Changelog
All notable changes to the ManyTypes4Py dataset will be documented in this file. Each version of the dataset can be downloaded on zenodo.
[0.7] 2021-08-24
- The dataset now contains clean and complete versions. The clean version has type-checked Python projects by mypy. The complete version contains all Python projects in the dataset same as v0.6.
- A list of type-checked Python source files are included in
type_checked_files.txt(using mypy).
[0.6] 2021-04-26
- Processed projects have now the type of variables inferred by Pyre.
- Visible type hints are added for the whole dataset. They are extracted recursively from the dataset's projects and their dependencies.
[0.5] 2021-03-12
- Added name-based visible type hints for processed projects in the
extracted_visible_typesfolder.
[0.4] - 2021-03-01
- Added type annotations coverage for processed projects and their source code files. See the
type_annot_covefield in JSON files.
[0.3] - 2021-01-29
- Fixed the path issue in
dataset_split.csvfile for some file paths that contains non-ASCII characters. - With the new fixes to the LibSA4py tool, more source code files were processed for included Python projects.
[0.2] - 2021-01-14
- Using the LibSA4py tool, all the Python projects were processed in JSON-formatted files.
- Using the CD4Py tool, the list of duplicate files in the dataset is provided in
duplicate_files.txtfile. - The dataset is split into train, validation and test sets by source code files. The list of files and their corresponding set is provided in
dataset_split.csvfile.
[0.1] - 2020-09-22
- Added a manifest file (
ManyTypes4PyDataset.spec) that contains the URL and hash commits of 5.4K Python projects that were cloned on Sep. 17th 2020 from GitHub.