A tool for extracting and archiving Last.fm music listening data - scrobbles.
Note:
- check out: the new Facets archiving Livebook
- previous analytics features have been migrated to coda, including on this day Livebook. Visit coda for future Lastfm analytics.
Download and create a file archive of Lastfm scrobble tracks via an Elixir
application or interactive Elixir
by invoking iex -S mix
command line action while in software home directory.
# archive all data of a default user specified in configuration
LastfmArchive.sync # subsequent calls download only latest scrobbles
# archive all data of any Lastfm user
# the data is stored in directory named after the user
LastfmArchive.sync("a_lastfm_user")
You can also deploy and use the tool in Livebook, as shown in various Livebook guides.
Scrobbles are downloaded via the Last.fm API and stored in the file archive on demand and on a daily basis. The software has a built-in cache to remember and resume from the previous downloads. It skips already downloaded scrobbles and does not make further requests to the API.
The stored data is in a raw Last.fm recenttracks
JSON format,
chunked into 200-track (max) gzip
compressed pages and stored within directories
corresponding to the days when tracks were scrobbled. The file archive in a main
directory specified in configuration - see below.
See Creating a file archive guide and sync/2
for various archiving options such overwrite
, year
, date
.
You can transform the file archive into other common storage formats such as CSV and columnar data structure such as Apache Parquet. These formats facilitate data interoperability, as well as OLAP, analytics use cases.
# transform the file archive into columnar Apache Parquet files
LastfmArchive.transform("a_lastfm_user", format: :parquet)
# to columnar Apache Arrow IPC files
LastfmArchive.transform("a_lastfm_user", format: :ipc_stream)
# CSV format also available
LastfmArchive.transform("a_lastfm_user", format: :csv)
Available formats:
- CSV (tab-delimited)
- Apache Arrow columnar format
- Apache Parquet columnar format
See Columnar data transforms guide and
transform/2
.
You can also transform the file archive into faceted (artists
, albums
, tracks
)
datasets.
LastfmArchive.transform("a_lastfm_user", format: :ipc_stream, facet: :artists)
See Facets archiving guide and
transform/2
.
The tool provides a read/2
function for retrieving data from the archive. It mainly relies on
Elixir Explorer data frame mechanisms
to underpin further data i/o, manipulation, analytics and visualisation.
The function returns a lazy t:Explorer.DataFrame.t/0
.
Scrobbles stored in the file archive can be read
with a day
or month
option:
# read a single-day scrobbles for the configured default user
LastfmArchive.read(day: ~D[2022-12-31])
# read a single-month scrobbles for a user with an arbitrary day of a month
LastfmArchive.read("a_lastfm_user", month: ~D[2022-12-01])
read/2
can
return a single-year or all scrobbles, i.e. the entire dataset from a columnar archive.
A columns
option is available to retrieve only a column subset.
# load all 2023 data from a Parquet archive
LastfmArchive.read("a_lastfm_user", format: :parquet, year: 2023)
# load all data from an Arrow IPC archive
LastfmArchive.read("a_lastfm_user", format: :ipc_stream)
# load data from specific columns
LastfmArchive.read("a_lastfm_user", format: :parquet, columns: [:id, :artist, :album])
read/2
can also
return the faceted datasets, e.g. all artists from a columnar archive.
LastfmArchive.read("a_lastfm_user", format: :ipc_stream, facet: :artists)
LastfmArchive
also provides the following Livebook interactive and step-by-step guides.
Creating a file archive guide for creating a local file archive consisting data fetched from the Last.fm API. It provides a heatmap and count visualisation for checking ongoing archiving status.
Columnar data transforms guide for transforming the local file archive to columnar data formats (Arrow, Parquet). It demonstrates how read/2
can be used to load single-year single-column data, as well as an entire dataset into data frame for various analytics.
See a sample output of this guide, showing top tracks analytics.
Facets archiving guide shows how the local file archive can generate faceted artists
, albums
, tracks
columnar datasets. It also demos how the datasets may be used. For example finding the new artists discovered on a particular date,
and visualising all artists, when their were first listened to and overall popularity.
To load all transformed CSV data from the archive into Solr:
# define a Solr endpoint with %Hui.URL{} struct
headers = [{"Content-type", "application/json"}]
url = %Hui.URL{url: "https://localhost:8983/solr/lastfm_archive", handler: "update", headers: headers}
LastfmArchive.load_archive("a_lastfm_user", url)
The function finds CSV files from the archive and send them to
Solr for ingestion one at a time. It uses Hui
client to interact
with Solr and the t:Hui.URL.t/0
struct for Solr endpoint specification.
This tool requires Elixir and Erlang, see installation details for various operating systems or Livebook.
lastfm_archive
is available in Hex,
the package can be installed by adding lastfm_archive
to your list of dependencies in mix.exs
:
def deps do
[
{:lastfm_archive, "~> 1.2"}
]
end
Documentation can be found at https://hexdocs.pm/lastfm_archive.
Add the following entries in your config - config/config.exs
. For example,
the following specifies an Lastfm user
and a main file location for
multiple user archives, ./lastfm_data/
relative to the software home directory.
You also need to specify an lastfm_api_key
in the config, so that the application can
access Lastfm API.
config :lastfm_archive,
user: "default_user", # the default user
data_dir: "./lastfm_data/", # main directory for multiple archives
lastfm_api_key: "api_key_provided_by_lastfm",
per_page: 200, # 200 is max no. of tracks per call permitted by Lastfm API
interval: 1000 # milliseconds between requests cf. Lastfm's max 5 reqs/s rate limit
# optional: Solr endpoint for Lastfm data loading
config :hui, :lastfm_archive,
url: "https://localhost:8983/solr/lastfm_archive",
handler: "update",
headers: [{"Content-type", "application/json"}]
See sync/2
for other configurable archiving options, e.g. interval
, per_page
.
See Hui
for more details on Solr configuration.
An api_key
must be configured to enable Lastfm API requests,
see https://www.last.fm/api ("Get an API account").