Files
Airplay-Predictor/src/playlists.rs
T

230 lines
14 KiB
Rust

//! Spotify editorial playlists:
//!
//! - Snapshots, from now on: the full track list of New Music Friday Denmark and Top 50 Denmark,
//! read from Spotify's public embed page (open.spotify.com/embed/playlist/{id}, no login).
//! Daily runs store a snapshot whenever the list changed.
//! - History, since January 2022: SuperFridayChart (superfridaychart.com), which records every
//! country's New Music Friday additions. Looked up per song by ISRC (from Deezer):
//! /track/{isrc} lists each Friday the song was added, with country, playlist, followers, position.
use crate::{db::Db, http::Http};
use anyhow::{anyhow, Result};
use chrono::{NaiveDate, Utc};
use regex::Regex;
use rusqlite::params;
use serde_json::Value;
pub const PLAYLISTS: &[(&str, &str)] = &[
("37i9dQZF1DWWuIAAkPM0Lu", "New Music Friday Denmark"),
("37i9dQZEVXbL3J0k32lWnN", "Top 50 - Denmark"),
];
pub fn create_tables(db: &Db) -> Result<()> {
db.conn.execute_batch(
"CREATE TABLE IF NOT EXISTS playlist_snapshots (
playlist_id TEXT NOT NULL, name TEXT NOT NULL, taken_at TEXT NOT NULL,
position INTEGER NOT NULL, track_id TEXT NOT NULL, title TEXT, artists TEXT, duration_ms INTEGER,
PRIMARY KEY (playlist_id, taken_at, position)
);
CREATE TABLE IF NOT EXISTS nmf_additions ( -- SuperFridayChart
isrc TEXT NOT NULL, friday TEXT NOT NULL, country TEXT NOT NULL,
playlist_id TEXT, followers INTEGER, position INTEGER,
PRIMARY KEY (isrc, friday, country)
);
CREATE TABLE IF NOT EXISTS nmf_lookups (
isrc TEXT PRIMARY KEY, found INTEGER NOT NULL, fetched_at TEXT NOT NULL
);",
)?;
Ok(())
}
/// A playlist's tracks in order: (track id, title, artists, duration ms).
fn read_embed(http: &Http, id: &str) -> Result<Vec<(String, String, String, Option<i64>)>> {
let html = http.get(&format!("https://open.spotify.com/embed/playlist/{id}"))?.ok_or_else(|| anyhow!("playlist {id} not found"))?;
let re = Regex::new(r#"(?s)<script id="__NEXT_DATA__"[^>]*>(.*?)</script>"#)?;
let data: Value = serde_json::from_str(&re.captures(&html).ok_or_else(|| anyhow!("no data in the embed page"))?[1])?;
let list = data["props"]["pageProps"]["state"]["data"]["entity"]["trackList"].as_array().cloned().unwrap_or_default();
Ok(list.iter().map(|t| (
t["uri"].as_str().unwrap_or_default().rsplit(':').next().unwrap_or_default().to_string(),
t["title"].as_str().unwrap_or_default().to_string(),
t["subtitle"].as_str().unwrap_or_default().replace('\u{a0}', " "),
t["duration"].as_i64(),
)).collect())
}
/// Stores a snapshot of each playlist if it differs from the last one stored.
pub fn snapshot(db: &Db, http: &Http, log: &mut Vec<String>) -> Result<usize> {
create_tables(db)?;
let mut stored = 0;
for &(id, name) in PLAYLISTS {
let tracks = match read_embed(http, id) {
Ok(t) if !t.is_empty() => t,
Ok(_) => { log.push(format!("playlist {name}: empty")); continue; }
Err(e) => { log.push(format!("playlist {name}: {e}")); continue; }
};
let last: Vec<String> = db.conn.prepare(
"SELECT track_id FROM playlist_snapshots WHERE playlist_id = ?1
AND taken_at = (SELECT max(taken_at) FROM playlist_snapshots WHERE playlist_id = ?1) ORDER BY position")?
