create graphs of listenbrainz data

listenbrainz.R

library(dplyr)
library(ggplot2)
library(jsonlite)
library(lubridate)
library(stringr)

# load data ---------------------------------------------------------------

data <- fromJSON("ercas_lb-2021-07-21.json") %>%
  flatten() %>%
  mutate(
    listened_at = with_tz(
      as_datetime(listened_at),
      system("readlink -f /etc/localtime | grep -o '[^/]*/[^/]*$'", intern = TRUE)
    )
  ) %>%
  filter(listened_at >= as_datetime("2020-01-01"))

# top artists -------------------------------------------------------------

n_artists <- 10

top_artists <- data$track_metadata.artist_name %>%
  table() %>%
  as.data.frame() %>%
  arrange(desc(Freq)) %>%
  .$. %>%
  head(n_artists)

data %>%
  mutate(
    Artist = ifelse(
      #track_metadata.artist_name %in% top_artists,
      #track_metadata.artist_name,
      str_detect(track_metadata.artist_name, paste(top_artists, collapse = "|")),
      str_extract(track_metadata.artist_name, paste(top_artists, collapse = "|")),
      "All others"
    )
  ) %>%
  ggplot() +
    aes(x = listened_at, fill = Artist) +
    geom_density(position = "stack", color = NA) +
    labs(
      x = "Date",
      y = "Density",
      title = sprintf("Top %d most listened-to artists over time", n_artists)
    ) +
    scale_fill_brewer(palette = "Spectral") +
    theme_minimal() +
    scale_y_continuous(breaks = NULL)

# top albums --------------------------------------------------------------

n_albums <- 10

top_albums <- data$track_metadata.release_name %>%
  table() %>%
  as.data.frame() %>%
  arrange(desc(Freq)) %>%
  .$. %>%
  head(n_albums)

data %>%
  mutate(
    Album = ifelse(
      track_metadata.release_name %in% top_albums,
      ifelse(
        str_starts(track_metadata.release_name, "The Idler Wheel"),
        "The Idler Wheel (...)",
        track_metadata.release_name
      ),
      "All others"
    )
  ) %>%
  ggplot() +
    aes(x = listened_at, fill = Album) +
    geom_density(position = "stack", color = NA) +
    labs(
      x = "Date",
      y = "Density",
      title = sprintf("Top %d most listened-to albums over time", n_albums)
    ) +
    scale_fill_brewer(palette = "Spectral") +
    theme_minimal() +
    scale_y_continuous(breaks = NULL)

# top listening times -----------------------------------------------------

weeks_since_epoch <- function(dt) {
  return(floor(as.numeric(dt) / (60*60*24*7)))
}

data %>%
  mutate(
    week = weeks_since_epoch(listened_at),
    hour = hour(listened_at)
  ) %>%
  group_by(week, hour) %>%
  summarize(
    week_of = floor_date(min(listened_at) - 1, "weeks") + 1,
    listens = n()
  ) %>%
  ggplot() +
    aes(week_of, hour, fill = log(listens)) +
    geom_tile() +
    scale_fill_viridis_c() +
    labs(
      x = "Week",
      y = "Hour",
      title = "Most active listening times (note log scale)"
    ) +
    scale_x_datetime(expand = c(0, 0)) +
    scale_y_continuous(expand = c(0, 0)) +
    theme(
      # viridis: https://github.com/BIDS/colormap/blob/master/colormaps.py#L788
      panel.background = element_rect(fill = rgb(0.267004, 0.004874, 0.329415)),
      panel.grid = element_blank()
    )