RelationRx/pyrelational

[Feature] simplify method names

thomasgaudelet opened this issue · 3 comments

thomasgaudelet commented

Is your feature request related to a problem? Please describe.
Names of method tend to be a bit long and redundant (e.g. active_learning_step in an active learning strategy)

Describe the solution you'd like
Proposal is to simplify names across the board, e.g.

  • active_learning_step -> step
  • active_learning_update -> update
  • full_active_learning_run -> run
  • etc.
alicerdv commented

active_learning_step (in Strategy) -> filtered_call

paulmorio commented

Looking at the main modules I focused on the public methods which may warrant some discussion for this. To me it seems that most modules are fine, but there are inconsistencies or improvements to be made in the DataManager and Pipeline classes.

I list the all the public methods below.

DataManager

  • set_target_value
  • get_train_set
  • get_validation_set
  • get_test_set
  • get_train_loader
  • get_validation_loader
  • get_test_loader
  • get_unlabelled_loader
  • get_labelled_loader
  • process_random -> clarify name (maybe: random_split?)
  • update_train_labels
  • percentage_labelled -> get_percentage_labelled
  • get_sample
  • get_sample_feature_vector
  • get_sample_feature_vectors
  • get_sample_labels
  • create_loader -> to private

Pipeline

  • theoretical_performance -> compute_theoretical_performance
  • current_performance -> compute_current_performance
  • compute_hit_ratio
  • active_learning_step -> step()
  • active_learning_update -> query()
  • full_active_learning_run -> run()
  • performance_history -> summary()
  • log_labelled_by

Strategy

  • active_learning_step -> suggest()

In most cases they can stay as is, but put them all there for some discussion if there needs to be a common style (like a get- prefix for all getters, etc.)

thomasgaudelet commented

addressed #85