Python，如何使用@ in类方法

Question

I try to use the @ in the class method. like this

class Dataset:
  @parse_func
  def get_next_batch(self):
      return self.generator.__next__()

and the parse function like this:

def parse_func(load_batch):
  def wrapper(**para):
    batch_files_path, batch_masks_path, batch_label = load_batch(**para)
    batch_images = []
    batch_masks = []
    for (file_path, mask_path) in zip(batch_files_path, batch_masks_path):
        image = cv2.imread(file_path)
        mask = cv2.imread(mask_path)
        batch_images.append(image)
        batch_masks.append(mask)
    return np.asarray(batch_images, np.float32), np.asarray(batch_masks, np.uint8), batch_label

  return wrapper

However, when I call dataset.get_next_batch() , it will raise a exception as followed.

Traceback (most recent call last): TypeError: wrapper() takes exactly 0 arguments (1 given)

Do you know why raise this error and any solution? Thank you very much!

Answer 1

The function wrapper(**kwargs) accepts named arguments only . However, in instance methods, the self is automatically passed as the first positional argument. Since your method does not accept positional arguments, it fails.

You could edit to wrapper(self, **kwargs) or, more general wrapper(*args, **kwargs) . However, the way you are using it, it is not clear what those arguments are.

Answer 2

Just simply change

def parse_func(load_batch):
  def wrapper(*para):
    batch_files_path, batch_masks_path, batch_label = load_batch(*para)
    batch_images = []
    batch_masks = []
    for (file_path, mask_path) in zip(batch_files_path, batch_masks_path):
        image = cv2.imread(file_path)
        mask = cv2.imread(mask_path)
        batch_images.append(image)
        batch_masks.append(mask)
    return np.asarray(batch_images, np.float32), np.asarray(batch_masks, np.uint8), batch_label

  return wrapper()

@ symbol mean a decorator function. Here, it means parse_func(get_next_batch) . So if the wrapper using the keyword params ( **para ), you just want to pass some params to the wrapper but you don't actually except for the self args. So here I replace the params to positional params *para .