Merge pull request #41 from Polochon-street/add-actually-useful-library-struct

Add a proper library / config example
This commit is contained in:
Polochon-street 2022-09-28 23:15:39 +02:00 committed by GitHub
commit bf428a62af
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
11 changed files with 4218 additions and 244 deletions

View File

@ -30,14 +30,16 @@ jobs:
run: cargo build --verbose
- name: Run tests
run: cargo test --verbose
- name: Run library tests
run: cargo test --verbose --features=library
- name: Run example tests
run: cargo test --verbose --examples
- name: Build benches
run: cargo +nightly-2022-02-16 bench --verbose --features=bench --no-run
- name: Build examples
run: cargo build --examples --verbose --features=serde
run: cargo build --examples --verbose --features=serde,library
- name: Lint
run: cargo clippy --examples --features=serde -- -D warnings
run: cargo clippy --examples --features=serde,library -- -D warnings
build-test-lint-windows:
name: Windows - build, test and lint

635
Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@ -11,7 +11,7 @@ keywords = ["audio", "analysis", "MIR", "playlist", "similarity"]
readme = "README.md"
[package.metadata.docs.rs]
features = ["bliss-audio-aubio-rs/rustdoc"]
features = ["bliss-audio-aubio-rs/rustdoc", "library"]
no-default-features = true
[features]
@ -23,6 +23,12 @@ ffmpeg-static = ["ffmpeg-next/static"]
python-bindings = ["bliss-audio-aubio-rs/fftw3"]
# Enable the benchmarks with `cargo +nightly bench --features=bench`
bench = []
library = [
"serde", "dep:rusqlite", "dep:dirs", "dep:tempdir",
"dep:anyhow", "dep:serde_ini", "dep:serde_json",
"dep:indicatif",
]
serde = ["dep:serde"]
[dependencies]
ripemd160 = "0.9.0"
@ -44,13 +50,35 @@ thiserror = "1.0.24"
bliss-audio-aubio-rs = "0.2.0"
strum = "0.21"
strum_macros = "0.21"
serde = { version = "1.0", optional = true, features = ["derive"] }
rcue = "0.1.1"
# Deps for the library feature
serde = { version = "1.0", optional = true, features = ["derive"] }
serde_json = { version = "1.0.59", optional = true }
serde_ini = { version = "0.2.0", optional = true }
tempdir = { version = "0.3.7", optional = true }
rusqlite = { version = "0.27.0", optional = true }
dirs = { version = "4.0.0", optional = true }
anyhow = { version = "1.0.58", optional = true }
indicatif = { version = "0.17.0", optional = true }
[dev-dependencies]
mime_guess = "2.0.3"
glob = "0.3.0"
anyhow = "1.0.45"
serde_json = "1.0.59"
clap = "2.33.3"
pretty_assertions = "1.2.1"
serde_json = "1.0.59"
[[example]]
name = "library"
required-features = ["library"]
[[example]]
name = "library_extra_info"
required-features = ["library"]
[[example]]
name = "playlist"
required-features = ["serde"]

