.. This file is part of Sympathy for Data.
.. Copyright (c) 2010-2012 Combine Control Systems AB
..
.. Sympathy for Data is free software: you can redistribute it and/or modify
.. it under the terms of the GNU General Public License as published by
.. the Free Software Foundation, version 3 of the License.
..
.. Sympathy for Data is distributed in the hope that it will be useful,
.. but WITHOUT ANY WARRANTY; without even the implied warranty of
.. MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
.. GNU General Public License for more details.
..
.. You should have received a copy of the GNU General Public License
.. along with Sympathy for Data.  If not, see <http://www.gnu.org/licenses/>.

.. _nodewriting:

Introduction
------------
Sympathy's standard library contains a lot of useful nodes and it is also
possible to add complete third-party libraries without writing any code
yourself. But sometimes you might come to a point when the node that you need
simply has not been written yet. One option is to write your own node.

All Sympathy nodes are written in Python, http://python.org, a powerful
scripting language whose main goal is to be easy to learn. Python has a very
powerful set of standard libraries, but the standard libraries are a bit
lacking when it comes to high performance numerical computations. Because of
this Sympathy comes with some third party libraries that are great for
numerical computations and data analysis:

`numpy`
  The basis for most of the other libraries mentioned here and therefore also the
  most useful library. It provides you with a generic data type for
  numerical data and some basic calculations on those data types. See
  http://wiki.scipy.org/Tentative_NumPy_Tutorial, or
  http://docs.scipy.org/doc/numpy/user/basics.html, or
  http://wiki.scipy.org/NumPy_for_Matlab_Users for some introduction.

`scipy`
  Provides functions for more advanced analysis such as numerical integration,
  solving differential equations, optimization, and working with sparse
  matrices. See http://docs.scipy.org/doc/scipy/reference/.

`pandas`
  See http://pandas.pydata.org/.


To create and edit nodes you will need some text editor or Python IDE. If you do
not already have a favorite editor/IDE, we recommend `Visual Studio Code
<https://code.visualstudio.com/>`__ or `PyCharm
<https://www.jetbrains.com/pycharm/>`__ (PyCharm Community Editon is free to
use).


.. _`node_wizard`:

Creating new nodes
------------------
The easiest way to get started writing your own node is to use the node wizard.
It will create an outline of a node code for you, so you can get right at
implementing the actual functionality of the node. To start the node wizard go
to *File->Wizards->New Node*.

On the first page of the wizard you can edit some descriptive meta data for
your new node. Please consider carefully what you write here as this affects
how well a user will be able to understand what your node does. See the section
:ref:`node_meta` for detailed info about the different fields.

The second page allows you to add input ports to your node. Enter a name, a
description and choose a data type, and press *Add*. On the next page you can
add output ports in the same way.

The next page of the wizard is where you choose in which library your node
should be created as well as where in the internal structure of that library
the node should be placed. The *Library Identifier* field should contain an
identifier for the whole library. It should be the same for every node in the
entire library. It should be of the form `<extension>.<organisation>`.
Something along the lines of `com.example`. The *Library Path* field should
be the absolute path to the library where you want the node, and the *Node Path*
should be the relative path inside that library to the folder where the node
should be placed. If your library is present in the file tree structure below
the text fields, you can also simply click on the folder where you want the
node and all fields on this page should be automatically filled. Note, libraries
that you create will not be placed in the library view automatically.

Click *Next* to proceed to the last page of the node wizard where you will be
presented with a preview of the node code. When finishing the wizard this code
will be written to a file that represents your node.

.. warning::
    You can in theory add new nodes to Sympathy's standard library (by moving
    the python files manually) or to some third-party library and have them
    appear in the Library view in Sympathy. This is not recommended though as
    it makes it much more difficult to manage library updates and such. In order
    to place nodes under a certain folder in the library view, read the section
    "Library tags" below.


