func TestNumbers(t *testing.T) {
	testNumbers(t, token.INT, intTests)
	testNumbers(t, token.FLOAT, floatTests)
	testNumbers(t, token.IMAG, imagTests)
}

var opTests = []string{
	// unary operations
	`+ 0 = 0`,
	`+ ? = ?`,
	`- 1 = -1`,
	`- ? = ?`,
	`^ 0 = -1`,
	`^ ? = ?`,

	`! true = false`,
	`! false = true`,
	`! ? = ?`,

	// etc.

	// binary operations

					eqn := fmt.Sprintf("%s_%s %s %s_%s != %d", t.name, valname(x), op.name, t.name, valname(y), r)
					fmt.Printf("\tif %s { println(\"bad: %s\") }\n", eqn, eqn)
				}
			}
		}
		// test unary ops
		for _, op := range unops {
			for _, x := range t.testvalues() {
				r := t.trunc(op.eval(x))
				eqn := fmt.Sprintf("%s %s_%s != %d", op.name, t.name, valname(x), r)

		n.Op.goString(indent+2, "Op: "))
}

// Unary is a unary operation in an expression.
type Unary struct {
	Op         AST
	Expr       AST
	Suffix     bool // true for ++ -- when used as postfix
	SizeofType bool // true for sizeof (type)
}

func (u *Unary) print(ps *printState) {
	op, _ := u.Op.(*Operator)
	expr := u.Expr

	// X.(type)
	// Lhs := X.(type)
	TypeSwitchGuard struct {
		Lhs *Name // nil means no Lhs :=
		X   Expr  // X.(type)
		expr
	}

	Operation struct {
		Op   Operator
		X, Y Expr // Y == nil means unary expression
		expr
	}

	// Fun(ArgList[0], ArgList[1], ...)
	CallExpr struct {
		Fun     Expr
		ArgList []Expr // nil means no arguments
		HasDots bool   // last argument is followed by ...
		expr

func underIs(typ Type, f func(Type) bool) bool {
	typ = Unalias(typ)
	if tpar, _ := typ.(*TypeParam); tpar != nil {
		return tpar.underIs(f)
	}
	return f(under(typ))
}

func (check *Checker) unary(x *operand, e *syntax.Operation) {
	check.expr(nil, x, e.X)
	if x.mode == invalid {
		return
	}

	op := e.Op
	switch op {
	case syntax.And:
		// spec: "As an exception to the addressability

  }

  const std::string type_index_name =
      ::tensorflow::port::MaybeAbiDemangle(a.TypeId().name());

  return ::tensorflow::errors::Internal(
      "No unary variant binary_op function found for op ADD Variant "
      "type_name: ",
      type_index_name, " for device type: ", cc_ctx->device()->name());
}

void TF_AddNVariant(TF_OpKernelContext* ctx,

            ConstantValueKind.Boolean -> CompileTimeType.BOOLEAN
            ConstantValueKind.String -> CompileTimeType.STRING

            else -> CompileTimeType.ANY
        }
    }

    // Unary operators
    private fun FirLiteralExpression.evaluate(function: FirSimpleFunction): FirLiteralExpression? {
        if (value == null) return null
        (value as? String)?.let { opr ->
            evalUnaryOp(

                                             #v1, \
                                             pred, \
                                             v1), on_failure)

// Unary predicate assertion macros.
#define EXPECT_PRED_FORMAT1(pred_format, v1) \
  GTEST_PRED_FORMAT1_(pred_format, v1, GTEST_NONFATAL_FAILURE_)
#define EXPECT_PRED1(pred, v1) \
  GTEST_PRED1_(pred, v1, GTEST_NONFATAL_FAILURE_)

// example tf.Sub operation.
def TF_CwiseBinary : NativeOpTrait<"TF::CwiseBinary">;

// Coefficient wise unary operation, for example tf.Sqrt operation.
def TF_CwiseUnary : NativeOpTrait<"TF::CwiseUnary">;

// op op X == op X (unary) / X op X == X (binary)
// This version requires TF_SameOperandsAndResultTypeResolveRef trait unlike
// the core implementation requiring SameOperandsAndResultType.

namespace mlir::quant {

using QuantMethod = tensorflow::quantization::QuantizationMethod::PresetMethod;

enum class OpType {
  kDynamicRangeOp,  // Dynamic Range kernels only have rhs attr.
  kUnaryOp,         // Unary ops have one min/max attr.
  kBinaryOp,        // Binary ops have lhs/rhs attr.
  kQuantizationOp,  // Quantization ops have input/output attr.
};

// For each op type, the following axis carries axis information: