Compare commits

...

16 Commits

Author SHA1 Message Date
pca006132 65d9b620fe Merge remote-tracking branch 'origin/master' 2021-01-27 16:00:47 +08:00
pca006132 075245f4c2 get list of unknowns 2021-01-27 16:00:06 +08:00
pca006132 0e9dea0e65 updated todo 2021-01-25 17:09:17 +08:00
pca006132 7d06c903e3 added get_expression_unknowns 2021-01-15 15:19:41 +08:00
pca006132 5fce6cf069 updated lifetime 2021-01-15 14:37:55 +08:00
pca006132 d466b7bc2b simplified lifetime 2021-01-11 11:12:37 +08:00
pca006132 b6e4e68587 started adding error types
* converted to string to avoid breaking interface, would be fixed later
* tests would not check error messages now, as messages are not
  finalized
2021-01-11 10:40:32 +08:00
pca006132 779288d685 added todo 2021-01-08 17:16:25 +08:00
pca006132 3ecf57a588 fixed bugs 2021-01-08 16:54:34 +08:00
pca006132 ebe1027ffa added resolve signature 2021-01-08 16:24:25 +08:00
pca006132 7c6349520c started getting signatures 2021-01-08 14:52:48 +08:00
pca006132 04f121403a moved type check code to submodule 2021-01-08 12:58:33 +08:00
pca006132 b51168a5ab added statement tests 2021-01-05 13:35:44 +08:00
pca006132 007843c1ef added aug assign for primitives 2021-01-05 13:21:39 +08:00
pca006132 ff41cdb000 implemented statement check 2021-01-05 12:17:45 +08:00
pca006132 e1efb47ad2 statement inference: assignment 2021-01-04 17:07:14 +08:00
15 changed files with 1320 additions and 135 deletions

33
Cargo.lock generated
View File

@ -288,6 +288,15 @@ dependencies = [
"proc-macro-hack", "proc-macro-hack",
] ]
[[package]]
name = "indoc"
version = "1.0.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e5a75aeaaef0ce18b58056d306c27b07436fbb34b8816c53094b76dd81803136"
dependencies = [
"unindent",
]
[[package]] [[package]]
name = "indoc-impl" name = "indoc-impl"
version = "0.3.6" version = "0.3.6"
@ -449,10 +458,12 @@ checksum = "0ee1c47aaa256ecabcaea351eae4a9b01ef39ed810004e298d2511ed284b1525"
name = "nac3core" name = "nac3core"
version = "0.1.0" version = "0.1.0"
dependencies = [ dependencies = [
"indoc 1.0.3",
"inkwell", "inkwell",
"num-bigint", "num-bigint",
"num-traits", "num-traits",
"rustpython-parser", "rustpython-parser",
"thiserror",
] ]
[[package]] [[package]]
@ -613,7 +624,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "bf6bbbe8f70d179260b3728e5d04eb012f4f0c7988e58c11433dd689cecaa72e" checksum = "bf6bbbe8f70d179260b3728e5d04eb012f4f0c7988e58c11433dd689cecaa72e"
dependencies = [ dependencies = [
"ctor", "ctor",
"indoc", "indoc 0.3.6",
"inventory", "inventory",
"libc", "libc",
"parking_lot", "parking_lot",
@ -830,6 +841,26 @@ dependencies = [
"winapi", "winapi",
] ]
[[package]]
name = "thiserror"
version = "1.0.23"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "76cc616c6abf8c8928e2fdcc0dbfab37175edd8fb49a4641066ad1364fdab146"
dependencies = [
"thiserror-impl",
]
[[package]]
name = "thiserror-impl"
version = "1.0.23"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9be73a2caec27583d0046ef3796c3794f868a5bc813db689eed00c7631275cd1"
dependencies = [
"proc-macro2",
"quote",
"syn",
]
[[package]] [[package]]
name = "thread_local" name = "thread_local"
version = "1.0.1" version = "1.0.1"

View File

@ -7,5 +7,10 @@ edition = "2018"
[dependencies] [dependencies]
num-bigint = "0.3" num-bigint = "0.3"
num-traits = "0.2" num-traits = "0.2"
thiserror = "1.0"
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" }
[dev-dependencies]
indoc = "1.0"

View File

@ -5,12 +5,7 @@ extern crate num_bigint;
extern crate inkwell; extern crate inkwell;
extern crate rustpython_parser; extern crate rustpython_parser;
pub mod expression_inference; pub mod type_check;
pub mod inference_core;
mod magic_methods;
pub mod primitives;
pub mod typedef;
pub mod context;
use std::error::Error; use std::error::Error;
use std::fmt; use std::fmt;

View File

@ -1,5 +1,5 @@
use super::super::typedef::*;
use super::TopLevelContext; use super::TopLevelContext;
use crate::typedef::*;
use std::boxed::Box; use std::boxed::Box;
use std::collections::HashMap; use std::collections::HashMap;
@ -22,14 +22,14 @@ pub struct InferenceContext<'a> {
primitives: Vec<Type>, primitives: Vec<Type>,
/// list of variable instances /// list of variable instances
variables: Vec<Type>, variables: Vec<Type>,
/// identifier to (type, readable) mapping. /// identifier to type mapping.
/// an identifier might be defined earlier but has no value (for some code path), thus not sym_table: HashMap<&'a str, Type>,
/// readable.
sym_table: HashMap<&'a str, (Type, bool)>,
/// resolution function reference, that may resolve unbounded identifiers to some type /// resolution function reference, that may resolve unbounded identifiers to some type
resolution_fn: Box<dyn FnMut(&str) -> Result<Type, String>>, resolution_fn: Box<dyn FnMut(&str) -> Result<Type, String>>,
/// stack /// stack
stack: ContextStack<'a>, stack: ContextStack<'a>,
/// return type
result: Option<Type>,
} }
// non-trivial implementations here // non-trivial implementations here
@ -56,6 +56,7 @@ impl<'a> InferenceContext<'a> {
var_defs: Vec::new(), var_defs: Vec::new(),
sym_def: Vec::new(), sym_def: Vec::new(),
}, },
result: None,
} }
} }
@ -63,7 +64,7 @@ impl<'a> InferenceContext<'a> {
/// variable assignment would be limited within the scope (not readable outside), and type /// variable assignment would be limited within the scope (not readable outside), and type
/// variable type guard would be limited within the scope. /// variable type guard would be limited within the scope.
/// returns the list of variables assigned within the scope, and the result of the function /// returns the list of variables assigned within the scope, and the result of the function
pub fn with_scope<F, R>(&mut self, f: F) -> (Vec<&'a str>, R) pub fn with_scope<F, R>(&mut self, f: F) -> (Vec<(&'a str, Type)>, R)
where where
F: FnOnce(&mut Self) -> R, F: FnOnce(&mut Self) -> R,
{ {
@ -84,8 +85,8 @@ impl<'a> InferenceContext<'a> {
let (_, level) = self.stack.sym_def.last().unwrap(); let (_, level) = self.stack.sym_def.last().unwrap();
if *level > self.stack.level { if *level > self.stack.level {
let (name, _) = self.stack.sym_def.pop().unwrap(); let (name, _) = self.stack.sym_def.pop().unwrap();
self.sym_table.remove(name).unwrap(); let ty = self.sym_table.remove(name).unwrap();
poped_names.push(name); poped_names.push((name, ty));
} else { } else {
break; break;
} }
@ -96,19 +97,15 @@ impl<'a> InferenceContext<'a> {
/// assign a type to an identifier. /// assign a type to an identifier.
/// may return error if the identifier was defined but with different type /// may return error if the identifier was defined but with different type
pub fn assign(&mut self, name: &'a str, ty: Type) -> Result<Type, String> { pub fn assign(&mut self, name: &'a str, ty: Type) -> Result<Type, String> {
if let Some((t, x)) = self.sym_table.get_mut(name) { if let Some(t) = self.sym_table.get_mut(name) {
if t == &ty { if t == &ty {
if !*x {
self.stack.sym_def.push((name, self.stack.level));
}
*x = true;
Ok(ty) Ok(ty)
} else { } else {
Err("different types".into()) Err("different types".into())
} }
} else { } else {
self.stack.sym_def.push((name, self.stack.level)); self.stack.sym_def.push((name, self.stack.level));
self.sym_table.insert(name, (ty.clone(), true)); self.sym_table.insert(name, ty.clone());
Ok(ty) Ok(ty)
} }
} }
@ -122,12 +119,8 @@ impl<'a> InferenceContext<'a> {
/// may return error if the identifier is not defined, and cannot be resolved with the /// may return error if the identifier is not defined, and cannot be resolved with the
/// resolution function. /// resolution function.
pub fn resolve(&mut self, name: &str) -> Result<Type, String> { pub fn resolve(&mut self, name: &str) -> Result<Type, String> {
if let Some((t, x)) = self.sym_table.get(name) { if let Some(t) = self.sym_table.get(name) {
if *x { Ok(t.clone())
Ok(t.clone())
} else {
Err("may not have value".into())
}
} else { } else {
self.resolution_fn.as_mut()(name) self.resolution_fn.as_mut()(name)
} }
@ -139,6 +132,10 @@ impl<'a> InferenceContext<'a> {
std::mem::swap(self.top_level.var_defs.get_mut(id.0).unwrap(), &mut def); std::mem::swap(self.top_level.var_defs.get_mut(id.0).unwrap(), &mut def);
self.stack.var_defs.push((id.0, def, self.stack.level)); self.stack.var_defs.push((id.0, def, self.stack.level));
} }
pub fn set_result(&mut self, result: Option<Type>) {
self.result = result;
}
} }
// trivial getters: // trivial getters:
@ -168,6 +165,9 @@ impl<'a> InferenceContext<'a> {
pub fn get_type(&self, name: &str) -> Option<Type> { pub fn get_type(&self, name: &str) -> Option<Type> {
self.top_level.get_type(name) self.top_level.get_type(name)
} }
pub fn get_result(&self) -> Option<Type> {
self.result.clone()
}
} }
impl TypeEnum { impl TypeEnum {
@ -215,7 +215,7 @@ impl TypeEnum {
} }
} }
pub fn get_base<'b: 'a, 'a>(&'a self, ctx: &'b InferenceContext) -> Option<&'b TypeDef> { pub fn get_base<'a>(&'a self, ctx: &'a InferenceContext) -> Option<&'a TypeDef> {
match self { match self {
TypeEnum::PrimitiveType(id) => Some(ctx.get_primitive_def(*id)), TypeEnum::PrimitiveType(id) => Some(ctx.get_primitive_def(*id)),
TypeEnum::ClassType(id) | TypeEnum::VirtualClassType(id) => { TypeEnum::ClassType(id) | TypeEnum::VirtualClassType(id) => {

