1
0
forked from M-Labs/nac3

cleanup: we don't actually need arena

This commit is contained in:
pca006132 2021-07-16 13:55:29 +08:00
parent c2d00aa762
commit 62736bd4bf
4 changed files with 113 additions and 117 deletions

30
Cargo.lock generated
View File

@ -75,12 +75,6 @@ version = "1.0.68"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4a72c244c1ff497a746a7e1fb3d14bd08420ecda70c8f25c7112f2781652d787" checksum = "4a72c244c1ff497a746a7e1fb3d14bd08420ecda70c8f25c7112f2781652d787"
[[package]]
name = "cfg-if"
version = "0.1.10"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4785bdd1c96b2a846b2bd7cc02e86b6b3dbf14e7e53446c4f54c92a361040822"
[[package]] [[package]]
name = "cfg-if" name = "cfg-if"
version = "1.0.0" version = "1.0.0"
@ -115,7 +109,7 @@ version = "2.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b98cf8ebf19c3d1b223e151f99a4f9f0690dca41414773390fc824184ac833e1" checksum = "b98cf8ebf19c3d1b223e151f99a4f9f0690dca41414773390fc824184ac833e1"
dependencies = [ dependencies = [
"cfg-if 1.0.0", "cfg-if",
"dirs-sys-next", "dirs-sys-next",
] ]
@ -151,22 +145,13 @@ version = "0.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "37ab347416e802de484e4d03c7316c48f1ecb56574dfd4a46a80f173ce1de04d" checksum = "37ab347416e802de484e4d03c7316c48f1ecb56574dfd4a46a80f173ce1de04d"
[[package]]
name = "generational-arena"
version = "0.2.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8e1d3b771574f62d0548cee0ad9057857e9fc25d7a3335f140c84f6acd0bf601"
dependencies = [
"cfg-if 0.1.10",
]
[[package]] [[package]]
name = "getrandom" name = "getrandom"
version = "0.1.16" version = "0.1.16"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8fc3cb4d91f53b50155bdcfd23f6a4c39ae1969c2ae85982b135750cccaf5fce" checksum = "8fc3cb4d91f53b50155bdcfd23f6a4c39ae1969c2ae85982b135750cccaf5fce"
dependencies = [ dependencies = [
"cfg-if 1.0.0", "cfg-if",
"libc", "libc",
"wasi 0.9.0+wasi-snapshot-preview1", "wasi 0.9.0+wasi-snapshot-preview1",
] ]
@ -177,7 +162,7 @@ version = "0.2.3"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7fcd999463524c52659517fe2cea98493cfe485d10565e7b0fb07dbba7ad2753" checksum = "7fcd999463524c52659517fe2cea98493cfe485d10565e7b0fb07dbba7ad2753"
dependencies = [ dependencies = [
"cfg-if 1.0.0", "cfg-if",
"libc", "libc",
"wasi 0.10.2+wasi-snapshot-preview1", "wasi 0.10.2+wasi-snapshot-preview1",
] ]
@ -280,7 +265,7 @@ version = "0.1.9"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "61124eeebbd69b8190558df225adf7e4caafce0d743919e5d6b19652314ec5ec" checksum = "61124eeebbd69b8190558df225adf7e4caafce0d743919e5d6b19652314ec5ec"
dependencies = [ dependencies = [
"cfg-if 1.0.0", "cfg-if",
] ]
[[package]] [[package]]
@ -386,7 +371,7 @@ version = "0.4.14"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "51b9bbe6c47d51fc3e1a9b945965946b4c44142ab8792c50835a980d362c2710" checksum = "51b9bbe6c47d51fc3e1a9b945965946b4c44142ab8792c50835a980d362c2710"
dependencies = [ dependencies = [
"cfg-if 1.0.0", "cfg-if",
] ]
[[package]] [[package]]
@ -400,7 +385,6 @@ name = "nac3core"
version = "0.1.0" version = "0.1.0"
dependencies = [ dependencies = [
"ena", "ena",
"generational-arena",
"indoc 1.0.3", "indoc 1.0.3",
"inkwell", "inkwell",
"itertools", "itertools",
@ -499,7 +483,7 @@ version = "0.8.3"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fa7a782938e745763fe6907fc6ba86946d72f49fe7e21de074e08128a99fb018" checksum = "fa7a782938e745763fe6907fc6ba86946d72f49fe7e21de074e08128a99fb018"
dependencies = [ dependencies = [
"cfg-if 1.0.0", "cfg-if",
"instant", "instant",
"libc", "libc",
"redox_syscall", "redox_syscall",
@ -852,7 +836,7 @@ version = "1.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3b114ece25254e97bf48dd4bfc2a12bad0647adacfe4cae1247a9ca6ad302cec" checksum = "3b114ece25254e97bf48dd4bfc2a12bad0647adacfe4cae1247a9ca6ad302cec"
dependencies = [ dependencies = [
"cfg-if 1.0.0", "cfg-if",
"proc-macro2", "proc-macro2",
"quote", "quote",
"syn", "syn",

