}

// Grammar for Operand:
//
// operand → multOp ( ("-" | "+") multOp )*
// multOp  → unary ( ("/" | "*" | "%") unary )*
// unary   → "-" unary | primary
// primary → Value | Variable | "(" expression ")"
//

// An Operand is a single term followed by an optional sequence of
// terms separated by +/-
type Operand struct {
	Left  *MultOp     `parser:"@@"`
	Right []*OpFactor `parser:"(@@)*"`
}

  xla::Shape shape = xla::ShapeUtil::MakeShape(xla::S32, {1});

  auto buffer = c_api_client->pjrt_c_client()->client->BufferFromHostBuffer(
      data.data(), shape.element_type(), shape.dimensions(),
      /*byte_strides=*/std::nullopt,
      xla::PjRtClient::HostBufferSemantics::kImmutableOnlyDuringCall, nullptr,
      c_api_client->pjrt_c_client()->client->addressable_devices()[0]);
  CHECK_OK(buffer.status());

                    int expected_max_outputs,
                    CancellationManager& cancellation_manager,
                    std::optional<int64_t> step_id = std::nullopt) const;

  void StartExecute(TFE_Context* context,
                    const std::vector<std::vector<TFE_TensorHandle*>>& inputs,
                    const char* operation_name, const TFE_OpAttrs* attributes,

dallo
dai
dagli
dall
dagl
dalla
dalle
di
del
dello
dei
degli
dell
degl
della
delle
in
nel
nello
nei
negli
nell
negl
nella
nelle
su
sul
sullo
sui
sugli
sull
sugl
sulla
sulle
per
tra
contro
io
tu
lui
lei
noi
voi
loro
mio
mia
miei
mie
tuo
tua
tuoi
tue
suo
sua
suoi
sue

	if err != nil {
		return err
	}
	for _, rt := range e.Right {
		err = rt.Right.aggregateRow(r, tableAlias)
		if err != nil {
			return err
		}
	}
	return nil
}

func (e *MultOp) aggregateRow(r Record, tableAlias string) error {
	err := e.Left.aggregateRow(r, tableAlias)
	if err != nil {
		return err
	}
	for _, rt := range e.Right {
		err = rt.Right.aggregateRow(r, tableAlias)

		if rerr != nil {
			return nil, rerr
		}
		err := lval.arithOp(op, rval)
		if err != nil {
			return nil, err
		}
	}
	return lval, nil
}

func (e *MultOp) evalNode(r Record, tableAlias string) (*Value, error) {
	lval, lerr := e.Left.evalNode(r, tableAlias)
	if lerr != nil || len(e.Right) == 0 {
		return lval, lerr
	}

            parallel_device.CopyToParallelDevice(
                context, absl::get<TFE_TensorHandle*>(input), status));
        if (TF_GetCode(status) != TF_OK) return absl::nullopt;
        parallel_inputs.push_back(parallel_tensor.get());
        implicitly_broadcast_tensors.emplace_back(std::move(parallel_tensor));
      } else {
        TF_SetStatus(
            status, TF_INVALID_ARGUMENT,

	return
}

func (e *Operand) analyze(s *Select) (result qProp) {
	result.combine(e.Left.analyze(s))
	for _, r := range e.Right {
		result.combine(r.Right.analyze(s))
	}
	return
}

func (e *MultOp) analyze(s *Select) (result qProp) {
	result.combine(e.Left.analyze(s))
	for _, r := range e.Right {
		result.combine(r.Right.analyze(s))
	}
	return
}

                    const TFE_OpAttrs* attributes, int expected_max_outputs,
                    CancellationManager& cancellation_manager,
                    absl::optional<int64_t> step_id = absl::nullopt);
  // Block until the previous `StartExecute` operation has executed. Forwards
  // the status from `TFE_Execute` and returns outputs if the status is OK.
  std::vector<TensorHandlePtr> Join(TF_Status* status);

      std::vector<int64_t>&& dims) const {
    return absl::make_optional(dims);
  }

  absl::optional<std::vector<int64_t>> unknown_shape() const {
    return absl::nullopt;
  }

  static constexpr int64_t kUnknownDim =
      shape_inference::InferenceContext::kUnknownDim;
  TF_Status* status_;
  TFE_ContextOptions* tfe_context_options_;
  TFE_Context* tfe_context_;
};