View File

@ -1,4 +1,4 @@
use crate::typedef::*; use super::super::typedef::*;
use std::collections::HashMap; use std::collections::HashMap;
use std::rc::Rc; use std::rc::Rc;
@ -130,7 +130,9 @@ impl<'a> TopLevelContext<'a> {
} }
pub fn get_type(&self, name: &str) -> Option<Type> { pub fn get_type(&self, name: &str) -> Option<Type> {
// TODO: handle parametric types // TODO: handle name visibility
// possibly by passing a function from outside to tell what names are allowed, and what are
// not...
self.sym_table.get(name).cloned() self.sym_table.get(name).cloned()
} }
} }

View File

@ -1,8 +1,8 @@
use crate::context::InferenceContext; use super::context::InferenceContext;
use crate::inference_core::resolve_call; use super::inference_core::resolve_call;
use crate::magic_methods::*; use super::magic_methods::*;
use crate::primitives::*; use super::primitives::*;
use crate::typedef::{Type, TypeEnum::*}; use super::typedef::{Type, TypeEnum::*};
use rustpython_parser::ast::{ use rustpython_parser::ast::{
Comparison, Comprehension, ComprehensionKind, Expression, ExpressionType, Operator, Comparison, Comprehension, ComprehensionKind, Expression, ExpressionType, Operator,
UnaryOperator, UnaryOperator,
@ -11,9 +11,9 @@ use std::convert::TryInto;
type ParserResult = Result<Option<Type>, String>; type ParserResult = Result<Option<Type>, String>;
pub fn infer_expr<'b: 'a, 'a>( pub fn infer_expr<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
expr: &'b Expression, expr: &'a Expression,
) -> ParserResult { ) -> ParserResult {
match &expr.node { match &expr.node {
ExpressionType::Number { value } => infer_constant(ctx, value), ExpressionType::Number { value } => infer_constant(ctx, value),
@ -83,9 +83,9 @@ fn infer_identifier(ctx: &mut InferenceContext, name: &str) -> ParserResult {
Ok(Some(ctx.resolve(name)?)) Ok(Some(ctx.resolve(name)?))
} }
fn infer_list<'b: 'a, 'a>( fn infer_list<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
elements: &'b [Expression], elements: &'a [Expression],
) -> ParserResult { ) -> ParserResult {
if elements.is_empty() { if elements.is_empty() {
return Ok(Some(ParametricType(LIST_TYPE, vec![BotType.into()]).into())); return Ok(Some(ParametricType(LIST_TYPE, vec![BotType.into()]).into()));
@ -105,9 +105,9 @@ fn infer_list<'b: 'a, 'a>(
Ok(Some(ParametricType(LIST_TYPE, vec![head.unwrap()]).into())) Ok(Some(ParametricType(LIST_TYPE, vec![head.unwrap()]).into()))
} }
fn infer_tuple<'b: 'a, 'a>( fn infer_tuple<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
elements: &'b [Expression], elements: &'a [Expression],
) -> ParserResult { ) -> ParserResult {
let types: Result<Option<Vec<_>>, String> = let types: Result<Option<Vec<_>>, String> =
elements.iter().map(|v| infer_expr(ctx, v)).collect(); elements.iter().map(|v| infer_expr(ctx, v)).collect();
@ -164,11 +164,11 @@ fn infer_bool_ops<'a>(ctx: &mut InferenceContext<'a>, values: &'a [Expression])
} }
} }
fn infer_bin_ops<'b: 'a, 'a>( fn infer_bin_ops<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
op: &Operator, op: &Operator,
left: &'b Expression, left: &'a Expression,
right: &'b Expression, right: &'a Expression,
) -> ParserResult { ) -> ParserResult {
let left = infer_expr(ctx, left)?.ok_or_else(|| "no value".to_string())?; let left = infer_expr(ctx, left)?.ok_or_else(|| "no value".to_string())?;
let right = infer_expr(ctx, right)?.ok_or_else(|| "no value".to_string())?; let right = infer_expr(ctx, right)?.ok_or_else(|| "no value".to_string())?;
@ -176,10 +176,10 @@ fn infer_bin_ops<'b: 'a, 'a>(
resolve_call(ctx, Some(left), fun, &[right]) resolve_call(ctx, Some(left), fun, &[right])
} }
fn infer_unary_ops<'b: 'a, 'a>( fn infer_unary_ops<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
op: &UnaryOperator, op: &UnaryOperator,
obj: &'b Expression, obj: &'a Expression,
) -> ParserResult { ) -> ParserResult {
let ty = infer_expr(ctx, obj)?.ok_or_else(|| "no value".to_string())?; let ty = infer_expr(ctx, obj)?.ok_or_else(|| "no value".to_string())?;
if let UnaryOperator::Not = op { if let UnaryOperator::Not = op {
@ -193,10 +193,10 @@ fn infer_unary_ops<'b: 'a, 'a>(
} }
} }
fn infer_compare<'b: 'a, 'a>( fn infer_compare<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
vals: &'b [Expression], vals: &'a [Expression],
ops: &'b [Comparison], ops: &'a [Comparison],
) -> ParserResult { ) -> ParserResult {
let types: Result<Option<Vec<_>>, _> = vals.iter().map(|v| infer_expr(ctx, v)).collect(); let types: Result<Option<Vec<_>>, _> = vals.iter().map(|v| infer_expr(ctx, v)).collect();
let types = types?; let types = types?;
@ -218,10 +218,10 @@ fn infer_compare<'b: 'a, 'a>(
Ok(Some(boolean)) Ok(Some(boolean))
} }
fn infer_call<'b: 'a, 'a>( fn infer_call<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
args: &'b [Expression], args: &'a [Expression],
function: &'b Expression, function: &'a Expression,
) -> ParserResult { ) -> ParserResult {
let types: Result<Option<Vec<_>>, _> = args.iter().map(|v| infer_expr(ctx, v)).collect(); let types: Result<Option<Vec<_>>, _> = args.iter().map(|v| infer_expr(ctx, v)).collect();
let types = types?; let types = types?;
@ -240,10 +240,10 @@ fn infer_call<'b: 'a, 'a>(
resolve_call(ctx, obj, fun.as_str(), &types.unwrap()) resolve_call(ctx, obj, fun.as_str(), &types.unwrap())
} }
fn infer_subscript<'b: 'a, 'a>( fn infer_subscript<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
a: &'b Expression, a: &'a Expression,
b: &'b Expression, b: &'a Expression,
) -> ParserResult { ) -> ParserResult {
let a = infer_expr(ctx, a)?.ok_or_else(|| "no value".to_string())?; let a = infer_expr(ctx, a)?.ok_or_else(|| "no value".to_string())?;
let t = if let ParametricType(LIST_TYPE, ls) = a.as_ref() { let t = if let ParametricType(LIST_TYPE, ls) = a.as_ref() {
@ -283,11 +283,11 @@ fn infer_subscript<'b: 'a, 'a>(
} }
} }
fn infer_if_expr<'b: 'a, 'a>( fn infer_if_expr<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
test: &'b Expression, test: &'a Expression,
body: &'b Expression, body: &'a Expression,
orelse: &'b Expression, orelse: &'a Expression,
) -> ParserResult { ) -> ParserResult {
let test = infer_expr(ctx, test)?.ok_or_else(|| "no value".to_string())?; let test = infer_expr(ctx, test)?.ok_or_else(|| "no value".to_string())?;
if test != ctx.get_primitive(BOOL_TYPE) { if test != ctx.get_primitive(BOOL_TYPE) {
@ -303,8 +303,8 @@ fn infer_if_expr<'b: 'a, 'a>(
} }
} }
fn infer_simple_binding<'a: 'b, 'b>( pub fn infer_simple_binding<'a>(
ctx: &mut InferenceContext<'b>, ctx: &mut InferenceContext<'a>,
name: &'a Expression, name: &'a Expression,
ty: Type, ty: Type,
) -> Result<(), String> { ) -> Result<(), String> {
@ -337,10 +337,10 @@ fn infer_simple_binding<'a: 'b, 'b>(
} }
} }
fn infer_list_comprehension<'b: 'a, 'a>( fn infer_list_comprehension<'a>(
ctx: &mut InferenceContext<'a>, ctx: &mut InferenceContext<'a>,
element: &'b Expression, element: &'a Expression,
comprehension: &'b Comprehension, comprehension: &'a Comprehension,
) -> ParserResult { ) -> ParserResult {
if comprehension.is_async { if comprehension.is_async {
return Err("async is not supported".into()); return Err("async is not supported".into());
@ -370,9 +370,10 @@ fn infer_list_comprehension<'b: 'a, 'a>(
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use super::*; use super::{
use crate::context::*; super::{context::*, typedef::*},
use crate::typedef::*; *,
};
use rustpython_parser::parser::parse_expression; use rustpython_parser::parser::parse_expression;
use std::collections::HashMap; use std::collections::HashMap;
use std::rc::Rc; use std::rc::Rc;
@ -704,7 +705,7 @@ mod test {
let ast = parse_expression("a == a == 1").unwrap(); let ast = parse_expression("a == a == 1").unwrap();
let result = infer_expr(&mut ctx, &ast); let result = infer_expr(&mut ctx, &ast);
assert_eq!(result, Err("not equal".into())); assert_eq!(result, Err("different types".into()));
let ast = parse_expression("True > False").unwrap(); let ast = parse_expression("True > False").unwrap();
let result = infer_expr(&mut ctx, &ast); let result = infer_expr(&mut ctx, &ast);

