Compare commits

...

3 Commits

Author SHA1 Message Date
0392c293ba Cleanup 2025-07-05 01:41:03 +03:00
61ba00ea58 Transform AST into MIR successfully, completing the chain 2025-07-05 01:34:57 +03:00
a66fc61c07 Fiddle with names a bit 2025-07-04 21:54:07 +03:00
16 changed files with 875 additions and 1142 deletions

View File

@ -6,7 +6,7 @@
# Do note this file is extremely simply for my own personal convenience
export .env
cargo run --example testcodegen && \
cargo run --example $1 && \
# clang hello.o -o main && \
ld -dynamic-linker /lib64/ld-linux-x86-64.so.2 \
-o main /usr/lib/crt1.o hello.o -lc && \

View File

@ -0,0 +1,11 @@
use reid::compile;
pub static FIBONACCI: &str = include_str!("./reid/fibonacci.reid");
fn main() {
let text = match compile(FIBONACCI) {
Ok(t) => t,
Err(e) => panic!("{}", e),
};
println!("{}", text);
}

View File

@ -3,12 +3,11 @@
fn main() {
let test = 9;
let simpleAdd = 2 + 2;
let simpleMult = 7 * 2; // 14
let arithmetic = 3 + 2 * 5 + 1 * 2; // 15
let simpleSub = 7 - 2; // 14
if simpleAdd < test {
return 3;
}
return arithmetic + simpleMult * arithmetic;
return arithmetic + simpleSub + boop;
}

View File

@ -11,115 +11,114 @@ fn main() {
kind: FunctionDefinitionKind::Local(
Block {
statements: vec![Statement(
StatementKind::If(IfExpression(
// If N < 3
Box::new(Expression(
ExpressionKind::BinOp(
BinaryOperator::Logic(LogicOperator::GreaterThan),
Box::new(Expression(
ExpressionKind::Variable(VariableReference(
TypeKind::I32,
"N".to_string(),
StmtKind::Expression(Expression(
ExprKind::If(IfExpression(
// If N < 3
Box::new(Expression(
ExprKind::BinOp(
BinaryOperator::Logic(LogicOperator::GreaterThan),
Box::new(Expression(
ExprKind::Variable(VariableReference(
TypeKind::I32,
"N".to_string(),
Default::default(),
)),
Default::default(),
)),
Default::default(),
)),
Box::new(Expression(
ExpressionKind::Literal(Literal::I32(2)),
Default::default(),
)),
),
Default::default(),
)),
// Then
Block {
statements: vec![],
return_expression: Some((
ReturnKind::HardReturn,
// return fibonacci(n-1) + fibonacci(n-2)
Box::new(Expression(
ExpressionKind::BinOp(
BinaryOperator::Add,
// fibonacci(n-1)
Box::new(Expression(
ExpressionKind::FunctionCall(FunctionCall {
name: fibonacci_name.clone(),
return_type: TypeKind::I32,
parameters: vec![Expression(
ExpressionKind::BinOp(
BinaryOperator::Minus,
Box::new(Expression(
ExpressionKind::Variable(
VariableReference(
TypeKind::I32,
fibonacci_n.clone(),
Default::default(),
),
),
Default::default(),
)),
Box::new(Expression(
ExpressionKind::Literal(Literal::I32(
1,
)),
Default::default(),
)),
),
Default::default(),
)],
}),
Default::default(),
)),
// fibonacci(n-2)
Box::new(Expression(
ExpressionKind::FunctionCall(FunctionCall {
name: fibonacci_name.clone(),
return_type: TypeKind::I32,
parameters: vec![Expression(
ExpressionKind::BinOp(
BinaryOperator::Minus,
Box::new(Expression(
ExpressionKind::Variable(
VariableReference(
TypeKind::I32,
fibonacci_n.clone(),
Default::default(),
),
),
Default::default(),
)),
Box::new(Expression(
ExpressionKind::Literal(Literal::I32(
2,
)),
Default::default(),
)),
),
Default::default(),
)],
}),
Default::default(),
)),
),
Default::default(),
)),
Box::new(Expression(
ExprKind::Literal(Literal::I32(2)),
Default::default(),
)),
),
Default::default(),
)),
range: Default::default(),
},
// No else-block
None,
// Then
Block {
statements: vec![],
return_expression: Some((
ReturnKind::Hard,
// return fibonacci(n-1) + fibonacci(n-2)
Box::new(Expression(
ExprKind::BinOp(
BinaryOperator::Add,
// fibonacci(n-1)
Box::new(Expression(
ExprKind::FunctionCall(FunctionCall {
name: fibonacci_name.clone(),
return_type: TypeKind::I32,
parameters: vec![Expression(
ExprKind::BinOp(
BinaryOperator::Minus,
Box::new(Expression(
ExprKind::Variable(
VariableReference(
TypeKind::I32,
fibonacci_n.clone(),
Default::default(),
),
),
Default::default(),
)),
Box::new(Expression(
ExprKind::Literal(Literal::I32(1)),
Default::default(),
)),
),
Default::default(),
)],
}),
Default::default(),
)),
// fibonacci(n-2)
Box::new(Expression(
ExprKind::FunctionCall(FunctionCall {
name: fibonacci_name.clone(),
return_type: TypeKind::I32,
parameters: vec![Expression(
ExprKind::BinOp(
BinaryOperator::Minus,
Box::new(Expression(
ExprKind::Variable(
VariableReference(
TypeKind::I32,
fibonacci_n.clone(),
Default::default(),
),
),
Default::default(),
)),
Box::new(Expression(
ExprKind::Literal(Literal::I32(2)),
Default::default(),
)),
),
Default::default(),
)],
}),
Default::default(),
)),
),
Default::default(),
)),
)),
meta: Default::default(),
},
// No else-block
None,
)),
Default::default(),
)),
Default::default(),
)],
// return 1
return_expression: Some((
ReturnKind::SoftReturn,
ReturnKind::Soft,
Box::new(Expression(
ExpressionKind::Literal(Literal::I32(1)),
ExprKind::Literal(Literal::I32(1)),
Default::default(),
)),
)),
range: Default::default(),
meta: Default::default(),
},
Default::default(),
),
@ -132,20 +131,20 @@ fn main() {
Block {
statements: vec![],
return_expression: Some((
ReturnKind::SoftReturn,
ReturnKind::Soft,
Box::new(Expression(
ExpressionKind::FunctionCall(FunctionCall {
ExprKind::FunctionCall(FunctionCall {
name: fibonacci_name.clone(),
return_type: TypeKind::I32,
parameters: vec![Expression(
ExpressionKind::Literal(Literal::I32(5)),
ExprKind::Literal(Literal::I32(5)),
Default::default(),
)],
}),
Default::default(),
)),
)),
range: Default::default(),
meta: Default::default(),
},
Default::default(),
),

109
reid/src/ast/mod.rs Normal file
View File

@ -0,0 +1,109 @@
use crate::token_stream::TokenRange;
pub mod parse;
pub mod process;
#[derive(Debug, Clone, Copy)]
pub struct Type(pub TypeKind, pub TokenRange);
#[derive(Debug, Clone, Copy)]
pub enum TypeKind {
I32,
}
#[derive(Debug, Clone)]
pub enum Literal {
I32(i32),
}
#[derive(Debug, Clone)]
pub struct Expression(pub ExpressionKind, pub TokenRange);
#[derive(Debug, Clone)]
pub enum ExpressionKind {
VariableName(String),
Literal(Literal),
Binop(BinaryOperator, Box<Expression>, Box<Expression>),
FunctionCall(Box<FunctionCallExpression>),
BlockExpr(Box<Block>),
IfExpr(Box<IfExpression>),
}
#[derive(Debug, Clone, Copy)]
pub enum BinaryOperator {
Add,
Minus,
Mult,
And,
LessThan,
}
impl BinaryOperator {
pub fn get_precedence(&self) -> i8 {
use BinaryOperator::*;
match &self {
Add => 10,
Minus => 10,
Mult => 20,
And => 100,
LessThan => 100,
}
}
}
#[derive(Debug, Clone)]
pub struct FunctionCallExpression(pub String, pub Vec<Expression>, pub TokenRange);
#[derive(Debug, Clone)]
pub struct IfExpression(pub Expression, pub Block, pub Option<Block>, pub TokenRange);
#[derive(Debug, Clone)]
pub struct LetStatement(pub String, pub Expression, pub TokenRange);
#[derive(Debug, Clone)]
pub struct ImportStatement(Vec<String>, pub TokenRange);
#[derive(Debug)]
pub struct FunctionDefinition(pub FunctionSignature, pub Block, pub TokenRange);
#[derive(Debug, Clone)]
pub struct FunctionSignature {
pub name: String,
pub args: Vec<(String, Type)>,
pub return_type: Option<Type>,
pub range: TokenRange,
}
#[derive(Debug, Clone, Copy)]
pub enum ReturnType {
Soft,
Hard,
}
#[derive(Debug, Clone)]
pub struct Block(
pub Vec<BlockLevelStatement>,
pub Option<(ReturnType, Expression)>,
pub TokenRange,
);
#[derive(Debug, Clone)]
pub enum BlockLevelStatement {
Let(LetStatement),
Import(ImportStatement),
Expression(Expression),
Return(ReturnType, Expression),
}
#[derive(Debug)]
pub enum TopLevelStatement {
Import(ImportStatement),
FunctionDefinition(FunctionDefinition),
}
#[derive(Debug)]
pub struct Module {
pub name: String,
pub top_level_statements: Vec<TopLevelStatement>,
}

View File

@ -1,3 +1,4 @@
use crate::ast::*;
use crate::{
lexer::Token,
token_stream::{Error, TokenRange, TokenStream},
@ -10,14 +11,6 @@ where
fn parse(stream: TokenStream) -> Result<Self, Error>;
}
#[derive(Debug, Clone)]
pub struct Type(pub TypeKind, pub TokenRange);
#[derive(Debug, Clone)]
pub enum TypeKind {
I32,
}
impl Parse for Type {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
let kind = if let Some(Token::Identifier(ident)) = stream.next() {
@ -33,24 +26,6 @@ impl Parse for Type {
}
}
#[derive(Debug, Clone)]
pub enum Literal {
I32(i32),
}
#[derive(Debug, Clone)]
pub struct Expression(pub ExpressionKind, pub TokenRange);
#[derive(Debug, Clone)]
pub enum ExpressionKind {
VariableName(String),
Literal(Literal),
Binop(BinaryOperator, Box<Expression>, Box<Expression>),
FunctionCall(Box<FunctionCallExpression>),
BlockExpr(Box<Block>),
IfExpr(Box<IfExpression>),
}
impl Parse for Expression {
fn parse(mut stream: TokenStream) -> Result<Expression, Error> {
let lhs = parse_primary_expression(&mut stream)?;
@ -142,16 +117,6 @@ fn parse_binop_rhs(
Ok(lhs)
}
#[derive(Debug, Clone, Copy)]
pub enum BinaryOperator {
Add,
Minus,
Mult,
And,
LessThan,
}
impl Parse for BinaryOperator {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
Ok(match (stream.next(), stream.peek()) {
@ -169,22 +134,6 @@ impl Parse for BinaryOperator {
}
}
impl BinaryOperator {
pub fn get_precedence(&self) -> i8 {
use BinaryOperator::*;
match &self {
Add => 10,
Minus => 10,
Mult => 20,
And => 100,
LessThan => 100,
}
}
}
#[derive(Debug, Clone)]
pub struct FunctionCallExpression(pub String, pub Vec<Expression>, pub TokenRange);
impl Parse for FunctionCallExpression {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
if let Some(Token::Identifier(name)) = stream.next() {
@ -213,23 +162,18 @@ impl Parse for FunctionCallExpression {
}
}
#[derive(Debug, Clone)]
pub struct IfExpression(pub Expression, pub Block, pub TokenRange);
impl Parse for IfExpression {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
stream.expect(Token::If)?;
Ok(IfExpression(
stream.parse()?,
stream.parse()?,
None,
stream.get_range().unwrap(),
))
}
}
#[derive(Debug, Clone)]
pub struct LetStatement(pub String, pub Expression, pub TokenRange);
impl Parse for LetStatement {
fn parse(mut stream: TokenStream) -> Result<LetStatement, Error> {
stream.expect(Token::LetKeyword)?;
@ -250,9 +194,6 @@ impl Parse for LetStatement {
}
}
#[derive(Debug, Clone)]
pub struct ImportStatement(Vec<String>, pub TokenRange);
impl Parse for ImportStatement {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
stream.expect(Token::ImportKeyword)?;
@ -278,9 +219,6 @@ impl Parse for ImportStatement {
}
}
#[derive(Debug)]
pub struct FunctionDefinition(pub FunctionSignature, pub Block, pub TokenRange);
impl Parse for FunctionDefinition {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
stream.expect(Token::FnKeyword)?;
@ -292,14 +230,6 @@ impl Parse for FunctionDefinition {
}
}
#[derive(Debug, Clone)]
pub struct FunctionSignature {
pub name: String,
pub args: Vec<(String, Type)>,
pub return_type: Option<Type>,
pub range: TokenRange,
}
impl Parse for FunctionSignature {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
if let Some(Token::Identifier(name)) = stream.next() {
@ -331,19 +261,6 @@ impl Parse for FunctionSignature {
}
}
#[derive(Debug, Clone, Copy)]
pub enum ReturnType {
Soft,
Hard,
}
#[derive(Debug, Clone)]
pub struct Block(
pub Vec<BlockLevelStatement>,
pub Option<(ReturnType, Expression)>,
pub TokenRange,
);
impl Parse for Block {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
let mut statements = Vec::new();
@ -382,14 +299,6 @@ impl Parse for Block {
}
}
#[derive(Debug, Clone)]
pub enum BlockLevelStatement {
Let(LetStatement),
Import(ImportStatement),
Expression(Expression),
Return(ReturnType, Expression),
}
impl Parse for BlockLevelStatement {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
use BlockLevelStatement as Stmt;
@ -417,12 +326,6 @@ impl Parse for BlockLevelStatement {
}
}
#[derive(Debug)]
pub enum TopLevelStatement {
Import(ImportStatement),
FunctionDefinition(FunctionDefinition),
}
impl Parse for TopLevelStatement {
fn parse(mut stream: TokenStream) -> Result<Self, Error> {
use TopLevelStatement as Stmt;

