From c9b64d56f3606fff45225db6567110a840b35e72 Mon Sep 17 00:00:00 2001
From: Silvan Mosberger <nca@infinisil.com>
Date: Sun, 3 Nov 2024 23:28:07 +0100
Subject: [PATCH] Add scripts to help with determining results

The random generate script is useful to test OpaVote/CIVS in advance

Please forgive me for using bash ':)
---
 nix/generate.sh | 76 +++++++++++++++++++++++++++++++++++++++++++++++++
 nix/process.sh  | 62 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 138 insertions(+)
 create mode 100755 nix/generate.sh
 create mode 100755 nix/process.sh

diff --git a/nix/generate.sh b/nix/generate.sh
new file mode 100755
index 00000000..7d32ebcb
--- /dev/null
+++ b/nix/generate.sh
@@ -0,0 +1,76 @@
+#!/usr/bin/env bash
+
+# Generates a randomised OpaVote ballot file (https://opavote.com/help/overview#blt-file-format) to stdout
+
+export BC_ENV_ARGS="-lq"
+
+# Every voter has a normal-distributed preference X
+
+# 25 voters and 10 candidates is the maximum for the free tier
+voterCount=450
+# 9 so that we have candidates -2, -1.5, ..., 0, ..., 1.5, 2, instead of messy numbers
+candidateCount=23
+seatCount=7
+# With 40%, a minority group should get a representative
+minorityPercentage=15
+
+
+# Voters are represented with a single number. They vote for the candidate nearest to their number
+voterValues=()
+pi=$(echo "scale=10; 4*a(1)" | bc -l)
+normal() {
+  bc <<< "$1 * sqrt(-2 * l($(( RANDOM )) / 32767 )) * c(2 * $pi * $(( RANDOM )) / 32767) + $2"
+}
+
+for i in $(seq "$voterCount"); do
+  voterValues+=( "$(normal 2 0)" )
+  # Check proportional representation works
+  #if (( 100 * i / voterCount <= minorityPercentage )); then
+  #  # The minority group votes with a wide standard deviation around candidate (-3)
+  #  # such that no single candidate gets more 1st ranks than the majority group.
+  #  # We want to test that this group still get a representative seat regardless
+  #  voterValues+=( "$(normal 2 -3)" )
+  #else
+  #  # The majority group all votes effectively for the same candidate (3) with only minor differences,
+  #  # which ensures that a number of candidates that could in theory fill all seats get most 1st and 2nd ranks,
+  #  # but we don't want that to happen because of proportional representation
+  #  voterValues+=( "$(normal 0.5 3)" )
+  #fi
+done
+
+# Sort the voters by their value number and therefore also ballots
+readarray -t voterValues < <(printf "%s\n" "${voterValues[@]}" | sort -n)
+declare -p voterValues >&2
+
+# Candidates are uniformly distributed from -4 to +4, somewhat matching what would happen in reality,
+# where people want to have a representative to vote for, and if they don't, they'd nominate themselves.
+candidates=()
+for (( i=1; i <= candidateCount; i++ )); do
+  candidates[i]=$(bc <<< "scale=1; ($i - 1) * 8 / ($candidateCount - 1) - 4")
+done
+declare -p candidates >&2
+
+# Start generating the file
+
+echo "$candidateCount $seatCount"
+
+for value in "${voterValues[@]}"; do
+  #echo "New voter: $value"
+  readarray -t rankedCandidates < <(
+    for candidate in "${!candidates[@]}"; do
+      # Calculate the distance between the voters value and the candidates
+      echo "$candidate $(bc <<< "($value - ${candidates[$candidate]}) ^ 2")"
+    done | sort -n -k2 | cut -d' ' -f1)
+
+  #declare -p rankedCandidates
+  weight=1
+  end=0
+  echo "$weight" "${rankedCandidates[@]}" "$end"
+done
+echo 0
+
+for candidate in "${candidates[@]}"; do
+  echo "\"$candidate\""
+done
+
+echo "\"Test Election\""
diff --git a/nix/process.sh b/nix/process.sh
new file mode 100755
index 00000000..7a67922d
--- /dev/null
+++ b/nix/process.sh
@@ -0,0 +1,62 @@
+#!/usr/bin/env bash
+
+# Converts an OpaVote BLT file (https://opavote.com/help/overview#blt-file-format) from stdin
+# to a CIVS ballot file (https://github.com/andrewcmyers/civs/blob/867d2c68b0259cac6e0269515b56edb1c49a8003/cgi-bin/languages/base_language.pm#L326-L344) to stdout
+
+# Summary of the CIVS format:
+# Each line contains the ranks of the N choices. Ranks are numbers from 1 to N
+# 2 4 3 1 5        A simple ballot ranking five choices. The top-ranked choice is the fourth one.
+
+# However, entirely undocumented right now, for the best-candidate (the opposite of combined weights/ratings) voter criteria, which we're using, the rankings are inverted!
+# https://github.com/andrewcmyers/civs/blob/867d2c68b0259cac6e0269515b56edb1c49a8003/cgi-bin/vote#L176-L178
+# https://github.com/andrewcmyers/civs/blob/867d2c68b0259cac6e0269515b56edb1c49a8003/cgi-bin/results#L468-L475
+
+set -euo pipefail
+
+read -r candidateCount _seatCount;
+echo "Candidate count: $candidateCount" >&2
+while read -ra ballot; do
+  if [[ "${#ballot[@]}" == 1 && "${ballot[0]}" == "0" ]]; then
+    echo "Read all ballots" >&2
+    break
+  else
+    weight=${ballot[0]}
+    unset 'ballot[0]'
+    end=${ballot[-1]}
+    unset 'ballot[-1]'
+    if [[ "$weight" != 1 || "$end" != 0 ]]; then
+      echo "Problematic ballot: First number (weight) is $weight (should be 1), last number (end marker) is $end (should be 0)" >&2
+      exit 1
+    fi
+
+    declare -a candidateRanks=()
+    for candidate in $(seq "${candidateCount}"); do
+      candidateRanks[candidate]=$candidateCount
+    done
+
+    # Inverts the ranking, see top of file!
+    rank=$candidateCount
+    for candidate in "${ballot[@]}"; do
+      candidateRanks[candidate]=$(( rank-- ))
+    done
+
+    echo "${ballot[@]}" "->" "${candidateRanks[@]}" >&2
+    echo "${candidateRanks[@]}"
+  fi
+done
+
+echo "Candidates (fill this into the candidate field):" >&2
+for candidate in $(seq "${candidateCount}"); do
+  read -r name
+  name=${name%\"}
+  name=${name#\"}
+  if [[ "$name" == "0" ]]; then
+    # For some reason CIVS doesn't store candidates named "0"
+    echo "0.0" >&2
+  else
+    echo "$name" >&2
+  fi
+done
+
+read -r title
+echo "Title: $title" >&2