Search anything:

User-defined Operators in Kaleidoscope

Binary Tree book by OpenGenus

Open-Source Internship opportunity by OpenGenus for programmers. Apply now.

User-defined operators take in a set of operands as input and return a result. We want to support functionalities such as division, logical negation, comparisons, etc.

Table of contents.

  1. Introduction.
  2. Supporting Binary operators.
  3. Supporting Unary Operators.
  4. Summary.
  5. References.


LLVM Control Flow: for loops.


Until now the language - Kaleidoscope generates optimized LLVM IR and supports JIT and control flow, we will further extend it to also include operators such as division, comparisons, logical negations, etc.

We will implement 'operator overloading' which is generalized compared to the implementation in languages such as C++. We will make it possible for a user to round out the set of supported operators.

Also, by using operator precedence parsing we can introduce new operators into the grammar of a programming language. As the JIT runs, the grammar is extended.

First, we will add unary and then binary operators. Let's first look at examples in Kaleidoscope;

# Logical unary not.
def unary!(v)
  if v then

# Define > with the same precedence as <.
def binary> 10 (LHS RHS)
  RHS < LHS;

# Binary "logical or", (note that it does not "short circuit")
def binary| 5 (LHS RHS)
  if LHS then
  else if RHS then

# Define = with slightly lower precedence than relationals.
def binary= 9 (LHS RHS)
  !(LHS < RHS | LHS > RHS);

Programming languages want to implement their standard run-time library in the language itself, We will be implementing important parts of Kaleidoscope in the library.

Implementation will involve first implementing support for user-defined binary operators followed by supporting unary operators.

Supporting Binary Operators.

We start by supporting binary/unary keywords;

enum Token {
  // operators
  tok_binary = -11,
  tok_unary = -12
static int gettok() {
    if (IdentifierStr == "for")
      return tok_for;
    if (IdentifierStr == "in")
      return tok_in;
    if (IdentifierStr == "binary")
      return tok_binary;
    if (IdentifierStr == "unary")
      return tok_unary;
    return tok_identifier;

Above we change the lexer so that it can support unary and binary keywords.

Next we extend PrototypeAST as follows;

/// PrototypeAST - This class represents the "prototype" for a function,
/// which captures its argument names as well as if it is an operator.
class PrototypeAST {
  std::string Name;
  std::vector<std::string> Args;
  bool IsOperator;
  unsigned Precedence;  // Precedence if a binary op.

  PrototypeAST(const std::string &name, std::vector<std::string> Args,
               bool IsOperator = false, unsigned Prec = 0)
  : Name(name), Args(std::move(Args)), IsOperator(IsOperator),
    Precedence(Prec) {}

  Function *codegen();
  const std::string &getName() const { return Name; }

  bool isUnaryOp() const { return IsOperator && Args.size() == 1; }
  bool isBinaryOp() const { return IsOperator && Args.size() == 2; }

  char getOperatorName() const {
    assert(isUnaryOp() || isBinaryOp());
    return Name[Name.size() - 1];

  unsigned getBinaryPrecedence() const { return Precedence; }

The above code allows us to represent definitions of the new operators in the def binary| 5 part of the function definition.
Now we know the name of the prototype and whether it is an operator or not and its precedence.

Now to parse user-defined operator prototype;

/// prototype
///   ::= id '(' id* ')'
///   ::= binary LETTER number? (id, id)
static std::unique_ptr<PrototypeAST> ParsePrototype() {
  std::string FnName;

  unsigned Kind = 0;  // 0 = identifier, 1 = unary, 2 = binary.
  unsigned BinaryPrecedence = 30;

  switch (CurTok) {
    return LogErrorP("Expected function name in prototype");
  case tok_identifier:
    FnName = IdentifierStr;
    Kind = 0;
  case tok_binary:
    if (!isascii(CurTok))
      return LogErrorP("Expected binary operator");
    FnName = "binary";
    FnName += (char)CurTok;
    Kind = 2;

    // Read the precedence if present.
    if (CurTok == tok_number) {
      if (NumVal < 1 || NumVal > 100)
        return LogErrorP("Invalid precedence: must be 1..100");
      BinaryPrecedence = (unsigned)NumVal;

  if (CurTok != '(')
    return LogErrorP("Expected '(' in prototype");

  std::vector<std::string> ArgNames;
  while (getNextToken() == tok_identifier)
  if (CurTok != ')')
    return LogErrorP("Expected ')' in prototype");

  // success.
  getNextToken();  // eat ')'.

  // Verify right number of names for operator.
  if (Kind && ArgNames.size() != Kind)
    return LogErrorP("Invalid number of operands for operator");

  return std::make_unique<PrototypeAST>(FnName, std::move(ArgNames), Kind != 0,

The next item is code generation, we need to generate LLVM IR for binary operators. For this we extend BinaryExprAST function as follows;

Value *BinaryExprAST::codegen() {
  Value *L = LHS->codegen();
  Value *R = RHS->codegen();
  if (!L || !R)
    return nullptr;

  switch (Op) {
  case '+':
    return Builder.CreateFAdd(L, R, "addtmp");
  case '-':
    return Builder.CreateFSub(L, R, "subtmp");
  case '*':
    return Builder.CreateFMul(L, R, "multmp");
  case '<':
    L = Builder.CreateFCmpULT(L, R, "cmptmp");
    // Convert bool 0/1 to double 0.0 or 1.0
    return Builder.CreateUIToFP(L, Type::getDoubleTy(TheContext),

  // If it wasn't a built-in binary operator, it must be a user-defined one. Emit
  // a call to it.
  Function *F = getFunction(std::string("binary") + Op);
  assert(F && "binary operator not found!");

  Value *Ops[2] = { L, R };
  return Builder.CreateCall(F, Ops, "binop");

Here, we've just added a default case for the existing binary node. The code looks up the appropriate operator from the symbol table and then generates a function call.

Now for the final part;

Function *FunctionAST::codegen() {
  // Transfer ownership of the prototype to the FunctionProtos map, but keep a
  // reference to it for use below.
  auto &P = *Proto;
  FunctionProtos[Proto->getName()] = std::move(Proto);
  Function *TheFunction = getFunction(P.getName());
  if (!TheFunction)
    return nullptr;

  // If this is an operator, install it.
  if (P.isBinaryOp())
    BinopPrecedence[P.getOperatorName()] = P.getBinaryPrecedence();

  // Create a new basic block to start insertion into.
  BasicBlock *BB = BasicBlock::Create(TheContext, "entry", TheFunction);

First, we register a user-defined operator in the precedence table before codegening a function.
The binary operator parsing logic handles this. This is all that is involved in extending the grammar.

Supporting Unary Operators.

First we need an AST for the same;

/// UnaryExprAST - Expression class for a unary operator.
class UnaryExprAST : public ExprAST {
  char Opcode;
  std::unique_ptr<ExprAST> Operand;

  UnaryExprAST(char Opcode, std::unique_ptr<ExprAST> Operand)
    : Opcode(Opcode), Operand(std::move(Operand)) {}

  Value *codegen() override;

The AST directly mirrors the binary operator AST node, except in this case it has a single child. The next step is to add the parsing logic;

/// unary
///   ::= primary
///   ::= '!' unary
static std::unique_ptr<ExprAST> ParseUnary() {
  // If the current token is not an operator, it must be a primary expr.
  if (!isascii(CurTok) || CurTok == '(' || CurTok == ',')
    return ParsePrimary();

  // If this is a unary operator, read it.
  int Opc = CurTok;
  if (auto Operand = ParseUnary())
    return std::make_unique<UnaryExprAST>(Opc, std::move(Operand));
  return nullptr;

First, if we encounter a unary operator during parsing, we eat it as a prefix and then parse the remaining piece as another unary operator.
This enables us to be able to handle multiple unary operators.

We change the previous callers for ParsePrimary to call ParseUnary, now we can call the later;

/// binoprhs
///   ::= ('+' unary)*
static std::unique_ptr<ExprAST> ParseBinOpRHS(int ExprPrec,
                                              std::unique_ptr<ExprAST> LHS) {
    // Parse the unary expression after the binary operator.
    auto RHS = ParseUnary();
    if (!RHS)
      return nullptr;
/// expression
///   ::= unary binoprhs
static std::unique_ptr<ExprAST> ParseExpression() {
  auto LHS = ParseUnary();
  if (!LHS)
    return nullptr;

  return ParseBinOpRHS(0, std::move(LHS));

Now we can parse unary operators and construct AST for them. The next step is to support prototypes so we can parse unary operator prototypes. We extend as follows;

/// prototype
///   ::= id '(' id* ')'
///   ::= binary LETTER number? (id, id)
///   ::= unary LETTER (id)
static std::unique_ptr<PrototypeAST> ParsePrototype() {
  std::string FnName;

  unsigned Kind = 0;  // 0 = identifier, 1 = unary, 2 = binary.
  unsigned BinaryPrecedence = 30;

  switch (CurTok) {
    return LogErrorP("Expected function name in prototype");
  case tok_identifier:
    FnName = IdentifierStr;
    Kind = 0;
  case tok_unary:
    if (!isascii(CurTok))
      return LogErrorP("Expected unary operator");
    FnName = "unary";
    FnName += (char)CurTok;
    Kind = 1;
  case tok_binary:

We name unary operators using names that include the operator character. This very useful during code generation.
Finally we add codegen support for unary operators;

Value *UnaryExprAST::codegen() {
  Value *OperandV = Operand->codegen();
  if (!OperandV)
    return nullptr;

  Function *F = getFunction(std::string("unary") + Opcode);
  if (!F)
    return LogErrorV("Unknown unary operator");

  return Builder.CreateCall(F, OperandV, "unop");

And we have successfully added support for unary and binary operators in Kaleidoscope.

We can test this using examples provided here.


User-defined operators take in a set of operands as input and return a result. They are defined by a user as input, it could be a comparison, a logical evaluation, addition, division, or any arithmetic.
In this article, we have extended Kaleidoscope to support user-defined variables.


LLVM Memory.

User-defined Operators in Kaleidoscope
Share this