feat: Introduce REF= and REFERENCE_TO

compiler/plc_ast/src/ast.rs

@@ -487,6 +487,9 @@ pub enum DataType {
     PointerType {
         name: Option<String>,
         referenced_type: Box<DataTypeDeclaration>,
+        auto_deref: bool,
+        /// Denotes whether the variable was declared as `REFERENCE TO`, e.g. `foo : REFERENCE TO DINT`
+        is_reference_to: bool,
     },
     StringType {
         name: Option<String>,
@@ -596,11 +599,14 @@ pub struct AstNode {
 #[derive(Debug, Clone, PartialEq)]
 pub enum AstStatement {
     EmptyStatement(EmptyStatement),
-    // a placeholder that indicates a default value of a datatype
+
+    // A placeholder which indicates a default value of a datatype
     DefaultValue(DefaultValue),
+
     // Literals
     Literal(AstLiteral),
     MultipliedStatement(MultipliedStatement),
+
     // Expressions
     ReferenceExpr(ReferenceExpr),
     Identifier(String),
@@ -612,15 +618,17 @@ pub enum AstStatement {
     ParenExpression(Box<AstNode>),
     RangeStatement(RangeStatement),
     VlaRangeStatement,
-    // Assignment
+
+    // TODO: Merge these variants with a `kind` field?
+    // Assignments
     Assignment(Assignment),
-    // OutputAssignment
     OutputAssignment(Assignment),
-    //Call Statement
+    RefAssignment(Assignment),
+
     CallStatement(CallStatement),
+
     // Control Statements
     ControlStatement(AstControlStatement),
-
     CaseCondition(Box<AstNode>),
     ExitStatement(()),
     ContinueStatement(()),
@@ -661,6 +669,9 @@ impl Debug for AstNode {
             AstStatement::OutputAssignment(Assignment { left, right }) => {
                 f.debug_struct("OutputAssignment").field("left", left).field("right", right).finish()
             }
+            AstStatement::RefAssignment(Assignment { left, right }) => {
+                f.debug_struct("ReferenceAssignment").field("left", left).field("right", right).finish()
+            }
             AstStatement::CallStatement(CallStatement { operator, parameters }) => f
                 .debug_struct("CallStatement")
                 .field("operator", operator)
@@ -1319,6 +1330,19 @@ impl AstFactory {
         )
     }
 
+    // TODO: Merge `create_assignment`, `create_output_assignment` and `create_ref_assignment`
+    //       once the the Assignment AstStatements have been merged and a `kind` field is available
+    //       I.e. something like `AstStatement::Assignment { data, kind: AssignmentKind { Normal, Output, Reference } }
+    //       and then fn create_assignment(kind: AssignmentKind, ...)
+    pub fn create_ref_assignment(left: AstNode, right: AstNode, id: AstId) -> AstNode {
+        let location = left.location.span(&right.location);
+        AstNode::new(
+            AstStatement::RefAssignment(Assignment { left: Box::new(left), right: Box::new(right) }),
+            id,
+            location,
+        )
+    }
+
     pub fn create_member_reference(member: AstNode, base: Option<AstNode>, id: AstId) -> AstNode {
         let location = base
             .as_ref()

compiler/plc_ast/src/visitor.rs

@@ -312,6 +312,15 @@ pub trait AstVisitor: Sized {
         stmt.walk(self)
     }
 
+    /// Visits an `RefAssignment` node.
+    /// Make sure to call `walk` on the `Assignment` node to visit its children.
+    /// # Arguments
+    /// * `stmt` - The unwraped, typed `Assignment` node to visit.
+    /// * `node` - The wrapped `AstNode` node to visit. Offers access to location information and AstId
+    fn visit_ref_assignment(&mut self, stmt: &Assignment, _node: &AstNode) {
+        stmt.walk(self)
+    }
+
     /// Visits a `CallStatement` node.
     /// Make sure to call `walk` on the `CallStatement` node to visit its children.
     /// # Arguments
@@ -556,6 +565,7 @@ impl Walker for AstNode {
             AstStatement::VlaRangeStatement => visitor.visit_vla_range_statement(node),
             AstStatement::Assignment(stmt) => visitor.visit_assignment(stmt, node),
             AstStatement::OutputAssignment(stmt) => visitor.visit_output_assignment(stmt, node),
+            AstStatement::RefAssignment(stmt) => visitor.visit_ref_assignment(stmt, node),
             AstStatement::CallStatement(stmt) => visitor.visit_call_statement(stmt, node),
             AstStatement::ControlStatement(stmt) => visitor.visit_control_statement(stmt, node),
             AstStatement::CaseCondition(stmt) => visitor.visit_case_condition(stmt, node),