544
reid/src/ast/process.rs Normal file
View File

@ -0,0 +1,544 @@
use std::collections::HashMap;
use crate::{
ast,
mir::{self, StmtKind, VariableReference},
token_stream::TokenRange,
};
#[derive(Clone)]
pub enum InferredType {
FromVariable(String, TokenRange),
FunctionReturn(String, TokenRange),
Static(mir::TypeKind, TokenRange),
OneOf(Vec<InferredType>),
Void(TokenRange),
DownstreamError(IntoMIRError, TokenRange),
}
fn all_ok<T, E>(result: Vec<Result<T, E>>) -> Option<Vec<T>> {
let mut res = Vec::with_capacity(result.len());
for item in result {
if let Ok(item) = item {
res.push(item);
} else {
return None;
}
}
Some(res)
}
impl InferredType {
fn collapse(
&self,
state: &mut State,
scope: &VirtualScope,
) -> Result<mir::TypeKind, IntoMIRError> {
match self {
InferredType::FromVariable(name, token_range) => {
if let Some(inferred) = scope.get_var(name) {
let temp = inferred.collapse(state, scope);
state.note(temp)
} else {
state.err(IntoMIRError::VariableNotDefined(name.clone(), *token_range))
}
}
InferredType::FunctionReturn(name, token_range) => {
if let Some(type_kind) = scope.get_return_type(name) {
Ok(*type_kind)
} else {
state.err(IntoMIRError::VariableNotDefined(name.clone(), *token_range))
}
}
InferredType::Static(type_kind, _) => Ok(*type_kind),
InferredType::OneOf(inferred_types) => {
let collapsed = all_ok(
inferred_types
.iter()
.map(|t| {
let temp = t.collapse(state, scope);
state.note(temp)
})
.collect(),
);
if let Some(list) = collapsed {
if let Some(first) = list.first() {
if list.iter().all(|i| i == first) {
Ok((*first).into())
} else {
state.err(IntoMIRError::ConflictingType(self.get_range()))
}
} else {
state.err(IntoMIRError::VoidType(self.get_range()))
}
} else {
state.err(IntoMIRError::DownstreamError(self.get_range()))
}
}
InferredType::Void(token_range) => state.err(IntoMIRError::VoidType(*token_range)),
InferredType::DownstreamError(e, _) => state.err(e.clone()),
}
}
fn get_range(&self) -> TokenRange {
match &self {
InferredType::FromVariable(_, token_range) => *token_range,
InferredType::FunctionReturn(_, token_range) => *token_range,
InferredType::Static(_, token_range) => *token_range,
InferredType::OneOf(inferred_types) => {
inferred_types.iter().map(|i| i.get_range()).sum()
}
InferredType::Void(token_range) => *token_range,
InferredType::DownstreamError(_, range) => *range,
}
}
}
pub struct VirtualVariable {
name: String,
inferred: InferredType,
meta: mir::Metadata,
}
pub struct VirtualFunctionSignature {
name: String,
return_type: mir::TypeKind,
parameter_types: Vec<mir::TypeKind>,
metadata: mir::Metadata,
}
pub enum VirtualStorageError {
KeyAlreadyExists(String),
}
pub struct VirtualStorage<T> {
storage: HashMap<String, Vec<T>>,
}
impl<T> VirtualStorage<T> {
fn set(&mut self, name: String, value: T) -> Result<(), VirtualStorageError> {
let result = if let Some(list) = self.storage.get_mut(&name) {
list.push(value);
Err(VirtualStorageError::KeyAlreadyExists(name.clone()))
} else {
self.storage.insert(name, vec![value]);
Ok(())
};
result
}
fn get(&self, name: &String) -> Option<&T> {
if let Some(list) = self.storage.get(name) {
list.first()
} else {
None
}
}
}
impl<T> Default for VirtualStorage<T> {
fn default() -> Self {
Self {
storage: Default::default(),
}
}
}
#[derive(Clone, Debug)]
pub enum IntoMIRError {
DuplicateVariable(String, TokenRange),
DuplicateFunction(String, TokenRange),
VariableNotDefined(String, TokenRange),
FunctionNotDefined(String, TokenRange),
DownstreamError(TokenRange),
ConflictingType(TokenRange),
VoidType(TokenRange),
}
pub struct VirtualScope {
variables: VirtualStorage<VirtualVariable>,
functions: VirtualStorage<VirtualFunctionSignature>,
}
impl VirtualScope {
pub fn set_var(&mut self, variable: VirtualVariable) -> Result<(), IntoMIRError> {
let range = variable.meta.range;
match self.variables.set(variable.name.clone(), variable) {
Ok(_) => Ok(()),
Err(VirtualStorageError::KeyAlreadyExists(n)) => {
Err(IntoMIRError::DuplicateVariable(n, range))
}
}
}
pub fn set_fun(&mut self, function: VirtualFunctionSignature) -> Result<(), IntoMIRError> {
let range = function.metadata.range;
match self.functions.set(function.name.clone(), function) {
Ok(_) => Ok(()),
Err(VirtualStorageError::KeyAlreadyExists(n)) => {
Err(IntoMIRError::DuplicateVariable(n, range))
}
}
}
pub fn get_var(&self, name: &String) -> Option<&InferredType> {
self.variables.get(name).map(|v| &v.inferred)
}
pub fn get_return_type(&self, name: &String) -> Option<&mir::TypeKind> {
self.functions.get(name).map(|v| &v.return_type)
}
}
impl Default for VirtualScope {
fn default() -> Self {
Self {
variables: Default::default(),
functions: Default::default(),
}
}
}
#[derive(Debug)]
pub struct State {
errors: Vec<IntoMIRError>,
fatal: bool,
}
impl State {
fn note<T: std::fmt::Debug>(
&mut self,
value: Result<T, IntoMIRError>,
) -> Result<T, IntoMIRError> {
dbg!(&value);
if let Err(e) = &value {
self.errors.push(e.clone());
}
value
}
fn err<T>(&mut self, error: IntoMIRError) -> Result<T, IntoMIRError> {
self.errors.push(error.clone());
Err(error)
}
}
impl Default for State {
fn default() -> Self {
Self {
errors: Default::default(),
fatal: false,
}
}
}
impl ast::Module {
pub fn process(&self) -> mir::Module {
let mut state = State::default();
let mut scope = VirtualScope::default();
for stmt in &self.top_level_statements {
match stmt {
FunctionDefinition(ast::FunctionDefinition(signature, _, range)) => {
state.note(scope.set_fun(VirtualFunctionSignature {
name: signature.name.clone(),
return_type: signature.return_type.into(),
parameter_types: signature.args.iter().map(|p| p.1.into()).collect(),
metadata: (*range).into(),
}));
}
_ => {}
}
}
let mut imports = Vec::new();
let mut functions = Vec::new();
use ast::TopLevelStatement::*;
for stmt in &self.top_level_statements {
match stmt {
Import(import) => {
for name in &import.0 {
imports.push(mir::Import(name.clone(), import.1.into()));
}
}
FunctionDefinition(ast::FunctionDefinition(signature, block, range)) => {
for (name, ptype) in &signature.args {
state.note(scope.set_var(VirtualVariable {
name: name.clone(),
inferred: InferredType::Static((*ptype).into(), *range),
meta: ptype.1.into(),
}));
}
dbg!(&signature);
if let Some(mir_block) = block.process(&mut state, &mut scope) {
let def = mir::FunctionDefinition {
name: signature.name.clone(),
parameters: signature
.args
.iter()
.cloned()
.map(|p| (p.0, p.1.into()))
.collect(),
kind: mir::FunctionDefinitionKind::Local(mir_block, (*range).into()),
};
functions.push(def);
}
}
}
}
dbg!(&state);
// TODO do something with state here
mir::Module {
name: self.name.clone(),
imports,
functions,
}
}
}
impl ast::Block {
pub fn process(&self, state: &mut State, scope: &mut VirtualScope) -> Option<mir::Block> {
let mut mir_statements = Vec::new();
for statement in &self.0 {
let (kind, range): (Option<mir::StmtKind>, TokenRange) = match statement {
ast::BlockLevelStatement::Let(s_let) => {
let res = s_let.1.infer_return_type().collapse(state, scope);
let collapsed = state.note(res);
let inferred = match &collapsed {
Ok(t) => InferredType::Static(*t, s_let.2),
Err(e) => InferredType::DownstreamError(e.clone(), s_let.2),
};
state
.note(scope.set_var(VirtualVariable {
name: s_let.0.clone(),
inferred,
meta: s_let.2.into(),
}))
.ok();
(
collapsed.ok().and_then(|t| {
s_let.1.process(state, scope).map(|e| {
mir::StmtKind::Let(
mir::VariableReference(t, s_let.0.clone(), s_let.2.into()),
e,
)
})
}),
s_let.2,
)
}
ast::BlockLevelStatement::Import(_) => todo!(),
ast::BlockLevelStatement::Expression(e) => (
e.process(state, scope).map(|e| StmtKind::Expression(e)),
e.1,
),
ast::BlockLevelStatement::Return(_, e) => (
e.process(state, scope).map(|e| StmtKind::Expression(e)),
e.1,
),
};
if let Some(kind) = kind {
mir_statements.push(mir::Statement(kind, range.into()));
} else {
state.fatal = true;
}
}
let return_expression = if let Some(r) = &self.1 {
if let Some(expr) = r.1.process(state, scope) {
Some((r.0.into(), Box::new(expr)))
} else {
state.fatal = true;
None?
}
} else {
None
};
Some(mir::Block {
statements: mir_statements,
return_expression,
meta: self.2.into(),
})
}
fn infer_return_type(&self) -> InferredType {
self.1
.as_ref()
.map(|(_, expr)| expr.infer_return_type())
.unwrap_or(InferredType::Void(self.2))
}
}
impl From<ast::ReturnType> for mir::ReturnKind {
fn from(value: ast::ReturnType) -> Self {
match value {
ast::ReturnType::Soft => mir::ReturnKind::Soft,
ast::ReturnType::Hard => mir::ReturnKind::Hard,
}
}
}
impl ast::Expression {
fn process(&self, state: &mut State, scope: &mut VirtualScope) -> Option<mir::Expression> {
let kind = match &self.0 {
ast::ExpressionKind::VariableName(name) => {
let ty = scope.get_var(name);
if let Some(ty) = ty {
let res = ty.collapse(state, scope);
state
.note(res)
.map(|result| {
mir::ExprKind::Variable(VariableReference(
result,
name.clone(),
self.1.into(),
))
})
.ok()
} else {
state
.err(IntoMIRError::VariableNotDefined(
name.clone(),
self.1.into(),
))
.ok()
}
}
ast::ExpressionKind::Literal(literal) => Some(mir::ExprKind::Literal(literal.mir())),
ast::ExpressionKind::Binop(binary_operator, lhs, rhs) => {
let mir_lhs = lhs.process(state, scope);
let mir_rhs = rhs.process(state, scope);
Some(mir::ExprKind::BinOp(
binary_operator.mir(),
Box::new(mir_lhs?),
Box::new(mir_rhs?),
))
}
ast::ExpressionKind::FunctionCall(fn_call_expr) => {
if let Some(fn_type) = scope.get_return_type(&fn_call_expr.0).cloned() {
let parameters = all_ok(
fn_call_expr
.1
.iter()
.map(|e| {
e.process(state, scope)
.ok_or(IntoMIRError::DownstreamError(self.1.into()))
})
.collect(),
);
if let Some(parameters) = parameters {
Some(mir::ExprKind::FunctionCall(mir::FunctionCall {
name: fn_call_expr.0.clone(),
return_type: fn_type,
parameters,
}))
} else {
None
}
} else {
state
.err(IntoMIRError::FunctionNotDefined(
fn_call_expr.0.clone(),
self.1,
))
.ok()
}
}
ast::ExpressionKind::BlockExpr(block) => {
block.process(state, scope).map(|b| mir::ExprKind::Block(b))
}
ast::ExpressionKind::IfExpr(if_expression) => {
let cond = if_expression.0.process(state, scope);
let then_block = if_expression.1.process(state, scope);
let else_block = if let Some(el) = &if_expression.2 {
Some(el.process(state, scope)?)
} else {
None
};
Some(mir::ExprKind::If(mir::IfExpression(
Box::new(cond?),
then_block?,
else_block,
)))
}
};
kind.map(|k| mir::Expression(k, self.1.into()))
}
fn infer_return_type(&self) -> InferredType {
use ast::ExpressionKind::*;
match &self.0 {
VariableName(name) => InferredType::FromVariable(name.clone(), self.1),
Literal(lit) => InferredType::Static(lit.mir().as_type(), self.1),
Binop(_, lhs, rhs) => {
InferredType::OneOf(vec![lhs.infer_return_type(), rhs.infer_return_type()])
}
FunctionCall(fncall) => InferredType::FunctionReturn(fncall.0.clone(), self.1),
BlockExpr(block) => block.infer_return_type(),
IfExpr(exp) => {
let mut types = vec![exp.1.infer_return_type()];
if let Some(e) = &exp.2 {
types.push(e.infer_return_type())
}
InferredType::OneOf(types)
}
}
}
}
impl ast::BinaryOperator {
fn mir(&self) -> mir::BinaryOperator {
match self {
ast::BinaryOperator::Add => mir::BinaryOperator::Add,
ast::BinaryOperator::Minus => mir::BinaryOperator::Minus,
ast::BinaryOperator::Mult => mir::BinaryOperator::Mult,
ast::BinaryOperator::And => mir::BinaryOperator::And,
ast::BinaryOperator::LessThan => {
mir::BinaryOperator::Logic(mir::LogicOperator::LessThan)
}
}
}
}
impl ast::Literal {
fn mir(&self) -> mir::Literal {
match *self {
ast::Literal::I32(v) => mir::Literal::I32(v),
}
}
}
impl From<ast::TypeKind> for mir::TypeKind {
fn from(value: ast::TypeKind) -> Self {
match value {
ast::TypeKind::I32 => mir::TypeKind::I32,
}
}
}
impl From<ast::Type> for mir::TypeKind {
fn from(value: ast::Type) -> Self {
value.0.into()
}
}
impl From<Option<ast::Type>> for mir::TypeKind {
fn from(value: Option<ast::Type>) -> Self {
match value {
Some(v) => v.into(),
None => mir::TypeKind::Void,
}
}
}