202
examples/library.rs Normal file
View File

@ -0,0 +1,202 @@
/// Basic example of how one would combine bliss with an "audio player",
/// through [Library].
///
/// For simplicity's sake, this example recursively gets songs from a folder
/// to emulate an audio player library, without handling CUE files.
use anyhow::Result;
use bliss_audio::library::{AppConfigTrait, BaseConfig, Library};
use clap::{App, Arg, SubCommand};
use glob::glob;
use serde::{Deserialize, Serialize};
use std::fs;
use std::path::{Path, PathBuf};
#[derive(Serialize, Deserialize, Clone, Debug)]
// A config structure, that will be serialized as a
// JSON file upon Library creation.
pub struct Config {
#[serde(flatten)]
// The base configuration, containing both the config file
// path, as well as the database path.
pub base_config: BaseConfig,
// An extra field, to store the music library path. Any number
// of arbitrary fields (even Serializable structures) can
// of course be added.
pub music_library_path: PathBuf,
}
impl Config {
pub fn new(
music_library_path: PathBuf,
config_path: Option<PathBuf>,
database_path: Option<PathBuf>,
) -> Result<Self> {
let base_config = BaseConfig::new(config_path, database_path)?;
Ok(Self {
base_config,
music_library_path,
})
}
}
// The AppConfigTrait must know how to access the base config.
impl AppConfigTrait for Config {
fn base_config(&self) -> &BaseConfig {
&self.base_config
}
fn base_config_mut(&mut self) -> &mut BaseConfig {
&mut self.base_config
}
}
// A trait allowing to implement methods for the Library,
// useful if you don't need to store extra information in fields.
// Otherwise, doing
// ```
// struct CustomLibrary {
// library: Library<Config>,
// extra_field: ...,
// }
// ```
// and implementing functions for that struct would be the way to go.
// That's what the [reference](https://github.com/Polochon-street/blissify-rs)
// implementation does.
trait CustomLibrary {
fn song_paths(&self) -> Result<Vec<String>>;
}
impl CustomLibrary for Library<Config> {
/// Get all songs in the player library
fn song_paths(&self) -> Result<Vec<String>> {
let music_path = &self.config.music_library_path;
let pattern = Path::new(&music_path).join("**").join("*");
Ok(glob(&pattern.to_string_lossy())?
.map(|e| fs::canonicalize(e.unwrap()).unwrap())
.filter(|e| match mime_guess::from_path(e).first() {
Some(m) => m.type_() == "audio",
None => false,
})
.map(|x| x.to_string_lossy().to_string())
.collect::<Vec<String>>())
}
}
// A simple example of what a CLI-app would look.
//
// Note that `Library::new` is used only on init, and subsequent
// commands use `Library::from_path`.
fn main() -> Result<()> {
let matches = App::new("library-example")
.version(env!("CARGO_PKG_VERSION"))
.author("Polochon_street")
.about("Example binary implementing bliss for an audio player.")
.subcommand(
SubCommand::with_name("init")
.about(
"Initialize a Library, both storing the config and analyzing folders
containing songs.",
)
.arg(
Arg::with_name("FOLDER")
.help("A folder containing the music library to analyze.")
.required(true),
)
.arg(
Arg::with_name("database-path")
.short("d")
.long("database-path")
.help(
"Optional path where to store the database file containing
the songs' analysis. Defaults to XDG_DATA_HOME/bliss-rs/bliss.db.",
)
.takes_value(true),
)
.arg(
Arg::with_name("config-path")
.short("c")
.long("config-path")
.help(
"Optional path where to store the config file containing
the library setup. Defaults to XDG_DATA_HOME/bliss-rs/config.json.",
)
.takes_value(true),
),
)
.subcommand(
SubCommand::with_name("update")
.about(
"Update a Library's songs, trying to analyze failed songs,
as well as songs not in the library.",
)
.arg(
Arg::with_name("config-path")
.short("c")
.long("config-path")
.help(
"Optional path where to load the config file containing
the library setup. Defaults to XDG_DATA_HOME/bliss-rs/config.json.",
)
.takes_value(true),
),
)
.subcommand(
SubCommand::with_name("playlist")
.about(
"Make a playlist, starting with the song at SONG_PATH, returning
the songs' paths.",
)
.arg(Arg::with_name("SONG_PATH").takes_value(true))
.arg(
Arg::with_name("config-path")
.short("c")
.long("config-path")
.help(
"Optional path where to load the config file containing
the library setup. Defaults to XDG_DATA_HOME/bliss-rs/config.json.",
)
.takes_value(true),
)
.arg(
Arg::with_name("playlist-length")
.short("l")
.long("playlist-length")
.help("Optional playlist length. Defaults to 20.")
.takes_value(true),
),
)
.get_matches();
if let Some(sub_m) = matches.subcommand_matches("init") {
let folder = PathBuf::from(sub_m.value_of("FOLDER").unwrap());
let config_path = sub_m.value_of("config-path").map(PathBuf::from);
let database_path = sub_m.value_of("database-path").map(PathBuf::from);
let config = Config::new(folder, config_path, database_path)?;
let mut library = Library::new(config)?;
library.analyze_paths(library.song_paths()?, true)?;
} else if let Some(sub_m) = matches.subcommand_matches("update") {
let config_path = sub_m.value_of("config-path").map(PathBuf::from);
let mut library: Library<Config> = Library::from_config_path(config_path)?;
library.update_library(library.song_paths()?, true)?;
} else if let Some(sub_m) = matches.subcommand_matches("playlist") {
let song_path = sub_m.value_of("SONG_PATH").unwrap();
let config_path = sub_m.value_of("config-path").map(PathBuf::from);
let playlist_length = sub_m
.value_of("playlist-length")
.unwrap_or("20")
.parse::<usize>()?;
let library: Library<Config> = Library::from_config_path(config_path)?;
let songs = library.playlist_from::<()>(song_path, playlist_length)?;
let song_paths = songs
.into_iter()
.map(|s| s.bliss_song.path.to_string_lossy().to_string())
.collect::<Vec<String>>();
for song in song_paths {
println!("{:?}", song);
}
}
Ok(())
}

