Docs - https://activitysim.github.io/sharrow/walkthrough/one-dim.html#data-trees-and-flows

[noornaddour]

Hi,

I hope this is not a bother, but I think there's a slight misalignment between the text and the example in:
https://activitysim.github.io/sharrow/walkthrough/one-dim.html#data-trees-and-flows

The tree defined here is not a tree in the mathematical sense. It has cycle(s), when showing at least the edges by label. The text above the example indicates that "all sharrow.DataTree objects are trees in the mathematical sense". The text says:
"Then, it’s time to prepare our data...This is a tree in the mathematical sense..."

Here's a cycle in the example just below that text: Tour[PERID] -> Person[PERID], then Person[household_id] -> hh[HHID], then HH[TAZ] -> dot_skims[dtaz], then ANTIDirectionOF[(dot_skims[time_period] -> tour['out_time_period']).

Ultimately, this class is defined based on networkx.MultiDiGraph, describing it as a tree in the "mathematical sense" doesn't really make much sense to me. Happy to be corrected.

[jpn--]

My apologies for your confusion. On thinking about your question, it's not really the datasets that are the nodes of the tree, but rather the dimensions on the datasets. For each dataset dimension, there needs to be only one unique path back to the root node, so that for that dataset we can map some single value to every position in the root data. Thus, there's no cycle in the tree in the example, because each of the 3 dimensions of dot_skims is its own leaf.

I've added some additional context in #76 that hopefully will help clarify.

[jpn--]

Closed by #76