jugalgajjar/MultiLang-Code-Parser-Dataset

MultiLang Code Parser Dataset (MLCPD)

Dataset Description

The MultiLang Code Parser Dataset (MLCPD) is a comprehensive multi-language code dataset designed to benchmark language-agnostic AI code parsers. It currently offers a filtered version of the StarCoder dataset, parsed with language-specific parsers, with future plans to unify outputs into a standard JSON format for complete AST representation.

Key Features

• Cleaned and Filtered Code: Samples have been processed to remove outliers in terms of line length and code size
• Quality Metrics: Each sample includes metadata about average line length and line count of code along with AST node count and error count
• Multi-language Support: 10 programming languages represented in separate subsets
• Consistent Format: All samples follow the same Parquet structure for easy processing

Dataset Size

The complete dataset is approximately 35GB in size. Individual language files vary in size, with the largest being C++ (5.85GB) and the smallest being Ruby (1.71GB).

Dataset Statistics

Language	Sample Count	Avg. Line Length	Avg. Line Count
C	700,821	28.08	61.76
C++	707,641	28.16	87.88
C#	705,203	29.53	44.26
Go	700,331	25.18	68.22
Java	711,922	30.85	54.40
JavaScript	687,775	27.69	44.15
Python	706,126	32.67	54.70
Ruby	703,473	27.35	27.41
Scala	702,833	35.30	44.38
TypeScript	695,597	29.18	36.89

Dataset Structure

The dataset is organized with separate Parquet files for each programming language:

• c_parsed_1.parquet ... c_parsed_4.parquet - C language samples
• cpp_parsed_1.parquet ... cpp_parsed_4.parquet - C++ language samples
• c_sharp_parsed_1.parquet ... c_sharp_parsed_4.parquet - C# language samples
• go_parsed_1.parquet ... go_parsed_4.parquet - Go language samples
• java_parsed_1.parquet ... java_parsed_4.parquet - Java language samples
• javascript_parsed_1.parquet ... javascript_parsed_4.parquet - JavaScript language samples
• python_parsed_1.parquet ... python_parsed_4.parquet - Python language samples
• ruby_parsed_1.parquet ... ruby_parsed_4.parquet - Ruby language samples
• scala_parsed_1.parquet ... scala_parsed_4.parquet - Scala language samples
• typescript_parsed_1.parquet ... typescript_parsed_4.parquet - TypeScript language samples

Within each file, data is stored with the following schema:

- language: string (the programming language of the code sample)
- code: string (the complete code content)
- avg_line_length: float (average character count per line)
- line_count: integer (total number of lines in the code)
- lang_specific_parse: string (tree-sitter parsed output of the code sample)
- ast_node_count: integer (total number of nodes in the AST)
- num_errors: integer (total number of errors in the code)

Each sample is stored as a row in the Parquet file with these four columns.

How to Access the Dataset

Using the Hugging Face datasets Library

This dataset is hosted on the Hugging Face Hub and can be easily accessed using the datasets library.

Install the Required Library

pip install datasets

Import Library

from datasets import load_dataset

Load the Entire Dataset

dataset = load_dataset(
    "jugalgajjar/MultiLang-Code-Parser-Dataset"
)

Load a Specific Language

dataset = load_dataset(
    "jugalgajjar/MultiLang-Code-Parser-Dataset",
    data_files="python_parsed_1.parquet"
)

Stream Data

dataset = load_dataset(
    "jugalgajjar/MultiLang-Code-Parser-Dataset",
    data_files="python_parsed_1.parquet",
    streaming=True
)

Access Data Content (After Downloading)

try:
    for example in dataset["train"].take(5):
        print(example)
        print("-"*25)
except Exception as e:
    print(f"An error occurred: {e}")

Manual Download

You can also manually download specific language files from the Hugging Face repository page:

1. Visit https://huggingface.co/datasets/jugalgajjar/MultiLang-Code-Parser-Dataset
2. Navigate to the "Files" tab
3. Click on the language file you want to download (e.g., python_parsed_1.parquet)
4. Use the download button to save the file locally

Dataset Creation

This dataset was created through the following process:

1. Original code samples were collected from the StarCoder dataset (URL)
2. Statistical analysis was performed to identify quality metrics
3. Outliers were removed using IQR (Interquartile Range) method
4. Samples were filtered to remove excessively long or short code examples
5. Data was normalized and standardized across languages
6. Metadata (average line length and line count) was calculated for each sample
7. Data was serialized in the efficient Parquet format for optimal storage and access speed
8. Code samples from each language were parsed using language-specific tree-sitter parsers
9. Metadata (AST node count and number of errors in the code) were recorded for each sample
10. Final data was split into four files and stored in the Parquet format

Citation

If you use this dataset in your research or project, please cite it as follows:

@misc{mlcpd2025,
  author = {Jugal Gajjar, Kamalasankari Subramaniakuppusamy, Kaustik Ranaware},
  title = {Filtered CodeStar Dataset Mini},
  year = {2025},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/datasets/jugalgajjar/MultiLang-Code-Parser-Dataset}}
}

License

This dataset is released under the MIT License. See the LICENSE file for more details.