The dot product implementation broke some workflows and isn't faster in every instance. The default behavior (aggregating data through a for loop over polgyons) is restored, with the dot product implementation accessible through an optional impl='dot_product' argument.
One test has been added to make sure the dot_product works in the basic aggregation case, but only one test has been changed. This may cause downstream issues in the future - therefore, recommend either
fully integrating the dot product as a possibility into the workflow (this would require additionally making every relevant test also be run on the dot product implementation, in addition to lots of code cleanup - there are some redundancies right now due to the tacked on nature of this fix)
revisiting under which circumstances the dot product implementation gives a substantial speed boost. If that is only the case in very limited circumstances, then perhaps it should be removed entirely.
The dot product implementation broke some workflows and isn't faster in every instance. The default behavior (aggregating data through a for loop over polgyons) is restored, with the dot product implementation accessible through an optional
impl='dot_product'
argument.One test has been added to make sure the dot_product works in the basic aggregation case, but only one test has been changed. This may cause downstream issues in the future - therefore, recommend either