More work on UIR

drimc_rs/src/ir_untyped.rs

@@ -4,8 +4,10 @@
 //! simplifies the complexity of the syntax tree, and is used to perform type inference, at which
 //! point it is translated into typed IR.
 
+use std::fmt::Display;
+
 use crate::{
-    syntax::{self, Identifier, SyntaxTree},
+    syntax::{self, Identifier, Literal, SyntaxTree},
     typeck::Type,
 };
 
@@ -143,6 +145,19 @@ enum Location {
 
     /// A compiler-generated temporary location.
     Temporary(u64),
+
+    /// A constant value.
+    Literal(Literal),
+}
+
+impl Display for Location {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Location::Named(id) => write!(f, "n`{}`", id),
+            Location::Temporary(t) => write!(f, "t`{}`", t),
+            Location::Literal(l) => write!(f, "c`{}`", l),
+        }
+    }
 }
 
 /// Converts a program's abstract syntax tree into untyped IR code.
@@ -204,7 +219,13 @@ fn convert_fn(
     mut arguments: Vec<syntax::Pattern>,
     definition: syntax::Expr,
 ) -> Vec<Instruction> {
-    if arguments.len() > 1 {
+    if arguments.is_empty() {
+        let ret_loc = temporary(counter);
+        eval_expr(counter, definition, &ret_loc)
+            .into_iter()
+            .chain(std::iter::once(Instruction::Return { target: ret_loc }))
+            .collect()
+    } else {
         let first = arguments.remove(0);
 
         let lambda_loc = temporary(counter);
@@ -220,12 +241,6 @@ fn convert_fn(
             },
             Instruction::Return { target: lambda_loc },
         ]
-    } else {
-        let ret_loc = temporary(counter);
-        eval_expr(counter, definition, &ret_loc)
-            .into_iter()
-            .chain(std::iter::once(Instruction::Return { target: ret_loc }))
-            .collect()
     }
 }
 
@@ -271,11 +286,6 @@ fn bind_pattern(counter: &mut u64, p: syntax::Pattern, l: &Location) -> Vec<Inst
 /// Emits intructions that evaluate the expression `e`, then place the result in location `l`.
 fn eval_expr(counter: &mut u64, e: syntax::Expr, l: &Location) -> Vec<Instruction> {
     match e {
-        syntax::Expr::UnaryOp {
-            kind,
-            val,
-            translation,
-        } => todo!(),
         syntax::Expr::BinaryOp {
             kind,
             left,
@@ -315,7 +325,13 @@ fn eval_expr(counter: &mut u64, e: syntax::Expr, l: &Location) -> Vec<Instructio
                 }))
                 .collect()
         }
-        syntax::Expr::VariableReference(_) => todo!(),
-        syntax::Expr::Literal(_) => todo!(),
+        syntax::Expr::VariableReference(name) => vec![Instruction::Collect {
+            target: l.clone(),
+            source: vec![Location::Named(name)],
+        }],
+        syntax::Expr::Literal(lit) => vec![Instruction::Collect {
+            target: l.clone(),
+            source: vec![Location::Literal(lit)],
+        }],
     }
 }

drimc_rs/src/main.rs

@@ -4,8 +4,8 @@ use std::{error::Error, fmt::Display, process::exit, str::FromStr};
 
 use clap::Parser;
 use drimc_rs::{
+    ir_untyped::ast_to_untyped_ir,
     parser::{parser, ParserError, ParserMeta},
-    typeck::typeck, ir_untyped::ast_to_untyped_ir,
 };
 
 /// Optimization levels.
@@ -98,6 +98,10 @@ impl FromStr for Optimization {
 /// Targets for compiling to.
 #[derive(Debug)]
 enum Target {
+    // -------------------- IR generation --------------------
+    /// Generate untyped intermediate representation (UIR) code.
+    UIR,
+
     // -------------------- C generation --------------------
 
     // Highest priority codegen, since it allows us to compile to the vast majority of possible
@@ -138,25 +142,26 @@ enum Target {
     Wasm,
 
     // -------------------- Other language generation --------------------
-    /// Directly generate Lua code.
+
     // Medium-priority codegen, since Lua is the only type of code that runs in e.g. plugins and
     // scripts in some software, and it is difficult to generate efficient Lua without direct
     // support from the compiler.
+    /// Directly generate Lua code.
     Lua,
 
