public Artist Recommend(SpotifyClient client, Playlist playlist)
{
//Find a common ancestor for a portion of the artists in the playlist.
IEnumerable<Track> playlistEntries = playlist.Tracks.Select(track => track.Track);
Dictionary<string, ArtistGroup> cachedRelatedArtists = new Dictionary<string, ArtistGroup>();
Dictionary<string, RecommendationCriteria> relatedArtistMap = new Dictionary<string, RecommendationCriteria>();
playlistEntries.ToList().ForEach(track =>
{
ArtistGroup relatedArtists = null;
if (!cachedRelatedArtists.ContainsKey(track.Artists.First().ID))
{
relatedArtists = client.GetRelatedArtists(track.Artists.First().ID); //Handle only the first artist for now...
cachedRelatedArtists.Add(track.Artists.First().ID, relatedArtists);
}
else
{
relatedArtists = cachedRelatedArtists[track.Artists.First().ID];
}
relatedArtists.Artists.ForEach(relatedArtist =>
{
if (!relatedArtistMap.ContainsKey(relatedArtist.ID))
{
RecommendationCriteria recommendationCriteria = new RecommendationCriteria();
relatedArtistMap.Add(relatedArtist.ID, recommendationCriteria);
recommendationCriteria.TargetedArtist = relatedArtist;
}
if (!relatedArtistMap[relatedArtist.ID].RelatedArtists.Where(artist => artist.ID == track.Artists.First().ID).Any())
relatedArtistMap[relatedArtist.ID].RelatedArtists.Add(track.Artists.First());
relatedArtistMap[relatedArtist.ID].Weight++;
});
});
//Other ideas:
//Recommend by the playlist contents.
// Evaluate whether the playlist is a collection of like-artists or individual songs of a particular type (e.g. energetic, fast?)
//Recommend by the artist selected.
//
//Recommend something completely new.
//Recommend a set of brand-new songs.
// Analyze all changelists to get an idea of what a person likes in particular.
// Easily solution is to find artists that they don't have that are similar.
// Could also randomly find songs from various areas:
// The current Top 10-40 on the charts.
// The list of a band that is playing near you live.
// Find a whole smattering of a different genre (trip hop, hip hop, grunge alternative).
//Recommend by a specific song.
// Find songs that were popular at the same time (may not need to be part of the same genre).
IOrderedEnumerable<KeyValuePair<string, RecommendationCriteria>> orderedDictionary = relatedArtistMap.OrderByDescending(keyValuePair => keyValuePair.Value.Weight)
.ThenByDescending(keyValuePair => keyValuePair.Value.RelatedArtists.Count());
KeyValuePair<string, RecommendationCriteria> recommendedEntry = orderedDictionary.First();
return recommendedEntry.Value.TargetedArtist;
}
public Artist Recommend(SpotifyClient client, Playlist playlist)
{
//Find a common ancestor for a portion of the artists in the playlist.
IEnumerable <Track> playlistEntries = playlist.Tracks.Select(track => track.Track);
Dictionary <string, ArtistGroup> cachedRelatedArtists = new Dictionary <string, ArtistGroup>();
Dictionary <string, RecommendationCriteria> relatedArtistMap = new Dictionary <string, RecommendationCriteria>();
playlistEntries.ToList().ForEach(track =>
{
ArtistGroup relatedArtists = null;
if (!cachedRelatedArtists.ContainsKey(track.Artists.First().ID))
{
relatedArtists = client.GetRelatedArtists(track.Artists.First().ID); //Handle only the first artist for now...
cachedRelatedArtists.Add(track.Artists.First().ID, relatedArtists);
}
else
{
relatedArtists = cachedRelatedArtists[track.Artists.First().ID];
}
relatedArtists.Artists.ForEach(relatedArtist =>
{
if (!relatedArtistMap.ContainsKey(relatedArtist.ID))
{
RecommendationCriteria recommendationCriteria = new RecommendationCriteria();
relatedArtistMap.Add(relatedArtist.ID, recommendationCriteria);
recommendationCriteria.TargetedArtist = relatedArtist;
}
if (!relatedArtistMap[relatedArtist.ID].RelatedArtists.Where(artist => artist.ID == track.Artists.First().ID).Any())
{
relatedArtistMap[relatedArtist.ID].RelatedArtists.Add(track.Artists.First());
}
relatedArtistMap[relatedArtist.ID].Weight++;
});
});
//Other ideas:
//Recommend by the playlist contents.
// Evaluate whether the playlist is a collection of like-artists or individual songs of a particular type (e.g. energetic, fast?)
//Recommend by the artist selected.
//
//Recommend something completely new.
//Recommend a set of brand-new songs.
// Analyze all changelists to get an idea of what a person likes in particular.
// Easily solution is to find artists that they don't have that are similar.
// Could also randomly find songs from various areas:
// The current Top 10-40 on the charts.
// The list of a band that is playing near you live.
// Find a whole smattering of a different genre (trip hop, hip hop, grunge alternative).
//Recommend by a specific song.
// Find songs that were popular at the same time (may not need to be part of the same genre).
IOrderedEnumerable <KeyValuePair <string, RecommendationCriteria> > orderedDictionary = relatedArtistMap.OrderByDescending(keyValuePair => keyValuePair.Value.Weight)
.ThenByDescending(keyValuePair => keyValuePair.Value.RelatedArtists.Count());
KeyValuePair <string, RecommendationCriteria> recommendedEntry = orderedDictionary.First();
return(recommendedEntry.Value.TargetedArtist);
}