.query_map([id], |r| r.get(0))?.collect::<Result<_, _>>()?;
if last == tracks.iter().map(|t| t.0.clone()).collect::<Vec<_>>() { continue; }
let now = Utc::now().format("%Y-%m-%dT%H:%M:%SZ").to_string();
let tx = db.conn.unchecked_transaction()?;
for (i, (track, title, artists, dur)) in tracks.iter().enumerate() {
tx.execute("INSERT OR REPLACE INTO playlist_snapshots VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8)",
params![id, name, now, i as i64 + 1, track, title, artists, dur])?;
}
tx.commit()?;
stored += 1;
}
Ok(stored)
}
/// New Music Friday additions per ISRC from SuperFridayChart (every ISRC Deezer gave us, once).
pub fn nmf_history(db: &Db, http: &Http, limit: usize, log: &mut Vec<String>) -> Result<usize> {
create_tables(db)?;
crate::deezer::create_tables(db)?;
db.conn.execute_batch("CREATE TABLE IF NOT EXISTS nmf_artist_tracks (spotify_artist_id TEXT NOT NULL, isrc TEXT NOT NULL, title TEXT,
friday TEXT NOT NULL, playlists INTEGER, followers INTEGER, PRIMARY KEY (spotify_artist_id, isrc, friday));")?;
// Title and artists of each looked-up track (added later; older rows have them empty).
for col in ["title TEXT", "artists TEXT"] {
let _ = db.conn.execute(&format!("ALTER TABLE nmf_lookups ADD COLUMN {col}"), []);
}
db.conn.execute_batch("CREATE TABLE IF NOT EXISTS nmf_global_chart (friday TEXT NOT NULL, rank INTEGER NOT NULL, isrc TEXT NOT NULL,
PRIMARY KEY (friday, rank));")?;
// Which tracks to read, most likely to be on NMF Denmark first:
// 0 artist-page placements of Danish acts (DR artist country), or one playlist the size of
// NMF Denmark that Friday (the most common single-playlist follower count among Danish acts)
// 1 SuperFridayChart's weekly global top 50
// 2 other artist-page placements on several playlists
// 3 every ISRC Deezer gave us
let todo: Vec<String> = db.conn.prepare(
"WITH danish AS (SELECT l.spotify_artist_id FROM nmf_artist_lookups l JOIN dr_artists a ON a.artist_id = l.dr_artist_id
WHERE json_extract(a.raw, '$.artist.country') = 'Danmark' AND l.spotify_artist_id IS NOT NULL),
dk_size AS (SELECT friday, followers FROM (
SELECT t.friday, t.followers, count(*) n, row_number() OVER (PARTITION BY t.friday ORDER BY count(*) DESC) rn
FROM nmf_artist_tracks t WHERE t.playlists = 1 AND t.spotify_artist_id IN (SELECT spotify_artist_id FROM danish)
GROUP BY t.friday, t.followers) WHERE rn = 1)
SELECT isrc FROM (
SELECT t.isrc, 0 AS o FROM nmf_artist_tracks t LEFT JOIN dk_size d ON d.friday = t.friday
WHERE t.spotify_artist_id IN (SELECT spotify_artist_id FROM danish)
OR (t.playlists = 1 AND abs(t.followers - d.followers) <= 0.02 * d.followers)
UNION ALL SELECT isrc, 1 FROM nmf_global_chart
UNION ALL SELECT isrc, 2 FROM nmf_artist_tracks WHERE playlists > 1
UNION ALL SELECT isrc, 3 FROM deezer_tracks WHERE coalesce(isrc, '') != '')
WHERE isrc NOT IN (SELECT isrc FROM nmf_lookups) GROUP BY isrc ORDER BY min(o)")?
