Spotify Stream History Analyis

# Calculate total listening hours by year
by_year <- stream %>%
  group_by(year = as.factor(year(date))) %>%
  summarise(h_played = sum(ms_played)/1000/60/60,
            n_tracks=n(),
            n_songs=n_distinct(n_tracks),
            n_artists=n_distinct(artist_name)) 

# Display table
by_year%>%
  transmute(
    Year=year,
    `Hours played`=round(h_played,2),
    `N tracks`= n_tracks,
    `N songs` = n_songs,
    `N artists` = n_artists
  )%>%
kable(caption="Spotify listened by year")

# Calculate total listening hours by year and device
by_device <- stream %>%
  group_by(Device, year = as.factor(year(date))) %>%
  summarise(h_played = sum(ms_played)/1000/60/60) 

# Create stacked bar chart of listening hours by device and year
plot_device<-by_device%>%
  ggplot(aes(x = year, y = h_played, fill = Device)) +
  geom_col(position = position_stack(reverse = TRUE)) +  # Stack bars with reverse order
  scale_fill_manual(values = pal, na.value = "grey80") + # Custom color palette
  theme_minimal() +
  ggtitle("Hours listening to Spotify") +
  ylab("")

# Display plot
plot_device

# Save plot
if(save_svg) ggsave(paste0("plot_device",".svg"), path = "output_files/")

# Create a dataframe with all days in the date range
all_days <- data.frame(
  day_date = seq(as.Date(floor_date(min(stream$date), unit="year")),
                as.Date(ceiling_date(max(stream$date), unit="year")-1), 
                by="days")) %>%
  mutate(day_n = row_number())

# Calculate first weekday of each year
first_week_day <- all_days %>%
  mutate(weekday = as.POSIXlt(day_date)$wday,
         weekday = ifelse(weekday==0, 7, weekday),
         year = year(day_date)) %>%
  group_by(year) %>%
  summarize(first_week_day = first(weekday)-1,
            days_year = n())

# Process streaming data by day
by_day_year <- stream %>%
  mutate(day_date = date(stream$date)) %>%
  right_join(all_days) %>%
  group_by(day_date, day_n) %>%
  # Calculate daily metrics
  summarise(
    h_played = sum(ms_played)/1000/60/60,  # Convert ms to hours
    n_tracks = n(),
    computer = mean(Device=="Computer", na.rm=TRUE),
    shuffle = mean(shuffle, na.rm=TRUE),
    incognito_mode = mean(incognito_mode, na.rm=TRUE),
    offline = mean(offline, na.rm=TRUE),
    h_top_artist = sum(ifelse(grepl(top_artist,artist_name), h_played, 0), na.rm=TRUE)
  ) %>%
  # Add calendar columns
  mutate(
    h_played = ifelse(is.na(h_played), 0, h_played),
    year = year(day_date),
    month = month(day_date),
    week = week(day_date),
    day_year = yday(day_date),
    weekday = as.POSIXlt(day_date)$wday,
    weekday = factor(ifelse(weekday==0, 7, weekday), levels=c(1:7))
  ) %>%
  left_join(first_week_day) %>%
  mutate(
    calendar_row = ceiling((first_week_day + day_year)/7)
  )

# Create calendar heatmap
plot_day_year<-by_day_year %>%
  ggplot(aes(x = weekday, y = -calendar_row, fill=h_played)) +
  geom_tile() +
  facet_grid(cols=vars(year), rows=vars(month), scales = "free") +
  theme_void() +
  scale_fill_gradientn(colors=c("grey95", pal[c(1,3,2)]), na.value ="grey95") +
  ggtitle("Hours listened per day") +
  theme(legend.title=element_blank()) +
  ylab("")

# Display plot
plot_day_year

# Save plot if save_svg is TRUE
if(save_svg) ggsave(paste0("plot_day_year",".svg"), width=7, height=8, path="output_files/")

# Other calendars - not included in the automatic report

# Plot for computer usage
by_day_year %>%
  ggplot(aes(x = weekday, y = -calendar_row, fill=computer)) +
  geom_tile() +
  facet_grid(cols=vars(year), rows=vars(month), scales = "free") +
  theme_void() +
  scale_fill_gradientn(colors=c("grey95", pal[c(1,3,2)]), na.value="grey95") +
  ggtitle("% of tracks listened on the computer per day") +
  ylab("")

