rdafn 1.2.0

Redistricting analytics for scoring ensembles of redistricting plans

rdafn

Compute Dave's Redistricting (DRA) analytics for an ensemble of redistricting plans.

Installation

I've converted this into a pip installable package:

$ pip install rdafn

As noted next, you probably also want to clone the companion rdadata repository.

Usage

There are both code and script examples of how to use this code in the sample directory. That directory also contains some sample results from the main scoring function analyze_plan(). The samples use data from the companion rdadata repo.

Notes

With four exceptions, analyze_plan() computes all the analytics that DRA does:

• For a variety of reasons, DRA's production TypeScript package dra-analytics does not calculate a few minor things that show up in the UI. The Python port rdapy does not either. This repo uses the latter, so those few things also aren't in the "scorecard" output.
• To keep the results simple, district-level results are suppressed. The scorecard is a simple flat dictionary of metric key/value pairs.
• To maximize throughput KIWYSI compactness is not calculated. The simple naive approach to performing compactness calculations is to dissolve precinct shapes into district shapes, but dissolve is very expensive operation. Analyzing a congressional plan for North Carolina take ~60 seconds. A much faster approach is to convert precinct shapes into topologies using TopoJSON like DRA does and then merging precincts into district shapes. That approach takes ~5 seconds, virtually all of the time being calling TopoJSON merge() from Python and marshalling the result back from JavaScript. I could have chosen to implement a Python native version of merge(). Instead, I chose to skip KIWYSI compactness (which requires actual district shapes) and just calculate the two main compactness metrics in DRA: Reock and Polsby-Popper. Together these only depend on district area, perimeter, and diameter, and with some pre-processing once per state (analogous to converting shapes into a topology) these values can be imputed without ever creating the district shapes. The result is that analyzing a congressional plan for North Carolina — calculating all the analytics — takes a small fraction of a second.
• Finally, we've already created the precinct contiguity graphs as part of finding root map candidates in my baseline GitHub repo, and we're also already using the graph in Todd's ensembles repo to support generating spanning trees. So, by definition, the plans in our ensembles are contiguous. Hence, we don't check that.

Testing

$ pytest --disable-warnings