View File

@ -0,0 +1,225 @@
/// Basic example of how one would combine bliss with an "audio player",
/// through [Library], showing how to put extra info in the database for
/// each song.
///
/// For simplicity's sake, this example recursively gets songs from a folder
/// to emulate an audio player library, without handling CUE files.
use anyhow::Result;
use bliss_audio::library::{AppConfigTrait, BaseConfig, Library};
use clap::{App, Arg, SubCommand};
use glob::glob;
use serde::{Deserialize, Serialize};
use std::fs;
use std::path::{Path, PathBuf};
#[derive(Serialize, Deserialize, Clone, Debug)]
/// A config structure, that will be serialized as a
/// JSON file upon Library creation.
pub struct Config {
#[serde(flatten)]
/// The base configuration, containing both the config file
/// path, as well as the database path.
pub base_config: BaseConfig,
/// An extra field, to store the music library path. Any number
/// of arbitrary fields (even Serializable structures) can
/// of course be added.
pub music_library_path: PathBuf,
}
impl Config {
pub fn new(
music_library_path: PathBuf,
config_path: Option<PathBuf>,
database_path: Option<PathBuf>,
) -> Result<Self> {
let base_config = BaseConfig::new(config_path, database_path)?;
Ok(Self {
base_config,
music_library_path,
})
}
}
// The AppConfigTrait must know how to access the base config.
impl AppConfigTrait for Config {
fn base_config(&self) -> &BaseConfig {
&self.base_config
}
fn base_config_mut(&mut self) -> &mut BaseConfig {
&mut self.base_config
}
}
// A trait allowing to implement methods for the Library,
// useful if you don't need to store extra information in fields.
// Otherwise, doing
// ```
// struct CustomLibrary {
// library: Library<Config>,
// extra_field: ...,
// }
// ```
// and implementing functions for that struct would be the way to go.
// That's what the [reference](https://github.com/Polochon-street/blissify-rs)
// implementation does.
trait CustomLibrary {
fn song_paths_info(&self) -> Result<Vec<(String, ExtraInfo)>>;
}
impl CustomLibrary for Library<Config> {
/// Get all songs in the player library, along with the extra info
/// one would want to store along with each song.
fn song_paths_info(&self) -> Result<Vec<(String, ExtraInfo)>> {
let music_path = &self.config.music_library_path;
let pattern = Path::new(&music_path).join("**").join("*");
Ok(glob(&pattern.to_string_lossy())?
.map(|e| fs::canonicalize(e.unwrap()).unwrap())
.filter_map(|e| {
mime_guess::from_path(&e).first().map(|m| {
(
e.to_string_lossy().to_string(),
ExtraInfo {
extension: e.extension().map(|e| e.to_string_lossy().to_string()),
file_name: e.file_name().map(|e| e.to_string_lossy().to_string()),
mime_type: format!("{}/{}", m.type_(), m.subtype()),
},
)
})
})
.collect::<Vec<(String, ExtraInfo)>>())
}
}
#[derive(Deserialize, Serialize, Debug, PartialEq, Clone, Default)]
// An (somewhat simple) example of what extra metadata one would put, along
// with song analysis data.
struct ExtraInfo {
extension: Option<String>,
file_name: Option<String>,
mime_type: String,
}
// A simple example of what a CLI-app would look.
//
// Note that `Library::new` is used only on init, and subsequent
// commands use `Library::from_path`.
fn main() -> Result<()> {
let matches = App::new("library-example")
.version(env!("CARGO_PKG_VERSION"))
.author("Polochon_street")
.about("Example binary implementing bliss for an audio player.")
.subcommand(
SubCommand::with_name("init")
.about(
"Initialize a Library, both storing the config and analyzing folders
containing songs.",
)
.arg(
Arg::with_name("FOLDER")
.help("A folder containing the music library to analyze.")
.required(true),
)
.arg(
Arg::with_name("database-path")
.short("d")
.long("database-path")
.help(
"Optional path where to store the database file containing
the songs' analysis. Defaults to XDG_DATA_HOME/bliss-rs/bliss.db.",
)
.takes_value(true),
)
.arg(
Arg::with_name("config-path")
.short("c")
.long("config-path")
.help(
"Optional path where to store the config file containing
the library setup. Defaults to XDG_DATA_HOME/bliss-rs/config.json.",
)
.takes_value(true),
),
)
.subcommand(
SubCommand::with_name("update")
.about(
"Update a Library's songs, trying to analyze failed songs,
as well as songs not in the library.",
)
.arg(
Arg::with_name("config-path")
.short("c")
.long("config-path")
.help(
"Optional path where to load the config file containing
the library setup. Defaults to XDG_DATA_HOME/bliss-rs/config.json.",
)
.takes_value(true),
),
)
.subcommand(
SubCommand::with_name("playlist")
.about(
"Make a playlist, starting with the song at SONG_PATH, returning
the songs' paths.",
)
.arg(Arg::with_name("SONG_PATH").takes_value(true))
.arg(
Arg::with_name("config-path")
.short("c")
.long("config-path")
.help(
"Optional path where to load the config file containing
the library setup. Defaults to XDG_DATA_HOME/bliss-rs/config.json.",
)
.takes_value(true),
)
.arg(
Arg::with_name("playlist-length")
.short("l")
.long("playlist-length")
.help("Optional playlist length. Defaults to 20.")
.takes_value(true),
),
)
.get_matches();
if let Some(sub_m) = matches.subcommand_matches("init") {
let folder = PathBuf::from(sub_m.value_of("FOLDER").unwrap());
let config_path = sub_m.value_of("config-path").map(PathBuf::from);
let database_path = sub_m.value_of("database-path").map(PathBuf::from);
let config = Config::new(folder, config_path, database_path)?;
let mut library = Library::new(config)?;
library.analyze_paths_extra_info(library.song_paths_info()?, true)?;
} else if let Some(sub_m) = matches.subcommand_matches("update") {
let config_path = sub_m.value_of("config-path").map(PathBuf::from);
let mut library: Library<Config> = Library::from_config_path(config_path)?;
library.update_library_extra_info(library.song_paths_info()?, true)?;
} else if let Some(sub_m) = matches.subcommand_matches("playlist") {
let song_path = sub_m.value_of("SONG_PATH").unwrap();
let config_path = sub_m.value_of("config-path").map(PathBuf::from);
let playlist_length = sub_m
.value_of("playlist-length")
.unwrap_or("20")
.parse::<usize>()?;
let library: Library<Config> = Library::from_config_path(config_path)?;
let songs = library.playlist_from::<ExtraInfo>(song_path, playlist_length)?;
let playlist = songs
.into_iter()
.map(|s| {
(
s.bliss_song.path.to_string_lossy().to_string(),
s.extra_info.mime_type,
)
})
.collect::<Vec<(String, String)>>();
for (path, mime_type) in playlist {
println!("{} <{}>", path, mime_type,);
}
}
Ok(())
}