.. _`node_code`:

The node code
-------------
Nodes are loaded from their definition files when Sympathy is started, and only
Python files with names starting with ``node_`` and ending with ``.py`` will
generate nodes. You can place the nodes in subfolders to group related nodes
together. Now, create a file called ``node_helloworld.py`` and open it in your
editor of choice.

Without further ado let us look at the code for a simple example node::

    from sympathy.api import node as synode
    from sympathy.api.nodeconfig import Port, Ports, Tag, Tags


    class HelloWorld(synode.Node):
        name = 'Hello world!'
        description = 'An amazing node!'
        nodeid = 'com.example.boblib.helloworld'
        tags = Tags(Tag.Development.Example)

        def execute(self, node_context):
            print('Hello world!')

Copy this code into the file ``node_helloworld.py``, reload the libraries in
Sympathy (*Control->Reload Library*) and add the node to a new workflow.

A node is defined as a Python class which inherits from
``sympathy.api.node.Node``. The name of the class is irrelevant. The
class definition starts with a description of the node, then you have to
define some variables that contain meta data about the node. Lastly, you
write the method that actually controls the behavior of the node (such as
``execute``). For all the details of what goes in a node class, please refer to
the :ref:`node_reference`.

You can place several such classes in the same python file, but only do this if
they are clearly related to one another.


Library tags
------------
In the example above you may have spotted the tags variable. Each node can be
part of a library tag. This specific one:

    tags = Tags(Tag.Development.Example)

will group the node into the *Development->Test* folder in the library hierarchy.

To see what different tags are available have a look in
*sylib/librarytag_sylib.py* or look at the code of any specific
node which uses the tag that you are interested in. If you don't specify a tag
the node will be shown under a folder called "Unknown" in the library view.


.. _node_ports:

Adding input and output ports
-----------------------------
The possibilities for a node with neither input nor output ports are quite
limited. To add a single Table output port to your node, add the class variable
``outputs`` as follows::

    import numpy as np

    from sympathy.api import node as synode
    from sympathy.api.nodeconfig import Ports, Port, Tags, Tag

    class FooTableNode(synode.Node):
        """Creates a foo Table"""

        name = 'Create foo Table'
        nodeid = 'com.example.boblib.footable'
        tags = Tags(Tag.Development.Example)

        outputs = Ports([Port.Table('Table of foo', name='foo')])

        def execute(self, node_context):
            outputfile = node_context.output['foo']
            outputfile.set_column_from_array('foo column', np.array([1, 2, 3]))

Also notice the new `import` statements at the head of the file. Reload the
library and add a new instance of your node to a workflow. You can see that it
now has an output port of the Table type.

Writing to the output file is as easy as adding those two lines to your
``execute`` method.

The object ``outputfile`` which is used in the example is of the class
:class:`table.File` Please refer to the :ref:`tableapi` to get more information
about how to interact with the Table data type. Once again, reload the
libraries, add the node to a flow, and execute it. With these changes the node
will produce an output table with a single column called *foo column*
containing the values 1, 2, and 3.

Inspect the output by double clicking on the output port of your node. It will
open in Sympathy's internal data viewer.

If you want your output to be a modified version of the input you can use the
``source`` method::

    import numpy as np

    from sympathy.api import node as synode
    from sympathy.api.nodeconfig import Ports, Port, Tags, Tag

    class AddBarNode(synode.Node):
        """Adds a bar column to a Table."""

        name = 'Add bar column'
        nodeid = 'com.example.boblib.addbar'
        tags = Tags(Tag.Development.Example)

        inputs = Ports([Port.Table('Input Table', name='foo')])
        outputs = Ports([Port.Table('Table with some added bar', name='foobar')])

        def execute(self, node_context):
            inputfile = node_context.input['foo']
            outputfile = node_context.output['foobar']
            outputfile.source(inputfile)
            number_of_rows = inputfile.number_of_rows()
            outputfile.set_column_from_array('bar',
                np.arange(number_of_rows, dtype=int))

