ProcessorABC

class coffea.processor.ProcessorABC

Bases: object

ABC for a generalized processor

The various data delivery mechanisms (spark, striped, uproot, uproot+futures, condor, …) receive such an object and the appropriate metadata to deliver NanoEvents to it. It is expected that the entire processor object can be serializable (check with save) No attempt should be made to track state inside an instance of ProcessorABC, it is to be treated simply as a bundle of methods.

Examples

A skeleton processor:

from coffea import hist, processor

class MyProcessor(processor.ProcessorABC):
    def __init__(self, flag=False):
        self._flag = flag

    def process(self, events):
        out = {"sumw": len(events)}

        # ...

        return {events.metadata.dataset: out}

    def postprocess(self, accumulator):
        pass

p = MyProcessor()

Methods Summary

postprocess(accumulator)	Final processing on aggregated accumulator
process(events)	Processes a single NanoEvents chunk

Methods Documentation

abstract postprocess(accumulator)

Final processing on aggregated accumulator

Do any final processing on the resulting accumulator object, it should be modified in-place

abstract process(events)

Processes a single NanoEvents chunk

Returns a filled accumulator object