%16 = "tf.Const"() {value = dense<1> : tensor<1xi32>} : () -> tensor<1xi32>

// CHECK: return [[z]]
func.func @while_cond_if(%cond: tensor<i1>, %x: tensor<i1>, %y: tensor<i1>, %z: tensor<i32>) -> (tensor<i1>) {
  %r = "tf.If"(%cond, %x, %y, %z) {then_branch = @then, else_branch = @else, is_stateless = true} : (tensor<i1>, tensor<i1>, tensor<i1>, tensor<i32>) -> tensor<i1>
  return %r : tensor<i1>
}

// CHECK-LABEL: func @while_body_if

      if (type_from_body.has_value()) return type_from_body;
    } else if (auto if_op = llvm::dyn_cast<TF::IfOp>(use.getOwner())) {
      auto then_branch = if_op.then_function();
      auto else_branch = if_op.else_function();
      assert(then_branch && else_branch);
      auto type_from_then = GetElementTypeFromAccess(
          then_branch.getArgument(use.getOperandNumber() - 1), module,
          infer_from_op);

    tensors, whose types are the same as what else_branch returns.
else_branch: A function that takes 'inputs' and returns a list of
    tensors.  whose types are the same as what then_branch returns.
  }];

  let arguments = (ins
    TF_Tensor:$cond,
    Variadic<TF_Tensor>:$input,

    FlatSymbolRefAttr:$then_branch,
    FlatSymbolRefAttr:$else_branch,

  // CHECK: "tf.If"(%[[COND]], %[[VAR]], %[[GVAR1]], %[[GVAR2]])
  %1 = "tf.If"(%cond, %ta#0) {
    then_branch = @then_branch, else_branch = @else_branch, device = "", is_stateless = false}
       : (tensor<i1>, tensor<!tf_type.resource>) -> tensor<!tf_type.resource>

  // CHECK: [[cond_res:%.*]]:2 = tfrt.cond [[cond]]
  // CHECK-SAME: @cond_true @cond_false(%arg0, %arg2) : (!tfrt.chain, !tfrt_fallback.tf_tensor)
  %2 = "tf.If"(%arg0, %arg1) {else_branch = @cond_false, then_branch = @cond_true, is_stateless = true} : (tensor<i1>, tensor<i32>) -> tensor<i32>
  // CHECK: [[out_ch:%.*]] = tfrt.merge.chains [[cond_res]]#0, %arg0 : !tfrt.chain, !tfrt.chain

    // CHECK: tf_device.return [[TRUE]] :
    tf_device.return %13 : tensor<i1>
  }) : () -> tensor<i1>
  func.return
}

// CHECK-LABEL: func @case_branch
func.func @case_branch(%arg0: !tf_res) -> tensor<i1> {
  // CHECK: [[TRUE:%.+]] = "tf.Const"
  // CHECK-SAME: value = dense<true> : tensor<i1>
  %0 = "tf.VarIsInitializedOp"(%arg0) : (!tf_res) -> tensor<i1>

  // Rewrite the branches.
  Region& then_branch = if_op.getThenBranch();
  Region& else_branch = if_op.getElseBranch();
  if (failed(DecomposeTensorListOpsInternal(
          &then_branch.front(), module, buffer_to_size,
          decomposed_partitioned_call_callees)))
    return failure();
  if (failed(DecomposeTensorListOpsInternal(
          &else_branch.front(), module, buffer_to_size,
          decomposed_partitioned_call_callees)))

  // Build the XLA computation func.
  // "const0" (bool)
  // "const1" (int32)
  // "if0" (pred = "const0", input = "const1", then_branch = "true_fn",
  //        else_branch = "false_fn")
  FunctionDefLibrary fdl;
  {
    tensorflow::Scope s = tensorflow::Scope::NewRootScope();
    Output arg = ops::_Arg(s.WithOpName("arg"), DT_INT32, 0);

namespace tensorflow {

TEST(RearrangeFunctionArgumentForFunctionTest, Basic) {
  FunctionDefLibrary fdl;
  {
    // Function for StatefulPartitionedCall's "f", If's
    // "then_branch"/"else_branch".
    // "arg0" (T=DT_RESOURCE), "arg1" (T=DT_BOOL)
    // "ret0" = "arg1"
    // "ret1" = "arg0"
    tensorflow::Scope s = tensorflow::Scope::NewRootScope();