-    /// Directly generate Python code.
     // Low-priority codegen; the same situation as Lua, but situations that require Python code with
     // no alternatives are much less common than situations that require Lua code with no
     // alternatives.
+    /// Directly generate Python code.
     Python,
 
-    /// Directly generate Go code.
     // Extremely low-priority codegen; almost no valid use cases.
+    /// Directly generate Go code.
     Go,
 
-    /// Directly generate Ada code.
     // Currently zero-priority codegen; no valid use cases whatsoever as far as I (Alex) can
     // determine.
+    /// Directly generate Ada code.
     Ada,
 }
 
@@ -166,6 +171,7 @@ impl Display for Target {
             f,
             "{}",
             match self {
+                Target::UIR => "uir",
                 Target::CSource => "c",
                 Target::Assembly => "asm",
                 Target::ObjectFile => "obj",
@@ -190,6 +196,7 @@ impl FromStr for Target {
         use Target::*;
 
         let targets = [
+            ("uir", "Untyped intermediate representation.", UIR),
             ("c", "C source code.", CSource),
             ("asm", "Assembly code.", Assembly),
             ("obj", "Object file.", ObjectFile),

drimc_rs/src/parser.rs

@@ -379,9 +379,7 @@ fn parse_expression<'a>(
                 .unwrap()
         });
 
-        let unary = parse_unary(m, application);
-
-        let binary = (0..=10).rev().fold(unary.boxed(), |p, precedence| {
+        let binary = (0..=10).rev().fold(application.boxed(), |p, precedence| {
             parse_binary(m, precedence, p).boxed()
         });
 
@@ -389,22 +387,6 @@ fn parse_expression<'a>(
     })
 }
 
-fn parse_unary(
-    _m: &ParserMeta,
-    base: impl Parser<char, Expr, Error = Simple<char>> + Clone,
-) -> impl Parser<char, Expr, Error = Simple<char>> + Clone {
-    pad(just("-").to("-"))
-        .repeated()
-        .then(base)
-        .map(|(ops, exp)| {
-            ops.into_iter().fold(exp, |exp, op| Expr::UnaryOp {
-                kind: op.to_string(),
-                val: Box::new(exp),
-                translation: "negate".to_string(),
-            })
-        })
-}
-
 fn parse_binary<'a>(
     m: &'a ParserMeta,
     prec: u32,

drimc_rs/src/syntax.rs

@@ -1,5 +1,7 @@
 //! Syntax tree for Drim code.
 
+use std::fmt::Display;
+
 use num_bigint::BigUint;
 
 /// A concrete syntax tree. This represents the full content of a Drim program, including all
@@ -90,18 +92,6 @@ pub struct TypeConstructor {
 /// The different kinds of expressions.
 #[derive(Clone, Debug)]
 pub enum Expr {
-    /// Unary operators, e.g., `-5`.
-    UnaryOp {
-        /// The text of the operator.
-        kind: String,
-
-        /// The value being operated upon.
-        val: Box<Expr>,
-
-        /// The function that the unary operator translates to.
-        translation: String,
-    },
-
     /// Binary operators, e.g., `5 + 5`.
     BinaryOp {
         /// The text of the operator.
@@ -258,6 +248,18 @@ pub struct Identifier {
     pub elems: Vec<String>,
 }
 
+impl Display for Identifier {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        for (n, e) in self.elems.iter().enumerate() {
+            if n != 0 {
+                write!(f, "::")?;
+            }
+            write!(f, "{}", e)?;
+        }
+        Ok(())
+    }
+}
+
 /// Literal values included in source code.
 #[derive(Clone, Debug)]
 pub enum Literal {
@@ -270,3 +272,13 @@ pub enum Literal {
     /// `123.456`
     Float(f64),
 }
+
+impl Display for Literal {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Literal::String(s) => write!(f, "{}", s),
+            Literal::Integer(i) => write!(f, "{}", i),
+            Literal::Float(fl) => write!(f, "{}", fl),
+        }
+    }
+}

drimc_rs/uir_test.drim

@@ -1,3 +1,3 @@
 // IR Test
 
-def nul (a, b) = ();
+def nul (a, b) = sin a;