KNIME

In KNIME, local flow variables can be connected through special "flow ports", identified as red circles. Each node, regardless of the number input ports included, has a input port for a local flow variable.

PyWorkflow

The StringNode is currently the only FlowNode we have defined for PyWorkflow and it uses standard ports/edges as the other Nodes do. Depending on how the new "flow ports" are implemented, some back-end logic may need to change. How it currently works (on execution) is:

1. Any preceding data is loaded. This includes a node_type != FlowNode condition so only Nodes with data (not flow variables) are included.
2. FlowNodes that are defined in a Node's option_replace attribute are retrieved and passed to the node.get_execution_options() method. This retrieves the default_value of each FlowNode and creates a new Parameter that the given Node will use.

Connecting a FlowNode to another Node through a flow port does not automatically substitute the value for a given parameter. Instead, it allows the FlowNode's variable to be selected as an option to replace a parameter. Parameters include all global variables, and the optional local flow variable, as options to select from.

One proposal for how to implement this visually is this wireframe

where a flow variable can be selected through a dropdown select and enabled/disabled by a checkbox. Alternatively, an empty or blank choice in the dropdown could signify no flow variable overrides.

For the front-end to retrieve all flow variable options, the endpoint GET /node/<node_id> can be used to retrieve the current user-specified options, and all currently defined (local and global) flow variables. Node information is stored in the retrieved_node attribute, flow variables are stored in the flow_variables of the JSON response.