View File

@ -1,6 +1,21 @@
use crate::context::InferenceContext; use super::context::InferenceContext;
use crate::typedef::{TypeEnum::*, *}; use super::typedef::{TypeEnum::*, *};
use std::collections::HashMap; use std::collections::HashMap;
use thiserror::Error;
#[derive(Error, Debug)]
enum SubstError {
#[error("different type variables after substitution")]
DifferentSubstVar(VariableId, VariableId),
#[error("cannot substitute unbounded type variable into bounded one")]
UnboundedTypeVar(VariableId, VariableId),
#[error("incompatible bound for type variables")]
IncompatibleBound(VariableId, VariableId),
#[error("only subtype of virtual class can be substituted into virtual class type")]
NotVirtualClassSubtype(Type, ClassId),
#[error("different types")]
DifferentTypes(Type, Type),
}
fn find_subst( fn find_subst(
ctx: &InferenceContext, ctx: &InferenceContext,
@ -8,8 +23,7 @@ fn find_subst(
sub: &mut HashMap<VariableId, Type>, sub: &mut HashMap<VariableId, Type>,
mut a: Type, mut a: Type,
mut b: Type, mut b: Type,
) -> Result<(), String> { ) -> Result<(), SubstError> {
// TODO: fix error messages later
if let TypeVariable(id) = a.as_ref() { if let TypeVariable(id) = a.as_ref() {
if let Some((assumption_id, t)) = valuation { if let Some((assumption_id, t)) = valuation {
if assumption_id == id { if assumption_id == id {
@ -33,14 +47,14 @@ fn find_subst(
return if id_a == id_b { return if id_a == id_b {
Ok(()) Ok(())
} else { } else {
Err("different variables".to_string()) Err(SubstError::DifferentSubstVar(*id_a, *id_b))
}; };
} }
let v_a = ctx.get_variable_def(*id_a); let v_a = ctx.get_variable_def(*id_a);
let v_b = ctx.get_variable_def(*id_b); let v_b = ctx.get_variable_def(*id_b);
if !v_b.bound.is_empty() { if !v_b.bound.is_empty() {
if v_a.bound.is_empty() { if v_a.bound.is_empty() {
return Err("unbounded a".to_string()); return Err(SubstError::UnboundedTypeVar(*id_a, *id_b));
} else { } else {
let diff: Vec<_> = v_a let diff: Vec<_> = v_a
.bound .bound
@ -48,7 +62,7 @@ fn find_subst(
.filter(|x| !v_b.bound.contains(x)) .filter(|x| !v_b.bound.contains(x))
.collect(); .collect();
if !diff.is_empty() { if !diff.is_empty() {
return Err("different domain".to_string()); return Err(SubstError::IncompatibleBound(*id_a, *id_b));
} }
} }
} }
@ -60,7 +74,7 @@ fn find_subst(
if v_a.bound.len() == 1 && v_a.bound[0].as_ref() == b.as_ref() { if v_a.bound.len() == 1 && v_a.bound[0].as_ref() == b.as_ref() {
Ok(()) Ok(())
} else { } else {
Err("different domain".to_string()) Err(SubstError::DifferentTypes(a.clone(), b.clone()))
} }
} }
(_, TypeVariable(id_b)) => { (_, TypeVariable(id_b)) => {
@ -69,7 +83,7 @@ fn find_subst(
sub.insert(*id_b, a.clone()); sub.insert(*id_b, a.clone());
Ok(()) Ok(())
} else { } else {
Err("different domain".to_string()) Err(SubstError::DifferentTypes(a.clone(), b.clone()))
} }
} }
(_, VirtualClassType(id_b)) => { (_, VirtualClassType(id_b)) => {
@ -82,7 +96,7 @@ fn find_subst(
parents = [*id_a].to_vec(); parents = [*id_a].to_vec();
} }
_ => { _ => {
return Err("cannot substitute non-class type into virtual class".to_string()); return Err(SubstError::NotVirtualClassSubtype(a.clone(), *id_b));
} }
}; };
while !parents.is_empty() { while !parents.is_empty() {
@ -92,11 +106,11 @@ fn find_subst(
let c = ctx.get_class_def(parents.remove(0)); let c = ctx.get_class_def(parents.remove(0));
parents.extend_from_slice(&c.parents); parents.extend_from_slice(&c.parents);
} }
Err("not subtype".to_string()) Err(SubstError::NotVirtualClassSubtype(a.clone(), *id_b))
} }
(ParametricType(id_a, param_a), ParametricType(id_b, param_b)) => { (ParametricType(id_a, param_a), ParametricType(id_b, param_b)) => {
if id_a != id_b || param_a.len() != param_b.len() { if id_a != id_b || param_a.len() != param_b.len() {
Err("different parametric types".to_string()) Err(SubstError::DifferentTypes(a.clone(), b.clone()))
} else { } else {
for (x, y) in param_a.iter().zip(param_b.iter()) { for (x, y) in param_a.iter().zip(param_b.iter()) {
find_subst(ctx, valuation, sub, x.clone(), y.clone())?; find_subst(ctx, valuation, sub, x.clone(), y.clone())?;
@ -108,7 +122,7 @@ fn find_subst(
if a == b { if a == b {
Ok(()) Ok(())
} else { } else {
Err("not equal".to_string()) Err(SubstError::DifferentTypes(a.clone(), b.clone()))
} }
} }
} }
@ -177,7 +191,7 @@ fn resolve_call_rec(
return Err("incorrect parameter number".to_string()); return Err("incorrect parameter number".to_string());
} }
for (a, b) in args.iter().zip(fun.args.iter()) { for (a, b) in args.iter().zip(fun.args.iter()) {
find_subst(ctx, valuation, &mut subst, a.clone(), b.clone())?; find_subst(ctx, valuation, &mut subst, a.clone(), b.clone()).map_err(|v| v.to_string())?;
} }
let result = fun.result.as_ref().map(|v| v.subst(&subst)); let result = fun.result.as_ref().map(|v| v.subst(&subst));
Ok(result.map(|result| { Ok(result.map(|result| {
@ -200,9 +214,11 @@ pub fn resolve_call(
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*; use super::{
use crate::context::TopLevelContext; super::{context::*, primitives::*},
use crate::primitives::*; *,
};
use std::matches;
use std::rc::Rc; use std::rc::Rc;