View File

@ -1,26 +1,16 @@
#[cfg(feature = "serde")]
use anyhow::Result;
#[cfg(feature = "serde")]
use bliss_audio::playlist::{closest_to_first_song, dedup_playlist, euclidean_distance};
#[cfg(feature = "serde")]
use bliss_audio::{analyze_paths, Song};
#[cfg(feature = "serde")]
use clap::{App, Arg};
#[cfg(feature = "serde")]
use glob::glob;
#[cfg(feature = "serde")]
use std::env;
#[cfg(feature = "serde")]
use std::fs;
#[cfg(feature = "serde")]
use std::io::BufReader;
#[cfg(feature = "serde")]
use std::path::{Path, PathBuf};
/* Analyzes a folder recursively, and make a playlist out of the file
* provided by the user. */
// How to use: ./playlist [-o file.m3u] [-a analysis.json] <folder> <file to start the playlist from>
#[cfg(feature = "serde")]
fn main() -> Result<()> {
let matches = App::new("playlist")
.version(env!("CARGO_PKG_VERSION"))
@ -103,8 +93,3 @@ fn main() -> Result<()> {
}
Ok(())
}
#[cfg(not(feature = "serde"))]
fn main() {
println!("You need the serde feature enabled to run this file.");
}

View File

@ -131,7 +131,7 @@ impl BlissCueFile {
let song = Song {
path: PathBuf::from(format!(
"{}/CUE_TRACK{:03}",
self.audio_file_path.to_string_lossy(),
self.cue_path.to_string_lossy(),
index,
)),
album: self.album.to_owned(),
@ -210,7 +210,7 @@ mod tests {
let songs = BlissCue::songs_from_path("data/testcue.cue").unwrap();
let expected = vec![
Ok(Song {
path: Path::new("data/testcue.flac/CUE_TRACK001").to_path_buf(),
path: Path::new("data/testcue.cue/CUE_TRACK001").to_path_buf(),
analysis: Analysis {
internal_analysis: [
0.38463724,
@ -250,7 +250,7 @@ mod tests {
..Default::default()
}),
Ok(Song {
path: Path::new("data/testcue.flac/CUE_TRACK002").to_path_buf(),
path: Path::new("data/testcue.cue/CUE_TRACK002").to_path_buf(),
analysis: Analysis {
internal_analysis: [
0.18622077,
@ -290,7 +290,7 @@ mod tests {
..Default::default()
}),
Ok(Song {
path: Path::new("data/testcue.flac/CUE_TRACK003").to_path_buf(),
path: Path::new("data/testcue.cue/CUE_TRACK003").to_path_buf(),
analysis: Analysis {
internal_analysis: [
0.0024261475,