# Plot for incognito mode usage
by_day_year %>%
  ggplot(aes(x = weekday, y = -calendar_row, fill=incognito_mode)) +
  geom_tile() +
  facet_grid(cols=vars(year), rows=vars(month), scales = "free") +
  theme_void() +
  scale_fill_gradientn(colors=c("grey95", pal[c(1,3,2)]), na.value="grey95") +
  ggtitle("% of tracks listened in incognito mode") +
  ylab("")

# Plot for offline mode
by_day_year %>%
  ggplot(aes(x = weekday, y = -calendar_row, fill=offline)) +
  geom_tile() +
  facet_grid(cols=vars(year), rows=vars(month), scales = "free") +
  theme_void() +
  scale_fill_gradientn(colors=c("grey95", pal[c(1,3,2)]), na.value="grey95") +
  ggtitle("% of tracks listened offline") +
  ylab("")

# Plot for top artist listening patterns
by_day_year %>%
  ggplot(aes(x = weekday, y = -calendar_row, fill=h_top_artist)) +
  geom_tile() +
  facet_grid(cols=vars(year), rows=vars(month), scales = "free") +
  theme_void() +
  scale_fill_gradientn(colors=c("grey95", pal[c(1,3,2)]), na.value="grey95") +
  ggtitle(paste0("Hours listened to ",top_artist," per day")) +
  ylab("")

# Create summary of number of times a track was listened in a day
by_top_track_day <- stream %>%
  mutate(day_date = date(stream$date)) %>%
  group_by(track_name, day_date) %>%
  summarise(n_tracks = n())%>%
  ungroup() %>%
  arrange(desc(n_tracks))


# Get top 10 days by hours played
by_top_day <- stream %>%
  mutate(day_date = date(stream$date)) %>%
  group_by(day_date) %>%
  summarise(
    n_tracks_all = n(),
    computer = mean(Device == "Computer"),
    shuffle = mean(shuffle),
    h_played = sum(ms_played)/1000/60/60
  ) %>%
  ungroup() %>%
  arrange(desc(h_played)) %>%
  slice_max(h_played, n = 10)
  
# Join the data and calculate final metrics
by_top_day <-by_top_day %>%
  left_join(by_top_track_day) %>%
  group_by(day_date) %>%
  summarize(mult_times = mean(n_tracks > 1)) %>%
  left_join(by_top_day) %>%
  arrange(desc(h_played)) %>%
  relocate(day_date, h_played, n_tracks = n_tracks_all)

# Display table
by_top_day%>%
  transmute(
    Date=day_date,
    Hours=round(h_played,2),
    `N tracks`=n_tracks,
    `% tracks replayed`=paste0(round(mult_times,4)*100, "%"),
    `% tracks on the computer`=paste0(round(computer,4)*100, "%"),
    `% tracks on shuffle`=paste0(round(computer,4)*100, "%"))%>%
  kable(caption="Top 10 days by listening hours")

# Create a dataframe with all possible hours for each day
all_hours <- all_days %>% expand(day_date, hour=1:24)


# Calculate average listening time by hour and year
by_hour_year <- stream %>%
  mutate(day_date = date(stream$date),
         hour = hour(date)) %>%
  right_join(all_hours) %>% 
  group_by(day_date, hour) %>%
  mutate(ms_played = ifelse(is.na(ms_played), 0, ms_played)) %>%
  summarise(min_played = sum(ms_played/1000/60)) %>%
  group_by(year = year(day_date), hour) %>%
  summarise(min_played = mean(min_played)) 
  

# Create visualization of average listening time by hour and year
plot_hour_year <- by_hour_year %>%
  ggplot(aes(x = year, y = hour, fill = min_played)) +
  geom_tile() +
  theme_void() +
  facet_grid(cols = vars(year), rows = vars(hour), scales = "free") +
  scale_fill_gradientn(colors = c("grey95", pal[c(1,3,2)])) +
  ggtitle("Minutes listened per hour of the day (on average)") +
  theme(legend.title = element_blank(),
        axis.text.x = element_blank(), 
        axis.ticks.x = element_blank())