View File

@ -10,7 +10,6 @@ num-traits = "0.2"
inkwell = { git = "https://github.com/TheDan64/inkwell", branch = "master", features = ["llvm10-0"] } inkwell = { git = "https://github.com/TheDan64/inkwell", branch = "master", features = ["llvm10-0"] }
rustpython-parser = { git = "https://github.com/RustPython/RustPython", branch = "master" } rustpython-parser = { git = "https://github.com/RustPython/RustPython", branch = "master" }
indoc = "1.0" indoc = "1.0"
generational-arena = "0.2"
ena = "0.14" ena = "0.14"
[dev-dependencies] [dev-dependencies]

View File

@ -5,7 +5,6 @@ extern crate num_bigint;
extern crate inkwell; extern crate inkwell;
extern crate rustpython_parser; extern crate rustpython_parser;
extern crate indoc; extern crate indoc;
extern crate generational_arena;
extern crate ena; extern crate ena;
mod typecheck; mod typecheck;

View File

@ -1,29 +1,20 @@
use ena::unify::{InPlaceUnificationTable, NoError, UnifyKey, UnifyValue}; use ena::unify::{InPlaceUnificationTable, NoError, UnifyKey, UnifyValue};
use generational_arena::{Arena, Index};
use std::cell::RefCell; use std::cell::RefCell;
use std::collections::HashMap; use std::collections::HashMap;
use std::fmt::Debug;
use std::iter::once; use std::iter::once;
use std::mem::swap; use std::mem::swap;
use std::ops::Deref;
use std::rc::Rc; use std::rc::Rc;
// Order:
// TVar
// |--> TSeq
// | |--> TTuple
// | `--> TList
// |--> TRecord
// | |--> TObj
// | `--> TVirtual
// `--> TCall
// `--> TFunc
#[derive(Copy, Clone, PartialEq, Eq, Debug)] #[derive(Copy, Clone, PartialEq, Eq, Debug)]
/// Handle for a type, implementated as a key in the unification table.
pub struct Type(u32); pub struct Type(u32);
#[derive(Copy, Clone, Debug, PartialEq, Eq)] #[derive(Clone)]
pub struct TypeIndex(Index); pub struct TypeCell(Rc<RefCell<TypeEnum>>);
impl UnifyValue for TypeIndex { impl UnifyValue for TypeCell {
type Error = NoError; type Error = NoError;
fn unify_values(_: &Self, value2: &Self) -> Result<Self, Self::Error> { fn unify_values(_: &Self, value2: &Self) -> Result<Self, Self::Error> {
// WARN: depends on the implementation details of ena. // WARN: depends on the implementation details of ena.
@ -31,12 +22,12 @@ impl UnifyValue for TypeIndex {
// and assign the type by `union_value(key, new_value)`, which set the // and assign the type by `union_value(key, new_value)`, which set the
// value as `unify_values(key.value, new_value)`. So, we need to return // value as `unify_values(key.value, new_value)`. So, we need to return
// the right one. // the right one.
Ok(*value2) Ok(value2.clone())
} }
} }
impl UnifyKey for Type { impl UnifyKey for Type {
type Value = TypeIndex; type Value = TypeCell;
fn index(&self) -> u32 { fn index(&self) -> u32 {
self.0 self.0
} }
@ -48,6 +39,14 @@ impl UnifyKey for Type {
} }
} }
impl Deref for TypeCell {
type Target = Rc<RefCell<TypeEnum>>;
fn deref(&self) -> &<Self as Deref>::Target {
&self.0
}
}
pub type Mapping<K, V = Type> = HashMap<K, V>; pub type Mapping<K, V = Type> = HashMap<K, V>;
pub type VarMap = Mapping<u32>; pub type VarMap = Mapping<u32>;
@ -104,18 +103,63 @@ pub enum TypeEnum {
}, },
} }
// Order:
// TVar
// |--> TSeq
// | |--> TTuple
// | `--> TList
