Remove CFB/NASCAR/PGA and streamline to 8 supported sports

- Remove College Football, NASCAR, and PGA from scraper and app
- Clean all data files (stadiums, games, pipeline reports)
- Update Sport.swift enum and all UI components
- Add sportstime.py CLI tool for pipeline management
- Add DATA_SCRAPING.md documentation
- Add WNBA/MLS/NWSL implementation documentation
- Scraper now supports: NBA, MLB, NHL, NFL, WNBA, MLS, NWSL, CBB

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
Trey t
2026-01-09 23:22:13 -06:00
parent f5e509a9ae
commit 8790d2ad73
35 changed files with 117819 additions and 65871 deletions

View File

@@ -2,7 +2,7 @@
// GameDAGRouter.swift
// SportsTime
//
// Time-expanded DAG + Beam Search algorithm for route finding.
// DAG-based route finding with multi-dimensional diversity.
//
// Key insight: This is NOT "which subset of N games should I attend?"
// This IS: "what time-respecting paths exist through a graph of games?"
@@ -10,11 +10,14 @@
// The algorithm:
// 1. Bucket games by calendar day
// 2. Build directed edges where time moves forward AND driving is feasible
// 3. Beam search: keep top K paths at each depth
// 4. Dominance pruning: discard inferior paths
// 3. Generate routes via beam search
// 4. Diversity pruning: ensure routes span full range of games, cities, miles, and days
//
// Complexity: O(days × beamWidth × avgNeighbors) 900 operations for 5-day, 78-game scenario
// (vs 2^78 for naive subset enumeration)
// The diversity system ensures users see:
// - Short trips (2-3 cities) AND long trips (5+ cities)
// - Quick trips (2-3 games) AND packed trips (5+ games)
// - Low mileage AND high mileage options
// - Short duration AND long duration trips
//
import Foundation
@@ -24,12 +27,11 @@ enum GameDAGRouter {
// MARK: - Configuration
/// Default beam width - how many partial routes to keep at each step
/// Increased to ensure we preserve diverse route lengths (short and long trips)
private static let defaultBeamWidth = 50
/// Default beam width during expansion
private static let defaultBeamWidth = 100
/// Maximum options to return (increased to provide more diverse trip lengths)
private static let maxOptions = 50
/// Maximum options to return (diverse sample)
private static let maxOptions = 75
/// Buffer time after game ends before we can depart (hours)
private static let gameEndBufferHours: Double = 3.0
@@ -37,21 +39,55 @@ enum GameDAGRouter {
/// Maximum days ahead to consider for next game (1 = next day only, 5 = allows multi-day drives)
private static let maxDayLookahead = 5
// MARK: - Route Profile
/// Captures the key metrics of a route for diversity analysis
private struct RouteProfile {
let route: [Game]
let gameCount: Int
let cityCount: Int
let totalMiles: Double
let tripDays: Int
// Bucket indices for stratified sampling
var gameBucket: Int { min(gameCount - 1, 5) } // 1, 2, 3, 4, 5, 6+
var cityBucket: Int { min(cityCount - 1, 5) } // 1, 2, 3, 4, 5, 6+
var milesBucket: Int {
switch totalMiles {
case ..<500: return 0 // Short
case 500..<1000: return 1 // Medium
case 1000..<2000: return 2 // Long
case 2000..<3000: return 3 // Very long
default: return 4 // Epic
}
}
var daysBucket: Int { min(tripDays - 1, 6) } // 1-7+ days
/// Composite key for exact deduplication
var uniqueKey: String {
route.map { $0.id.uuidString }.joined(separator: "-")
}
}
// MARK: - Public API
/// Finds best routes through the game graph using DAG + beam search.
/// Finds routes through the game graph with multi-dimensional diversity.
///
/// This replaces the exponential GeographicRouteExplorer with a polynomial-time algorithm.