compiler/plc_diagnostics/src/diagnostics/diagnostics_registry.rs

@@ -199,6 +199,8 @@ lazy_static! {
         E095,   Error,      include_str!("./error_codes/E095.md"),  // Action call without `()`
         E096,   Warning,    include_str!("./error_codes/E096.md"),  // Integer Condition
         E097,   Error,      include_str!("./error_codes/E097.md"),  // Invalid Array Range
+        E098,   Error,      include_str!("./error_codes/E098.md"),  // Invalid `REF=` assignment
+        E099,   Error,      include_str!("./error_codes/E099.md"),  // Invalid `REFERENCE TO` declaration
     );
 }
 

compiler/plc_diagnostics/src/diagnostics/error_codes/E098.md

@@ -0,0 +1,19 @@
+# Invalid REF= assignment
+
+`REF=` assignments are considered valid if the left-hand side of the assignment is a pointer variable 
+and the right-hand side is a variable of the type that is being referenced.
+
+For example assignments such as the following are invalid
+
+```smalltalk
+VAR
+    foo     : DINT;
+    bar     : DINT;
+    qux     : SINT;
+    refFoo  : REFERENCE TO DINT;
+END_VAR
+
+refFoo  REF= 5;         // `5` is not a variable
+foo     REF= bar;       // `foo` is not a pointer
+refFoo  REF= qux;       // `refFoo` and `qux` have different types, DINT vs SINT
+```

compiler/plc_diagnostics/src/diagnostics/error_codes/E099.md

@@ -0,0 +1,8 @@
+# Invalid `REFERENCE TO` declaration
+
+`REFERENCE TO` variable declarations are considered valid if the referenced type is not of the following form
+* `foo : REFERENCE TO REFERENCE TO (* ... *)`
+* `foo : ARRAY[...] OF REFERENCE TO (* ... *)`
+* `foo : REF_TO REFERENCE TO (* ... *)`
+
+Furthermore `REFERENCE_TO` variables must not be initialized in their declaration, e.g. `foo : REFERENCE TO DINT := bar`.

src/codegen/generators/expression_generator.rs

@@ -2507,14 +2507,13 @@ impl<'ink, 'b> ExpressionCodeGenerator<'ink, 'b> {
     /// - `access` the ReferenceAccess of the reference to generate
     /// - `base` the "previous" segment of an optional qualified reference-access
     /// - `original_expression` the original ast-statement used to report Diagnostics
-    fn generate_reference_expression(
+    pub(crate) fn generate_reference_expression(
         &self,
         access: &ReferenceAccess,
         base: Option<&AstNode>,
         original_expression: &AstNode,
     ) -> Result<ExpressionValue<'ink>, Diagnostic> {
         match (access, base) {
-
             // expressions like `base.member`, or just `member`
             (ReferenceAccess::Member(member), base) => {
                 let base_value = base.map(|it| self.generate_expression_value(it)).transpose()?;

src/codegen/generators/statement_generator.rs

@@ -125,7 +125,9 @@ impl<'a, 'b> StatementCodeGenerator<'a, 'b> {
             AstStatement::Assignment(data, ..) => {
                 self.generate_assignment_statement(&data.left, &data.right)?;
             }
-
+            AstStatement::RefAssignment(data, ..) => {
+                self.generate_ref_assignment(&data.left, &data.right)?;
+            }
             AstStatement::ControlStatement(ctl_statement, ..) => {
                 self.generate_control_statement(ctl_statement)?
             }
@@ -234,6 +236,31 @@ impl<'a, 'b> StatementCodeGenerator<'a, 'b> {
         }
     }
 
