=========================================
Using RapidWright Directly in Python 3
=========================================

TL;DR
==============================
.. code-block:: bash

	pip install rapidwright


Introduction
==============================
Although RapidWright is written in Java, there is significant interest
to access it from Python.  Python has many features that make it a
great choice for rapid prototyping and scripting solutions.  In fact,
RapidWright ships with `Jython <https://www.jython.org/>`_ (Python
implemented in Java) to provide an authentic Python experience.

Despite RapidWright's Jython integration, for real-world Python
development, the world has transitioned to Python 3 and depend
on packages that have native implementations which are incompatible
with Jython. This has generally excluded RapidWright (with the
exclusion of the experimental `GraalVM's Python
<https://www.graalvm.org/reference-manual/python/>`_) from working
directly with Python 3.

However, there is a Python package called `JPype
<https://jpype.readthedocs.io/en/latest/>`_ that enable Python to call
Java packages directly as if they were native APIs. This tutorial
shows you how RapidWright can take advantage of this package to use
RapidWright directly in your Python projects.


Python Virtual Environments 
==============================

A highly recommeneded way to develop in Python is to use `Virtual
Environments <https://docs.python.org/3/tutorial/venv.html>`_.  Python
Virtual Environments allow you to isolate your Python modules and
installation from the default system installation.  As each project can
have a variety of specific needs and version dependencies, having a
dedicated Virtual Environment per project can make for a smoother
development experience and minimize conflicts.


Pre-requisites
===============
  - Python 3
  - Java 1.8 or later


Setting up a Virtual Python Environment
======================================== 

The Python module used to create a virtual environment is called
``venv``.  For more details about configuring a virtual
environment, please refer to the ```venv`` documentation
<https://www.graalvm.org/reference-manual/python/>`_. The default
settings of a virtual environment can be set up with the following
command:

.. code-block:: bash

	python3 -m venv venv

This will create a directory called ``venv`` which will
contain the essential ingredients for a Python interpreter and its
environment.  To activate the virtual environment, run:

.. code-block:: bash

	source venv/bin/activate

or on Windows, run:

.. code-block:: bat

	venv\Scripts\activate

In either case your terminal prompt should now have a prefix
``(venv)``.  To leave or deactivate the virtual environment,
simply run:

.. code-block:: bash

	deactivate


Running RapidWright in the Virtual Environment
===================================================
Now that the virtual environment is setup, we can begin to experiment
with RapidWright.  As mentioned in the introduction, ``Jpype1`` is
listed as a dependency, so if we simply run:

.. code-block:: bash

	pip install rapidwright

It will be installed automatically.  Then we can run Python:

.. code-block:: bash

	python

To use RapidWright inside the Python interpreter (or a script), all we need to do is simply:

.. code-block:: python

	import rapidwright

On the very first invocation of this import, it will take a few
seconds to get things set up.  After the first time, it will be faster.
	
At this point, you can import java classes to allow you to access any
RapidWright Java API:

.. code-block:: python

	from com.xilinx.rapidwright.device import Device
	device = Device.getDevice(Device.AWS_F1)

At this point you can also get tab-completion on the Java classes, for example:
	
.. code-block:: python

	>>> device.
	device.AWS_F1                         device.getClass(                      device.getSLRByConfigOrderIndex(
	device.DEVICE_FILE_VERSION            device.getClockRegion(                device.getSLRs(
	device.FRAMEWORK_NAME                 device.getClockRegionFromTile(        device.getSeries(
	device.FRAMEWORK_NAME_AND_VERSION     device.getClockRegions(               device.getSite(
	device.KCU105                         device.getColumns(                    device.getSiteFromPackagePin(
	device.PYNQ_Z1                        device.getDevice(                     device.getSitePin(
	device.QUIET_MESSAGE                  device.getDeviceName(                 device.getSiteTypeCount(
	device.RAPIDWRIGHT_MINOR_VERSION      device.getDeviceVersion(              device.getTile(
	device.RAPIDWRIGHT_QUARTER_VERSION    device.getFamilyType(                 device.getTileTypeCount(
	device.RAPIDWRIGHT_VERSION            device.getMasterSLR(                  device.getTiles(
	device.RAPIDWRIGHT_YEAR_VERSION       device.getName(                       device.getWire(
	device.RW_QUIET_MESSAGE               device.getNode(                       device.hashCode(
	device.a(                             device.getNumOfClockRegionRows(       device.notify(
	device.equals(                        device.getNumOfClockRegionsColumns(   device.notifyAll(
	device.getActivePackage(              device.getNumOfSLRs(                  device.quietReflectiveAccessWarning(
	device.getAllCompatibleSites(         device.getPIP(                        device.releaseDeviceReferences(
	device.getAllSitesOfType(             device.getPackage(                    device.setActivePackage(
	device.getAllTiles(                   device.getPackages(                   device.toString(
	device.getArchitecture(               device.getRows(                       device.wait(
	device.getAvailableDevices(           device.getSLR(                        
	>>> device.
	
Which is quite handy.  Object return types are translated for
primitive types (int, String, ...), but Java objects are preserved and
can be accessed via APIs as well:

.. code-block:: python

	>>> device.getName()
	'xcvu9p'
	>>> device.getTiles()
	<java array 'com.xilinx.rapidwright.device.Tile[][]'>


Although there is limited interaction, you can also run RapidWright
GUI applications from Python:

.. code-block:: python

	>>> from com.xilinx.rapidwright.device.browser import DeviceBrowser
	>>> DeviceBrowser.main([])

.. figure:: images/DeviceBrowserFromPython.png
	:width: 550px
	:align: center
	
	Screen capture of RapidWright's Device Browser called from Python


We expect this integration capability with Python to help increase
RapidWright's applicability to a wider number of projects.  There are
more opportunities for integration as well, so stay tuned!


Java Development and Python
===================================
When you install the Python RapidWright package, it downloads the
standalone jar so it can run without any extra setup.  However, if you
already have a git repo checked out and compiled, you can tell the
Python RapidWright package to point to your local install by setting
the following environment variables:

.. code-block:: bash

	RAPIDWRIGHT_PATH=<path_to_RapidWright_directory_checked_out_from_GitHub>
	CLASSPATH=$RAPIDWRIGHT_PATH/bin:$RAPIDWRIGHT_PATH/jars/*
 
This way, the Python RapidWright will use your development copy of
RapidWright.

Things to Know When Using RapidWright in Python
=================================================================

Equality
------------
In Java, there are two main ways to check for equality:

1. Reference equality, ``==`` operator
2. Object equality, ``equals()`` method

Reference equality essentially checks if two objects point to the same
reference or location in memory.  Whereas ``equals()`` invokes the
method on referenced object's class definition.

Jpype has chosen to map the Python ``==`` operator to use the Java
``equals()`` method and the Java ``==`` is not directly accessible.
More on this can be found in `Jpype documentation
<https://jpype.readthedocs.io/en/latest/userguide.html#classes>`_.