/// Returns a curated sample that spans the full range of:
/// - Number of games (2-game quickies to 6+ game marathons)
/// - Number of cities (2-city to 6+ city routes)
/// - Total miles (short drives to cross-country epics)
/// - Trip duration (weekend getaways to week-long adventures)
///
/// - Parameters:
/// - games: All games to consider, in any order (will be sorted internally)
/// - games: All games to consider
/// - stadiums: Dictionary mapping stadium IDs to Stadium objects
/// - constraints: Driving constraints (number of drivers, max hours per day)
/// - anchorGameIds: Games that MUST appear in every valid route (for Scenario B)
/// - constraints: Driving constraints (max hours per day)
/// - anchorGameIds: Games that MUST appear in every valid route
/// - allowRepeatCities: If false, each city can only appear once in a route
/// - beamWidth: How many partial routes to keep at each depth (default 30)
/// - beamWidth: How many partial routes to keep during expansion
///
/// - Returns: Array of valid game combinations, sorted by score (most games, least driving)
/// - Returns: Array of diverse route options
///
static func findRoutes(
games: [Game],
@@ -65,21 +101,18 @@ enum GameDAGRouter {
// Edge cases
guard !games.isEmpty else { return [] }
if games.count == 1 {
// Single game - just return it if it satisfies anchors
if anchorGameIds.isEmpty || anchorGameIds.contains(games[0].id) {
return [games]
}
return []
}
if games.count == 2 {
// Two games - check if both are reachable
let sorted = games.sorted { $0.startTime < $1.startTime }
if canTransition(from: sorted[0], to: sorted[1], stadiums: stadiums, constraints: constraints) {
if anchorGameIds.isSubset(of: Set(sorted.map { $0.id })) {
return [sorted]
}
}
// Can't connect them - return individual games if they satisfy anchors
if anchorGameIds.isEmpty {
return [[sorted[0]], [sorted[1]]]
}
@@ -95,18 +128,9 @@ enum GameDAGRouter {
guard !sortedDays.isEmpty else { return [] }
// Step 3: Initialize beam with first day's games
// Step 3: Initialize beam with first few days' games as starting points
var beam: [[Game]] = []
if let firstDayGames = buckets[sortedDays[0]] {
for game in firstDayGames {
beam.append([game])
}
}
// Also include option to skip first day entirely and start later
// (handled by having multiple starting points in beam)
for dayIndex in sortedDays.dropFirst().prefix(maxDayLookahead - 1) {
for dayIndex in sortedDays.prefix(maxDayLookahead) {
if let dayGames = buckets[dayIndex] {
for game in dayGames {
beam.append([game])
@@ -114,9 +138,8 @@ enum GameDAGRouter {
}
}
// Step 4: Expand beam day by day
for (_, dayIndex) in sortedDays.dropFirst().enumerated() {
for dayIndex in sortedDays.dropFirst() {
let todaysGames = buckets[dayIndex] ?? []
var nextBeam: [[Game]] = []
@@ -124,36 +147,34 @@ enum GameDAGRouter {
guard let lastGame = path.last else { continue }
let lastGameDay = dayIndexFor(lastGame.startTime, referenceDate: sortedGames[0].startTime)
// Only consider games on this day or within lookahead
// Skip if this day is too far ahead for this route
if dayIndex > lastGameDay + maxDayLookahead {
// This path is too far behind, keep it as-is
nextBeam.append(path)
continue
}
// Try adding each of today's games
for candidate in todaysGames {
// Check for repeat city violation during route building
// Check for repeat city violation
if !allowRepeatCities {
let candidateCity = stadiums[candidate.stadiumId]?.city ?? ""
let pathCities = Set(path.compactMap { stadiums[$0.stadiumId]?.city })
if pathCities.contains(candidateCity) {
continue // Skip - would violate allowRepeatCities
continue
}
}
if canTransition(from: lastGame, to: candidate, stadiums: stadiums, constraints: constraints) {
let newPath = path + [candidate]
nextBeam.append(newPath)
nextBeam.append(path + [candidate])
}
}
// Also keep the path without adding a game today (allows off-days)
// Keep the path without adding a game today
nextBeam.append(path)
}
// Dominance pruning + beam truncation
beam = pruneAndTruncate(nextBeam, beamWidth: beamWidth, stadiums: stadiums)
// Diversity-aware pruning during expansion
beam = diversityPrune(nextBeam, stadiums: stadiums, targetCount: beamWidth)
}
// Step 5: Filter routes that contain all anchors
@@ -162,21 +183,15 @@ enum GameDAGRouter {
return anchorGameIds.isSubset(of: pathGameIds)
}
// Step 6: Ensure geographic diversity in results
// Group routes by their primary region (city with most games)
// Then pick the best route from each region
// Step 6: Final diversity selection
let finalRoutes = selectDiverseRoutes(routesWithAnchors, stadiums: stadiums, maxCount: maxOptions)
print("🔍 DAG: Input games=\(games.count), beam final=\(beam.count), withAnchors=\(routesWithAnchors.count), final=\(finalRoutes.count)")
if let best = finalRoutes.first {
print("🔍 DAG: Best route has \(best.count) games")
}
print("🔍 DAG: Input games=\(games.count), beam=\(beam.count), withAnchors=\(routesWithAnchors.count), final=\(finalRoutes.count)")
return finalRoutes
}
/// Compatibility wrapper that matches GeographicRouteExplorer's interface.