// |--> TRecord
// | |--> TObj
// | `--> TVirtual
// `--> TCall
// `--> TFunc
// We encode the types as natural numbers, and subtyping relation as divisibility.
// If a | b, b <: a.
// We assign unique prime numbers (1 to TVar, everything is a subtype of it) to each type:
// TVar = 1
// |--> TSeq = 2
// | |--> TTuple = 3
// | `--> TList = 5
// |--> TRecord = 7
// | |--> TObj = 11
// | `--> TVirtual = 13
// `--> TCall = 17
// `--> TFunc = 21
//
// And then, based on the subtyping relation, multiply them together...
// TVar = 1
// |--> TSeq = 2 * TVar
// | |--> TTuple = 3 * TSeq * TVar
// | `--> TList = 5 * TSeq * TVar
// |--> TRecord = 7 * TVar
// | |--> TObj = 11 * TRecord * TVar
// | `--> TVirtual = 13 * TRecord * TVar
// `--> TCall = 17 * TVar
// `--> TFunc = 21 * TCall * TVar
impl TypeEnum { impl TypeEnum {
fn get_int(&self) -> i32 { fn get_int(&self) -> i32 {
const TVAR: i32 = 1;
const TSEQ: i32 = 2;
const TTUPLE: i32 = 3;
const TLIST: i32 = 5;
const TRECORD: i32 = 7;
const TOBJ: i32 = 11;
const TVIRTUAL: i32 = 13;
const TCALL: i32 = 17;
const TFUNC: i32 = 21;
match self { match self {
TypeEnum::TVar { .. } => 1, TypeEnum::TVar { .. } => TVAR,
TypeEnum::TSeq { .. } => 5, TypeEnum::TSeq { .. } => TSEQ * TVAR,
TypeEnum::TTuple { .. } => 10, TypeEnum::TTuple { .. } => TTUPLE * TSEQ * TVAR,
TypeEnum::TList { .. } => 15, TypeEnum::TList { .. } => TLIST * TSEQ * TVAR,
TypeEnum::TRecord { .. } => 7, TypeEnum::TRecord { .. } => TRECORD * TVAR,
TypeEnum::TObj { .. } => 14, TypeEnum::TObj { .. } => TOBJ * TRECORD * TVAR,
TypeEnum::TVirtual { .. } => 21, TypeEnum::TVirtual { .. } => TVIRTUAL * TRECORD * TVAR,
TypeEnum::TCall { .. } => 11, TypeEnum::TCall { .. } => TCALL * TVAR,
TypeEnum::TFunc { .. } => 22, TypeEnum::TFunc { .. } => TFUNC * TCALL * TVAR,
} }
} }
@ -126,7 +170,7 @@ impl TypeEnum {
(a % b) == 0 (a % b) == 0
} }
pub fn get_kind_name(&self) -> &'static str { pub fn get_type_name(&self) -> &'static str {
// this function is for debugging only... // this function is for debugging only...
// a proper to_str implementation requires the context // a proper to_str implementation requires the context
match self { match self {
@ -143,6 +187,12 @@ impl TypeEnum {
} }
} }
impl Debug for TypeCell {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.write_str(&self.borrow().get_type_name())
}
}
pub struct ObjDef { pub struct ObjDef {
name: String, name: String,
fields: Mapping<String>, fields: Mapping<String>,
@ -150,7 +200,6 @@ pub struct ObjDef {
pub struct Unifier { pub struct Unifier {
unification_table: RefCell<InPlaceUnificationTable<Type>>, unification_table: RefCell<InPlaceUnificationTable<Type>>,
type_arena: RefCell<Arena<Rc<RefCell<TypeEnum>>>>,
obj_def_table: Vec<ObjDef>, obj_def_table: Vec<ObjDef>,
} }
@ -158,53 +207,43 @@ impl Unifier {
pub fn new() -> Unifier { pub fn new() -> Unifier {
Unifier { Unifier {
unification_table: RefCell::new(InPlaceUnificationTable::new()), unification_table: RefCell::new(InPlaceUnificationTable::new()),
type_arena: RefCell::new(Arena::new()),
obj_def_table: Vec::new(), obj_def_table: Vec::new(),
} }
} }
/// Register a type to the unifier.
/// Returns a key in the unification_table.
