DaskLocalCluster
================

Purpose
-------

The ``DaskLocalCluster`` is a wrapper function around a :py:class:`distributed.LocalCluster` configured
with batchlet's dask parameters. Its purpose is to keep the :ref:`worker-configuration resolution logic <dask-worker-configuration-from-resources>` consistent
with :py:class:`DaskSlurmCluster`. This keeps behavior consistent between local and SLURM execution modes.

Working
-------

The function maps batchlet parameters to :py:class:`distributed.LocalCluster` arguments:

* ``workers_per_node`` -> ``n_workers``
* ``threads_per_worker`` -> ``threads_per_worker``
* ``memory_per_worker`` -> ``memory_limit``
* ``worker_scratch_directory`` -> ``local_directory``
* ``resources_per_worker`` -> worker ``resources``

The parameters ``nodes`` and ``use_entry_node`` are ignored for local mode.

Usage
-----

.. code-block:: python

    from ska_sdp_batchlet.utils.dask_cluster.local_cluster import DaskLocalCluster

    with DaskLocalCluster(
        workers_per_node=4,
        threads_per_worker=2,
        memory_per_worker="2GB",
        resources_per_worker={"CPU": 2},
        worker_scratch_directory="/tmp/dask-workers",
        name="BatchletLocal",
    ) as cluster:
        client = cluster.get_client()
        # submit dask tasks with client

API
---

.. autofunction:: ska_sdp_batchlet.utils.dask_cluster.local_cluster.DaskLocalCluster
   :no-index: