This PR fixes a performance regression in Pandana v0.5.
The network.get_node_ids() function had switched from using scikit-learn nearest neighbors to scipy nearest neighbors, to streamline dependencies. But it turns out the performance is bad when you scale up to hundreds of thousands of locations or more. Thanks @cvanoli for identifying this.
This PR switches it back to the prior implementation.
This PR fixes a performance regression in Pandana v0.5.
The
network.get_node_ids()
function had switched from usingscikit-learn
nearest neighbors toscipy
nearest neighbors, to streamline dependencies. But it turns out the performance is bad when you scale up to hundreds of thousands of locations or more. Thanks @cvanoli for identifying this.This PR switches it back to the prior implementation.
Release logistics: