pipeline.pipeline_function

pipeline_function - class decorator

Source on GitHub

pipeline.objects.decorators.pipeline_function(
    function: Any = None,
  *,
  run_once: str = None,
  on_startup: bool = False,
)

Description

The pipeline_function decorator is used on runnable functions to allow them to be used by the Pipeline context manager.

It also prevents the wrapped functions execution with it is called inside of the context manager. Instead of calling the wrapped function a GraphNode object is created and entered into the Graph. The functions can still be used as normal outside of the context manager. The two instances where you can run the function when wrapped are described below:

  • Inside of the Pipeline context manager
    • Only Variable object types can be passed in.
    • kwargs are currently not treated separately to args, variables are passed in by position (kwarg support coming).
    • * for explicit kwarg inputs is not supported.
  • Normal execution outside of the Pipeline context manager
    • Any args or kwargs can be passed in as normal.

It is essential to use proper typing descriptions on an function. Typing is used by the Graph engine to determine if an input or output of the function is of a compatible type when running.

Parameters

  • function (Any, optional) - The function to be wrapped, this is implicitly passed in you do not manually pass this.
  • run_once (str, optional) - If true, the pipeline_function will only be called once if the pipeline is sequentially run.
  • on_startup (bool, optional) - If true then the function is called at the start of the pipeline regardless of where it is declared in the Pipeline context manager.
    • Note: the only Variable object that can be passed in if on_startup=True is the PipelineFile object currently. More will be supported in the future.

Examples

from pipeline import Pipeline, Variable, pipeline_function


@pipeline_function
def add_numbers(a: int, b: int) -> int:
    return a + b
    
c = add_numbers(3, 4)
# c = 7

with Pipeline("add-numbers") as pipeline:
    a = Variable(type_class=int, is_input=True)
    b = Variable(type_class=int, is_input=True)
    pipeline.add_variables(a, b)

    c = add_numbers(a, b)
        # c is equal to a Variable object returned from add_numbers

    pipeline.output(c)