/// This allows drop-in replacement in ScenarioAPlanner and ScenarioBPlanner.
static func findAllSensibleRoutes(
from games: [Game],
stadiums: [UUID: Stadium],
@@ -184,9 +199,7 @@ enum GameDAGRouter {
allowRepeatCities: Bool = true,
stopBuilder: ([Game], [UUID: Stadium]) -> [ItineraryStop]
) -> [[Game]] {
// Use default driving constraints
let constraints = DrivingConstraints.default
return findRoutes(
games: games,
stadiums: stadiums,
@@ -196,9 +209,236 @@ enum GameDAGRouter {
)
}
// MARK: - Multi-Dimensional Diversity Selection
/// Selects routes that maximize diversity across all dimensions.
/// Uses stratified sampling to ensure representation of:
/// - Short trips (2-3 games) AND long trips (5+ games)
/// - Few cities (2-3) AND many cities (5+)
/// - Low mileage AND high mileage
/// - Short duration AND long duration
private static func selectDiverseRoutes(
_ routes: [[Game]],
stadiums: [UUID: Stadium],
maxCount: Int
) -> [[Game]] {
guard !routes.isEmpty else { return [] }
// Build profiles for all routes
let profiles = routes.map { route in
buildProfile(for: route, stadiums: stadiums)
}
// Remove duplicates
var uniqueProfiles: [RouteProfile] = []
var seenKeys = Set<String>()
for profile in profiles {
if !seenKeys.contains(profile.uniqueKey) {
seenKeys.insert(profile.uniqueKey)
uniqueProfiles.append(profile)
}
}
// Stratified selection: ensure representation across all buckets
var selected: [RouteProfile] = []
var selectedKeys = Set<String>()
// Pass 1: Ensure at least one route per game count bucket (2, 3, 4, 5, 6+)
let byGames = Dictionary(grouping: uniqueProfiles) { $0.gameBucket }
for bucket in byGames.keys.sorted() {
if selected.count >= maxCount { break }
if let candidates = byGames[bucket]?.sorted(by: { $0.totalMiles < $1.totalMiles }) {
if let best = candidates.first, !selectedKeys.contains(best.uniqueKey) {
selected.append(best)
selectedKeys.insert(best.uniqueKey)
}
}
}
// Pass 2: Ensure at least one route per city count bucket (2, 3, 4, 5, 6+)
let byCities = Dictionary(grouping: uniqueProfiles) { $0.cityBucket }
for bucket in byCities.keys.sorted() {
if selected.count >= maxCount { break }
if let candidates = byCities[bucket]?.filter({ !selectedKeys.contains($0.uniqueKey) }) {
if let best = candidates.sorted(by: { $0.totalMiles < $1.totalMiles }).first {
selected.append(best)
selectedKeys.insert(best.uniqueKey)
}
}
}
// Pass 3: Ensure at least one route per mileage bucket
let byMiles = Dictionary(grouping: uniqueProfiles) { $0.milesBucket }
for bucket in byMiles.keys.sorted() {
if selected.count >= maxCount { break }
if let candidates = byMiles[bucket]?.filter({ !selectedKeys.contains($0.uniqueKey) }) {
if let best = candidates.sorted(by: { $0.gameCount > $1.gameCount }).first {
selected.append(best)
selectedKeys.insert(best.uniqueKey)
}
}
}
// Pass 4: Ensure at least one route per duration bucket
let byDays = Dictionary(grouping: uniqueProfiles) { $0.daysBucket }
for bucket in byDays.keys.sorted() {
if selected.count >= maxCount { break }
if let candidates = byDays[bucket]?.filter({ !selectedKeys.contains($0.uniqueKey) }) {
if let best = candidates.sorted(by: { $0.gameCount > $1.gameCount }).first {
selected.append(best)
selectedKeys.insert(best.uniqueKey)
}
}
}
// Pass 5: Fill remaining slots with diverse combinations
// Create composite buckets for more granular diversity
let remaining = uniqueProfiles.filter { !selectedKeys.contains($0.uniqueKey) }
let byComposite = Dictionary(grouping: remaining) { profile in
"\(profile.gameBucket)-\(profile.cityBucket)-\(profile.milesBucket)"
}
// Round-robin from composite buckets
var compositeKeys = Array(byComposite.keys).sorted()
