if (!status.ok())
    GTEST_SKIP() << "NewWritableFile() not supported: " << status;

  const std::string test_data("asdf");
  status = new_file->Append(test_data);
  if (!status.ok()) GTEST_SKIP() << "Append() not supported: " << status;
  status = new_file->Flush();
  if (!status.ok()) GTEST_SKIP() << "Flush() not supported: " << status;
  status = new_file->Close();

  *result = filename_;
  return OkStatus();
}

Status ModularWritableFile::Append(StringPiece data) {
  if (ops_->append == nullptr)
    return errors::Unimplemented(tensorflow::strings::StrCat(
        "Append() not implemented for ", filename_));

  UniquePtrTo_TF_Status plugin_status(TF_NewStatus(), TF_DeleteStatus);
  ops_->append(file_.get(), data.data(), data.size(), plugin_status.get());

_tf_api_dir = _os.path.dirname(_os.path.dirname(_API_MODULE.__file__))
_current_module = _sys.modules[__name__]

if not hasattr(_current_module, "__path__"):
  __path__ = [_tf_api_dir]
elif _tf_api_dir not in __path__:
  __path__.append(_tf_api_dir)

# Hook external TensorFlow modules.

# Load tensorflow-io-gcs-filesystem if enabled
if (_os.getenv("TF_USE_MODULAR_FILESYSTEM", "0") == "true" or
    _os.getenv("TF_USE_MODULAR_FILESYSTEM", "0") == "1"):

    control_output_nodes.push_back(&control_outputs[i]->node);
  }

  // Do the actual function creation.
  DCHECK(append_hash_to_fn_name <= 1);
  tensorflow::FunctionDef fdef;
  status->status = tensorflow::GraphToFunctionDef(
      fn_body->graph, fn_name, append_hash_to_fn_name != 0,
      /*set_stateful_from_nodes=*/true,
      /*copy_placeholder_attrs_from_nodes=*/true, body_nodes, input_tensors,

  ASSERT_TF_OK(status_);
  tf_writable_file::Append(file, content, 4, status_);
  ASSERT_TF_OK(status_);
  length = tf_writable_file::Tell(file, status_);
  ASSERT_EQ(length, 4);
  ASSERT_TF_OK(status_);
  tf_writable_file::Flush(file, status_);
  ASSERT_TF_OK(status_);
  ASSERT_TRUE(
      CompareWithServer(filepath, 0, 4, &gcs_file->gcs_client, status_));

  tf_writable_file::Append(file, content + 4, 4, status_);

int64_t Read(const TF_RandomAccessFile* file, uint64_t offset, size_t n,
             char* buffer, TF_Status* status);
}  // namespace tf_random_access_file

namespace tf_writable_file {
void Cleanup(TF_WritableFile* file);
void Append(const TF_WritableFile* file, const char* buffer, size_t n,
            TF_Status* status);
int64_t Tell(const TF_WritableFile* file, TF_Status* status);
void Flush(const TF_WritableFile* file, TF_Status* status);

//            name (OpDef.name) regexp [A-Z][A-Za-z0-9_.\\-/]*.
//            If `append_hash_to_fn_name` is false, `fn_name` must be distinct
//            from other function and operation names (at least those
//            registered in graphs where this function will be used).
//  append_hash_to_fn_name - Must be 0 or 1. If set to 1, the actual name

      # it, and print that list for debugging. Not really clear if this is
      # helpful since the only examples I've seen are enormous.
      bazel query "rdeps(kind(py_test, $(cat $BATS_TEST_TMPDIR/deps)), $dep, 1)"
    done < $BATS_TEST_TMPDIR/missing_deps
    exit 1
  fi
}

# The Python package is not allowed to depend on any CUDA packages.

// DT_RESOURCE. The string serialization is converted back to a
// ResourceHandle during Session run where the TF_Tensor is converted to a
// Tensor.
// TFv2 does not depend on this conversion. There is no matching
// TF_TensorFromTensorV1 because the conversion to string is performed by the
// python side of Session.
static Status TF_TensorToTensorV1(const TF_Tensor* src, Tensor* dst) {

  ops->writable_file_ops = static_cast<TF_WritableFileOps*>(
      plugin_memory_allocate(TF_WRITABLE_FILE_OPS_SIZE));
  ops->writable_file_ops->cleanup = tf_writable_file::Cleanup;
  ops->writable_file_ops->append = tf_writable_file::Append;
  ops->writable_file_ops->tell = tf_writable_file::Tell;
  ops->writable_file_ops->flush = tf_writable_file::Flush;
  ops->writable_file_ops->sync = tf_writable_file::Sync;