Benchmarking node runtime

Node hooks can be used to keep track of node execution durations. Using hooks to achieve this has the advantage that the timing does not add cognitive complexity to the dataset transformations. It also avoids repetitive boilerplate code inside the nodes.

timer_hook.py:

>>> from time import perf_counter
>>>
>>> from ordeq import Node, NodeHook
>>>
>>> class TimerHook(NodeHook):
...     """Hook to time nodes."""
...
...     def __init__(self):
...         self.node_timing = {}
...
...     def before_node_run(self, node: Node) -> None:
...         self.node_timing[hash(node)] = perf_counter()
...
...     def after_node_run(self, node: Node) -> None:
...         start = self.node_timing[hash(node)]
...         end = perf_counter()
...         function_name = node.func.__name__
...         print(f"Executing `{function_name}` took {end - start:.2f}s")
>>>
>>> timer = TimerHook()

Usage:

>>> from ordeq import node, run
>>>
>>> @node(inputs=[...], outputs=[...])
... def my_transformation_function(data):
...     ...
>>> run(my_transformation_function, hooks=[timer])  # doctest: +SKIP