.. _interface-analysis_drivers-python:

""""""
python
""""""


Run Python through a Pybind11-based direct interface - requires a special Dakota build


.. toctree::
   :hidden:
   :maxdepth: 1

   interface-analysis_drivers-python-numpy


**Specification**

- *Alias:* None

- *Arguments:* None


**Child Keywords:**

+-------------------------+--------------------+--------------------+---------------------------------------------+
| Required/Optional       | Description of     | Dakota Keyword     | Dakota Keyword Description                  |
|                         | Group              |                    |                                             |
+=========================+====================+====================+=============================================+
| Optional                                     | `numpy`__          | Enable the use of numpy in Dakota's Python  |
|                                              |                    | interface                                   |
+----------------------------------------------+--------------------+---------------------------------------------+

.. __: interface-analysis_drivers-python-numpy.html



**Description**


Dakota supports :ref:`a library-linked interface to Python <advint:existingdirect>`, but it must be
explicitly enabled when compiling Dakota from source.  Contact the
Dakota users mailing list for assistance building and using Dakota
with these interfaces.

The :ref:`interface-analysis_drivers<interface-analysis_drivers>` keyword specifies a
Python module:function which implements the parameter to
response mapping.  List data structures are the default,
but NumPy is also supported, if enabled in the build.

Batch evaluations ( :ref:`interface-batch<interface-batch>`) are supported through a
list of dictionaries.