Skip to content
Switch branches/tags
Go to file
Cannot retrieve contributors at this time


Dask APIs generally follow from upstream APIs:

.. toctree::
   :maxdepth: 1

   Array <array-api.rst>
   DataFrame <dataframe-api.rst>
   Bag <bag-api.rst>
   Machine Learning <>
   Delayed <delayed-api.rst>
   Futures <futures>

Additionally, Dask has its own functions to start computations, persist data in memory, check progress, and so forth that complement the APIs above. These more general Dask functions are described below:

.. currentmodule:: dask

.. autosummary::

These functions work with any scheduler. More advanced operations are available when using the newer scheduler and starting a :obj:`dask.distributed.Client` (which, despite its name, runs nicely on a single machine). This API provides the ability to submit, cancel, and track work asynchronously, and includes many functions for complex inter-task workflows. These are not necessary for normal operation, but can be useful for real-time or advanced operation.

This more advanced API is available in the Dask distributed documentation

.. autofunction:: annotate
.. autofunction:: compute
.. autofunction:: is_dask_collection
.. autofunction:: optimize
.. autofunction:: persist
.. autofunction:: visualize


Dask has a few helpers for generating demo datasets

.. currentmodule:: dask.datasets

.. autofunction:: make_people
.. autofunction:: timeseries


Dask has some public utility methods. These are primarily used for parsing configuration values.

.. currentmodule:: dask.utils

.. autofunction:: format_bytes
.. autofunction:: format_time
.. autofunction:: parse_bytes
.. autofunction:: parse_timedelta