Frequently asked questions

Is Larkspur a dataframe library?

No. If you need pivots, joins across large tables, or a query planner, reach for a full analytics package. Larkspur is for straightforward row-at-a-time shaping where a big runtime would be overkill.

How does it stay fast?

It does the minimum: a generator per stage and standard-library readers. There is no scheduler and no copying of whole datasets. For CPU-bound work you can still split the input and run several processes.

Does it support parallelism?

Not internally. Larkspur pipelines are ordinary iterators, so wrap them in multiprocessing or a job runner if you need more than one core.

What about type coercion?

Values arrive as strings from CSV. Convert them explicitly in a derive step — being explicit avoids the guessing that bites people later.

What is the license?

MIT-0. Use it, vendor it, ship it. Attribution is appreciated but not required.