auto input = block->getArgument(0);
    auto size = block->getArgument(3);

    // Subtract `input_rank` by 1 to get the item's rank, which is used as
    // `partial_position_shape`.
    auto input_rank = rewriter->create<TF::RankOp>(
        loc, tensorflow::GetTypeFromTFTensorShape({}, shape_dtype), input);
    auto partial_position_shape = rewriter->create<TF::SubOp>(

```python
    # Scalar indices (output is rank(params) - 1).
    output[a_0, ..., a_n, b_0, ..., b_n] =
      params[a_0, ..., a_n, indices, b_0, ..., b_n]

    # Vector indices (output is rank(params)).
    output[a_0, ..., a_n, i, b_0, ..., b_n] =
      params[a_0, ..., a_n, indices[i], b_0, ..., b_n]

    # Higher rank indices (output is rank(params) + rank(indices) - 1).

    int64_t rank = lhs_type.getRank();
    auto batch_dimensions = llvm::to_vector<4>(llvm::seq<int64_t>(0, rank - 2));
    auto lhs_contracting_dimensions = llvm::to_vector<4>(
        llvm::ArrayRef({op.getAdjX() ? rank - 2 : rank - 1}));
    auto rhs_contracting_dimensions = llvm::to_vector<4>(
        llvm::ArrayRef({op.getAdjY() ? rank - 1 : rank - 2}));

  if (topology_proto.mesh_shape_size() != kTPUTopologyRank)
    return absl::InvalidArgumentError(absl::StrCat(
        "'", kTopologyAttr, "' 'mesh_shape' must be rank ", kTPUTopologyRank,
        ", got rank ", topology_proto.mesh_shape_size()));

  for (auto mesh_shape_dim : llvm::enumerate(topology_proto.mesh_shape()))
    if (mesh_shape_dim.value() <= 0)
      return absl::InvalidArgumentError(

    module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
      func.func @main() -> tensor<!tf_type.variant<tensor<64x1xbf16>>> {
      // unknown rank
      %elem_shape = "tf.Const"() <{value = dense<-1> : tensor<i32>}> {device = "/job:localhost/replica:0/task:0/device:CPU:0"} : () -> tensor<i32>
      // zero reserved elements

           << "rank greater than 2.\n";
  }

  if (kernel_tensor_shape.size() <= 2) {
    return op.emitOpError()
           << "kernel tensor argument is " << kernel_tensor_shape.size()
           << " which is invalid, since kernel tensor argument must has a "
           << "rank greater than 2.\n";
  }

        as `tf.tpu.experimental.embedding.serving_embedding_lookup` which can
        take arbitrary rank of dense and sparse tensor. For ragged tensor,
        though the input tensor remains to be rank 2, the activations now can be
        rank 2 or above by specifying the output shape in the feature config or
        via the build method.

*   Add

			return len(m1.Headers) > len(m2.Headers)
			// 4: number of query matches
		default:
			return len(m1.QueryParams) > len(m2.QueryParams)
		}
	})
}

// getURIRank ranks a URI match type. Exact > Prefix > Regex
func getURIRank(match *istio.HTTPMatchRequest) int {
	if match.Uri == nil {
		return -1
	}
	switch match.Uri.MatchType.(type) {
	case *istio.StringMatch_Exact:
		return 3