fn get_inference_context(ctx: TopLevelContext) -> InferenceContext { fn get_inference_context(ctx: TopLevelContext) -> InferenceContext {
@ -243,25 +259,25 @@ mod tests {
Ok(Some(ctx.get_primitive(FLOAT_TYPE))) Ok(Some(ctx.get_primitive(FLOAT_TYPE)))
); );
assert_eq!( assert!(matches!(
resolve_call(&ctx, None, "float", &[ctx.get_primitive(BOOL_TYPE)]), resolve_call(&ctx, None, "float", &[ctx.get_primitive(BOOL_TYPE)]),
Err("different domain".to_string()) Err(..)
); ));
assert_eq!( assert!(matches!(
resolve_call(&ctx, None, "float", &[]), resolve_call(&ctx, None, "float", &[]),
Err("incorrect parameter number".to_string()) Err(..)
); ));
assert_eq!( assert_eq!(
resolve_call(&ctx, None, "float", &[v1]), resolve_call(&ctx, None, "float", &[v1]),
Ok(Some(ctx.get_primitive(FLOAT_TYPE))) Ok(Some(ctx.get_primitive(FLOAT_TYPE)))
); );
assert_eq!( assert!(matches!(
resolve_call(&ctx, None, "float", &[v2]), resolve_call(&ctx, None, "float", &[v2]),
Err("different domain".to_string()) Err(..)
); ));
} }
#[test] #[test]
@ -308,40 +324,40 @@ mod tests {
Ok(Some(int64.clone())) Ok(Some(int64.clone()))
); );
assert_eq!( assert!(matches!(
resolve_call(&ctx, Some(int32), "__add__", &[int64]), resolve_call(&ctx, Some(int32), "__add__", &[int64]),
Err("not equal".to_string()) Err(..)
); ));
// with type variables // with type variables
assert_eq!( assert_eq!(
resolve_call(&ctx, Some(v1.clone()), "__add__", &[v1.clone()]), resolve_call(&ctx, Some(v1.clone()), "__add__", &[v1.clone()]),
Ok(Some(v1.clone())) Ok(Some(v1.clone()))
); );
assert_eq!( assert!(matches!(
resolve_call(&ctx, Some(v0.clone()), "__add__", &[v2.clone()]), resolve_call(&ctx, Some(v0.clone()), "__add__", &[v2.clone()]),
Err("unbounded type var".to_string()) Err(..)
); ));
assert_eq!( assert!(matches!(
resolve_call(&ctx, Some(v1.clone()), "__add__", &[v0]), resolve_call(&ctx, Some(v1.clone()), "__add__", &[v0]),
Err("different domain".to_string()) Err(..)
); ));
assert_eq!( assert!(matches!(
resolve_call(&ctx, Some(v1.clone()), "__add__", &[v2]), resolve_call(&ctx, Some(v1.clone()), "__add__", &[v2]),
Err("different domain".to_string()) Err(..)
); ));
assert_eq!( assert!(matches!(
resolve_call(&ctx, Some(v1.clone()), "__add__", &[v3.clone()]), resolve_call(&ctx, Some(v1.clone()), "__add__", &[v3.clone()]),
Err("different domain".to_string()) Err(..)
); ));
assert_eq!( assert!(matches!(
resolve_call(&ctx, Some(v3.clone()), "__add__", &[v1]), resolve_call(&ctx, Some(v3.clone()), "__add__", &[v1]),
Err("no such function".to_string()) Err(..)
); ));
assert_eq!( assert!(matches!(
resolve_call(&ctx, Some(v3.clone()), "__add__", &[v3]), resolve_call(&ctx, Some(v3.clone()), "__add__", &[v3]),
Err("no such function".to_string()) Err(..)
); ));
} }
#[test] #[test]
@ -393,10 +409,10 @@ mod tests {
resolve_call(&ctx, None, "foo", &[v2.clone(), v2.clone(), v3.clone()]), resolve_call(&ctx, None, "foo", &[v2.clone(), v2.clone(), v3.clone()]),
Ok(Some(v2.clone())) Ok(Some(v2.clone()))
); );
assert_eq!( assert!(matches!(
resolve_call(&ctx, None, "foo", &[v2.clone(), v3.clone(), v3.clone()]), resolve_call(&ctx, None, "foo", &[v2.clone(), v3.clone(), v3.clone()]),
Err("different variables".to_string()) Err(..)
); ));
assert_eq!( assert_eq!(
resolve_call( resolve_call(
@ -416,15 +432,15 @@ mod tests {
), ),
Ok(Some(v2.clone())) Ok(Some(v2.clone()))
); );
assert_eq!( assert!(matches!(
resolve_call( resolve_call(
&ctx, &ctx,
None, None,
"foo1", "foo1",
&[ParametricType(TUPLE_TYPE, vec![v2, v3.clone(), v3]).into()] &[ParametricType(TUPLE_TYPE, vec![v2, v3.clone(), v3]).into()]
), ),
Err("different variables".to_string()) Err(..)
); ));
} }
#[test] #[test]
@ -478,15 +494,15 @@ mod tests {
), ),
Ok(None) Ok(None)
); );
assert_eq!( assert!(matches!(
resolve_call( resolve_call(
&ctx, &ctx,
Some(ParametricType(LIST_TYPE, vec![v0]).into()), Some(ParametricType(LIST_TYPE, vec![v0]).into()),
"append", "append",
&[v1] &[v1]
), ),
Err("different variables".to_string()) Err(..)
); ));
} }
#[test] #[test]
@ -560,10 +576,10 @@ mod tests {
Ok(None) Ok(None)
); );
assert_eq!( assert!(matches!(
resolve_call(&ctx, None, "foo", &[ClassType(bar).into()]), resolve_call(&ctx, None, "foo", &[ClassType(bar).into()]),
Err("not subtype".to_string()) Err(..)
); ));
assert_eq!( assert_eq!(
resolve_call(&ctx, None, "foo1", &[ClassType(foo1).into()]), resolve_call(&ctx, None, "foo1", &[ClassType(foo1).into()]),
@ -575,10 +591,10 @@ mod tests {
Ok(None) Ok(None)
); );
assert_eq!( assert!(matches!(
resolve_call(&ctx, None, "foo1", &[ClassType(foo).into()]), resolve_call(&ctx, None, "foo1", &[ClassType(foo).into()]),
Err("not subtype".to_string()) Err(..)
); ));
// virtual class substitution // virtual class substitution
assert_eq!( assert_eq!(
@ -593,9 +609,9 @@ mod tests {
resolve_call(&ctx, None, "foo", &[VirtualClassType(foo2).into()]), resolve_call(&ctx, None, "foo", &[VirtualClassType(foo2).into()]),
Ok(None) Ok(None)
); );
assert_eq!( assert!(matches!(
resolve_call(&ctx, None, "foo", &[VirtualClassType(bar).into()]), resolve_call(&ctx, None, "foo", &[VirtualClassType(bar).into()]),
Err("not subtype".to_string()) Err(..)
); ));
} }
} }