View File

@ -102,17 +102,14 @@ impl<'ctx> Scope<'ctx> {
impl mir::Statement {
pub fn codegen<'ctx>(&self, scope: &mut Scope<'ctx>) -> Option<Value<'ctx>> {
match &self.0 {
mir::StatementKind::Let(VariableReference(_, name, _), expression) => {
mir::StmtKind::Let(VariableReference(_, name, _), expression) => {
let value = expression.codegen(scope).unwrap();
scope.stack_values.insert(name.clone(), value);
None
}
mir::StatementKind::If(if_expression) => if_expression.codegen(scope),
mir::StatementKind::Import(_) => todo!(),
mir::StatementKind::Expression(expression) => {
let value = expression.codegen(scope).unwrap();
Some(value)
}
// mir::StmtKind::If(if_expression) => if_expression.codegen(scope),
mir::StmtKind::Import(_) => todo!(),
mir::StmtKind::Expression(expression) => expression.codegen(scope),
}
}
}
@ -178,15 +175,15 @@ impl mir::IfExpression {
impl mir::Expression {
pub fn codegen<'ctx>(&self, scope: &mut Scope<'ctx>) -> Option<Value<'ctx>> {
match &self.0 {
mir::ExpressionKind::Variable(varref) => {
mir::ExprKind::Variable(varref) => {
let v = scope
.stack_values
.get(&varref.1)
.expect("Variable reference not found?!");
Some(v.clone())
}
mir::ExpressionKind::Literal(lit) => Some(lit.codegen(scope.context)),
mir::ExpressionKind::BinOp(binop, lhs_exp, rhs_exp) => {
mir::ExprKind::Literal(lit) => Some(lit.codegen(scope.context)),
mir::ExprKind::BinOp(binop, lhs_exp, rhs_exp) => {
let lhs = lhs_exp.codegen(scope).expect("lhs has no return value");
let rhs = rhs_exp.codegen(scope).expect("rhs has no return value");
Some(match binop {
@ -203,7 +200,7 @@ impl mir::Expression {
}
})
}
mir::ExpressionKind::FunctionCall(call) => {
mir::ExprKind::FunctionCall(call) => {
let params = call
.parameters
.iter()
@ -215,8 +212,8 @@ impl mir::Expression {
.expect("function not found!");
Some(scope.block.call(callee, params, "call").unwrap())
}
mir::ExpressionKind::If(if_expression) => if_expression.codegen(scope),
mir::ExpressionKind::Block(block) => {
mir::ExprKind::If(if_expression) => if_expression.codegen(scope),
mir::ExprKind::Block(block) => {
let mut inner_scope = scope.with_block(scope.function.block("inner"));
if let Some(ret) = block.codegen(&mut inner_scope) {
inner_scope.block.br(&scope.block);
@ -249,11 +246,11 @@ impl mir::Block {
if let Some((kind, expr)) = &self.return_expression {
let ret = expr.codegen(&mut scope).unwrap();
match kind {
mir::ReturnKind::HardReturn => {
mir::ReturnKind::Hard => {
scope.block.ret(&ret).unwrap();
None
}
mir::ReturnKind::SoftReturn => Some(ret),
mir::ReturnKind::Soft => Some(ret),
}
} else {
None
@ -276,6 +273,7 @@ impl TypeKind {
match &self {
TypeKind::I32 => TypeEnum::Integer(context.type_i32()),
TypeKind::I16 => TypeEnum::Integer(context.type_i16()),
TypeKind::Void => panic!("Void not a supported type"),
}
}
}