View File

@ -15,6 +15,14 @@
//! is as easy as computing distances between that song and the rest, and ordering
//! the songs by distance, ascending.
//!
//! If you want to implement a bliss plugin for an already existing audio
//! player, the [Library] struct is a collection of goodies that should prove
//! useful (it contains utilities to store analyzed songs in a self-contained
//! database file, to make playlists directly from the database, etc).
//! [blissify](https://github.com/Polochon-street/blissify-rs/) for both
//! an example of how the [Library] struct works, and a real-life demo of bliss
//! implemented for [MPD](https://www.musicpd.org/).
//!
//! # Examples
//!
//! ### Analyze & compute the distance between two songs
@ -59,6 +67,8 @@
#![warn(rustdoc::missing_doc_code_examples)]
mod chroma;
pub mod cue;
#[cfg(feature = "library")]
pub mod library;
mod misc;
pub mod playlist;
mod song;
@ -262,11 +272,11 @@ mod tests {
#[test]
fn test_analyze_paths() {
let paths = vec![
PathBuf::from("./data/s16_mono_22_5kHz.flac"),
PathBuf::from("./data/testcue.cue"),
PathBuf::from("./data/white_noise.flac"),
PathBuf::from("definitely-not-existing.foo"),
PathBuf::from("not-existing.foo"),
"./data/s16_mono_22_5kHz.flac",
"./data/testcue.cue",
"./data/white_noise.flac",
"definitely-not-existing.foo",
"not-existing.foo",
];
let mut results = analyze_paths(&paths)
.map(|x| match &x.1 {
@ -304,21 +314,9 @@ mod tests {
)),
),
(true, PathBuf::from("./data/s16_mono_22_5kHz.flac"), None),
(
true,
PathBuf::from("./data/testcue.flac/CUE_TRACK001"),
None,
),
(
true,
PathBuf::from("./data/testcue.flac/CUE_TRACK002"),
None,
),
(
true,
PathBuf::from("./data/testcue.flac/CUE_TRACK003"),
None,
),
(true, PathBuf::from("./data/testcue.cue/CUE_TRACK001"), None),
(true, PathBuf::from("./data/testcue.cue/CUE_TRACK002"), None),
(true, PathBuf::from("./data/testcue.cue/CUE_TRACK003"), None),
(true, PathBuf::from("./data/white_noise.flac"), None),
];

