Random sample

random selects rows uniformly from the entire file. The sample approximately keeps the proportions of the file.

tabalyst sample customers.csv --sample-method random --percent 5 --seed 42

Key points

• Seed: add --seed when another run must select the same rows.
• Size: --rows or --percent, exactly one.
• Cost: one reading of the file, and only the requested rows are stored in memory.
• Output: customers.sample.csv beside the source, in file order.

Limits

• A rare value can be missed: a value that appears once in a million rows is unlikely to be drawn. Use a smart sample to keep every value of the small columns and the odd cases.
• To keep the proportions of one field, or of a few, use a stratified sample.
• --field, --max-strata, --yes and --no-html are errors with this method (exit code 2).

See Tabalyst Sample for sizes, output locations and CSV formats.