View File

@ -1,13 +1,11 @@
use old_codegen::{form_context, from_statements};
use reid_lib::Context;
use crate::{lexer::Token, parser::TopLevelStatement, token_stream::TokenStream};
use crate::{ast::TopLevelStatement, lexer::Token, token_stream::TokenStream};
mod ast;
mod codegen;
mod lexer;
pub mod mir;
mod old_codegen;
mod parser;
// mod llvm_ir;
pub mod codegen;
mod token_stream;
// TODO:
@ -41,8 +39,21 @@ pub fn compile(source: &str) -> Result<String, ReidError> {
statements.push(statement);
}
let mut context = form_context();
let mut module = from_statements(&mut context, statements).unwrap();
let text = module.print_to_string().unwrap();
Ok(text.to_owned())
let ast_module = ast::Module {
name: "test".to_owned(),
top_level_statements: statements,
};
dbg!(&ast_module);
let mir_module = ast_module.process();
dbg!(&mir_module);
let mut context = Context::new();
let cogegen_module = mir_module.codegen(&mut context);
Ok(match cogegen_module.module.print_to_string() {
Ok(v) => v,
Err(e) => panic!("Err: {:?}", e),
})
}

View File

@ -1,18 +1,34 @@
/// In this module are defined structs that are used for performing passes on
/// Reid. It contains a simplified version of Reid which must already be
/// type-checked beforehand.
use std::collections::HashMap;
use types::*;
use crate::token_stream::TokenRange;
pub mod types;
#[derive(Clone, Copy)]
#[derive(Debug, Clone, Copy)]
pub struct Metadata {
pub range: TokenRange,
}
impl From<TokenRange> for Metadata {
fn from(value: TokenRange) -> Self {
Metadata { range: value }
}
}
impl Default for Metadata {
fn default() -> Self {
Metadata {
range: Default::default(),
}
}
}
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum TypeKind {
I32,
I16,
Void,
}
impl TypeKind {
@ -23,14 +39,14 @@ impl TypeKind {
}
}
#[derive(Clone, Copy)]
#[derive(Debug, Clone, Copy)]
pub enum Literal {
I32(i32),
I16(i16),
}
impl Literal {
fn as_type(self: &Literal) -> TypeKind {
pub fn as_type(self: &Literal) -> TypeKind {
match self {
Literal::I32(_) => TypeKind::I32,
Literal::I16(_) => TypeKind::I16,
@ -55,15 +71,18 @@ pub enum LogicOperator {
#[derive(Debug, Clone, Copy)]
pub enum ReturnKind {
HardReturn,
SoftReturn,
Hard,
Soft,
}
pub struct VariableReference(pub TypeKind, pub String, pub TokenRange);
#[derive(Debug)]
pub struct VariableReference(pub TypeKind, pub String, pub Metadata);
pub struct Import(pub String, pub TokenRange);
#[derive(Debug)]
pub struct Import(pub String, pub Metadata);
pub enum ExpressionKind {
#[derive(Debug)]
pub enum ExprKind {
Variable(VariableReference),
Literal(Literal),
BinOp(BinaryOperator, Box<Expression>, Box<Expression>),
@ -72,47 +91,55 @@ pub enum ExpressionKind {
Block(Block),
}
pub struct Expression(pub ExpressionKind, pub TokenRange);
#[derive(Debug)]
pub struct Expression(pub ExprKind, pub Metadata);
/// Condition, Then, Else
#[derive(Debug)]
pub struct IfExpression(pub Box<Expression>, pub Block, pub Option<Block>);
#[derive(Debug)]
pub struct FunctionCall {
pub name: String,
pub return_type: TypeKind,
pub parameters: Vec<Expression>,
}
#[derive(Debug)]
pub struct FunctionDefinition {
pub name: String,
pub parameters: Vec<(String, TypeKind)>,
pub kind: FunctionDefinitionKind,
}
#[derive(Debug)]
pub enum FunctionDefinitionKind {
/// Actual definition block and surrounding signature range
Local(Block, TokenRange),
Local(Block, Metadata),
/// Return Type
Extern(TypeKind),
}
#[derive(Debug)]
pub struct Block {
/// List of non-returning statements
pub statements: Vec<Statement>,
pub return_expression: Option<(ReturnKind, Box<Expression>)>,
pub range: TokenRange,
pub meta: Metadata,
}
pub struct Statement(pub StatementKind, pub TokenRange);
#[derive(Debug)]
pub struct Statement(pub StmtKind, pub Metadata);
pub enum StatementKind {
#[derive(Debug)]
pub enum StmtKind {
/// Variable name+type, evaluation
Let(VariableReference, Expression),
If(IfExpression),
Import(Import),
Expression(Expression),
}
#[derive(Debug)]
pub struct Module {
pub name: String,
pub imports: Vec<Import>,

View File

@ -16,26 +16,25 @@ impl ReturnType for Block {
fn return_type(&self) -> Result<TypeKind, ReturnTypeOther> {
self.return_expression
.as_ref()
.ok_or(ReturnTypeOther::NoBlockReturn(self.range.clone()))
.ok_or(ReturnTypeOther::NoBlockReturn(self.meta.range))
.and_then(|(_, stmt)| stmt.return_type())
}
}
impl ReturnType for Statement {
fn return_type(&self) -> Result<TypeKind, ReturnTypeOther> {
use StatementKind::*;
use StmtKind::*;
match &self.0 {
Expression(e) => e.return_type(),
If(e) => e.return_type(),
Import(_) => Err(ReturnTypeOther::Import(self.1)),
Let(_, _) => Err(ReturnTypeOther::Let(self.1)),
Import(_) => Err(ReturnTypeOther::Import(self.1.range)),
Let(_, _) => Err(ReturnTypeOther::Let(self.1.range)),
}
}
}
impl ReturnType for Expression {
fn return_type(&self) -> Result<TypeKind, ReturnTypeOther> {
use ExpressionKind::*;
use ExprKind::*;
match &self.0 {
Literal(lit) => Ok(lit.as_type()),
Variable(var) => var.return_type(),