2850
src/library.rs Normal file

File diff suppressed because it is too large Load Diff

View File

@ -7,6 +7,8 @@
//! They will yield different styles of playlists, so don't hesitate to
//! experiment with them if the default (euclidean distance for now) doesn't
//! suit you.
// TODO on the `by_key` functions: maybe Fn(&T) -> &Song is enough? Compared
// to -> Song
use crate::{BlissError, BlissResult, Song, NUMBER_FEATURES};
use ndarray::{Array, Array1, Array2, Axis};
use ndarray_stats::QuantileExt;
@ -47,6 +49,23 @@ pub fn closest_to_first_song(
songs.sort_by_cached_key(|song| n32(first_song.custom_distance(song, &distance)));
}
/// Sort `songs` in place by putting songs close to `first_song` first
/// using the `distance` metric.
///
/// Sort songs with a key extraction function, useful for when you have a
/// structure like `CustomSong { bliss_song: Song, something_else: bool }`
pub fn closest_to_first_song_by_key<F, T>(
first_song: &T,
#[allow(clippy::ptr_arg)] songs: &mut Vec<T>,
distance: impl DistanceMetric,
key_fn: F,
) where
F: Fn(&T) -> Song,
{
let first_song = key_fn(first_song);
songs.sort_by_cached_key(|song| n32(first_song.custom_distance(&key_fn(song), &distance)));
}
/// Sort `songs` in place using the `distance` metric and ordering by
/// the smallest distance between each song.
///
@ -71,6 +90,43 @@ pub fn song_to_song(first_song: &Song, songs: &mut Vec<Song>, distance: impl Dis
*songs = new_songs;
}
/// Sort `songs` in place using the `distance` metric and ordering by
/// the smallest distance between each song.
///
/// If the generated playlist is `[song1, song2, song3, song4]`, it means
/// song2 is closest to song1, song3 is closest to song2, and song4 is closest
/// to song3.
///
/// Note that this has a tendency to go from one style to the other very fast,
/// and it can be slow on big libraries.
///
/// Sort songs with a key extraction function, useful for when you have a
/// structure like `CustomSong { bliss_song: Song, something_else: bool }`
// TODO: maybe Clone is not needed?
pub fn song_to_song_by_key<F, T: std::cmp::PartialEq + Clone>(
first_song: &T,
songs: &mut Vec<T>,
distance: impl DistanceMetric,
key_fn: F,
) where
F: Fn(&T) -> Song,
{
let mut new_songs: Vec<T> = Vec::with_capacity(songs.len());
let mut bliss_song = key_fn(&first_song.to_owned());
while !songs.is_empty() {
let distances: Array1<f32> = Array::from_shape_fn(songs.len(), |i| {
bliss_song.custom_distance(&key_fn(&songs[i]), &distance)
});
let idx = distances.argmin().unwrap();
let song = songs[idx].to_owned();
bliss_song = key_fn(&songs[idx]).to_owned();
new_songs.push(song.to_owned());
songs.retain(|s| s != &song);
}
*songs = new_songs;
}
/// Remove duplicate songs from a playlist, in place.
///
/// Two songs are considered duplicates if they either have the same,
@ -86,6 +142,29 @@ pub fn dedup_playlist(songs: &mut Vec<Song>, distance_threshold: Option<f32>) {
dedup_playlist_custom_distance(songs, distance_threshold, euclidean_distance);
}
/// Remove duplicate songs from a playlist, in place.
///
/// Two songs are considered duplicates if they either have the same,
/// non-empty title and artist name, or if they are close enough in terms
/// of distance.
///
/// Dedup songs with a key extraction function, useful for when you have a
/// structure like `CustomSong { bliss_song: Song, something_else: bool }` you
/// want to deduplicate.
///
/// # Arguments
///
/// * `songs`: The playlist to remove duplicates from.
/// * `distance_threshold`: The distance threshold under which two songs are
/// considered identical. If `None`, a default value of 0.05 will be used.
/// * `key_fn`: A function used to retrieve the bliss [Song] from `T`.
pub fn dedup_playlist_by_key<T, F>(songs: &mut Vec<T>, distance_threshold: Option<f32>, key_fn: F)
where
F: Fn(&T) -> Song,
{
dedup_playlist_custom_distance_by_key(songs, distance_threshold, euclidean_distance, key_fn);
}
/// Remove duplicate songs from a playlist, in place, using a custom distance
/// metric.
///
@ -115,6 +194,45 @@ pub fn dedup_playlist_custom_distance(
});
}
/// Remove duplicate songs from a playlist, in place, using a custom distance
/// metric.
///
/// Two songs are considered duplicates if they either have the same,
/// non-empty title and artist name, or if they are close enough in terms
/// of distance.
///
/// Dedup songs with a key extraction function, useful for when you have a
/// structure like `CustomSong { bliss_song: Song, something_else: bool }`
/// you want to deduplicate.
///
/// # Arguments
///
/// * `songs`: The playlist to remove duplicates from.
/// * `distance_threshold`: The distance threshold under which two songs are
/// considered identical. If `None`, a default value of 0.05 will be used.
/// * `distance`: A custom distance metric.
/// * `key_fn`: A function used to retrieve the bliss [Song] from `T`.
pub fn dedup_playlist_custom_distance_by_key<F, T>(
songs: &mut Vec<T>,
distance_threshold: Option<f32>,
distance: impl DistanceMetric,
key_fn: F,
) where
F: Fn(&T) -> Song,
{
songs.dedup_by(|s1, s2| {
let s1 = key_fn(s1);
let s2 = key_fn(s2);
n32(s1.custom_distance(&s2, &distance)) < distance_threshold.unwrap_or(0.05)
|| (s1.title.is_some()
&& s2.title.is_some()
&& s1.artist.is_some()
&& s2.artist.is_some()
&& s1.title == s2.title
&& s1.artist == s2.artist)
});
}
/// Return a list of albums in a `pool` of songs that are similar to
/// songs in `group`, discarding songs that don't belong to an album.
/// It basically makes an "album" playlist from the `pool` of songs.
@ -203,6 +321,114 @@ pub fn closest_album_to_group(group: Vec<Song>, pool: Vec<Song>) -> BlissResult<
Ok(playlist)
}
/// Return a list of albums in a `pool` of songs that are similar to
/// songs in `group`, discarding songs that don't belong to an album.
/// It basically makes an "album" playlist from the `pool` of songs.
///
/// `group` should be ordered by track number.
///
/// Songs from `group` would usually just be songs from an album, but not
/// necessarily - they are discarded from `pool` no matter what.
///
/// Order songs with a key extraction function, useful for when you have a
/// structure like `CustomSong { bliss_song: Song, something_else: bool }`
/// you want to order.
///
/// # Arguments
///
/// * `group` - A small group of songs, e.g. an album.
/// * `pool` - A pool of songs to find similar songs in, e.g. a user's song
/// library.
/// * `key_fn`: A function used to retrieve the bliss [Song] from `T`.
///
/// # Returns
///
/// A vector of T, including `group` at the beginning, that you
/// most likely want to plug in your audio player by using something like
/// `ret.map(|song| song.path.to_owned()).collect::<Vec<String>>()`.
// TODO: maybe Clone is not needed?
pub fn closest_album_to_group_by_key<T: PartialEq + Clone, F>(
group: Vec<T>,
pool: Vec<T>,
key_fn: F,
) -> BlissResult<Vec<T>>
where
F: Fn(&T) -> Song,
{
let mut albums_analysis: HashMap<String, Array2<f32>> = HashMap::new();
let mut albums = Vec::new();
// Remove songs from the group from the pool.
let pool = pool
.into_iter()
.filter(|s| !group.contains(s))
.collect::<Vec<_>>();
for song in &pool {
let song = key_fn(song);
if let Some(album) = song.album {
if let Some(analysis) = albums_analysis.get_mut(&album as &str) {
analysis
.push_row(song.analysis.as_arr1().view())
.map_err(|e| {
BlissError::ProviderError(format!("while computing distances: {}", e))
})?;
} else {
let mut array = Array::zeros((1, song.analysis.as_arr1().len()));
array.assign(&song.analysis.as_arr1());
albums_analysis.insert(album.to_owned(), array);
}
}
}
let mut group_analysis = Array::zeros((group.len(), NUMBER_FEATURES));
for (song, mut column) in group.iter().zip(group_analysis.axis_iter_mut(Axis(0))) {
let song = key_fn(song);
column.assign(&song.analysis.as_arr1());
}
let first_analysis = group_analysis
.mean_axis(Axis(0))
.ok_or_else(|| BlissError::ProviderError(String::from("Mean of empty slice")))?;
for (album, analysis) in albums_analysis.iter() {
let mean_analysis = analysis
.mean_axis(Axis(0))
.ok_or_else(|| BlissError::ProviderError(String::from("Mean of empty slice")))?;
let album = album.to_owned();
albums.push((album, mean_analysis.to_owned()));
}
albums.sort_by_key(|(_, analysis)| n32(euclidean_distance(&first_analysis, analysis)));
let mut playlist = group;
for (album, _) in albums {
let mut al = pool
.iter()
.filter(|s| {
let s = key_fn(s);
s.album.is_some() && s.album.as_ref().unwrap() == &album.to_string()
})
.map(|s| s.to_owned())
.collect::<Vec<T>>();
al.sort_by(|s1, s2| {
let s1 = key_fn(s1);
let s2 = key_fn(s2);
let track_number1 = s1
.track_number
.to_owned()
.unwrap_or_else(|| String::from(""));
let track_number2 = s2
.track_number
.to_owned()
.unwrap_or_else(|| String::from(""));
if let Ok(x) = track_number1.parse::<i32>() {
if let Ok(y) = track_number2.parse::<i32>() {
return x.cmp(&y);
}
}
s1.track_number.cmp(&s2.track_number)
});
playlist.extend_from_slice(&al);
}
Ok(playlist)
}
#[cfg(test)]
mod test {
use super::*;
@ -210,6 +436,12 @@ mod test {
use ndarray::arr1;
use std::path::Path;
#[derive(Debug, Clone, PartialEq)]
struct CustomSong {
something: bool,
bliss_song: Song,
}
#[test]
fn test_dedup_playlist_custom_distance() {
let first_song = Song {
@ -316,6 +548,91 @@ mod test {
fourth_song.to_owned(),
]
);
let first_song = CustomSong {
bliss_song: first_song,
something: true,
};
let second_song = CustomSong {
bliss_song: second_song,
something: true,
};
let first_song_dupe = CustomSong {
bliss_song: first_song_dupe,
something: true,
};
let third_song = CustomSong {
bliss_song: third_song,
something: true,
};