+    /// Generates IR for a `REF=` assignment, which is syntactic sugar for an assignment where the
+    /// right-hand side is wrapped in a `REF(...)` call. Specifically `foo REF= bar` and
+    /// `foo := REF(bar)` are the same.
+    ///
+    /// Note: Although somewhat similar to the [`generate_assignment_statement`] function, we can't
+    /// apply the code here because the left side of a `REF=` assignment is flagged as auto-deref.
+    /// For `REF=` assignments we don't want (and can't) deref without generating incorrect IR.
+    pub fn generate_ref_assignment(&self, left: &AstNode, right: &AstNode) -> Result<(), Diagnostic> {
+        let exp = self.create_expr_generator();
+        let ref_builtin = self.index.get_builtin_function("REF").expect("REF must exist");
+
+        let AstStatement::ReferenceExpr(data) = &left.stmt else {
+            unreachable!("should be covered by a validation")
+        };
+
+        let left_ptr_val = {
+            let expr = exp.generate_reference_expression(&data.access, data.base.as_deref(), left)?;
+            expr.get_basic_value_enum().into_pointer_value()
+        };
+        let right_expr_val = ref_builtin.codegen(&exp, &[&right], right.get_location())?;
+
+        self.llvm.builder.build_store(left_ptr_val, right_expr_val.get_basic_value_enum());
+        Ok(())
+    }
+
     /// generates an assignment statement _left_ := _right_
     ///
     /// `left_statement` the left side of the assignment

src/codegen/tests/statement_codegen_test.rs

@@ -184,3 +184,127 @@ fn floating_point_type_casting() {
 
     insta::assert_snapshot!(result);
 }
+
+#[test]
+fn ref_assignment() {
+    let result = codegen(
+        r#"
+        FUNCTION main
+        VAR
+            a : REF_TO DINT;
+            b : DINT;
+        END_VAR
+            a REF= b;
+        END_PROGRAM
+        "#,
+    );
+
+    insta::assert_snapshot!(result, @r###"
+    ; ModuleID = 'main'
+    source_filename = "main"
+
+    define void @main() section "fn-$RUSTY$main:v" {
+    entry:
+      %a = alloca i32*, align 8
+      %b = alloca i32, align 4
+      store i32* null, i32** %a, align 8
+      store i32 0, i32* %b, align 4
+      store i32* %b, i32** %a, align 8
+      ret void
+    }
+    "###);
+}
+
+#[test]
+fn reference_to_assignment() {
+    let auto_deref = codegen(
+        r#"
+        FUNCTION main
+            VAR
+                a : REFERENCE TO DINT;
+            END_VAR
+            a := 5;
+        END_FUNCTION
+        "#,
+    );
+
+    let manual_deref = codegen(
+        r#"
+        FUNCTION main
+            VAR
+                a : REF_TO DINT;
+            END_VAR
+            a^ := 5;
+        END_FUNCTION
+        "#,
+    );
+
+    // We want to assert that `a := 5` and `a^ := 5` yield identical IR
+    assert_eq!(auto_deref, manual_deref);
+
+    insta::assert_snapshot!(auto_deref, @r###"
+    ; ModuleID = 'main'
+    source_filename = "main"
+
+    define void @main() section "fn-$RUSTY$main:v" {
+    entry:
+      %a = alloca i32*, align 8
+      store i32* null, i32** %a, align 8
+      %deref = load i32*, i32** %a, align 8
+      store i32 5, i32* %deref, align 4
+      ret void
+    }
+    "###);
+}
+
+#[test]
+fn reference_to_string_assignment() {
+    let auto_deref = codegen(
+        r#"
+        FUNCTION main
+            VAR
+                a : REFERENCE TO STRING;
+            END_VAR
+
+            a := 'hello';
+        END_FUNCTION
+        "#,
+    );
+
+    let manual_deref = codegen(
+        r#"
+        FUNCTION main
+            VAR
+                a : REF_TO STRING;
+            END_VAR
+
+            a^ := 'hello';
+        END_FUNCTION
+        "#,
+    );
+
+    // We want to assert that `a := 'hello'` and `a^ := 'hello'` yield identical IR
+    assert_eq!(auto_deref, manual_deref);
+
+    insta::assert_snapshot!(auto_deref, @r###"
+    ; ModuleID = 'main'
+    source_filename = "main"
+
+    @utf08_literal_0 = private unnamed_addr constant [6 x i8] c"hello\00"
+
+    define void @main() section "fn-$RUSTY$main:v" {
+    entry:
+      %a = alloca [81 x i8]*, align 8
+      store [81 x i8]* null, [81 x i8]** %a, align 8
+      %deref = load [81 x i8]*, [81 x i8]** %a, align 8
+      %0 = bitcast [81 x i8]* %deref to i8*
+      call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %0, i8* align 1 getelementptr inbounds ([6 x i8], [6 x i8]* @utf08_literal_0, i32 0, i32 0), i32 6, i1 false)
+      ret void
+    }
+
+    ; Function Attrs: argmemonly nofree nounwind willreturn
+    declare void @llvm.memcpy.p0i8.p0i8.i32(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i32, i1 immarg) #0
+
+    attributes #0 = { argmemonly nofree nounwind willreturn }
+    "###);
+}

src/index.rs

@@ -1130,6 +1130,7 @@ impl Index {
         if segments.is_empty() {
             return None;
         }
+
         //For the first element, if the context does not contain that element, it is possible that the element is also a global variable
         let init = match context {
             Some(context) => self

src/index/tests/index_tests.rs

@@ -1253,6 +1253,7 @@ fn pointer_and_in_out_pointer_should_not_conflict() {
             name: "__main_x".to_string(),
             inner_type_name: "INT".to_string(),
             auto_deref: false,
+            is_reference_to: false,
         }
     );
 
@@ -1264,6 +1265,7 @@ fn pointer_and_in_out_pointer_should_not_conflict() {
             name: "__auto_pointer_to_INT".to_string(),
             inner_type_name: "INT".to_string(),
             auto_deref: true,
+            is_reference_to: false,
         }
     );
 }
@@ -1303,6 +1305,7 @@ fn pointer_and_in_out_pointer_should_not_conflict_2() {
             name: "__main_x".to_string(),
             inner_type_name: "INT".to_string(),
             auto_deref: false,
+            is_reference_to: false,
         }
     );
 
@@ -1314,6 +1317,7 @@ fn pointer_and_in_out_pointer_should_not_conflict_2() {
             name: "__auto_pointer_to_INT".to_string(),
             inner_type_name: "INT".to_string(),
             auto_deref: true,
+            is_reference_to: false,
         }
     );
 }

src/index/tests/snapshots/rusty__index__tests__index_tests__pre_processing_generates_inline_pointer_to_pointer-2.snap

@@ -10,6 +10,8 @@ UserTypeDeclaration {
         referenced_type: DataTypeReference {
             referenced_type: "__foo_inline_pointer_",
         },
+        auto_deref: false,
+        is_reference_to: false,
     },
     initializer: None,
     scope: Some(

src/index/tests/snapshots/rusty__index__tests__index_tests__pre_processing_generates_inline_pointer_to_pointer.snap

@@ -10,6 +10,8 @@ UserTypeDeclaration {
         referenced_type: DataTypeReference {
             referenced_type: "INT",
         },
+        auto_deref: false,
+        is_reference_to: false,
     },
     initializer: None,
     scope: Some(

src/index/tests/snapshots/rusty__index__tests__index_tests__pre_processing_generates_inline_pointers.snap

@@ -10,6 +10,8 @@ UserTypeDeclaration {
         referenced_type: DataTypeReference {
             referenced_type: "INT",
         },
+        auto_deref: false,
+        is_reference_to: false,
     },
     initializer: None,
     scope: Some(