View File

@ -1,287 +0,0 @@
use std::borrow::BorrowMut;
use std::ffi::{CStr, CString};
use std::mem;
use std::ptr::null_mut;
use llvm_sys::analysis::LLVMVerifyModule;
use llvm_sys::transforms::pass_manager_builder::{
self, LLVMOpaquePassManagerBuilder, LLVMPassManagerBuilderCreate,
LLVMPassManagerBuilderSetOptLevel,
};
use llvm_sys::{
core::*, prelude::*, LLVMBasicBlock, LLVMBuilder, LLVMContext, LLVMModule, LLVMType, LLVMValue,
};
use crate::parser;
fn into_cstring<T: Into<String>>(value: T) -> CString {
let string = value.into();
unsafe { CString::from_vec_with_nul_unchecked((string + "\0").into_bytes()) }
}
#[derive(thiserror::Error, Debug)]
pub enum Error {
#[error("Type mismatch: {0:?} vs {1:?}")]
TypeMismatch(IRType, IRType),
}
#[derive(Clone, Debug, Eq, PartialEq)]
pub enum IRType {
I32,
Boolean,
}
impl IRType {
fn in_context(&self, context: &IRContext) -> *mut LLVMType {
use IRType::*;
unsafe {
return match self {
I32 => LLVMInt32TypeInContext(context.context),
Boolean => LLVMInt1TypeInContext(context.context),
};
}
}
}
#[derive(Clone)]
pub struct IRValue(pub IRType, *mut LLVMValue);
impl IRValue {
pub fn from_literal(literal: &parser::Literal, module: &IRModule) -> Self {
use parser::Literal;
match literal {
Literal::I32(v) => {
let ir_type = IRType::I32;
unsafe {
let ir_value = LLVMConstInt(
ir_type.in_context(module.context),
mem::transmute(*v as i64),
1,
);
return IRValue(ir_type, ir_value);
}
}
};
}
}
pub struct IRContext {
context: *mut LLVMContext,
builder: *mut LLVMBuilder,
}
impl IRContext {
pub fn new() -> IRContext {
unsafe {
// Set up a context, module and builder in that context.
let context = LLVMContextCreate();
let builder = LLVMCreateBuilderInContext(context);
IRContext { context, builder }
}
}
pub fn module<'a>(&'a mut self, name: String) -> IRModule<'a> {
IRModule::new(self, name)
}
}
impl Drop for IRContext {
fn drop(&mut self) {
// Clean up. Values created in the context mostly get cleaned up there.
unsafe {
LLVMDisposeBuilder(self.builder);
LLVMContextDispose(self.context);
}
}
}
pub struct IRModule<'a> {
context: &'a mut IRContext,
module: *mut LLVMModule,
}
impl<'a> IRModule<'a> {
fn new<'b: 'a>(context: &'b mut IRContext, name: String) -> IRModule<'a> {
unsafe {
let module =
LLVMModuleCreateWithNameInContext(into_cstring(name).as_ptr(), context.context);
IRModule { context, module }
}
}
pub fn print_to_string(&mut self) -> Result<&str, std::str::Utf8Error> {
unsafe {
LLVMVerifyModule(
self.module,
llvm_sys::analysis::LLVMVerifierFailureAction::LLVMPrintMessageAction,
null_mut(),
);
CStr::from_ptr(LLVMPrintModuleToString(self.module)).to_str()
}
}
}
impl<'a> Drop for IRModule<'a> {
fn drop(&mut self) {
// Clean up. Values created in the context mostly get cleaned up there.
unsafe {
LLVMDisposeModule(self.module);
}
}
}
pub struct IRFunction<'a, 'b> {
pub module: &'b IRModule<'a>,
/// The actual function
value: *mut LLVMValue,
}
impl<'a, 'b> IRFunction<'a, 'b> {
pub fn new(name: &String, module: &'b IRModule<'a>) -> IRFunction<'a, 'b> {
unsafe {
// TODO, fix later!
let return_type = LLVMInt32TypeInContext(module.context.context);
let mut argts = [];
let func_type =
LLVMFunctionType(return_type, argts.as_mut_ptr(), argts.len() as u32, 0);
let function = LLVMAddFunction(module.module, into_cstring(name).as_ptr(), func_type);
IRFunction {
module,
value: function,
}
}
}
}
pub struct IRBlock<'a, 'b, 'c> {
pub function: &'c IRFunction<'a, 'b>,
blockref: *mut LLVMBasicBlock,
}
impl<'a, 'b, 'c> IRBlock<'a, 'b, 'c> {
pub fn new(function: &'c IRFunction<'a, 'b>, name: &CStr) -> IRBlock<'a, 'b, 'c> {
unsafe {
let blockref =
LLVMCreateBasicBlockInContext(function.module.context.context, name.as_ptr());
IRBlock { function, blockref }
}
}
pub fn add(
&mut self,
IRValue(lhs_t, lhs_v): IRValue,
IRValue(rhs_t, rhs_v): IRValue,
) -> Result<IRValue, Error> {
unsafe {
LLVMPositionBuilderAtEnd(self.function.module.context.builder, self.blockref);
if lhs_t == rhs_t {
Ok(IRValue(
lhs_t,
LLVMBuildAdd(
self.function.module.context.builder,
lhs_v,
rhs_v,
c"tmpadd".as_ptr(),
),
))
} else {
Err(Error::TypeMismatch(lhs_t, rhs_t))
}
}
}
pub fn mult(
&mut self,
IRValue(lhs_t, lhs_v): IRValue,
IRValue(rhs_t, rhs_v): IRValue,
) -> Result<IRValue, Error> {
unsafe {
LLVMPositionBuilderAtEnd(self.function.module.context.builder, self.blockref);
if lhs_t == rhs_t {
Ok(IRValue(
lhs_t,
LLVMBuildMul(
self.function.module.context.builder,
lhs_v,
rhs_v,
c"tmpadd".as_ptr(),
),
))
} else {
Err(Error::TypeMismatch(lhs_t, rhs_t))
}
}
}
pub fn less_than(
&mut self,
IRValue(lhs_t, lhs_v): IRValue,
IRValue(rhs_t, rhs_v): IRValue,
) -> Result<IRValue, Error> {
unsafe {
LLVMPositionBuilderAtEnd(self.function.module.context.builder, self.blockref);
if lhs_t == rhs_t {
Ok(IRValue(
IRType::Boolean,
LLVMBuildICmp(
self.function.module.context.builder,
llvm_sys::LLVMIntPredicate::LLVMIntULT,
lhs_v,
rhs_v,
c"IntULT".as_ptr(),
),
))
} else {
Err(Error::TypeMismatch(lhs_t, rhs_t))
}
}
}
pub fn add_return(&mut self, value: Option<IRValue>) {
unsafe {
LLVMPositionBuilderAtEnd(self.function.module.context.builder, self.blockref);
if let Some(IRValue(_, value)) = value {
LLVMBuildRet(self.function.module.context.builder, value);
} else {
LLVMBuildRetVoid(self.function.module.context.builder);
}
}
}
pub fn branch(
&mut self,
IRValue(_, condition): IRValue,
then_block: &mut IRBlock,
else_block: &mut IRBlock,
) {
unsafe {
LLVMPositionBuilderAtEnd(self.function.module.context.builder, self.blockref);
LLVMBuildCondBr(
self.function.module.context.builder,
condition,
then_block.blockref,
else_block.blockref,
);
}
}
pub fn move_into(&mut self, block: &mut IRBlock) {
unsafe {
LLVMPositionBuilderAtEnd(self.function.module.context.builder, self.blockref);
LLVMBuildBr(self.function.module.context.builder, block.blockref);
}
}
}
impl<'a, 'b, 'c> Drop for IRBlock<'a, 'b, 'c> {
fn drop(&mut self) {
unsafe {
LLVMAppendExistingBasicBlock(self.function.value, self.blockref);
}
}
}

View File

@ -1,176 +0,0 @@
mod llvm;
use std::collections::HashMap;
use llvm::{Error, IRBlock, IRContext, IRFunction, IRModule, IRValue};
use crate::{
parser::{
Block, BlockLevelStatement, Expression, ExpressionKind, FunctionDefinition, IfExpression,
LetStatement, ReturnType,
},
TopLevelStatement,
};
pub fn form_context() -> IRContext {
IRContext::new()
}
pub fn from_statements(
context: &mut IRContext,
statements: Vec<TopLevelStatement>,
) -> Result<IRModule, Error> {
let mut module = context.module("testmod".to_owned());
let mut scope = ScopeData::new();
for statement in statements {
statement.codegen(&mut scope, &mut module);
}
Ok(module)
}
impl TopLevelStatement {
fn codegen(&self, scope: &mut ScopeData, module: &mut IRModule) {
match self {
Self::FunctionDefinition(func) => func.codegen(scope, module),
Self::Import(_) => panic!("not implemented"),
}
}
}
impl FunctionDefinition {
fn codegen(&self, scope: &mut ScopeData, module: &mut IRModule) {
let FunctionDefinition(signature, block, _) = self;
let ir_function = IRFunction::new(&signature.name, module);
let ir_block = IRBlock::new(&ir_function, c"entry");
let mut scope = scope.inner(ir_block);
if let Some((_, val)) = block.codegen(&mut scope) {
scope.block.add_return(Some(val));
} else {
scope.block.add_return(None);
}
}
}
impl Block {
#[must_use]
fn codegen(&self, scope: &mut Scope) -> Option<(ReturnType, IRValue)> {
for statement in &self.0 {
statement.codegen(scope);
}
if let Some((ret_type, return_exp)) = &self.1 {
let value = return_exp.codegen(scope);
Some((*ret_type, value))
} else {
None
}
}
}
impl BlockLevelStatement {
fn codegen(&self, scope: &mut Scope) {
use BlockLevelStatement::*;
match self {
Expression(exp) | Return(ReturnType::Soft, exp) => {
exp.codegen(scope);
}
Let(LetStatement(name, exp, _)) => {
let val = exp.codegen(scope);
scope.data.insert(name, val);
}
Return(ReturnType::Hard, _) => panic!("hard returns here should not be possible.."),
Import(_) => panic!("block level import not supported"),
}
}
}
impl Expression {
fn codegen(&self, scope: &mut Scope) -> IRValue {
let Expression(kind, _) = self;
use ExpressionKind::*;
match kind {
Literal(lit) => IRValue::from_literal(lit, &scope.block.function.module),
VariableName(v) => scope.data.fetch(v),
Binop(op, lhs, rhs) => {
let lhs = lhs.codegen(scope);
let rhs = rhs.codegen(scope);
use crate::parser::BinaryOperator::*;
match op {
Add => scope.block.add(lhs, rhs).unwrap(),
Mult => scope.block.mult(lhs, rhs).unwrap(),
LessThan => scope.block.less_than(lhs, rhs).unwrap(),
_ => panic!("operator not supported: {:?}", op),
}
}
IfExpr(ifx) => {
let IfExpression(expr, block, _) = ifx.as_ref();
let condition = expr.codegen(scope);
let mut thenb = IRBlock::new(scope.block.function, c"then");
let mut afterb = IRBlock::new(scope.block.function, c"merge");
scope.block.branch(condition, &mut thenb, &mut afterb);
scope.block = afterb;
let mut then = scope.inner(thenb);
match block.codegen(&mut then) {
Some((ReturnType::Hard, v)) => then.block.add_return(Some(v)),
_ => then.block.move_into(&mut scope.block),
}
IRValue::from_literal(&crate::parser::Literal::I32(1), scope.block.function.module)
}
BlockExpr(_) => panic!("block expr not supported"),
FunctionCall(_) => panic!("function call expr not supported"),
}
}
}
#[derive(Clone)]
struct ScopeData {
vars: HashMap<String, IRValue>,
}
impl ScopeData {
fn new() -> ScopeData {
ScopeData {
vars: HashMap::new(),
}
}
fn with_block<'a, 'b, 'c>(self, block: IRBlock<'a, 'b, 'c>) -> Scope<'a, 'b, 'c> {
Scope { data: self, block }
}
fn inner<'a, 'b, 'c>(&self, block: IRBlock<'a, 'b, 'c>) -> Scope<'a, 'b, 'c> {
self.clone().with_block(block)
}
fn fetch(&self, name: &String) -> IRValue {
match self.vars.get(name) {
Some(val) => val.clone(),
_ => panic!("No such variable in scope: {}", name),
}
}
fn insert(&mut self, name: &String, value: IRValue) {
match self.vars.insert(name.clone(), value) {
Some(_) => panic!("{} was already defined in scope", name),
_ => {}
}
}
}
struct Scope<'a, 'b, 'c> {
data: ScopeData,
block: IRBlock<'a, 'b, 'c>,
}
impl<'a, 'b, 'c> Scope<'a, 'b, 'c> {
fn inner(&self, block: IRBlock<'a, 'b, 'c>) -> Scope<'a, 'b, 'c> {
self.data.clone().with_block(block)
}
}

View File

@ -1,222 +0,0 @@
use std::collections::{hash_map, HashMap};
use crate::{
ast::{
BinaryOperator, Block, BlockLevelStatement, Expression, FunctionCallExpression,
FunctionDefinition, FunctionSignature, ReturnType, TopLevelStatement,
},
llvm_ir::{self, IRBlock, IRFunction, IRModule, IRValue, IRValueType},
};
#[derive(Clone)]
pub struct ScopeData {
named_vars: HashMap<String, IRValue>,
defined_functions: HashMap<String, (FunctionSignature, Option<IRFunction>)>,
}
impl ScopeData {
pub fn inner<'a, 'b>(&self, block: &'b mut IRBlock<'a>) -> Scope<'a, 'b> {
Scope {
block,
data: self.clone(),
}
}
pub fn var(&self, name: &String) -> Option<&IRValue> {
self.named_vars.get(name)
}
pub fn set_var(&mut self, name: &str, val: IRValue) -> Result<(), Error> {
if let hash_map::Entry::Vacant(e) = self.named_vars.entry(name.to_owned()) {
e.insert(val);
Ok(())
} else {
Err(Error::VariableAlreadyDefined(name.to_owned()))
}
}
pub fn function(
&mut self,
name: &String,
) -> Option<&mut (FunctionSignature, Option<IRFunction>)> {
self.defined_functions.get_mut(name)
}
pub fn set_function_signature(
&mut self,
name: &str,
sig: FunctionSignature,
ir: IRFunction,
) -> Result<(), Error> {
if let hash_map::Entry::Vacant(e) = self.defined_functions.entry(name.to_owned()) {
e.insert((sig, Some(ir)));
Ok(())
} else {
Err(Error::VariableAlreadyDefined(name.to_owned()))
}
}
}
pub struct Scope<'a, 'b> {
pub block: &'b mut IRBlock<'a>,
pub data: ScopeData,
}
impl<'a, 'b> Scope<'a, 'b> {
pub fn inner<'c>(&'c mut self) -> Scope<'a, 'c> {
Scope {
block: self.block,
data: self.data.clone(),
}
}
}
pub fn codegen_from_statements(statements: Vec<TopLevelStatement>) -> Result<IRModule, Error> {
let mut module = IRModule::new("testmod");
let mut scope = ScopeData {
defined_functions: HashMap::new(),
named_vars: HashMap::new(),
};
for statement in &statements {
match statement {
TopLevelStatement::FunctionDefinition(FunctionDefinition(sig, _)) => {
let function = module.create_func(&sig.name, IRValueType::I32);
scope.set_function_signature(&sig.name.clone(), sig.clone(), function)?;
}
TopLevelStatement::Import(_) => {}
}
}
for statement in &statements {
statement.codegen(&mut module, &mut scope)?;
}
Ok(module)
}
impl TopLevelStatement {
pub fn codegen(&self, module: &mut IRModule, root_data: &mut ScopeData) -> Result<(), Error> {
match self {
TopLevelStatement::FunctionDefinition(FunctionDefinition(sig, block)) => {
if let Some((_, ir)) = root_data.function(&sig.name) {
if let Some(ir_function) = ir.take() {
let mut ir_block = module.create_block();
let mut scope = root_data.inner(&mut ir_block);
let (_, value) = match block.codegen(&mut scope)? {
Some(v) => v,
None => panic!("Void-return type function not yet implemented!"),
};
ir_function.add_definition(value, ir_block);
} else {
Err(Error::FunctionAlreadyDefined(sig.name.clone()))?
}
} else {
panic!("Function was not declared before it's definition")
}
}
TopLevelStatement::Import(_) => {}
}
Ok(())
}
}
impl Block {
pub fn codegen(&self, scope: &mut Scope) -> Result<Option<(ReturnType, IRValue)>, Error> {
for statement in &self.0 {
statement.codegen(scope)?;
}
let value = if let Some((rt, exp)) = &self.1 {
Some((*rt, exp.codegen(scope)?))
} else {
None
};
Ok(value)
}
}
impl BlockLevelStatement {
pub fn codegen(&self, scope: &mut Scope) -> Result<(), Error> {
match self {
BlockLevelStatement::Let(let_statement) => {
let val = let_statement.1.codegen(scope)?;
scope.data.set_var(&let_statement.0, val)?;
Ok(())
}
BlockLevelStatement::Return(_) => panic!("Should never happen"),
BlockLevelStatement::Import(_) => Ok(()), // TODO: To implement
BlockLevelStatement::Expression(e) => {
let _value = e.codegen(scope)?;
Ok(())
}
}
}
}
impl Expression {
pub fn codegen(&self, scope: &mut Scope) -> Result<IRValue, Error> {
use Expression::*;
match self {
Binop(op, lhs, rhs) => match op {
BinaryOperator::Add => {
let lhs = lhs.codegen(scope)?;
let rhs = rhs.codegen(scope)?;
Ok(scope.block.add(lhs, rhs)?)
}
BinaryOperator::Mult => {
let lhs = lhs.codegen(scope)?;
let rhs = rhs.codegen(scope)?;
Ok(scope.block.mul(lhs, rhs)?)
}
_ => panic!("Other binary operators not supported yet!"),
},
BlockExpr(block) => {
let mut inner = scope.inner();
Ok(match block.codegen(&mut inner)? {
Some((r_type, value)) => match r_type {
ReturnType::Soft => value,
ReturnType::Hard => {
panic!("Hard returns in inner blocks not supported yet")
}
},
None => panic!("Void-return type block not yet implemented!"),
})
}
FunctionCall(fc) => {
let FunctionCallExpression(name, _) = &**fc;
if let Some((sig, _)) = scope.data.function(name) {
Ok(scope.block.function_call(sig)?)
} else {
Err(Error::UndefinedFunction(name.clone()))?
}
}
VariableName(name) => scope
.data
.var(name)
.cloned()
.ok_or(Error::UndefinedVariable(name.clone())),
Literal(lit) => Ok(scope.block.get_const(lit)),
IfExpr(_) => panic!("if expressions not yet supported"),
}
}
}
#[derive(thiserror::Error, Debug)]
pub enum Error {
#[error("Variable '{0}' already defined")]
VariableAlreadyDefined(String),
#[error("Variable '{0}' not yet defined")]
UndefinedVariable(String),
#[error("Function '{0}' not defined")]
UndefinedFunction(String),
#[error("Function '{0}' already defined")]
FunctionAlreadyDefined(String),
#[error(transparent)]
Deeper(#[from] llvm_ir::Error),
}