.query_map([], |r| r.get(0))?.collect::<Result<_, _>>()?;
let meta = Regex::new(r#"<meta name="description" content="Discover the Spotify New Music Friday playlists that added the track &quot;(.*?)&quot; from (.*?)\.">"#)?;
let day = Regex::new(r#"<h2[^>]*>\s*(?:Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday), ([A-Za-z]+ \d{2}, \d{4})\s*</h2>"#)?;
let row = Regex::new(r#"(?s)<p[^>]*>([^<]+)</p>\s*<a href="https://open\.spotify\.com/playlist/([A-Za-z0-9]{22})".*?<span[^>]*>([\d,]+) followers</span>\s*<span[^>]*>Position (\d+)</span>"#)?;
let mut done = 0;
for isrc in todo.into_iter().take(limit) {
let html = match http.get(&format!("https://www.superfridaychart.com/track/{isrc}")) {
Ok(h) => h,
Err(e) => { log.push(format!("superfridaychart {isrc}: {e}")); continue; }
};
let mut found = 0;
let (mut title, mut artists) = (None, None);
if let Some(html) = html {
if let Some(m) = meta.captures(&html) {
let un = |x: &str| x.replace("&amp;", "&").replace("&#039;", "'").replace("&quot;", "\"");
title = Some(un(&m[1]));
artists = Some(un(&m[2]));
}
// Each Friday is an <h2> date followed by its table of playlists.
let heads: Vec<(usize, NaiveDate)> = day.captures_iter(&html)
.filter_map(|c| Some((c.get(0)?.start(), NaiveDate::parse_from_str(&c[1], "%B %d, %Y").ok()?)))
.collect();
for (k, (start, friday)) in heads.iter().enumerate() {
let end = heads.get(k + 1).map(|h| h.0).unwrap_or(html.len());
for r in row.captures_iter(&html[*start..end]) {
db.conn.execute("INSERT OR REPLACE INTO nmf_additions VALUES (?1, ?2, ?3, ?4, ?5, ?6)",
params![isrc, friday.to_string(), r[1].trim(), &r[2], r[3].replace(',', "").parse::<i64>().ok(), r[4].parse::<i64>().ok()])?;
found += 1;
}
}
}
db.conn.execute("INSERT OR REPLACE INTO nmf_lookups (isrc, found, fetched_at, title, artists) VALUES (?1, ?2, ?3, ?4, ?5)",
params![isrc, found, Utc::now().to_rfc3339(), title, artists])?;
done += 1;
}
Ok(done)
}
/// New Music Friday placements per artist (SuperFridayChart's artist page, by the Spotify artist
/// id DR links on its artist page): every track of the artist added to any NMF playlist since
/// 2022, with the Friday, number of playlists and their followers. Complete per artist, so songs
/// DR never played are in it too. Which countries: the track page (`nmf_history`), fetched for
/// placements inside the modelling window.
pub fn nmf_artists(db: &Db, http: &Http, limit: usize, log: &mut Vec<String>) -> Result<usize> {
create_tables(db)?;
db.conn.execute_batch(
"CREATE TABLE IF NOT EXISTS nmf_artist_tracks (
spotify_artist_id TEXT NOT NULL, isrc TEXT NOT NULL, title TEXT, friday TEXT NOT NULL,
playlists INTEGER, followers INTEGER,
PRIMARY KEY (spotify_artist_id, isrc, friday)
);
CREATE TABLE IF NOT EXISTS nmf_artist_lookups (
dr_artist_id TEXT PRIMARY KEY, spotify_artist_id TEXT, tracks INTEGER, fetched_at TEXT NOT NULL
);",
)?;
let link = Regex::new(r"open\.spotify\.com/artist/([A-Za-z0-9]{22})")?;
let todo: Vec<(String, String)> = db.conn.prepare(
"SELECT artist_id, raw FROM dr_artists WHERE artist_id NOT IN (SELECT dr_artist_id FROM nmf_artist_lookups)")?
.query_map([], |r| Ok((r.get(0)?, r.get(1)?)))?.collect::<Result<_, _>>()?;
// One row per track: the track link, then its date, playlist count and followers.
let row = Regex::new(r#"(?s)/track/([A-Z0-9]{12})"[^>]*>([^<]*)</a>.*?([A-Z][a-z]{2} \d{2}, \d{4})\s*</?[^>]*>?.*?(\d+) playlists.*?([\d,]+) followers"#)?;
let mut done = 0;
for (dr_id, raw) in todo.into_iter().take(limit) {
let sid = link.captures(&raw).map(|c| c[1].to_string());
let mut n = 0;
if let Some(sid) = &sid {
match http.get(&format!("https://www.superfridaychart.com/artist/{sid}")) {
Ok(Some(html)) => {
// Split at each track link so a row's fields can't run into the next track.
let starts: Vec<usize> = html.match_indices("superfridaychart.com/track/").map(|(i, _)| i).collect();
for (k, s) in starts.iter().enumerate() {
let block = &html[s.saturating_sub(1)..starts.get(k + 1).copied().unwrap_or(html.len())];
let Some(c) = row.captures(block) else { continue };
let Ok(d) = NaiveDate::parse_from_str(&c[3], "%b %d, %Y") else { continue };
db.conn.execute("INSERT OR REPLACE INTO nmf_artist_tracks VALUES (?1, ?2, ?3, ?4, ?5, ?6)",
params![sid, &c[1], c[2].trim().replace("&amp;", "&").replace("&#039;", "'"), d.to_string(),
c[4].parse::<i64>().ok(), c[5].replace(',', "").parse::<i64>().ok()])?;
n += 1;
}
}
Ok(None) => {}
Err(e) => { log.push(format!("superfridaychart artist {sid}: {e}")); continue; }
}
}
db.conn.execute("INSERT OR REPLACE INTO nmf_artist_lookups VALUES (?1, ?2, ?3, ?4)", params![dr_id, sid, n as i64, Utc::now().to_rfc3339()])?;
done += 1;
}
Ok(done)
}
/// SuperFridayChart's weekly global chart (top 50 tracks across all NMF playlists), every Friday
/// listed on its front page: candidate ISRCs for the NMF Denmark lists.
pub fn nmf_global_charts(db: &Db, http: &Http, log: &mut Vec<String>) -> Result<usize> {
db.conn.execute_batch("CREATE TABLE IF NOT EXISTS nmf_global_chart (friday TEXT NOT NULL, rank INTEGER NOT NULL, isrc TEXT NOT NULL,
PRIMARY KEY (friday, rank));")?;
let front = http.get("https://www.superfridaychart.com/")?.unwrap_or_default();
let mut fridays: Vec<String> = Regex::new(r"/chart/(\d{4}-\d{2}-\d{2})")?.captures_iter(&front).map(|c| c[1].to_string()).collect();
fridays.sort(); fridays.dedup();
let track = Regex::new(r"/track/([A-Z0-9]{12})")?;
let mut n = 0;
for f in fridays {
let have: i64 = db.conn.query_row("SELECT count(*) FROM nmf_global_chart WHERE friday = ?1", [&f], |r| r.get(0))?;
if have > 0 { continue; }
let html = match http.get(&format!("https://www.superfridaychart.com/chart/{f}")) {
Ok(Some(h)) => h, Ok(None) => continue, Err(e) => { log.push(format!("superfridaychart chart {f}: {e}")); continue; }
};
let mut seen = Vec::new();
for c in track.captures_iter(&html) { if !seen.contains(&c[1].to_string()) { seen.push(c[1].to_string()); } }
for (i, isrc) in seen.iter().enumerate() {
db.conn.execute("INSERT OR REPLACE INTO nmf_global_chart VALUES (?1, ?2, ?3)", params![f, i as i64 + 1, isrc])?;
}
n += 1;
}
Ok(n)
}