var indices: [String: Int] = [:]
while selected.count < maxCount && !compositeKeys.isEmpty {
var addedAny = false
for key in compositeKeys {
if selected.count >= maxCount { break }
let idx = indices[key] ?? 0
if let candidates = byComposite[key], idx < candidates.count {
let profile = candidates[idx]
if !selectedKeys.contains(profile.uniqueKey) {
selected.append(profile)
selectedKeys.insert(profile.uniqueKey)
addedAny = true
}
indices[key] = idx + 1
}
}
if !addedAny { break }
}
// Pass 6: If still need more, add remaining sorted by efficiency
if selected.count < maxCount {
let stillRemaining = uniqueProfiles
.filter { !selectedKeys.contains($0.uniqueKey) }
.sorted { efficiency(for: $0) > efficiency(for: $1) }
for profile in stillRemaining.prefix(maxCount - selected.count) {
selected.append(profile)
}
}
return selected.map { $0.route }
}
/// Diversity-aware pruning during beam expansion.
/// Keeps routes that span the diversity space rather than just high-scoring ones.
private static func diversityPrune(
_ paths: [[Game]],
stadiums: [UUID: Stadium],
targetCount: Int
) -> [[Game]] {
// Remove exact duplicates first
var uniquePaths: [[Game]] = []
var seen = Set<String>()
for path in paths {
let key = path.map { $0.id.uuidString }.joined(separator: "-")
if !seen.contains(key) {
seen.insert(key)
uniquePaths.append(path)
}
}
guard uniquePaths.count > targetCount else { return uniquePaths }
// Build profiles
let profiles = uniquePaths.map { buildProfile(for: $0, stadiums: stadiums) }
// Group by game count to ensure length diversity
let byGames = Dictionary(grouping: profiles) { $0.gameBucket }
let slotsPerBucket = max(2, targetCount / max(1, byGames.count))
var selected: [RouteProfile] = []
var selectedKeys = Set<String>()
// Take from each game count bucket
for bucket in byGames.keys.sorted() {
if let candidates = byGames[bucket] {
// Within bucket, prioritize geographic diversity
let byCities = Dictionary(grouping: candidates) { $0.cityBucket }
var bucketSelected = 0
for cityBucket in byCities.keys.sorted() {
if bucketSelected >= slotsPerBucket { break }
if let cityCandidates = byCities[cityBucket] {
for profile in cityCandidates.prefix(2) {
if !selectedKeys.contains(profile.uniqueKey) {
selected.append(profile)
selectedKeys.insert(profile.uniqueKey)
bucketSelected += 1
if bucketSelected >= slotsPerBucket { break }
}
}
}
}
}
}
// Fill remaining with efficiency-sorted paths
if selected.count < targetCount {
let remaining = profiles.filter { !selectedKeys.contains($0.uniqueKey) }
.sorted { efficiency(for: $0) > efficiency(for: $1) }
for profile in remaining.prefix(targetCount - selected.count) {
selected.append(profile)
}
}
return selected.map { $0.route }
}
/// Builds a profile for a route.
private static func buildProfile(for route: [Game], stadiums: [UUID: Stadium]) -> RouteProfile {
let gameCount = route.count
let cities = Set(route.compactMap { stadiums[$0.stadiumId]?.city })
let cityCount = cities.count
// Calculate total miles
var totalMiles: Double = 0
for i in 0..<(route.count - 1) {
totalMiles += estimateDistanceMiles(from: route[i], to: route[i + 1], stadiums: stadiums)
}
// Calculate trip duration in days
let tripDays: Int
if let firstGame = route.first, let lastGame = route.last {
let calendar = Calendar.current
let days = calendar.dateComponents([.day], from: firstGame.startTime, to: lastGame.startTime).day ?? 1
tripDays = max(1, days + 1)
} else {
tripDays = 1
}
return RouteProfile(
route: route,
gameCount: gameCount,
cityCount: cityCount,
totalMiles: totalMiles,
tripDays: tripDays
)
}
/// Calculates efficiency score (games per hour of driving).