All the other basic port data types are also available in the ``Port`` class,
such as ``ADAF``, ``Datasource``, and ``Text``. Try changing your port to some
other type and add it again to a flow (do not forget to reload libraries first)
to see the port data type change. You can also just as easily add several input
or output ports to a node::

    inputs = Ports([Port.Datasource('Input foo file', name='foofile'),
                    Port.ADAFs('All the data', name='alldata')])
    outputs = Ports([Port.Table('Table with baz', name='baz'),
                     Port.ADAF('The best data', name='outdata')])

Note though that the different data types have different APIs whose references
can be found here: :ref:`datatypeapis`.

If you need ports of some type which does not have its own method in
:class:`Port` (such as generic types or lambdas) see :ref:`custom_ports`.


.. _node_parameters:

Adding a configuration GUI
--------------------------
Many of the nodes in the standard library have some configuration settings that
affect the details of what the node does when executed. For example in
:ref:`Random Table` you can choose how big the generated table should be.

Going back to the original Hello world node, let us now offer the user a choice
of what greeting to print.

.. TODO: Link

Parameters are defined in the class variable ``parameters``. Create a new
parameters object by calling the function ``synode.parameters``. Then add all
the parameters with methods such as ``set_string``. In our example it would
look something like this::

    from sympathy.api import node as synode
    from sympathy.api.nodeconfig import Tags, Tag


    class HelloWorldNode(synode.Node):
        """Prints a custom greeting to the node output."""

        name = 'Hello world!'
        description = 'An amazing node!'
        nodeid = 'com.example.boblib.helloworld'
        tags = Tags(Tag.Development.Example)

        parameters = synode.parameters()
        parameters.set_string(
            'greeting',
            value='Hello world!',
            label='Greeting:',
            description='Choose what kind of greeting the node will print.')

        def execute(self, node_context):
            greeting = node_context.parameters['greeting'].value
            print(greeting)

Once again try reloading the library and readding the node to a flow. You will
notice that you can now configure the node. A configuration GUI has been
automatically created from your parameter definition. As you can see the
``label`` argument is shown next to the line edit field and the ``description``
is shown as a tooltip. Try changing the greeting in the configuration

.. figure:: screenshot_hello_parameter.png
   :alt: Parameter gui example
   :align: center

You can add parameters of other types than strings as well by using the methods
``set_boolean``, ``set_integer``, ``set_float``, ``set_list``. Most of them
have the same arguments as ``set_string``, but lists are a bit different. A
simple example of storing a list might look like this::

    parameters.set_list(
        'toppings', label='Pizza toppings',
        description='Choose what toppings you want on your pizza.',
        list=['Cheese', 'Tomato sauce', 'Pineapple',
              'Ham', 'Anchovies', 'Mushrooms'],
        value_names=['Cheese', 'Tomato sauce'],
        editor=synode.editors.multilist_editor())

This list is named "toppings" and has the available options specified by the
``list`` argument. The ``value_names`` argument specifies which options in the
list that are selected by default. The ``editor`` argument is used to specify
that we want this list to be shown in a list view with multiple selection.

See :ref:`parameter_helper_reference` for more details or see
:ref:`All Parameters Example` for more examples of how to use all the different
parameter types and editors.


.. _node_errors:

Errors and warnings
-------------------
Any uncaught exceptions that occur in your code will be shown as *Exceptions*
in the error view. The stack traces in the details can be very valuable while
developing nodes, but are pretty incomprehensible for most users. Because of
this you should always try to eliminate the possibility of such uncaught
exceptions. If an error occurs which the node cannot recover from you should
instead try to raise an instance of one of the classes defined in
``sympathy.api.exceptions``. Here is an example that uses
``SyConfigurationError``::

    from sympathy.api.exceptions import SyConfigurationError
    from sympathy.api import node as synode
    from sympathy.api.nodeconfig import Tags, Tag

    class HelloWorldNode(synode.Node):
        """Prints a custom greeting to the node output."""

