Datasets:

paiml
/

python-doctest-corpus-test

license: apache-2.0
task_categories:
  - text-generation
language:
  - en
tags:
  - code
  - python
  - doctest
  - transpilation
  - rust
size_categories:
  - n<1K

Python Doctest Corpus

A curated corpus of Python doctest examples designed for training Python-to-Rust transpilers and testing code translation systems.

Dataset Description

This dataset contains Python function signatures, doctest inputs, and expected outputs that serve as high-quality training data for:

Transpilation training: Teaching models to translate Python patterns to Rust
Test validation: Verifying that transpiled code produces correct outputs
Code understanding: Benchmarking code comprehension systems

Schema

Column	Type	Description
`source`	string	Source project identifier
`version`	string	Git commit SHA or version
`module`	string	Module path (e.g., `example_math.math_tool`)
`function`	string	Function name being tested
`input`	string	Python doctest input (e.g., `>>> square(5)`)
`expected`	string	Expected output value
`signature`	string	Full Python function signature with type hints

Quality Score

Quality score validated using alimentar quality score --profile doctest-corpus:

Grade: B (90.3%)
Critical Checks: All passing
High Priority Checks: All passing

Example

def square(x: int) -> int:
    """
    >>> square(5)
    25
    >>> square(0)
    0
    """
    return x * x

Becomes:

{
  "source": "reprorusted-python-cli",
  "function": "square",
  "input": ">>> square(5)",
  "expected": "25",
  "signature": "def square(x: int) -> int"
}

Usage

from datasets import load_dataset

dataset = load_dataset("paiml/python-doctest-corpus")

for example in dataset["train"]:
    print(f"{example['function']}: {example['input']} -> {example['expected']}")

License

Apache 2.0

Citation

@dataset{python_doctest_corpus,
  author = {PAIML},
  title = {Python Doctest Corpus},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/datasets/paiml/python-doctest-corpus}
}