View File

@ -1,203 +0,0 @@
use std::ffi::{CStr, CString};
use std::mem;
use llvm_sys::{core::*, prelude::*, LLVMBuilder, LLVMContext, LLVMModule};
use crate::ast::{FunctionSignature, Literal};
macro_rules! cstr {
($string:expr) => {
core::ffi::CStr::from_bytes_with_nul_unchecked(concat!($string, "\0").as_bytes()).as_ptr()
};
}
#[derive(Clone, Debug)]
#[must_use = "value contains raw pointer and must be inserted somewhere"]
pub struct IRValue(IRValueType, LLVMValueRef);
#[derive(Clone, Debug, PartialEq, Eq)]
pub enum IRValueType {
I32,
}
impl IRValueType {
unsafe fn get_llvm_type(&self, module: &mut IRModule) -> LLVMTypeRef {
match *self {
Self::I32 => LLVMInt32TypeInContext(module.context),
}
}
}
fn into_cstring<T: Into<String>>(value: T) -> CString {
let string = value.into();
unsafe { CString::from_vec_with_nul_unchecked((string + "\0").into_bytes()) }
}
pub struct IRModule {
context: *mut LLVMContext,
module: *mut LLVMModule,
builder: *mut LLVMBuilder,
}
impl IRModule {
pub fn new<T: Into<String>>(name: T) -> IRModule {
unsafe {
// Set up a context, module and builder in that context.
let context = LLVMContextCreate();
let module = LLVMModuleCreateWithNameInContext(into_cstring(name).as_ptr(), context);
let builder = LLVMCreateBuilderInContext(context);
IRModule {
context,
module,
builder,
}
}
}
pub fn create_block(&mut self) -> IRBlock {
IRBlock::create("entry", self)
}
pub fn create_func<T: Into<String>>(
&mut self,
name: T,
return_type: IRValueType,
) -> IRFunction {
unsafe {
let mut argts = [];
let func_type = LLVMFunctionType(
return_type.get_llvm_type(self),
argts.as_mut_ptr(),
argts.len() as u32,
0,
);
let anon_func = LLVMAddFunction(self.module, into_cstring(name).as_ptr(), func_type);
IRFunction {
value: IRValue(return_type, anon_func),
}
}
}
pub fn print_to_string(&mut self) -> Result<&str, std::str::Utf8Error> {
unsafe { CStr::from_ptr(LLVMPrintModuleToString(self.module)).to_str() }
}
}
impl Drop for IRModule {
fn drop(&mut self) {
// Clean up. Values created in the context mostly get cleaned up there.
unsafe {
LLVMDisposeBuilder(self.builder);
LLVMDisposeModule(self.module);
LLVMContextDispose(self.context);
}
}
}
#[derive(Clone, Debug)]
pub struct IRFunction {
value: IRValue,
}
impl IRFunction {
pub fn add_definition(self, ret: IRValue, block: IRBlock) {
unsafe {
LLVMAppendExistingBasicBlock(self.value.1, block.blockref);
LLVMBuildRet(block.module.builder, ret.1);
}
}
}
pub struct IRBlock<'a> {
module: &'a mut IRModule,
blockref: LLVMBasicBlockRef,
}
impl<'a> IRBlock<'a> {
fn create<T: Into<String>>(name: T, codegen: &'a mut IRModule) -> IRBlock<'a> {
unsafe {
let blockref =
LLVMCreateBasicBlockInContext(codegen.context, into_cstring(name).as_ptr());
LLVMPositionBuilderAtEnd(codegen.builder, blockref);
IRBlock {
module: codegen,
blockref,
}
}
}
pub fn get_const(&mut self, literal_type: &Literal) -> IRValue {
unsafe {
match *literal_type {
Literal::I32(v) => IRValue(
IRValueType::I32,
LLVMConstInt(
LLVMInt32TypeInContext(self.module.context),
mem::transmute(v as i64),
1,
),
),
}
}
}
pub fn add(&mut self, lhs: IRValue, rhs: IRValue) -> Result<IRValue, Error> {
unsafe {
if lhs.0 == rhs.0 {
Ok(IRValue(
lhs.0,
LLVMBuildAdd(self.module.builder, lhs.1, rhs.1, cstr!("tmpadd")),
))
} else {
Err(Error::TypeMismatch(lhs.0, rhs.0))
}
}
}
pub fn mul(&mut self, lhs: IRValue, rhs: IRValue) -> Result<IRValue, Error> {
unsafe {
if lhs.0 == rhs.0 {
Ok(IRValue(
lhs.0,
LLVMBuildMul(self.module.builder, lhs.1, rhs.1, cstr!("tmpadd")),
))
} else {
Err(Error::TypeMismatch(lhs.0, rhs.0))
}
}
}
pub fn function_call(&mut self, callee: &FunctionSignature) -> Result<IRValue, Error> {
unsafe {
let function = LLVMGetNamedFunction(
self.module.module,
into_cstring(callee.name.clone()).as_ptr(),
);
let ret_t = LLVMInt32TypeInContext(self.module.context);
let mut argts = [];
let mut args = [];
let fun_t = LLVMFunctionType(ret_t, argts.as_mut_ptr(), argts.len() as u32, 0);
let call = LLVMBuildCall2(
self.module.builder,
fun_t,
function,
args.as_mut_ptr(),
args.len() as u32,
into_cstring(&callee.name).as_ptr(),
);
Ok(IRValue(IRValueType::I32, call))
}
}
}
#[derive(Debug, thiserror::Error)]
pub enum Error {
#[error("Type Mismatch: {0:?} {1:?}")]
TypeMismatch(IRValueType, IRValueType),
}

View File

@ -1,6 +1,6 @@
use crate::{
ast::parse::Parse,
lexer::{FullToken, Position, Token},
parser::Parse,
};
pub struct TokenStream<'a, 'b> {
@ -177,6 +177,27 @@ impl Default for TokenRange {
}
}
impl std::ops::Add for TokenRange {
type Output = TokenRange;
fn add(self, rhs: Self) -> Self::Output {
TokenRange {
start: self.start.min(rhs.start),
end: self.end.min(rhs.end),
}
}
}
impl std::iter::Sum for TokenRange {
fn sum<I: Iterator<Item = Self>>(mut iter: I) -> Self {
let mut start = iter.next().unwrap_or(Default::default());
for item in iter {
start = start + item;
}
start
}
}
#[derive(thiserror::Error, Debug)]
pub enum Error {
#[error("Expected {} at Ln {}, Col {}, got {:?}", .0, (.2).1, (.2).0, .1)]