README.md

metadata

language_creators:
  - found
language:
  - code
license:
  - cc-by-nc-nd-4.0
multilinguality:
  - multilingual
pretty_name: RepoBench-Retrieval
source_datasets:
  - original
task_categories:
  - text-retrieval
task_ids:
  - document-retrieval

Dataset Card for RepoBench-R

Dataset Description

• Homepage: https://github.com/Leolty/repobench
• Paper: https://arxiv.org/abs/2306.03091

Dataset Summary

RepoBench-R (Retrieval) is a subtask of RepoBench(GitHub, arXiv), targeting the retrieval component of a repository-level auto-completion system, focusing on retrieving the most relevant code snippet from a project repository for next-line code prediction.

Settings

• cff: short for cross_file_first, indicating the cross-file module in next line is first used in the current file.

• cfr: short for cross_file_random, indicating the cross-file module in next line is not first used in the current file.

Supported Tasks

The dataset has 4 subsets:

• python_cff: python dataset with cff setting.
• python_cfr: python dataset with cfr setting.
• java_cff: java dataset with cff setting.
• java_cfr: java dataset with cfr setting.

Each subset has 4 splits:

• train_easy: training set with easy difficulty, where the number of code snippets in the context $kk$ satisfies .
• train_hard: training set with hard difficulty, where the number of code snippets in the context $kk$ satisfies $k\ge 10k\; \backslash geq\; 10$.
• test_easy: testing set with easy difficulty.
• test_hard: testing set with hard difficulty.

Loading Data

For example, if you want to load the test cross_file_first python dataset with easy difficulty, you can use the following code:

from datasets import load_dataset

dataset = load_dataset("tianyang/repobench-r", "python_cff", split="test_easy")

Note: The split argument is optional. If not provided, the entire dataset (including, train and test data with easy and hard level) will be loaded.

Dataset Structure

{
  "repo_name": "repository name of the data point",
  "file_path": "path/to/file",
  "context": [
      "snippet 1",
      "snippet 2",
      // ...
      "snippet k"
  ],
  "import_statement": "all import statements in the file",
  "gold_snippet_idex": 2, // the index of the gold snippet in the context list, 0~k-1
  "code": "the code for next-line prediction",
  "next_line": "the next line of the code"
}

Licensing Information

CC BY-NC-ND 4.0

Citation Information

@misc{liu2023repobench,
      title={RepoBench: Benchmarking Repository-Level Code Auto-Completion Systems}, 
      author={Tianyang Liu and Canwen Xu and Julian McAuley},
      year={2023},
      eprint={2306.03091},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Contributions

Thanks to @Leolty for adding this dataset.