public void CalculateTraitsWithMods(BeatmapTraits nomodTraits, Mods mods, GameMode mode, BeatmapTraits expectedTraits)
{
// Act
var traits = BeatmapTraitsLogic.CalculateTraitsWithMods(nomodTraits, mods, mode);
// Assert
traits.Should().BeEquivalentTo(expectedTraits,
o => o.Using <double>(ctx => ctx.Subject.Should().BeApproximately(ctx.Expectation, 0.01)).WhenTypeIs <double>());
}
public async Task <IReadOnlyList <Recommendation> > GetRecommendationsAsync(IProgress <double> progressHandler)
{
var progress = 0.0;
progressHandler.Report(progress);
// Get user's top plays
var ownTopPlays = (await _dataService.GetUserTopPlaysAsync(UserId, GameMode))
.OrderByDescending(p => p.PerformancePoints) // sort by PP
.ToArray();
// Get user's top maps
var ownTopPlaysBeatmapIds = ownTopPlays.Select(p => p.BeatmapId).ToArray();
// Truncate user's top plays
ownTopPlays = ownTopPlays.Take(30).ToArray();
// If user doesn't have any plays - throw
if (!ownTopPlays.Any())
{
throw new RecommendationsUnavailableException($"User [{UserId}] doesn't have any top plays in [{GameMode}].");
}
// Set boundaries for recommendations based on PP
var minPP = ownTopPlays.Average(p => p.PerformancePoints);
var maxPP = minPP * 1.25;
// Prepare buffer for plays which will serve as base for recommendations
var candidatePlays = new List <Play>();
// Go through user's top plays
await ownTopPlays.ParallelForEachAsync(async ownTopPlay =>
{
// Get the map's top plays
var mapTopPlays =
(await _dataService.GetBeatmapTopPlaysAsync(ownTopPlay.BeatmapId, GameMode, ownTopPlay.Mods))
.OrderBy(p => Math.Abs(p.PerformancePoints - ownTopPlay.PerformancePoints)) // order by PP similarity
.Take(20) // only take top 20
.ToArray();
// Progress
progressHandler.Report(progress += 0.05 / ownTopPlays.Length);
// Go through those top plays
await mapTopPlays.ParallelForEachAsync(async mapTopPlay =>
{
// Get top plays of that user
var otherUserTopPlays = (await _dataService.GetUserTopPlaysAsync(mapTopPlay.PlayerId, GameMode))
.Where(p => p.Rank >= PlayRank.S) // only S ranks
.Where(p => p.PerformancePoints >= minPP) // limit by minPP
.Where(p => p.PerformancePoints <= maxPP) // limit by maxPP
.OrderBy(p => Math.Abs(p.PerformancePoints - ownTopPlay.PerformancePoints)) // order by PP similarity
.Take(20); // only take top 20
// Add these plays to candidates
candidatePlays.AddRange(otherUserTopPlays);
// Progress
progressHandler.Report(progress += 0.85 / mapTopPlays.Length / ownTopPlays.Length);
});
});
// Group candidate plays by beatmap
var candidatePlaysGroups = candidatePlays
.GroupBy(p => p.BeatmapId) // group
.Where(g => !ownTopPlaysBeatmapIds.Contains(g.Key)) // filter out maps that the user has top plays on
.OrderByDescending(g => g.Count()) // sort by number of times the map appears in candidate plays
.Take(200) // only take top 200
.ToArray();
// Assemble recommendations
var result = new List <Recommendation>();
await candidatePlaysGroups.ParallelForEachAsync(async group =>
{
var count = group.Count();
// Get median play based on PP
var play = group.OrderBy(p => p.PerformancePoints).ElementAt(count / 2);
// Get beatmap data
var beatmap = await _dataService.GetBeatmapAsync(play.BeatmapId, GameMode);
// Calculate traits with mods
var traitsWithMods = BeatmapTraitsLogic.CalculateTraitsWithMods(beatmap.Traits, play.Mods, GameMode);
// Add recommendation to the list
var recommendation = new Recommendation(beatmap, count, play.Mods, traitsWithMods, play.Accuracy, play.PerformancePoints);
result.Add(recommendation);
// Progress
progressHandler.Report(progress += 0.1 / candidatePlaysGroups.Length);
});
// Return recommendations sorted by weight
return(result.OrderByDescending(r => r.Weight).ToArray());
}
