Pabumeasures

Pabumeasures implements the project performance measures from Evaluation of Project Performance in Participatory Budgeting (Boehmer et al., arXiv:2312.14723) for the GreedyAV, Greedy over Cost, Phragmén, and Method of Equal Shares rules with both cost and approval-based utility functions. The package is fully compatible with pabutools, using its Instance, ApprovalProfile, and BudgetAllocation types.

Installation

Prerequisites

• CMake,
• Google OR-Tools (C++) – install via a package manager (e.g., brew) or download a binary distribution.

Building

Pabumeasures uses dynamic linking to reduce build times, so you might need to make the OR-Tools headers and libraries discoverable at both build-time and runtime. The path to OR-Tools must be the installation root containing the lib and include subdirectories.

Linux

export CMAKE_PREFIX_PATH="/path/to/ortools"
export LD_LIBRARY_PATH="$CMAKE_PREFIX_PATH/lib:$LD_LIBRARY_PATH"

macOS

export CMAKE_PREFIX_PATH="/path/to/ortools"
export DYLD_LIBRARY_PATH="$CMAKE_PREFIX_PATH/lib:$DYLD_LIBRARY_PATH"

Windows

set CMAKE_PREFIX_PATH=C:\path\to\ortools
set PATH=%CMAKE_PREFIX_PATH%\lib;%PATH%

You can then install pabumeasures from PyPI:

pip install pabumeasures

Documentation

The general workflow is: create or import PB instances using pabutools, then compute rule outcomes and measures using pabumeasures.

Example

from pabumeasures import Measure, mes_cost, mes_cost_measure
from pabutools.election import ApprovalBallot, ApprovalProfile, Instance, Project

p1 = Project("p1", 1)
p2 = Project("p2", 1)
p3 = Project("p3", 3)

b1 = ApprovalBallot([p1, p2])
b2 = ApprovalBallot([p1, p2, p3])
b3 = ApprovalBallot([p3])

instance = Instance([p1, p2, p3], budget_limit=3)
profile = ApprovalProfile([b1, b2, b3])

mes_cost(instance, profile)  # returns BudgetAllocation([p1, p2])
mes_cost_measure(instance, profile, p3, Measure.ADD_APPROVAL_OPTIMIST)  # returns 1

Available functions

Rules

All rule functions return a BudgetAllocation and accept an optional tie_breaking: ProjectComparator (default: ProjectComparator.ByCostAsc).

Measures

Measure functions have no pabutools equivalent.

All measure functions return an int (or None if the measure is not defined) and accept an optional tie_breaking: ProjectComparator. The measure argument is one of (see Boehmer et al. for formal definitions):