let fourth_song = CustomSong {
bliss_song: fourth_song,
something: true,
};
let fifth_song = CustomSong {
bliss_song: fifth_song,
something: true,
};
let mut playlist = vec![
first_song.to_owned(),
first_song_dupe.to_owned(),
second_song.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
fifth_song.to_owned(),
];
dedup_playlist_custom_distance_by_key(&mut playlist, None, euclidean_distance, |s| {
s.bliss_song.to_owned()
});
assert_eq!(
playlist,
vec![
first_song.to_owned(),
second_song.to_owned(),
fourth_song.to_owned(),
],
);
let mut playlist = vec![
first_song.to_owned(),
first_song_dupe.to_owned(),
second_song.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
fifth_song.to_owned(),
];
dedup_playlist_custom_distance_by_key(&mut playlist, Some(20.), cosine_distance, |s| {
s.bliss_song.to_owned()
});
assert_eq!(playlist, vec![first_song.to_owned()]);
let mut playlist = vec![
first_song.to_owned(),
first_song_dupe.to_owned(),
second_song.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
fifth_song.to_owned(),
];
dedup_playlist_by_key(&mut playlist, Some(20.), |s| s.bliss_song.to_owned());
assert_eq!(playlist, vec![first_song.to_owned()]);
let mut playlist = vec![
first_song.to_owned(),
first_song_dupe.to_owned(),
second_song.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
fifth_song.to_owned(),
];
dedup_playlist_by_key(&mut playlist, None, |s| s.bliss_song.to_owned());
assert_eq!(
playlist,
vec![
first_song.to_owned(),
second_song.to_owned(),
fourth_song.to_owned(),
]
);
}
#[test]
@ -358,20 +675,64 @@ mod test {
};
let mut songs = vec![
first_song.to_owned(),
third_song.to_owned(),
first_song_dupe.to_owned(),
second_song.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
];
song_to_song(&first_song, &mut songs, euclidean_distance);
assert_eq!(
songs,
vec![
first_song,
first_song.to_owned(),
first_song_dupe.to_owned(),
second_song.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
],
);
let first_song = CustomSong {
bliss_song: first_song,
something: true,
};
let second_song = CustomSong {
bliss_song: second_song,
something: true,
};
let first_song_dupe = CustomSong {
bliss_song: first_song_dupe,
something: true,
};
let third_song = CustomSong {
bliss_song: third_song,
something: true,
};
let fourth_song = CustomSong {
bliss_song: fourth_song,
something: true,
};
let mut songs: Vec<CustomSong> = vec![
first_song.to_owned(),
first_song_dupe.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
second_song.to_owned(),
];
song_to_song_by_key(&first_song, &mut songs, euclidean_distance, |s| {
s.bliss_song.to_owned()
});
assert_eq!(
songs,
vec![
first_song,
first_song_dupe,
second_song,
third_song,
fourth_song
fourth_song,
],
);
}
@ -431,6 +792,46 @@ mod test {
fifth_song.to_owned(),
];
closest_to_first_song(&first_song, &mut songs, euclidean_distance);
let first_song = CustomSong {
bliss_song: first_song,
something: true,
};
let second_song = CustomSong {
bliss_song: second_song,
something: true,
};
let first_song_dupe = CustomSong {
bliss_song: first_song_dupe,
something: true,
};
let third_song = CustomSong {
bliss_song: third_song,
something: true,
};
let fourth_song = CustomSong {
bliss_song: fourth_song,
something: true,
};
let fifth_song = CustomSong {
bliss_song: fifth_song,
something: true,
};
let mut songs: Vec<CustomSong> = vec![
first_song.to_owned(),
first_song_dupe.to_owned(),
second_song.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
fifth_song.to_owned(),
];
closest_to_first_song_by_key(&first_song, &mut songs, euclidean_distance, |s| {
s.bliss_song.to_owned()
});
assert_eq!(
songs,
vec![
@ -538,5 +939,46 @@ mod test {
],
closest_album_to_group(group, pool.to_owned()).unwrap(),
);
let first_song = CustomSong {
bliss_song: first_song,
something: true,
};
let second_song = CustomSong {
bliss_song: second_song,
something: true,
};
let third_song = CustomSong {
bliss_song: third_song,
something: true,
};
let fourth_song = CustomSong {
bliss_song: fourth_song,
something: true,
};
let fifth_song = CustomSong {
bliss_song: fifth_song,
something: true,
};
let pool = vec![
first_song.to_owned(),
fourth_song.to_owned(),
third_song.to_owned(),
second_song.to_owned(),
fifth_song.to_owned(),
];
let group = vec![first_song.to_owned(), third_song.to_owned()];
assert_eq!(
vec![
first_song.to_owned(),
third_song.to_owned(),
fourth_song.to_owned(),
second_song.to_owned()
],
closest_album_to_group_by_key(group, pool.to_owned(), |s| s.bliss_song.to_owned())
.unwrap(),
);
}
}

View File

@ -62,6 +62,7 @@ pub struct Song {
/// Song's album's artist name, read from the metadata
pub album_artist: Option<String>,
/// Song's tracked number, read from the metadata
/// TODO normalize this into an integer
pub track_number: Option<String>,
/// Song's genre, read from the metadata (`""` if empty)
pub genre: Option<String>,