private static func efficiency(for profile: RouteProfile) -> Double {
let drivingHours = profile.totalMiles / 60.0 // 60 mph average
guard drivingHours > 0 else { return Double(profile.gameCount) * 100 }
return Double(profile.gameCount) / drivingHours
}
// MARK: - Day Bucketing
/// Groups games by calendar day index (0 = first day of trip, 1 = second day, etc.)
private static func bucketByDay(games: [Game]) -> [Int: [Game]] {
guard let firstGame = games.first else { return [:] }
let referenceDate = firstGame.startTime
@@ -211,24 +451,15 @@ enum GameDAGRouter {
return buckets
}
/// Calculates the day index for a date relative to a reference date.
private static func dayIndexFor(_ date: Date, referenceDate: Date) -> Int {
let calendar = Calendar.current
let refDay = calendar.startOfDay(for: referenceDate)
let dateDay = calendar.startOfDay(for: date)
let components = calendar.dateComponents([.day], from: refDay, to: dateDay)
return components.day ?? 0
return calendar.dateComponents([.day], from: refDay, to: dateDay).day ?? 0
}
// MARK: - Transition Feasibility
/// Determines if we can travel from game A to game B.
///
/// Requirements:
/// 1. B starts after A (time moves forward)
/// 2. We have enough days between games to complete the drive
/// 3. We can arrive at B before B starts
///
private static func canTransition(
from: Game,
to: Game,
@@ -236,289 +467,62 @@ enum GameDAGRouter {
constraints: DrivingConstraints
) -> Bool {
// Time must move forward
guard to.startTime > from.startTime else {
return false
}
guard to.startTime > from.startTime else { return false }
// Same stadium = always feasible (no driving needed)
// Same stadium = always feasible
if from.stadiumId == to.stadiumId { return true }
// Get stadiums
guard let fromStadium = stadiums[from.stadiumId],
let toStadium = stadiums[to.stadiumId] else {
// Missing stadium info - can't calculate distance, reject to be safe
print("⚠️ DAG: Stadium lookup failed - from:\(stadiums[from.stadiumId] != nil) to:\(stadiums[to.stadiumId] != nil)")
return false
}
let fromCoord = fromStadium.coordinate
let toCoord = toStadium.coordinate
// Calculate driving time
let distanceMiles = TravelEstimator.haversineDistanceMiles(
from: CLLocationCoordinate2D(latitude: fromCoord.latitude, longitude: fromCoord.longitude),
to: CLLocationCoordinate2D(latitude: toCoord.latitude, longitude: toCoord.longitude)
from: CLLocationCoordinate2D(latitude: fromStadium.coordinate.latitude, longitude: fromStadium.coordinate.longitude),
to: CLLocationCoordinate2D(latitude: toStadium.coordinate.latitude, longitude: toStadium.coordinate.longitude)
) * 1.3 // Road routing factor
let drivingHours = distanceMiles / 60.0 // Average 60 mph
let drivingHours = distanceMiles / 60.0
// Calculate available driving time between games
// After game A ends (+ buffer), how much time until game B starts (- buffer)?
// Calculate available time
let departureTime = from.startTime.addingTimeInterval(gameEndBufferHours * 3600)
let deadline = to.startTime.addingTimeInterval(-3600) // 1 hour buffer before game
let availableSeconds = deadline.timeIntervalSince(departureTime)
let availableHours = availableSeconds / 3600.0
let availableHours = deadline.timeIntervalSince(departureTime) / 3600.0
// Calculate how many driving days we have
// Each day can have maxDailyDrivingHours of driving
// Calculate driving days available
let calendar = Calendar.current
let fromDay = calendar.startOfDay(for: from.startTime)
let toDay = calendar.startOfDay(for: to.startTime)
let daysBetween = calendar.dateComponents([.day], from: fromDay, to: toDay).day ?? 0
let daysBetween = calendar.dateComponents(
[.day],
from: calendar.startOfDay(for: from.startTime),
to: calendar.startOfDay(for: to.startTime)
).day ?? 0
// Available driving hours = days between * max per day
// (If games are same day, daysBetween = 0, but we might still have hours available)
let maxDrivingHoursAvailable: Double
if daysBetween == 0 {
// Same day - only have hours between games
maxDrivingHoursAvailable = max(0, availableHours)
} else {
// Multi-day - can drive each day
maxDrivingHoursAvailable = Double(daysBetween) * constraints.maxDailyDrivingHours
}
let maxDrivingHoursAvailable = daysBetween == 0
? max(0, availableHours)
: Double(daysBetween) * constraints.maxDailyDrivingHours
// Check if we have enough driving time
guard drivingHours <= maxDrivingHoursAvailable else {
return false
}
// Also verify we can arrive before game starts (sanity check)
guard availableHours >= drivingHours else {
return false
}
return true
return drivingHours <= maxDrivingHoursAvailable && drivingHours <= availableHours
}
// MARK: - Geographic Diversity
// MARK: - Distance Estimation
/// Selects diverse routes from the candidate set.