# Display plot
plot_hour_year

# Save plot
if(save_svg) ggsave(paste0("plot_hour_year",".svg"), width = 7, height = 5, path = "output_files/")

# Calculate statistics by year
by_year <- stream %>%
  group_by(year = year(date)) %>%
  summarise(
    min_played_year = sum(ms_played)/1000/60,
    n_tracks_year = n(),
    n_songs_year = n_distinct(track_name)
  )

# Calculate statistics by artist and year
by_artist_year <- stream %>%
  group_by(artist_name, year = year(date)) %>%
  summarise(
    min_played = sum(ms_played/1000/60),
    n_tracks = n(),
    n_songs = n_distinct(track_name),
    p_tracks = n()/sum(n_tracks)
  ) %>%
  # Split multiple artists into separate rows
  separate_wider_delim(
    artist_name, 
    ", ",
    names = paste0("artist_name", 1:5),
    too_few = "align_start",
    too_many = "drop"
  ) %>%
  pivot_longer(
    cols = starts_with("artist_name"),
    names_to = "artist_name_n",
    values_to = "artist_name"
  ) %>%
  left_join(by_year) %>%
  filter(!is.na(artist_name)) %>%
  group_by(artist_name, year) %>%
  summarise(
    min_played = sum(min_played),
    n_tracks = sum(n_tracks),
    n_songs = sum(n_songs),
    p_tracks = n_tracks/n_tracks_year,
    p_min_played = min_played/min_played_year,
    p_songs = n_songs/n_songs_year
  ) %>%
  arrange(year,desc(n_tracks))

by_artist_year %>%
  group_by(Year = year) %>%
  slice_max(n_tracks, n = 3)%>%
  transmute(
    Artist = artist_name,
    `Minutes played`= round(min_played,2),
    `N tracks` = n_tracks,
    `N songs` = n_songs,
    `% of the annual time` = paste0(round(p_min_played,4)*100, "%"),
    `% of the annual tracks` = paste0(round(p_tracks,4)*100, "%"),
    `% of the annual songs` = paste0(round(p_songs,4)*100, "%"),
  )%>%
  kable(caption = "Top 3 artists by year")

# Get top 40 artists by year
top_artist_year <- by_artist_year %>%
  ungroup() %>%
  slice_max(n_tracks, n = 40) %>%
  group_by(artist_name) %>%
  summarise(n_tracks_all = sum(n_tracks),
            min_played_all = sum(min_played),
            n_songs_all=sum(n_songs)) %>%
  select(artist_name,min_played_all, n_tracks_all, n_songs_all) %>%
  distinct()

# Calculate statistics by artist
by_artist_all <- stream %>%
  # Split multiple artists into separate rows
  separate_wider_delim(
    artist_name, 
    ", ",
    names = paste0("artist_name", 1:5),
    too_few = "align_start",
    too_many = "drop"
  ) %>%
  pivot_longer(
    cols = starts_with("artist_name"),
    names_to = "artist_name_n",
    values_to = "artist_name"
  ) %>%
  group_by(artist_name) %>%
  summarise(
    min_played_all = sum(ms_played/1000/60),
    n_tracks_all = n(),
    n_songs_all = n_distinct(track_name),
    p_tracks_all = n()/sum(n())
  ) %>%
  filter(!is.na(artist_name)) %>%
  arrange(desc(n_tracks_all))

# Get top 40 artists overall
top_artist_all<-by_artist_all %>%
  slice_max(n_tracks_all, n = 40) %>%
  select(artist_name,min_played_all, n_tracks_all, n_songs_all)


# Display table
top_artist_all[1:10,]%>%
  transmute(Artist=artist_name,
            `Minutes played`= round(min_played_all,2),
            `N tracks` = n_tracks_all,
            `N songs` = n_songs_all)%>%
  kable(caption = "Top 10 artists (all years)")

# Get top 40 artists overall or by year
top_artist_global<-unique(c(top_artist_year$artist_name, top_artist_all$artist_name))


# Join top artists data and filter top artists
by_top_artist <- by_artist_year %>%
  filter(artist_name%in%top_artist_global) %>%
  left_join(by_artist_all) %>%
  arrange(desc(n_tracks)) 

# Read and parse playlist data from JSON file
playlist <- fromJSON("input_files/Playlist1.json", flatten = TRUE)[[1]]

playlist_tracks <- data.frame()

