Dask — параллельные вычисления над большими данными

★ 7.0 · data

dask is a Claude Code skill that scales pandas and NumPy workflows to datasets larger than available RAM using parallel and distributed computing. It requires Python 3.10+ and dask 2025.1+; install via `dask[complete]` for the distributed scheduler and dashboard, with s3fs or gcsfs for cloud paths (s3://, gs://). The skill covers three core abstractions: DataFrame for parallel tabular operations across partitions, Array for chunked NumPy-style computation on scientific datasets (HDF5, Zarr, NetCDF), and Bag for streaming processing of unstructured data like logs and JSON records. Allowed tools — Read, Write, Edit, and Bash — let Claude Code reference bundled documentation, generate code, and run it directly. It is the right choice when migrating a pandas prototype to a production cluster pipeline or processing dozens of CSV and Parquet files on a single machine without loading the full dataset into memory.