DroopArchitecture

Droop Architecture

Droop design goals

The primary goal of Droop is to be able to implement STV counting rules in a native, self-contained, transparent manner.

• A counting rule implementation is native in the sense that it is implemented in its own terms, such as its legislative specification, rather than reinterpreted in the terms of the counting program.
• While the Droop framework provides useful low-level tools that enable efficient and concise implementation of a counting rule, the implementation is self-contained, in the sense that all of its logic is contained in the particular rule implementation.
• Once a rule implementation is complete, it should be transparent in the sense that it can be seen to be correct, largely as a consequence of the above two properties.
• A further consequence of these design goals is that rule implementations are independent. Logic changes or corrections to one rule have no impact on the other rules.
• To verify initial and continuing correctness, the Droop framework includes a complete set of unit tests.

One consequence of the primary goal is that Droop occasionally ignores Python's DRY ("don't repeat yourself") maxim, in favor of rule self-containment.

Droop framework components

The Droop framework consists of a collection of classes and functions that provide housekeeping for STV rule implementations as well as an API to external interfaces. The major components of the framework include:

• class ElectionProfile: ballot-file parsing
• class Election: top-level counting logic and rule support
  • class Ballot: internal ballot representation
• class Candidate: internal candidate representation
• class Candidates: the set of candidates
• class ElectionRecord: maintain and report a record of an election
• class ElectionRule: a specific counting rule
• class MethodMeek, MethodWIGM: generic support for Meek- & WIGM-based rules
• extensible arithmetic support
  • class Fixed: fixed-point arithmetic
  • class Guarded: guarded-precision fixed-point arithmetic provides an efficient approximation to exact arithmetic
  • class Rational: exact-value arithmetic
• uniform communication of help info from rules to UI
• uniform communication of election parameters from UI to rules
• unit testing
• several built-in rules
• a command-line interface

Together with drop-in rule implementations, the Droop framework acts as an STV counting engine, with external interfaces for ballot data, election parameters and result reporting.