# Loop through each playlist and extract tracks
for(i in 1:nrow(playlist)){
  playlist_tracks_i <- playlist$items[[i]]
  playlist_tracks_i$playlist_name <- playlist$name[i]
  playlist_tracks <- bind_rows(playlist_tracks, playlist_tracks_i)
}

# Read library data from JSON file
library <- fromJSON("input_files/YourLibrary.json", flatten = TRUE)[[1]]

# Process playlist tracks data
playlist_tracks <- playlist_tracks %>%
  mutate(
    added_date = as_datetime(addedDate, tz = "UTC"),
    artist_name = track.artistName,
    track_name = track.trackName,
    album_name = track.albumName
  )

# Get most common tracks in playlists
playlist_top_tracks <- playlist_tracks %>%
  group_by(track_name,artist_name) %>%
  summarise(n_playlists = n()) %>%
  arrange(desc(n_playlists))%>%
  ungroup()

# Display table
playlist_top_tracks[1:10,]%>%
  transmute(
    `Track`=track_name,
    `Artist` = artist_name,
    `N playlists` = n_playlists,
  )%>%
  kable(caption="Tracks included in more playlists")

# Get most common artists in playlists
playlist_top_artist<-playlist_tracks%>%
  group_by(artist_name)%>%
  summarise(n_tracks=n(),
            n_songs=n_distinct(track_name),
            n_playlists=n_distinct(playlist_name))%>%
  arrange(desc(n_playlists))%>%
  ungroup()

# Display table
playlist_top_artist[1:10,]%>%
  transmute(
    `Artist` = artist_name,
    `N tracks` = n_tracks,
    `N songs` = n_songs,
    `N playlists` = n_playlists,
  )%>%
  kable(caption="Artists included in more playlists")

# Get most common artists in library (saved songs)
library_tracks<-library%>%
  mutate(artist_name = artist,
         track_name = track,
         album_name = album)

library_top_artist<-library_tracks%>%
  group_by(artist_name)%>%
  summarise(n_songs=n_distinct(track_name))%>%
  arrange(desc(n_songs))%>%
  ungroup()

# Display table
library_top_artist[1:10,]%>%
  transmute(
    `Artist` = artist_name,
    `N songs` = n_songs,
  )%>%
  kable(caption="Artists with more songs in library")

# Combine tracks from playlists and library 
saved_tracks <- unique(c(playlist_top_tracks$track_name, library_tracks$track_name))

# Combine artists from playlists and library
saved_artist <- unique(c(playlist_top_artist$artist_name, library_top_artist$artist_name))

# Plot proportion of tracks in playlists by year
plot_saved_track <- stream %>%
  mutate(`In my playlists` = ifelse(track_name %in% saved_tracks, "Yes", "No")) %>%
  group_by(`In my playlists`, year = as.factor(year(date))) %>%
  summarise(n = n()) %>%
  ggplot(aes(x = year, y = n, fill = `In my playlists`)) +
  geom_col(position = "fill") +
  scale_fill_manual(values = pal, na.value = "grey80") +
  scale_y_continuous(labels = function(x) paste0(x*100, "%")) +
  theme_minimal() +
  ggtitle("Were the tracks (streaming history) in the current playlists?") +
  ylab("")

# Display plot
plot_saved_track

# Save plot
if(save_svg) ggsave(paste0("plot_saved_track",".svg"), path = "output_files/")

# Plot proportion of artists in playlists by year
plot_saved_artist <- stream %>%
  mutate(`In my playlists`=ifelse(artist_name%in%saved_artist,"Yes","No"))%>%
  group_by(`In my playlists`, year = as.factor(year(date))) %>%
  summarise(n = n()) %>%
  ggplot(aes(x = year, y = n, fill = `In my playlists`)) +
  geom_col(position = "fill") +
  scale_fill_manual(values = pal, na.value = "grey80") +
  scale_y_continuous(labels = function(x) paste0(x*100, "%"))+
  theme_minimal() +
  ggtitle("Were the artists (streaming history) in the current playlists?")+
  ylab("")


plot_saved_artist

# Save plot
if(save_svg) ggsave(paste0("plot_saved_artist",".svg"), path = "output_files/")