View File

@ -0,0 +1,9 @@
pub mod context;
pub mod expression_inference;
pub mod inference_core;
mod magic_methods;
pub mod primitives;
pub mod statement_check;
pub mod typedef;
pub mod signature;

View File

@ -1,7 +1,9 @@
use super::context::*;
use super::typedef::{TypeEnum::*, *}; use super::typedef::{TypeEnum::*, *};
use crate::context::*;
use std::collections::HashMap; use std::collections::HashMap;
pub const PRIMITIVES: [&str; 6] = ["int32", "int64", "float", "bool", "list", "tuple"];
pub const TUPLE_TYPE: ParamId = ParamId(0); pub const TUPLE_TYPE: ParamId = ParamId(0);
pub const LIST_TYPE: ParamId = ParamId(1); pub const LIST_TYPE: ParamId = ParamId(1);
@ -17,8 +19,11 @@ fn impl_math(def: &mut TypeDef, ty: &Type) {
result: result.clone(), result: result.clone(),
}; };
def.methods.insert("__add__", fun.clone()); def.methods.insert("__add__", fun.clone());
def.methods.insert("__iadd__", fun.clone());
def.methods.insert("__sub__", fun.clone()); def.methods.insert("__sub__", fun.clone());
def.methods.insert("__isub__", fun.clone());
def.methods.insert("__mul__", fun.clone()); def.methods.insert("__mul__", fun.clone());
def.methods.insert("__imul__", fun.clone());
def.methods.insert( def.methods.insert(
"__neg__", "__neg__",
FnDef { FnDef {
@ -33,9 +38,21 @@ fn impl_math(def: &mut TypeDef, ty: &Type) {
result: Some(PrimitiveType(FLOAT_TYPE).into()), result: Some(PrimitiveType(FLOAT_TYPE).into()),
}, },
); );
if ty.as_ref() == &PrimitiveType(FLOAT_TYPE) {
def.methods.insert(
"__itruediv__",
FnDef {
args: vec![ty.clone()],
result: Some(PrimitiveType(FLOAT_TYPE).into()),
},
);
}
def.methods.insert("__floordiv__", fun.clone()); def.methods.insert("__floordiv__", fun.clone());
def.methods.insert("__ifloordiv__", fun.clone());
def.methods.insert("__mod__", fun.clone()); def.methods.insert("__mod__", fun.clone());
def.methods.insert("__pow__", fun); def.methods.insert("__imod__", fun.clone());
def.methods.insert("__pow__", fun.clone());
def.methods.insert("__ipow__", fun);
} }
fn impl_bits(def: &mut TypeDef, ty: &Type) { fn impl_bits(def: &mut TypeDef, ty: &Type) {