pub fn add_ty(&self, a: TypeEnum) -> Type { pub fn add_ty(&self, a: TypeEnum) -> Type {
let index = self.type_arena.borrow_mut().insert(Rc::new(a.into()));
self.unification_table self.unification_table
.borrow_mut() .borrow_mut()
.new_key(TypeIndex(index)) .new_key(TypeCell(Rc::new(a.into())))
} }
/// Get the TypeEnum of a type.
pub fn get_ty(&self, a: Type) -> Rc<RefCell<TypeEnum>> { pub fn get_ty(&self, a: Type) -> Rc<RefCell<TypeEnum>> {
let mut table = self.unification_table.borrow_mut(); let mut table = self.unification_table.borrow_mut();
let arena = self.type_arena.borrow(); table.probe_value(a).0
arena.get(table.probe_value(a).0).unwrap().clone()
} }
pub fn unify(&self, mut a: Type, mut b: Type) -> Result<(), String> { pub fn unify(&self, mut a: Type, mut b: Type) -> Result<(), String> {
let (mut i_a, mut i_b) = {
let mut table = self.unification_table.borrow_mut();
(table.probe_value(a), table.probe_value(b))
};
if i_a == i_b {
return Ok(());
}
let (mut ty_a_cell, mut ty_b_cell) = { let (mut ty_a_cell, mut ty_b_cell) = {
let arena = self.type_arena.borrow(); let mut table = self.unification_table.borrow_mut();
( if table.unioned(a, b) {
arena.get(i_a.0).unwrap().clone(), return Ok(());
arena.get(i_b.0).unwrap().clone(), }
) (table.probe_value(a), table.probe_value(b))
}; };
let (ty_a, ty_b) = { let (ty_a, ty_b) = {
// simplify our pattern matching... // simplify our pattern matching...
if ty_a_cell.borrow().kind_le(&ty_b_cell.borrow()) { if ty_a_cell.borrow().kind_le(&ty_b_cell.borrow()) {
swap(&mut a, &mut b); swap(&mut a, &mut b);
swap(&mut i_a, &mut i_b);
swap(&mut ty_a_cell, &mut ty_b_cell); swap(&mut ty_a_cell, &mut ty_b_cell);
} }
(ty_a_cell.borrow(), ty_b_cell.borrow()) (ty_a_cell.borrow(), ty_b_cell.borrow())
}; };
self.occur_check(i_a, b)?; self.occur_check(a, b)?;
match &*ty_a { match &*ty_a {
TypeEnum::TVar { .. } => { TypeEnum::TVar { .. } => {
// TODO: type variables bound check... // TODO: type variables bound check...
@ -356,30 +395,26 @@ impl Unifier {
} }
fn set_a_to_b(&self, a: Type, b: Type) { fn set_a_to_b(&self, a: Type, b: Type) {
// unify a and b together, and set the value to b's value this would // unify a and b together, and set the value to b's value.
// also deallocate a's previous value in the arena to save space...
let mut table = self.unification_table.borrow_mut(); let mut table = self.unification_table.borrow_mut();
let i_a = table.probe_value(a); let ty_b = table.probe_value(b);
let i_b = table.probe_value(b);
table.union(a, b); table.union(a, b);
table.union_value(a, i_b); table.union_value(a, ty_b);
self.type_arena.borrow_mut().remove(i_a.0);
} }
fn report_kind_error(&self, a: &TypeEnum, b: &TypeEnum) -> Result<(), String> { fn report_kind_error(&self, a: &TypeEnum, b: &TypeEnum) -> Result<(), String> {
Err(format!( Err(format!(
"Cannot unify {} with {}", "Cannot unify {} with {}",
a.get_kind_name(), a.get_type_name(),
b.get_kind_name() b.get_type_name()
)) ))
} }
fn occur_check(&self, a: TypeIndex, b: Type) -> Result<(), String> { fn occur_check(&self, a: Type, b: Type) -> Result<(), String> {
let i_b = self.unification_table.borrow_mut().probe_value(b); if self.unification_table.borrow_mut().unioned(a, b) {
