Dev uses Claude Code beachside to prove geo/climate data is secretly SQL-shaped
A new feature called to_dataset() was added to Xarray-SQL, an array database library, allowing N-dimensional array data to be treated as tabular data. Using this feature, a benchmark was published showing that common operations in geo and climate science at the 100 TB scale are actually relational (SQL-style) operations in disguise. The work, done using Claude Code while on a beach vacation, culminated in a demo that implements a directly inside SQL, further proving the data model holds up in practice.
Key points
- Xarray-SQL added a to_dataset() feature enabling conversion between array and tabular data representations
- A benchmark demonstrates that 100 TB-scale geo and climate science operations are effectively relational (SQL) operations
- The benchmark and a SQL-based demo were built using Claude Code during a beach vacation
- The project is and published on GitHub