View File

@ -0,0 +1,503 @@
/// obtain class and function signature from AST
use super::context::TopLevelContext;
use super::primitives::*;
use super::typedef::*;
use rustpython_parser::ast::{
ComprehensionKind, ExpressionType, Statement, StatementType, StringGroup,
};
use std::collections::HashMap;
// TODO: fix condition checking, return error message instead of panic...
fn typename_from_expr<'a>(typenames: &mut Vec<&'a str>, expr: &'a ExpressionType) {
match expr {
ExpressionType::Identifier { name } => typenames.push(&name),
ExpressionType::String { value } => match value {
StringGroup::Constant { value } => typenames.push(&value),
_ => unimplemented!(),
},
ExpressionType::Subscript { a, b } => {
typename_from_expr(typenames, &b.node);
typename_from_expr(typenames, &a.node)
}
_ => unimplemented!(),
}
}
fn typename_from_fn<'a>(typenames: &mut Vec<&'a str>, fun: &'a StatementType) {
match fun {
StatementType::FunctionDef { args, returns, .. } => {
for arg in args.args.iter() {
if let Some(ann) = &arg.annotation {
typename_from_expr(typenames, &ann.node);
}
}
if let Some(returns) = &returns {
typename_from_expr(typenames, &returns.node);
}
}
_ => unreachable!(),
}
}
fn name_from_expr<'a>(expr: &'a ExpressionType) -> &'a str {
match &expr {
ExpressionType::Identifier { name } => &name,
ExpressionType::String { value } => match value {
StringGroup::Constant { value } => &value,
_ => unimplemented!(),
},
_ => unimplemented!(),
}
}
fn type_from_expr<'a>(ctx: &'a TopLevelContext, expr: &'a ExpressionType) -> Result<Type, String> {
match expr {
ExpressionType::Identifier { name } => {
ctx.get_type(name).ok_or_else(|| "no such type".into())
}
ExpressionType::String { value } => match value {
StringGroup::Constant { value } => {
ctx.get_type(&value).ok_or_else(|| "no such type".into())
}
_ => unimplemented!(),
},
ExpressionType::Subscript { a, b } => {
if let ExpressionType::Identifier { name } = &a.node {
match name.as_str() {
"list" => {
let ty = type_from_expr(ctx, &b.node)?;
Ok(TypeEnum::ParametricType(LIST_TYPE, vec![ty]).into())
}
"tuple" => {
if let ExpressionType::Tuple { elements } = &b.node {
let ty_list: Result<Vec<_>, _> = elements
.iter()
.map(|v| type_from_expr(ctx, &v.node))
.collect();
Ok(TypeEnum::ParametricType(TUPLE_TYPE, ty_list?).into())
} else {
Err("unsupported format".into())
}
}
_ => Err("no such parameterized type".into()),
}
} else {
// we require a to be an identifier, for a[b]
Err("unsupported format".into())
}
}
_ => Err("unsupported format".into()),
}
}
pub fn get_typenames<'a>(stmts: &'a [Statement]) -> (Vec<&'a str>, Vec<&'a str>) {
let mut classes = Vec::new();
let mut typenames = Vec::new();
for stmt in stmts.iter() {
match &stmt.node {
StatementType::ClassDef {
name, body, bases, ..
} => {
// check if class is not duplicated...
// and annotations
classes.push(&name[..]);
for base in bases.iter() {
let name = name_from_expr(&base.node);
typenames.push(name);
}
// may check if fields/functions are not duplicated
for stmt in body.iter() {
match &stmt.node {
StatementType::AnnAssign { annotation, .. } => {
typename_from_expr(&mut typenames, &annotation.node)
}
StatementType::FunctionDef { .. } => {
typename_from_fn(&mut typenames, &stmt.node);
}
_ => unimplemented!(),
}
}
}
StatementType::FunctionDef { .. } => {
// may check annotations
typename_from_fn(&mut typenames, &stmt.node);
}
_ => (),
}
}
let mut unknowns = Vec::new();
for n in typenames {
if !PRIMITIVES.contains(&n) && !classes.contains(&n) && !unknowns.contains(&n) {
unknowns.push(n);
}
}
(classes, unknowns)
}
fn resolve_function<'a>(
ctx: &'a TopLevelContext,
fun: &'a StatementType,
method: bool,
) -> Result<FnDef, String> {
if let StatementType::FunctionDef { args, returns, .. } = &fun {
let args = if method {
args.args[1..].iter()
} else {
args.args.iter()
};
let args: Result<Vec<_>, _> = args
.map(|arg| type_from_expr(ctx, &arg.annotation.as_ref().unwrap().node))
.collect();
let args = args?;
let result = match returns {
Some(v) => Some(type_from_expr(ctx, &v.node)?),
None => None,
};
Ok(FnDef { args, result })
} else {
unreachable!()
}
}
fn get_expr_unknowns<'a>(
defined: &mut Vec<&'a str>,
unknowns: &mut Vec<&'a str>,
expr: &'a ExpressionType,
) {
match expr {
ExpressionType::BoolOp { values, .. } => {
for v in values.iter() {
get_expr_unknowns(defined, unknowns, &v.node)
}
}
ExpressionType::Binop { a, b, .. } => {
get_expr_unknowns(defined, unknowns, &a.node);
get_expr_unknowns(defined, unknowns, &b.node);
}
ExpressionType::Subscript { a, b } => {
get_expr_unknowns(defined, unknowns, &a.node);
get_expr_unknowns(defined, unknowns, &b.node);
}
ExpressionType::Unop { a, .. } => {
get_expr_unknowns(defined, unknowns, &a.node);
}
ExpressionType::Compare { vals, .. } => {
for v in vals.iter() {
get_expr_unknowns(defined, unknowns, &v.node)
}
}
ExpressionType::Attribute { value, .. } => {
get_expr_unknowns(defined, unknowns, &value.node);
}
ExpressionType::Call { function, args, .. } => {
get_expr_unknowns(defined, unknowns, &function.node);
for v in args.iter() {
get_expr_unknowns(defined, unknowns, &v.node)
}
}
ExpressionType::List { elements } => {
for v in elements.iter() {
get_expr_unknowns(defined, unknowns, &v.node)
}
}
ExpressionType::Tuple { elements } => {
for v in elements.iter() {
get_expr_unknowns(defined, unknowns, &v.node)
}
}
ExpressionType::Comprehension { kind, generators } => {
if generators.len() != 1 {
unimplemented!()
}
let g = &generators[0];
get_expr_unknowns(defined, unknowns, &g.iter.node);
let mut scoped = defined.clone();
get_expr_unknowns(defined, &mut scoped, &g.target.node);
for if_expr in g.ifs.iter() {
get_expr_unknowns(&mut scoped, unknowns, &if_expr.node);
}
match kind.as_ref() {
ComprehensionKind::List { element } => {
get_expr_unknowns(&mut scoped, unknowns, &element.node);
}
_ => unimplemented!(),
}
}
ExpressionType::Slice { elements } => {
for v in elements.iter() {
get_expr_unknowns(defined, unknowns, &v.node);
}
}
ExpressionType::Identifier { name } => {
if !defined.contains(&name.as_str()) && !unknowns.contains(&name.as_str()) {
unknowns.push(name);
}
}