if a == i_b {
return Err("Recursive type is prohibited.".to_owned()); return Err("Recursive type is prohibited.".to_owned());
} }
let ty = self.type_arena.borrow().get(i_b.0).unwrap().clone(); let ty = self.unification_table.borrow_mut().probe_value(b);
let ty = ty.borrow(); let ty = ty.borrow();
match &*ty { match &*ty {
@ -433,11 +468,7 @@ impl Unifier {
} }
pub fn subst(&self, a: Type, mapping: &VarMap) -> Option<Type> { pub fn subst(&self, a: Type, mapping: &VarMap) -> Option<Type> {
let index = self.unification_table.borrow_mut().probe_value(a); let ty_cell = self.unification_table.borrow_mut().probe_value(a);
let ty_cell = {
let arena = self.type_arena.borrow();
arena.get(index.0).unwrap().clone()
};
let ty = ty_cell.borrow(); let ty = ty_cell.borrow();
// this function would only be called when we instantiate functions. // this function would only be called when we instantiate functions.
// function type signature should ONLY contain concrete types and type // function type signature should ONLY contain concrete types and type
@ -445,15 +476,9 @@ impl Unifier {
// should be safe to not implement the substitution for those variants. // should be safe to not implement the substitution for those variants.
match &*ty { match &*ty {
TypeEnum::TVar { id } => mapping.get(&id).cloned(), TypeEnum::TVar { id } => mapping.get(&id).cloned(),
TypeEnum::TSeq { map } => self.subst_map(map, mapping).map(|m| { TypeEnum::TSeq { map } => self
let index = self .subst_map(map, mapping)
.type_arena .map(|m| self.add_ty(TypeEnum::TSeq { map: m })),
.borrow_mut()
.insert(Rc::new(TypeEnum::TSeq { map: m }.into()));
self.unification_table
.borrow_mut()
.new_key(TypeIndex(index))
}),
TypeEnum::TTuple { ty } => { TypeEnum::TTuple { ty } => {
let mut new_ty = None; let mut new_ty = None;
for (i, t) in ty.iter().enumerate() { for (i, t) in ty.iter().enumerate() {
@ -482,9 +507,7 @@ impl Unifier {
// parameter list, we don't need to substitute the fields. // parameter list, we don't need to substitute the fields.
// This is also used to prevent infinite substitution... // This is also used to prevent infinite substitution...
let need_subst = params.values().any(|v| { let need_subst = params.values().any(|v| {
let index = self.unification_table.borrow_mut().probe_value(*v); let ty_cell = self.unification_table.borrow_mut().probe_value(*v);
let arena = self.type_arena.borrow();
let ty_cell = arena.get(index.0).unwrap();
let ty = ty_cell.borrow(); let ty = ty_cell.borrow();
if let TypeEnum::TVar { id } = &*ty { if let TypeEnum::TVar { id } = &*ty {
mapping.contains_key(id) mapping.contains_key(id)
@ -558,21 +581,12 @@ impl Unifier {
if a == b { if a == b {
return true; return true;
} }
let (i_a, i_b) = {
let mut table = self.unification_table.borrow_mut();
(table.probe_value(a), table.probe_value(b))
};
if i_a == i_b {
return true;
}
let (ty_a, ty_b) = { let (ty_a, ty_b) = {
let arena = self.type_arena.borrow(); let mut table = self.unification_table.borrow_mut();
( if table.unioned(a, b) {
arena.get(i_a.0).unwrap().clone(), return true;
arena.get(i_b.0).unwrap().clone(), }
) (table.probe_value(a), table.probe_value(b))
}; };
let ty_a = ty_a.borrow(); let ty_a = ty_a.borrow();