/// Ensures diversity by BOTH route length (city count) AND primary city.
/// This guarantees users see 2-city trips alongside 5+ city trips.
private static func selectDiverseRoutes(
_ routes: [[Game]],
stadiums: [UUID: Stadium],
maxCount: Int
) -> [[Game]] {
guard !routes.isEmpty else { return [] }
// Group routes by city count (route length)
var routesByLength: [Int: [[Game]]] = [:]
for route in routes {
let cityCount = Set(route.compactMap { stadiums[$0.stadiumId]?.city }).count
routesByLength[cityCount, default: []].append(route)
}
// Sort routes within each length by score
for (length, lengthRoutes) in routesByLength {
routesByLength[length] = lengthRoutes.sorted {
scorePath($0, stadiums: stadiums) > scorePath($1, stadiums: stadiums)
}
}
// Allocate slots to each length category
// Goal: ensure at least 1 route per length category if available
let sortedLengths = routesByLength.keys.sorted()
let minPerLength = max(1, maxCount / max(1, sortedLengths.count))
var selectedRoutes: [[Game]] = []
var selectedIds = Set<String>()
// First pass: take best route(s) from each length category
for length in sortedLengths {
if selectedRoutes.count >= maxCount { break }
if let lengthRoutes = routesByLength[length] {
let toTake = min(minPerLength, lengthRoutes.count, maxCount - selectedRoutes.count)
for route in lengthRoutes.prefix(toTake) {
let key = route.map { $0.id.uuidString }.joined(separator: "-")
if !selectedIds.contains(key) {
selectedRoutes.append(route)
selectedIds.insert(key)
}
}
}
}
// Second pass: fill remaining slots, prioritizing geographic diversity
if selectedRoutes.count < maxCount {
// Group remaining routes by primary city
var remainingByCity: [String: [[Game]]] = [:]
for route in routes {
let key = route.map { $0.id.uuidString }.joined(separator: "-")
if !selectedIds.contains(key) {
let city = getPrimaryCity(for: route, stadiums: stadiums)
remainingByCity[city, default: []].append(route)
}
}
// Sort by score within each city
for (city, cityRoutes) in remainingByCity {
remainingByCity[city] = cityRoutes.sorted {
scorePath($0, stadiums: stadiums) > scorePath($1, stadiums: stadiums)
}
}
// Round-robin from each city
let sortedCities = remainingByCity.keys.sorted { city1, city2 in
let score1 = remainingByCity[city1]?.first.map { scorePath($0, stadiums: stadiums) } ?? 0
let score2 = remainingByCity[city2]?.first.map { scorePath($0, stadiums: stadiums) } ?? 0
return score1 > score2
}
var cityIndices: [String: Int] = [:]
while selectedRoutes.count < maxCount {
var addedAny = false
for city in sortedCities {
if selectedRoutes.count >= maxCount { break }
let idx = cityIndices[city] ?? 0
if let cityRoutes = remainingByCity[city], idx < cityRoutes.count {
let route = cityRoutes[idx]
let key = route.map { $0.id.uuidString }.joined(separator: "-")
if !selectedIds.contains(key) {
selectedRoutes.append(route)
selectedIds.insert(key)
addedAny = true
}
cityIndices[city] = idx + 1
}
}
if !addedAny { break }
}
}
return selectedRoutes
}
/// Gets the primary city for a route (where most games are played).
private static func getPrimaryCity(for route: [Game], stadiums: [UUID: Stadium]) -> String {
var cityCounts: [String: Int] = [:]
for game in route {
let city = stadiums[game.stadiumId]?.city ?? "Unknown"
cityCounts[city, default: 0] += 1
}
return cityCounts.max(by: { $0.value < $1.value })?.key ?? "Unknown"
}
// MARK: - Scoring and Pruning
/// Scores a path. Higher = better.