ExpressionType::IfExpression { test, body, orelse } => {
get_expr_unknowns(defined, unknowns, &test.node);
get_expr_unknowns(defined, unknowns, &body.node);
get_expr_unknowns(defined, unknowns, &orelse.node);
}
_ => (),
};
}
struct ExprPattern<'a>(&'a ExpressionType, Vec<usize>, bool);
impl<'a> ExprPattern<'a> {
fn new(expr: &'a ExpressionType) -> ExprPattern {
let mut pattern = ExprPattern(expr, Vec::new(), true);
pattern.find_leaf();
pattern
}
fn pointed(&mut self) -> &'a ExpressionType {
let mut current = self.0;
for v in self.1.iter() {
if let ExpressionType::Tuple { elements } = current {
current = &elements[*v].node
} else {
unreachable!()
}
}
current
}
fn find_leaf(&mut self) {
let mut current = self.pointed();
while let ExpressionType::Tuple { elements } = current {
if elements.is_empty() {
break;
}
current = &elements[0].node;
self.1.push(0);
}
}
fn inc(&mut self) -> bool {
loop {
if self.1.is_empty() {
return false;
}
let ind = self.1.pop().unwrap() + 1;
let parent = self.pointed();
if let ExpressionType::Tuple { elements } = parent {
if ind < elements.len() {
self.1.push(ind);
self.find_leaf();
return true;
}
} else {
unreachable!()
}
}
}
}
impl<'a> Iterator for ExprPattern<'a> {
type Item = &'a ExpressionType;
fn next(&mut self) -> Option<Self::Item> {
if self.2 {
self.2 = false;
Some(self.pointed())
} else if self.inc() {
Some(self.pointed())
} else {
None
}
}
}
fn get_stmt_unknowns<'a>(
defined: &mut Vec<&'a str>,
unknowns: &mut Vec<&'a str>,
stmts: &'a [Statement],
) {
for stmt in stmts.iter() {
match &stmt.node {
StatementType::Return { value } => {
if let Some(value) = value {
get_expr_unknowns(defined, unknowns, &value.node);
}
}
StatementType::Assign { targets, value } => {
get_expr_unknowns(defined, unknowns, &value.node);
for target in targets.iter() {
for node in ExprPattern::new(&target.node).into_iter() {
if let ExpressionType::Identifier { name } = node {
let name = name.as_str();
if !defined.contains(&name) {
defined.push(name);
}
} else {
get_expr_unknowns(defined, unknowns, node);
}
}
}
}
StatementType::AugAssign { target, value, .. } => {
get_expr_unknowns(defined, unknowns, &target.node);
get_expr_unknowns(defined, unknowns, &value.node);
}
StatementType::AnnAssign { target, value, .. } => {
get_expr_unknowns(defined, unknowns, &target.node);
if let Some(value) = value {
get_expr_unknowns(defined, unknowns, &value.node);
}
}
StatementType::Expression { expression } => {
get_expr_unknowns(defined, unknowns, &expression.node);
}
StatementType::Global { names } => {
for name in names.iter() {
let name = name.as_str();
if !unknowns.contains(&name) {
unknowns.push(name);
}
}
}
StatementType::If { test, body, orelse }
| StatementType::While { test, body, orelse } => {
get_expr_unknowns(defined, unknowns, &test.node);
// we are not very strict at this point...
// some identifiers not treated as unknowns may not be resolved
// but should be checked during type inference
get_stmt_unknowns(defined, unknowns, body.as_slice());
if let Some(orelse) = orelse {
get_stmt_unknowns(defined, unknowns, orelse.as_slice());
}
}
StatementType::For { is_async, target, iter, body, orelse } => {
if *is_async {
unimplemented!()
}
get_expr_unknowns(defined, unknowns, &iter.node);
for node in ExprPattern::new(&target.node).into_iter() {
if let ExpressionType::Identifier { name } = node {
let name = name.as_str();
if !defined.contains(&name) {
defined.push(name);
}
} else {
get_expr_unknowns(defined, unknowns, node);
}
}
get_stmt_unknowns(defined, unknowns, body.as_slice());
if let Some(orelse) = orelse {
get_stmt_unknowns(defined, unknowns, orelse.as_slice());
}
}
_ => (),
}
}
}
pub fn resolve_signatures<'a>(ctx: &mut TopLevelContext<'a>, stmts: &'a [Statement]) {
for stmt in stmts.iter() {
match &stmt.node {
StatementType::ClassDef {
name, bases, body, ..
} => {
let mut parents = Vec::new();
for base in bases.iter() {
let name = name_from_expr(&base.node);
let c = ctx.get_type(name).unwrap();
let id = if let TypeEnum::ClassType(id) = c.as_ref() {
*id
} else {
unreachable!()
};
parents.push(id);
}
let mut fields = HashMap::new();
let mut functions = HashMap::new();
for stmt in body.iter() {
match &stmt.node {
StatementType::AnnAssign {
target, annotation, ..
} => {
let name = name_from_expr(&target.node);
let ty = type_from_expr(ctx, &annotation.node).unwrap();
fields.insert(name, ty);
}
StatementType::FunctionDef { name, .. } => {
functions.insert(
&name[..],
resolve_function(ctx, &stmt.node, true).unwrap(),
);
}
_ => unimplemented!(),
}
}
let class = ctx.get_type(name).unwrap();
let class = if let TypeEnum::ClassType(id) = class.as_ref() {
ctx.get_class_def_mut(*id)
} else {
unreachable!()
};
class.parents.extend_from_slice(&parents);
class.base.fields.clone_from(&fields);
class.base.methods.clone_from(&functions);
}
StatementType::FunctionDef { name, .. } => {
ctx.add_fn(&name[..], resolve_function(ctx, &stmt.node, false).unwrap());
}
_ => unimplemented!(),
}
}
}
#[cfg(test)]
mod tests {
use super::*;
use indoc::indoc;
use rustpython_parser::parser::{parse_program, parse_statement};
#[test]
fn test_get_classes() {
let ast = parse_program(indoc! {"
class Foo:
a: int32
b: Test
def test(self, a: int32) -> Test2:
return self.b
class Bar(Foo, 'FooBar'):
def test2(self, a: list[Foo]) -> Test2:
return self.b
def test3(self, a: list[FooBar2]) -> Test2:
return self.b
" })
.unwrap();
let (mut classes, mut unknowns) = get_typenames(&ast.statements);
let classes_count = classes.len();
let unknowns_count = unknowns.len();
classes.sort();
unknowns.sort();
assert_eq!(classes.len(), classes_count);
assert_eq!(unknowns.len(), unknowns_count);
assert_eq!(&classes, &["Bar", "Foo"]);
assert_eq!(&unknowns, &["FooBar", "FooBar2", "Test", "Test2"]);
}
#[test]
fn test_assignment() {
let ast = parse_statement(indoc! {"
((a, b), c[i]) = core.foo(x, get_y())
" })
.unwrap();
let mut defined = Vec::new();
let mut unknowns = Vec::new();
get_stmt_unknowns(&mut defined, &mut unknowns, ast.as_slice());
defined.sort();
unknowns.sort();
assert_eq!(defined.as_slice(), &["a", "b"]);
assert_eq!(unknowns.as_slice(), &["c", "core", "get_y", "i", "x"]);
}
}

