Skip to content

🤗 Datasets

MigrationBench provides three datasets on Hugging Face.

All repositories are available on GitHub under MIT or Apache-2.0 licenses.

Available Datasets

Dataset Size Description
MigrationBench-Full 5,102 Full dataset. Each repo has unit test.
MigrationBench-Selected 300 Curated challenging subset from Full
MigrationBench-UTG 4,814 Unit test generation dataset. Each repo doesn't have unit test. Disjoint with java-full

Metadata

  1. repo (str): The original repo URL without the https://github.com/ prefix
  2. base_commit (str): Base commit id
    • At this commit with java 8 and maven 3.9.6, the repository is able to (1) compile and (2) pass existing unit tests and integration tests if any
    • It is the starting point for code migration from java 8 to LTS versions
  3. num_java_files (int): Number of *.java files in the repository at base_commit, similarly for all other num_* columns
  4. num_loc (int): Lines of code for the repository
  5. num_pom_xml (int): Number of modules (pom.xml files) in the repository
  6. num_src_test_java_files (int): Number of *.java files in the dedicated src/test/ directory
  7. num_test_cases (int): Number of test cases, based on running the mvn -f test . command in the root directory
    • Non negative values indicate number of test cases is parsed correctly from the output
    • Negative values means it's unable to parse the output: [INFO] Results: (-2) or [INFO] Tests run: (-1) regex is missing
  8. license (str): The license of the repository, either MIT or Apache2.0 for the whole dataset

Loading Datasets

Install Hugging Face Datasets library:

pip install datasets

Load and use the datasets:

from datasets import load_dataset

# Load java-selected dataset
dataset = load_dataset("AmazonScience/migration-bench-java-selected")

# Iterate through repositories
for item in dataset['test']:
    print(f"Repository: {item['repo']}")

Stay tuned for updates!