Code nodes¶

User Function (user-function)¶

The User Function node executes a Python function entered in its multiline code editor. The editor uses locally bundled CodeMirror 5 with Python syntax highlighting, line numbers, indentation, line wrapping, and bracket matching. The source must define one regular Execute function. Its positional parameter names become optional input ports on that node. Its fixed-length tuple[...] return annotation determines the number and types of generated output ports. Use no annotation (or Any) for an Any port, or annotate parameters with str, bool, int, float, list[str], list[int], or list[float] to create a typed input or output. For example:

def Execute(name: str, scale: float) -> tuple[str, float]:
    return (name.upper(), scale * 2)

This creates name: Text and scale: Float inputs plus output1: Text and output2: Float. Use tuple[()] for no outputs. The runtime requires the returned tuple length and values to match the declared output ports.

The Web Editor checks this declaration without executing it and updates the node ports as the source changes.

This is trusted local Python execution, not a sandbox. Changing the code marks the node and all downstream nodes invalid until they are run again.

ImageJ Groovy Script (imagej-script)¶

Enter a complete Groovy script in the code editor. The node has only the code parameter; SciJava #@ declarations generate named input and output ports. For example:

#@ Dataset image
#@output Dataset result

result = image

ImageJ Groovy Script File (imagej-script-file)¶

Choose a .groovy file with the filename parameter, which is also exported as an optional FilePath input port. This node has no code-editor parameter. Its named inputs and outputs are inferred from the selected file's SciJava #@ declarations. An external change to the file requires refreshing the node to update its visible ports. A filename supplied through the port overrides the configured filename at execution time; when it refers to a different script, refresh the node's ports before connecting its script inputs and outputs.

Both ImageJ script nodes use the same Groovy execution and conversion rules:

Image ports (Dataset, Img, RandomAccessibleInterval) carry NumPy arrays. Common scalar declarations are typed; other types use Any. SciJava service parameters are injected by Fiji and do not appear as graph inputs. Install the optional extra with pip install 'flowgraph[imagej]' and provide a Java runtime before execution; Fiji is initialized in headless mode. The first execution may download Fiji dependencies. These nodes run trusted user code, not a sandbox.