View File

@ -0,0 +1,572 @@
use super::context::InferenceContext;
use super::expression_inference::{infer_expr, infer_simple_binding};
use super::inference_core::resolve_call;
use super::magic_methods::binop_assign_name;
use super::primitives::*;
use super::typedef::{Type, TypeEnum::*};
use rustpython_parser::ast::*;
pub fn check_stmts<'b: 'a, 'a>(
ctx: &mut InferenceContext<'a>,
stmts: &'b [Statement],
) -> Result<bool, String> {
for stmt in stmts.iter() {
match &stmt.node {
StatementType::Assign { targets, value } => {
check_assign(ctx, targets.as_slice(), &value)?;
}
StatementType::AugAssign { target, op, value } => {
check_aug_assign(ctx, &target, op, &value)?;
}
StatementType::If { test, body, orelse } => {
if check_if(ctx, test, body.as_slice(), orelse)? {
return Ok(true);
}
}
StatementType::While { test, body, orelse } => {
check_while_stmt(ctx, test, body.as_slice(), orelse)?;
}
StatementType::For {
is_async,
target,
iter,
body,
orelse,
} => {
if *is_async {
return Err("async for is not supported".to_string());
}
check_for_stmt(ctx, target, iter, body.as_slice(), orelse)?;
}
StatementType::Return { value } => {
let result = ctx.get_result();
let t = if let Some(value) = value {
infer_expr(ctx, value)?
} else {
None
};
return if t == result {
Ok(true)
} else {
Err("return type mismatch".to_string())
};
}
StatementType::Continue | StatementType::Break => {
continue;
}
_ => return Err("not supported".to_string()),
}
}
Ok(false)
}
fn get_target_type<'b: 'a, 'a>(
ctx: &mut InferenceContext<'a>,
target: &'b Expression,
) -> Result<Type, String> {
match &target.node {
ExpressionType::Subscript { a, b } => {
let int32 = ctx.get_primitive(INT32_TYPE);
if infer_expr(ctx, &a)? == Some(int32) {
let b = get_target_type(ctx, &b)?;
if let ParametricType(LIST_TYPE, t) = b.as_ref() {
Ok(t[0].clone())
} else {
Err("subscript is only supported for list".to_string())
}
} else {
Err("subscript must be int32".to_string())
}
}
ExpressionType::Attribute { value, name } => {
let t = get_target_type(ctx, &value)?;
let base = t.get_base(ctx).ok_or_else(|| "no attributes".to_string())?;
Ok(base
.fields
.get(name.as_str())
.ok_or_else(|| "no such attribute")?
.clone())
}
ExpressionType::Identifier { name } => Ok(ctx.resolve(name.as_str())?),
_ => Err("not supported".to_string()),
}
}
fn check_stmt_binding<'b: 'a, 'a>(
ctx: &mut InferenceContext<'a>,
target: &'b Expression,
ty: Type,
) -> Result<(), String> {
match &target.node {
ExpressionType::Identifier { name } => {
if name.as_str() == "_" {
Ok(())
} else {
match ctx.resolve(name.as_str()) {
Ok(t) if t == ty => Ok(()),
Err(_) => {
ctx.assign(name.as_str(), ty).unwrap();
Ok(())
}
_ => Err("conflicting type".into()),
}
}
}
ExpressionType::Tuple { elements } => {
if let ParametricType(TUPLE_TYPE, ls) = ty.as_ref() {
if ls.len() != elements.len() {
return Err("incorrect pattern length".into());
}
for (x, y) in elements.iter().zip(ls.iter()) {
check_stmt_binding(ctx, x, y.clone())?;
}
Ok(())
} else {
Err("pattern matching supports tuple only".into())
}
}
_ => {
let t = get_target_type(ctx, target)?;
if ty == t {
Ok(())
} else {
Err("type mismatch".into())
}
}
}
}
fn check_assign<'b: 'a, 'a>(
ctx: &mut InferenceContext<'a>,
targets: &'b [Expression],
value: &'b Expression,
) -> Result<(), String> {
let ty = infer_expr(ctx, value)?.ok_or_else(|| "no value".to_string())?;
for t in targets.iter() {
check_stmt_binding(ctx, t, ty.clone())?;
}
Ok(())
}
fn check_aug_assign<'b: 'a, 'a>(
ctx: &mut InferenceContext<'a>,
target: &'b Expression,
op: &'b Operator,
value: &'b Expression,
) -> Result<(), String> {
let left = infer_expr(ctx, target)?.ok_or_else(|| "no value".to_string())?;
let right = infer_expr(ctx, value)?.ok_or_else(|| "no value".to_string())?;
let fun = binop_assign_name(op);
resolve_call(ctx, Some(left), fun, &[right])?;
Ok(())
}
fn check_if<'b: 'a, 'a>(
ctx: &mut InferenceContext<'a>,
test: &'b Expression,
body: &'b [Statement],
orelse: &'b Option<Suite>,
) -> Result<bool, String> {
let boolean = ctx.get_primitive(BOOL_TYPE);
let t = infer_expr(ctx, test)?;
if t == Some(boolean) {
let (names, result) = ctx.with_scope(|ctx| check_stmts(ctx, body));
let returned = result?;
if let Some(orelse) = orelse {
let (names2, result) = ctx.with_scope(|ctx| check_stmts(ctx, orelse.as_slice()));
let returned = returned && result?;
for (name, ty) in names.iter() {
for (name2, ty2) in names2.iter() {
if *name == *name2 && ty == ty2 {
ctx.assign(name, ty.clone()).unwrap();
}
}
}
Ok(returned)
} else {
Ok(false)
}
} else {
Err("condition should be bool".to_string())
}
}
fn check_while_stmt<'b: 'a, 'a>(
ctx: &mut InferenceContext<'a>,
test: &'b Expression,
body: &'b [Statement],
orelse: &'b Option<Suite>,
) -> Result<bool, String> {
let boolean = ctx.get_primitive(BOOL_TYPE);
let t = infer_expr(ctx, test)?;
if t == Some(boolean) {
// to check what variables are defined, we would have to do a graph analysis...
// not implemented now
let (_, result) = ctx.with_scope(|ctx| check_stmts(ctx, body));
result?;
if let Some(orelse) = orelse {
let (_, result) = ctx.with_scope(|ctx| check_stmts(ctx, orelse.as_slice()));
result?;
}
// to check whether the loop returned on every possible path, we need to analyse the graph,
// not implemented now
Ok(false)
} else {
Err("condition should be bool".to_string())
}
}
fn check_for_stmt<'b: 'a, 'a>(
ctx: &mut InferenceContext<'a>,
target: &'b Expression,
iter: &'b Expression,
body: &'b [Statement],
orelse: &'b Option<Suite>,
) -> Result<bool, String> {
let ty = infer_expr(ctx, iter)?.ok_or_else(|| "no value".to_string())?;
if let ParametricType(LIST_TYPE, ls) = ty.as_ref() {
let (_, result) = ctx.with_scope(|ctx| {
infer_simple_binding(ctx, target, ls[0].clone())?;
check_stmts(ctx, body)
});
result?;
if let Some(orelse) = orelse {
let (_, result) = ctx.with_scope(|ctx| check_stmts(ctx, orelse.as_slice()));
result?;
}
// to check whether the loop returned on every possible path, we need to analyse the graph,
// not implemented now
Ok(false)
} else {
Err("only list can be iterated over".to_string())
}
}
#[cfg(test)]
mod test {
use super::{super::context::*, *};
use indoc::indoc;
use rustpython_parser::parser::parse_program;
fn get_inference_context(ctx: TopLevelContext) -> InferenceContext {
InferenceContext::new(ctx, Box::new(|_| Err("unbounded identifier".into())))
}
#[test]
fn test_assign() {
let ctx = basic_ctx();
let mut ctx = get_inference_context(ctx);
let ast = parse_program(indoc! {"
a = 1
b = a * 2
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
a = 1
b = b * 2
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("unbounded identifier".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
let ast = parse_program(indoc! {"
b = a = 1
b = b * 2
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
b = a = 1
b = [a]
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("conflicting type".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
}
#[test]
fn test_if() {
let ctx = basic_ctx();
let mut ctx = get_inference_context(ctx);
let ast = parse_program(indoc! {"
a = 1
b = a * 2
if b > a:
c = 1
else:
c = 0
d = c
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
a = 1
b = a * 2
if b > a:
c = 1
else:
d = 0
d = c
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("unbounded identifier".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
let ast = parse_program(indoc! {"
a = 1
b = a * 2
if b > a:
c = 1
d = c
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("unbounded identifier".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
let ast = parse_program(indoc! {"
a = 1
b = a * 2
if a:
b = 0
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("condition should be bool".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
let ast = parse_program(indoc! {"
a = 1
b = a * 2
if b > a:
c = 1
c = [1]
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
a = 1
b = a * 2
if b > a:
c = 1
else:
c = 0
c = [1]
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("conflicting type".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
}
#[test]
fn test_while() {
let ctx = basic_ctx();
let mut ctx = get_inference_context(ctx);
let ast = parse_program(indoc! {"
a = 1
b = 1
while a < 10:
a += 1
b *= a
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
a = 1
b = 1
while a < 10:
a += 1
b *= a
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
a = 1
b = 1
while a < 10:
a += 1
b *= a
else:
a += 1
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
a = 1
b = 1
while a:
a += 1
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("condition should be bool".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
let ast = parse_program(indoc! {"
a = 1
b = 1
while a < 10:
a += 1
c = a*2
else:
c = a*2
b = c
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("unbounded identifier".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
}
#[test]
fn test_for() {
let ctx = basic_ctx();
let mut ctx = get_inference_context(ctx);
let ast = parse_program(indoc! {"
b = 1
for a in [0, 1, 2, 3, 4, 5]:
b *= a
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
b = 1
for a, a1 in [(0, 1), (2, 3), (4, 5)]:
b *= a
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
}
#[test]
fn test_return() {
let ctx = basic_ctx();
let mut ctx = get_inference_context(ctx);
let ast = parse_program(indoc! {"
b = 1
return
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(true), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
b = 1
if b > 0:
return
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
b = 1
if b > 0:
return
else:
return
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(true), check_stmts(ctx, ast.statements.as_slice()));
});
let ast = parse_program(indoc! {"
b = 1
while b > 0:
return
else:
return
" })
.unwrap();
ctx.with_scope(|ctx| {
// with sophisticated analysis, this one should be Ok(true)
// but with our simple implementation, this is Ok(false)
// as we don't analyse the control flow
assert_eq!(Ok(false), check_stmts(ctx, ast.statements.as_slice()));
});
ctx.set_result(Some(ctx.get_primitive(INT32_TYPE)));
let ast = parse_program(indoc! {"
b = 1
return 1
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(Ok(true), check_stmts(ctx, ast.statements.as_slice()));
});
ctx.set_result(Some(ctx.get_primitive(INT32_TYPE)));
let ast = parse_program(indoc! {"
b = 1
return [1]
" })
.unwrap();
ctx.with_scope(|ctx| {
assert_eq!(
Err("return type mismatch".to_string()),
check_stmts(ctx, ast.statements.as_slice())
);
});
}
}

34
todo.txt Normal file
View File

@ -0,0 +1,34 @@
Errors:
- Not supported
- Only * is supported
- Expected * in *, but got *
- Divergent type in (construct), (location 1), (location 2)
- Unknown field
- Unbounded variable
- Different variable
- Different domain
- * is not subclass of *
- Type not equal
- Incorrect number of parameters
Symbol Resolution:
- Add all files with annotated class/functions.
- Find class references, load them all in TopLevelContext.
- Find unbounded identifiers in the functions.
- If it is a function/class name, record its object ID.
- Otherwise, load its value. (check to see if specified with `global`)
(Function implemented in python, with rust binding to add value to global
variable dictionary)
Global variable dictionary:
- Primitives, including integers, floats, bools, etc.
- Primitive lists.
- Numpy multi-dimensional array, with value + dimension vectors.
- Reference array, with integer index referring to other things.
- Symbol table: python id -> reference id.
TopLevelContext/InferenceContext:
- Restrict visibility by user defined function.