Skip to main content

Creating a custom node

Division: an example​

Suppose we wanted to contribute a node that divides two items elementwise (for the case of vector inputs, for instance). Although we could do this with the built-in invert and multiply nodes, we want to create this node for convenience.

Creating the source files​

To start, we create the file DIVIDE/divide.py inside /PYTHON/nodes/TRANSFORMERS/ARITHMETIC/. Each node must have its own folder.

We can then create our new function using the features discussed here.

divide.py
import numpy as np
from flojoy import flojoy, DataContainer

@flojoy
def DIVIDE(dc, params):

    a = dc[0].y
    b = dc[1].y

    result = np.divide(a,b) #ensure elementwise
    return DataContainer(
        type='ordered_pair',
        x={'a': a, 'b': b},
        y = result
    )

Creating the manifest​

To register our new node with Flojoy, let's make a new manifest file in PYTHON/nodes/MANIFEST.

divide.manifest.yaml
COMMAND:
  - { name: 'Div', key: 'DIVIDE', type: 'ARITHMETIC' }

Creating Custom Component ( Frontend )​

In Flojoy, you can create custom component for newly created nodes (i.e. shape and node connections). The custom components are located in /src/feature/flow_chart_panel/components/custom-nodes folder. Create a custom component for the newly created nodes and register the design in /src/configs/NodeConfigs.ts file. In this case, its a ARITHMETIC type node, so you register custom component as ARITHMETIC: YOUR_CUSTOM_COMPONENT. If you don't register the newly created node type,it will render the DefaultNode component.

NodeConfigs.ts
import MyCustomComponent from '@src/feature/flow_chart_panel/components/custom-nodes/YOUR_CUSTOM_COMPONENT';

export const nodeConfigs = {
    default: DefaultNode,
    ARITHMETIC: MyCustomComponent,
};

Registering the new function with Flojoy​

info

This is now performed at startup of Flojoy.

To update the databases with the functionalities of the nodes (including your new custom node), run the following in the root directory:

python3 write_python_metadata.py

Almost done! Housekeeping time​

Let's make sure your code is properly formatted!

We use black as our formatter for Python, which you can install by running

pip3 install black

or

pip3 install -r requirements.txt

on the root of the nodes repo.

Once the formatter is installed, simply run

black .

on the root of the nodes repo and all your Python files will be properly formatted!

tip

It is always a good idea to setup format on save on the editor of your choice!

Congratulations! You've created your first custom node.​

When creating custom nodes, make sure to go through the following steps:

  • Did I make my new function correctly?
    • Did I add the flojoy decorator to my function?
    • Did I pass two arguments to my function, the DataContainer inputs and the parameters params from the manifest?
  • Did I create a manifest file, correctly adding the correct category key?
  • Did I generate the manifest for the node?
  • Did I update the Python metadata?

Common Errors:​

  • [2023-05-17 08:29:33.105-RQ-watch] AttributeError: module 'nodes.GENERATORS.SIMULATIONS.TESTING.TESTING' has no attribute 'TESTING'

This likely means your function name does not match the Key in your manifest.yaml file.

  • [2023-05-17 08:59:25.876-RQ-watch] cmd = node["cmd"] KeyError: 'cmd'

This likely means you have to run python3 generate_manifest.py in the root Flojoy directory.