        name = 'Hello world!'
        description = 'An amazing node!'
        nodeid = 'com.example.boblib.helloworld'
        tags = Tags(Tag.Development.Example)

        parameters = synode.parameters()
        parameters.set_string(
            'greeting',
            value='Hello World!',
            label='Greeting:',
            description='Choose what kind of greeting the node will print.')

        def execute(self, node_context):
            greeting = node_context.parameters['greeting'].value
            if len(greeting) >= 200:
                raise SyConfigurationError('Too long a greeting!')
            print(greeting)

This will produce a more user friendly error message.

If you simply want to warn the user of something that *might* be a concern but
which does not stop the node from performing its task, use the function
``sympathy.api.exceptions.sywarn``::

    from sympathy.api.exceptions import sywarn
    from sympathy.api import node as synode
    from sympathy.api.nodeconfig import Tags, Tag

    class HelloWorldNode(synode.Node):
        """Prints a custom greeting to the node output."""

        name = 'Hello world!'
        description = 'An amazing node!'
        nodeid = 'com.example.boblib.helloworld'
        tags = Tags(Tag.Development.Example)

        parameters = synode.parameters()
        parameters.set_string(
            'greeting',
            value='Hello world!',
            label='Greeting:',
            description='Choose what kind of greeting the node will print.')

        def execute(self, node_context):
            greeting = node_context.parameters['greeting'].value
            if len(greeting) >= 100:
                sywarn("That's a very long greeting. Perhaps too wordy?")
            print(greeting)

See :ref:`error window` for more info about how the error view shows different
types of output. See the :ref:`Error Example` node for another example.

Documenting nodes
-----------------
Documentation for your library can be built into a convenient html format by
sympathy. The automatic node documentation uses docstrings from node classes and
description fields (top-level node description, node port descriptions and
descriptions of node parameters). Since Sympathy uses `Sphinx
<http://www.sphinx-doc.org/en/master/index.html>`_ for producing html
documentation for nodes and the platform, it requires docstrings to be written
in Sphinx compliant reStructuredText format. See also the :ref:`node_reference`
for more information.

Additionally, documentation for other functions and classes can benefit from
using numpy docstring format, see `A Guide to NumPy/SciPy Documentation
<https://docs.scipy.org/doc/numpy/docs/howto_document.html>`_.

The documentation can then be built from the command-line:

.. code-block:: bash

   sympathy doc --library-dir <PATH_TO_NODE_LIB>

For more information about the different command-line options, see
:ref:`start_options`.

If you are distributing your library to other users, consider building its
documentation yourself and including that with the distributed library. This
will save your end-users the hassle of building the documentation themselves.


Reusable nodes
--------------
Follow these simple guidelines to make sure that your node is as reusable as
possible.

- Break down the task into the smallest parts that are useful by themselves and
  write nodes for each of those, instead of writing one monolithic "fix
  everything" node. Take some inspiration from the Unix philosophy; every node
  should "do only one thing, and do it well".
- Try to work on the most natural data type for the problem that you are trying
  to solve. When in doubt go with Table since it is the simplest and most
  widely applicable data type.
- Do not hard code site specific stuff into your nodes. Instead add
  preprocessing steps or configuration options as needed.
- Add documentation for your node, describing what the node does, what the
  configuration options are, and whether there any constraints on the input
  data.
- When you write the code for your node, remember that how you write it can
  make a huge difference. If others can read and easily understand what your
  code does it can continue to be developed by others. As a starting point you
  should try to follow the Python style guide (PEP8_) as much as possible.

.. _PEP8: https://www.python.org/dev/peps/pep-0008/

If your nodes are very useful and do not include any secrets you may be able to
donate it to Combine_ for inclusion in the standard library. This is only
possible if the node is considered reusable.

.. _Combine: https://www.sympathyfordata.com