Move away from %reload_kedro IPython magic?

[astrojuanlu]

%reload_kedro has done its job at giving a 1-liner to get started with Kedro on notebooks. However,

• It's only compatible in IPython-based environments, hence not really available in Marimo, Hex, or other notebook environments (see discussion in kedro marimo + better programmatic setup for non jupyter/ipython envs #4440)
• It doesn't really work that well in places like Databricks, see for example https://community.databricks.com/t5/data-engineering/sparkcontext-lost-when-running-sh-script-py/td-p/67180
• It obscures performance issues Investigate performance of %reload_kedro on Databricks #4042
• It has weird bugs like Weird behavior of %reload_kedro when --params contain spaces #4813
• It has historically distracted us from providing a smooth way of programmatically running Kedro Make the Session more lightweight and re-entrant #4503 (comment)

Is it worth maintaining %reload_kedro when we could provide a 2-liner that is not tied to IPython in any way?

Something like

from kedro import load_kedro

context, catalog, session, pipelines = load_kedro()

or some other alternative? (which is almost there by the way!)

kedro/kedro/ipython/__init__.py

Lines 137 to 144 in 78b5846

	get_ipython().push( # type: ignore[no-untyped-call]
	variables={
	"context": context,
	"catalog": catalog,
	"session": session,
	"pipelines": pipelines,
	}
	)

)

[datajoely]

This would be great, particularly from this namespace