl_pow(unifier, store, int32_t, &[int32_t], int32_t); impl_pow(unifier, store, int32_t, &[int32_t], int32_t);