Structured Code Optimization, Bug-Fix, and AST Parsing Dataset for ML Training

Jamie Davis provides a dataset of structured JSON objects pairing raw source code with optimized equivalents, bug fixes, and invalid syntax examples. The dataset includes pre-computed complexity scores, execution tracking, and input-output verification arrays. It was last updated on 2026-05-28 and is engineered to train automated program repair tools, parsers, and static analyzers.

Use Cases

• Train automated program repair models based on the pairing of buggy and fixed code examples.
• Develop code parsers and static analyzers based on the provided Abstract Syntax Tree (AST) parsing examples.
• Benchmark code optimization algorithms based on the structured input-output verification arrays.
• Train models to detect and correct invalid syntax based on the provided invalid syntax examples.

Strengths

• Dataset is structured with pre-computed complexity scores and verification arrays.
• Data is specifically engineered for training automated program repair tools and parsers.
• The dataset is published under the open CC-BY-4.0 license.

Limitations

• The dataset is very small at 4.5 KB, indicating limited scope.
• Row count and column-level documentation are unknown, limiting suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Jamie Davis via figshare

Freshness

Last updated 2026-05-28 18:57:39.