jenyag/repo-code-completion

Repository Level Code Completion Dataset for Evaluation

This is a dataset of repository snapshots before a commit where a python file has been added. One needs to complete added file with given content of repository composed in different ways.

How to load the data

1. via load_dataset:

   from datasets import load_dataset
   
   data_files =  # choose from the table below
   dataset = load_dataset("jenyag/repo-code-completion", data_files=data_files, split="train")
   

Options for data_files:

	all_context	non_py_context	py_context
function class mask half composer	data/function_class_mask_half_composer/all_context/test-*	data/function_class_mask_half_composer/non_py_context/test-*	data/function_class_mask_half_composer/py_context/test-*
imports first composer	data/imports_first_composer/all_context/test-*	data/imports_first_composer/non_py_context/test-*	data/imports_first_composer/py_context/test-*
alphabetical composer	data/alphabetical_composer/all_context/test-*	data/alphabetical_composer/non_py_context/test-*	data/alphabetical_composer/py_context/test-*
naive composer	data/naive_composer/all_context/test-*	data/naive_composer/non_py_context/test-*	data/naive_composer/py_context/test-*
path distance composer	data/path_distance_composer/all_context/test-*	data/path_distance_composer/non_py_context/test-*	data/path_distance_composer/py_context/test-*
file length composer	data/file_length_composer/all_context/test-*	data/file_length_composer/non_py_context/test-*	data/file_length_composer/py_context/test-*
half memory composer	data/half_memory_composer/all_context/test-*	data/half_memory_composer/non_py_context/test-*	data/half_memory_composer/py_context/test-*
function class mask one composer	data/function_class_mask_one_composer/all_context/test-*	data/function_class_mask_one_composer/non_py_context/test-*	data/function_class_mask_one_composer/py_context/test-*

How to get the full context for the specific line

for datapoint in dataset:
  project_context = datapoint['project_context']  # The project context may be quite long
  for file_context_dict, ground_truth in zip(datapoint['file_context'], datapoint['gt']):
    file_context = file_context_dict['content']
    full_context = project_context + file_context