How to utilize the HierarchicalForecast reconciling- algorithm when having a "ragged hierarchy" / "unbalanced hierarchy".

[theforecastnoob]

I'm wondering how to utilize Nixtla's hierarchical forecasting library when using an unbalanced hierarchy, for example, see added image, when we have varying lower levels for some products.

(image from stackexchange, https://stats.stackexchange.com/a/412355)

One idea I had was to separate the hierarchy into sub-hierarchies, e.g., start with reconciling the lowest parent node: Chocolate Milk & its lower levels, then use the reconciled Chocolate milk for it's parent node Flavored Milk, and so on, until I'm at the top level.

Seems kinda slow to do this way, and I know the hts (R) library can handle this situation, but I want to do this in Python.

Would appreciate your thoughts about this problem! :)