[FEATURE]

La richiesta riguarda l’API @deeplake.compute, quindi trova innanzitutto l’implementazione del decorator e il percorso di trasformazione di esempio che scrive in sample_out. Poi leggi i test/la documentazione disponibili sul comportamento di compute o delle trasformazioni di Dataset, in particolare eventuali controlli sulla propagazione dei campi, e riproduci l’issue con un caso che aggiunga un solo campo di output. Verifica che il lavoro sia completato confermando che un percorso in stile pipe_missing preserva automaticamente i campi di input non modificati e aggiungendo/aggiornando un test di regressione per questo comportamento.

🚨🚨 Feature Request

• Related to an existing Issue
• A new implementation (Improvement, Extension)

Is your feature request related to a problem?

@deeplake.compute
def resize(sample_in, sample_out, new_size):
    ## First two arguments are always default arguments containing:
    #     1st argument is an element of the input iterable (list, dataset, array,...)
    #     2nd argument is a dataset sample
    ## Third argument is the required size for the output images
    
    # Append the label and image to the output sample
    sample_out.labels.append(sample_in.labels.numpy())
    sample_out.images.append(np.array(Image.fromarray(sample_in.images.numpy()).resize(new_size)))
    sample_out.info.append(sample_in.info.dict())
    
    return sample_out

In the above example when new samples are computed we need to make sure all tensors are passed to the sample_out even if it's unmodified. It will be really helpful if we can pipe them directly without explicitly having to get them as numpy or dict.

Description of the possible solution

@deeplake.compute
def resize(sample_in, sample_out, new_size):

    
    # Append the label and image to the output sample
    sample_out.labels.append(sample_in.labels)
    sample_out.images.append(np.array(Image.fromarray(sample_in.images.numpy()).resize(new_size)))
    sample_out.info.append(sample_in.info)
    return sample_out

An alternative solution to the problem can look like

@deeplake.compute
def resize(sample_in, sample_out, new_size):

    
    # Append the label and image to the output sample
    sample_out.images.append(np.array(Image.fromarray(sample_in.images.numpy()).resize(new_size)))
    sample_out.pipe_missing(sample_in)
    return sample_out

it could also be an additional decorator or an option in the existing decorator.

@deeplake.compute(pipe_missing=True)
def resize(sample_in, sample_out, new_size):

    
    # Append the label and image to the output sample
    sample_out.images.append(np.array(Image.fromarray(sample_in.images.numpy()).resize(new_size)))
    return sample_out