Gittables - grouped by same repo owner and table format

GitTables constitutes a large-scale corpus comprising one million relational tables (in parquet format) extracted from CSV files hosted on public GitHub repositories. According to Hulsebos et al. (2023), these tables resemble actual tables in real-world databases. The parquet files are grouped together if have the same columns and have the same repo owner. Such grouping achieves two goals: (i) it results in larger individual tables (i.e., more rows), and (ii) it facilitates schema inference. This dataset represents the grouped tables from the GitTables dataset. Practically, for every grouping a new subfolder is created in the dataset.