/// Prefers: more games, less driving, geographic coherence
private static func scorePath(_ path: [Game], stadiums: [UUID: Stadium]) -> Double {
// Handle empty or single-game paths
guard path.count > 1 else {
return Double(path.count) * 100.0
}
let gameCount = Double(path.count)
// Calculate total driving
var totalDriving: Double = 0
for i in 0..<(path.count - 1) {
totalDriving += estimateDrivingHours(from: path[i], to: path[i + 1], stadiums: stadiums)
}
// Score: heavily weight game count, penalize driving
return gameCount * 100.0 - totalDriving * 2.0
}
/// Estimates driving hours between two games.
private static func estimateDrivingHours(
private static func estimateDistanceMiles(
from: Game,
to: Game,
stadiums: [UUID: Stadium]
) -> Double {
// Same stadium = 0 driving
if from.stadiumId == to.stadiumId { return 0 }
guard let fromStadium = stadiums[from.stadiumId],
let toStadium = stadiums[to.stadiumId] else {
return 5.0 // Fallback: assume 5 hours
return 300 // Fallback estimate
}
let fromCoord = fromStadium.coordinate
let toCoord = toStadium.coordinate
let distanceMiles = TravelEstimator.haversineDistanceMiles(
from: CLLocationCoordinate2D(latitude: fromCoord.latitude, longitude: fromCoord.longitude),
to: CLLocationCoordinate2D(latitude: toCoord.latitude, longitude: toCoord.longitude)
return TravelEstimator.haversineDistanceMiles(
from: CLLocationCoordinate2D(latitude: fromStadium.coordinate.latitude, longitude: fromStadium.coordinate.longitude),
to: CLLocationCoordinate2D(latitude: toStadium.coordinate.latitude, longitude: toStadium.coordinate.longitude)
) * 1.3
return distanceMiles / 60.0
}
/// Prunes dominated paths and truncates to beam width.
/// Maintains diversity by both ending city AND route length to ensure short trips aren't eliminated.
private static func pruneAndTruncate(
_ paths: [[Game]],
beamWidth: Int,
stadiums: [UUID: Stadium]
) -> [[Game]] {
// Remove exact duplicates
var uniquePaths: [[Game]] = []
var seen = Set<String>()
for path in paths {
let key = path.map { $0.id.uuidString }.joined(separator: "-")
if !seen.contains(key) {
seen.insert(key)
uniquePaths.append(path)
}
}
// Group paths by unique city count (route length)
// This ensures we keep short trips (2 cities) alongside long trips (5+ cities)
var pathsByLength: [Int: [[Game]]] = [:]
for path in uniquePaths {
let cityCount = Set(path.compactMap { stadiums[$0.stadiumId]?.city }).count
pathsByLength[cityCount, default: []].append(path)
}
// Sort paths within each length group by score
for (length, lengthPaths) in pathsByLength {
pathsByLength[length] = lengthPaths.sorted {
scorePath($0, stadiums: stadiums) > scorePath($1, stadiums: stadiums)
}
}
// Allocate beam slots proportionally to length groups, with minimum per group
let sortedLengths = pathsByLength.keys.sorted()
let minPerLength = max(2, beamWidth / max(1, sortedLengths.count))
var pruned: [[Game]] = []
// First pass: take minimum from each length group
for length in sortedLengths {
if let lengthPaths = pathsByLength[length] {
let toTake = min(minPerLength, lengthPaths.count)
pruned.append(contentsOf: lengthPaths.prefix(toTake))
}
}
// Second pass: fill remaining slots with best paths overall
if pruned.count < beamWidth {
let remaining = beamWidth - pruned.count
let prunedIds = Set(pruned.map { $0.map { $0.id.uuidString }.joined(separator: "-") })
// Get all paths not yet added, sorted by score
var additional = uniquePaths.filter {
!prunedIds.contains($0.map { $0.id.uuidString }.joined(separator: "-"))
}
additional.sort { scorePath($0, stadiums: stadiums) > scorePath($1, stadiums: stadiums) }
pruned.append(contentsOf: additional.prefix(remaining))
}
// Final truncation
return Array(pruned.prefix(beamWidth))
}
}